Research ArticleEffect of Data Assimilation Using WRF-3DVAR for Heavy RainPrediction on the Northeastern Edge of the Tibetan Plateau

Junhua Yang1 Keqin Duan12 Jinkui Wu1 Xiang Qin1 Peihong Shi1

Huancai Liu1 Xiaolong Xie3 Xiao Zhang1 and Jianyong Sun1

1State Key Laboratory of Cryospheric Sciences Cold and Arid Regions Environmental and Engineering Research InstituteChinese Academy of Sciences Lanzhou 730000 China2Tourism and Environmental College Shanxi Normal University Xian 710000 China3Resource and Environment College of Lanzhou University Lanzhou Gansu 730000 China

Correspondence should be addressed to Keqin Duan kqduanlzbaccn

Received 25 June 2014 Revised 17 September 2014 Accepted 22 September 2014

Academic Editor Eduardo Garcıa-Ortega

Copyright copy 2015 Junhua Yang et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

The numerical weather prediction (NWP) is gaining more attention in providing high-resolution rainfall forecasts in the arid andsemiarid region However the modeling accuracy is negatively affected by errors in the initial conditions Here we investigatethe potential of data assimilation in improving the NWP rainfall forecasts in the northeastern Tibetan Plateau Three of three-dimensional variational (3DVar) data assimilation experiments were designed on running the advanced research weather researchforecast (WRF) model Two heavy rain events selected with different rainfall distribution in space and time are utilized to examinethe improvement for rainfall forecast after data assimilation For the spatial distribution the improvement of rainfall accumulationand area is obvious for the both two events But for the temporal variation the improvement is more obvious for the event witheven rainfall distribution in time while the effect of data assimilation is not ideal for the rainfall event with uneven distributionin space and time It is noteworthy that for both the spatial and temporal distribution of rainfall satellite radiances have greatereffect on rainfall forecasts than surface and upper-air meteorological observations in this high-altitude region Moreover the dataassimilation experiments provide more detail information to the initial fields

1 Introduction

Precipitation is a crucial component in the hydrological cycleof the Earth and has a profound influence on climate andhydrology at regional to global scales [1] A high-resolutionheavy rainfall forecast plays an important role in accurateflood forecasting andwater resource regulation [2] especiallyin the arid and semiarid region of northwest China whereprecipitation is crucial for water resources and the heavy rainoften results in flash flood in summer Therefore providingaccurate rainfall forecasts over northwest China using thenumerical weather prediction (NWP) is of prime importanceWeather research and forecast (WRF) is the latest generationmesoscale NWP model Recent studies have shown thatthe WRF model has good potential in capturing rainfallfeatures such as rainfall timing location and evolution [3ndash5]

However for producing accurate values for rainfall quantitiesthe results are not ideal due to the low-quality initial condi-tions [6] which can be improved by data assimilation [7]

With the improvement of the NWP models severaldata assimilation skills such as three and four-dimensionalvariational methods (3DVar4DVar) ensemble Kalman filers(EnKF) and latent heat nudging (LHN) have been developedin the methods of variational and ensemble [8 9] Althoughthe 4D-Var and EnKF methods show great potential theystill suffer from unaffordable computer costs for operationalNWPs In continuous cycling mode 3DVar performs betterin producing rational analyses of hydrometeorological fieldswith greater computational efficiency than 4DVar EnKF andLHN [10 11]

Real-time observations have generally been used inthe assimilation systems and shown to improve markedly

Hindawi Publishing CorporationAdvances in MeteorologyVolume 2015 Article ID 294589 14 pageshttpdxdoiorg1011552015294589

2 Advances in Meteorology

WPS domain configuration

46∘N

44∘N

42∘N

40∘N

38∘N

36∘N

34∘N

32∘N

30∘N

90∘E 95∘E 100∘E 105∘E 110∘E

d02

d01

Gansu Province

Qinghai Province

Figure 1 Domains for the ARW-WRF forecasts The outer box is the coarse grid with a resolution of 12 km (d01) the inner box is the nestedgrid (d02) with a resolution of 4 km

the performance of NWP models in spite of the poor initialconditions provided by the global NWP models [12] In thecase of high temporal and spatial resolution satellite datathe European Center for Medium-Range Weather Forecasts(ECMWF) pioneered the direct assimilation of microwaveradiance data affected by precipitation first in a 1 + 4Dvarassimilation approach [13 14] and later in an implementationof all-sky radiance assimilation in the operational 4Dvarsystem [15 16] Additionally many investigations have shownthat rainfall forecasts from theNWPmodels can be improvednoticeably with the assimilation of radar reflectivity [17ndash19]or radar-derived precipitation data [20 21] Nevertheless theremote sensing data need to be validated against ground truth[22] and surface observations have a wealth of informationthat can be used to simulate mesoscale weather phenomena[23] Therefore the assimilation of surface observations in aNWP model is likely to improve the model performance aswell Previous studies have beenmade to improve simulationsof weather parameters by using direct or modeled surfaceobservations These include the assimilation of tempera-ture water vapor mixing ratio and winds [24] and 2mpotential temperature 2m dew point temperature and 10mwind observations [25] into the NWP model to determineplanetary boundary layer (PBL) profiles and to analyse thesurface cold pool respectively The results show a markedimprovement in the model simulations after the assimilationIn the WRF assimilation system previous research hasalso shown that by assimilating only the satellite data theimprovement in precipitation forecasts is not as great as whenthe assimilation of satellite data is combined with that ofsurface observations [26 27]

In this study the WRF-3Dvar system is used to explorethe effect of the assimilation of NCAR surface and upper-air observations and of AMSU-A and AMSU-B microwaveradiance data on forecasts of heavy precipitation on thenortheastern Tibetan Plateau Two heavy rain events thatoccurred in June 2013 were selected in order to evaluate

the improvement for rainfall forecast after data assimilationThe paper is organized as follows Section 2 gives a conciseoverview of the method and experiment design Section 3provides information on the study area and the data Theresults of the data assimilation experiments are presented andevaluated in Section 4 and conclusions are given in Section 5

2 Method and Experimental Design

21 WRF Model Set-Up The numerical data assimilationexperiments in this study are conducted using the AdvancedResearch WRF model Version 35 WRF is a nonhydrostaticprimitive-equation mesoscale meteorological model withadvanced dynamics physics and numerical schemes (detailsof the model is at httpwwwmmmucaredu) As shown inFigure 1 the model domains are two-way nested with 12 km(208 times 182) and 4 km (235 times 182) horizontal spacing Eachdomain has 28 vertical pressure levels with the top levelset at 50 hPa The WRF physical parameterization schemesused in this study include the Purdue Lin microphysicalparameterization Rapid Radiative Transfer Model (RRTM)longwave radiation Dudhia shortwave radiation Monin-Obukhov surface layer Noah land surface Mellor-Yamada-Janjic (MYJ) planetary boundary layer scheme and Grell-Devenyi (GD) cumulus scheme The projection method isLambert

22 3D-Var Data Assimilation Data assimilation is the tech-nique by which observations are combined with a NWPproduct (called the first guess or background field) andtheir respective error statistics to provide an improved esti-mate (the analysis) of the atmospheric (or oceanic) stateVariational (Var) data assimilation achieves this through

Advances in Meteorology 3

the iterative minimization of a prescribed cost (or penalty)function [28]

119869 (119909) =1

2(119909 minus 119909

119887)119879

119861minus1(119909 minus 119909

119887)

+1

2(119910 minus 119910

0)119879

119877minus1(119910 minus 119910

0)

(1)

where 119909 is the analysis to be found that minimizes thecost function 119869(119909) 119909119887 is the first guess of the NWP model1199100 is the assimilated observation and 119910 = 119867(119909) is the

model-derived observation transformed from the analysis 119909by the observation operator 119867 for comparison against 1199100The solution for the cost function given by (1) representsa posteriori maximum likelihood (minimum variance) esti-mate of the true state given the two sources of a priori datathe first guess 119909119887 and the observation 1199100 [29] The fit toindividual observation points is weighted by the estimatesof their errors that is 119861 and 119877 which are the backgrounderror covariance matrix and the observation error covariancematrix respectively

The WRF-3DVar system developed by Barker et al [10]is used in this study in tandem with the WRF model forassimilating the satellite radiance data and the traditionalobservations The performance of the data-assimilation sys-tem largely depends on the plausibility of the backgrounderror covariance (BE) that is thematrix119861 in (1) In this studythe ldquoCV5rdquo background error option is used with the controlvariables of stream function unbalanced temperature unbal-anced potential velocity unbalanced surface pressure andpseudo relative humidity The background error covariancematrix is generated via the National Meteorological Center(NMC) method [30] for our own forecasting domain

23 Experimental Design Four sets of experiments have beenconducted using two domains As shown in Table 1 themodel simulation without data assimilation will be referredto as the control experiment (CTRL) Three assimilationexperiments are designed using different observation param-eters from different data sources In the DA-OBS experimentmeasurements of pressure geopotential height temperaturedewpoint temperature wind direction and speed fromNCARare assimilated while in theDA-SAT experiment onlysatellite radiance data are assimilated Finally in the DA-BOTH experiment both NCAR observations and AMSU-A(B) radiance data are assimilated

3 Study Area and Data

As shown in Figure 2 the northeastern of the TibetanPlateau (94∘391015840ndash103∘271015840E 35∘511015840ndash40∘311015840N) range of elevationbetween 758m and 5725m asl is the head of many inlandrivers which play an important role in the hydrology andagriculture in the downstream arid region A total of 43national observation stations have been used to verify thespatial distribution of the simulated precipitation amongwhich the Wulan station has the highest elevation of 3800mandDunhuang has the lowest altitude of 1137mTheobserved

Table 1 Details of the observed data used in the assimilationexperiments

Experiment name Assimilated dataCTRL No data assimilation

DA-OBS NCAR surface and upper-airobservation

DA-SAT AMSU-A and AMSU-B radiancedata

DA-BOTH NCAR observation andAMSU-A (B) radiance data

precipitation at Laohugou station (elevation 4200m) hasbeen used to evaluate the temporal variations of the simulatedrainfall which is measured by T200B every half hour

The initial and boundary conditions necessary to run theWRF are the ERA-Interim data at 1∘ times 1∘ grid resolutionobtained from the European Centre for Medium-RangeWeather Forecasts (ECMWF) rather than the NCEP-NCARFinal Analysis (FNL) data This is because several studieshave demonstrated that the reliability of ERA-Reanalysis datais higher than the NCEP data in China [31 32] In themodel integration the coordinates of the central point are381∘N and 989∘E WRF-3DVar experiments have been con-ducted with modified initial conditions which were obtainedby assimilating other measured data The assimilated dataincludes NCAR surface and upper-air observations andAMSUA and AMSUB radiance data The surface and upper-air data assimilated in this study are obtained from theldquods3370rdquo in the NCAR archives which contain measure-ments of pressure geopotential height temperature dewpoint temperature wind direction and speed from fixed andmobile landsea stations The data are initially downloadedin PREPBUFR format and can be assimilated directly intoWRFDA The AMSU-A and AMSU-B satellite radiance data(NOAA-1516171819) from theNOAAATOVS instrumentscan be read in WRFDA via CRTM202 which is in BUFRformat

4 Results and Discussion

41 Impacts of Data Assimilation on the Precipitation Forecast

411 Rainfall Event In June 2013 two heavy rain eventsoccurred in the northeast of the Tibetan Plateau Thedurations of the events and the maximummean rainfallaccumulation observed by rain gauge network are shownin Table 2 The two events are of different types accordingto the evenness of rainfall distribution in time and spaceFigure 3 illustrates the spatial distribution of the rainfallaccumulation for the durations of the two events whileFigure 4 presents the time series bars and the cumulativecurves of the observed precipitation for the two events atLaohugou station By comparing the evenness of the rainfalldistribution in time and space in Figures 3 and 4 it can befound that Event A has even rainfall distribution neither inspace nor in time The rainfall distribution of Event B is alsouneven in space but continuous and almost constant in time

4 Advances in Meteorology

Table 2 Durations maximummean rainfall accumulation and spatialtemporal evenness of the heavy rain events

Event ID Start time End time Spatial evenness Temporal evenness Maximum precipitation Mean precipitationA 1862013 000 1962013 2400 146 144 425 67B 762013 000 762013 2400 115 063 235 53

LaohugouRain-gauge stations

River

94∘E

40∘N

38∘N

36∘N

96∘E 98∘E 100∘E 102∘E 104∘E

Figure 2 Location map showing the northeastern edge of the Tibetan Plateau and 44 observation stations

03

0

11

17

25

39

52

71

114

228

425

40∘N

38∘N

36∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(a)

0

04

12

19

24

27

59

81

132

17

235

40∘N

38∘N

36∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(b)

Figure 3 Spatial distribution of the rainfall accumulation for the durations of the two events shown in Thiessen polygons (a) Event A (b)Event B

As for the meteorological characteristics of the storm eventsaccording to the weather analysis charts Event A is likelycaused by very strong local convections and Event B may bea stratiform storm

The rainfall evennessunevenness of the two heavy rainevents can be further verified quantitatively by using thecoefficient of variability (CV)

CV = radic 1119873

119873

sum

119894=1

(119909119894

119909minus 1)

2

(2)

where 119909119894 is the rainfall accumulation of each rain gauge 119894(when calculating the evenness in space) or the average areal

rainfall at each time step 119894 (for the evenness in time) 119909 isthe mean value of 119909119894 and 119873 is the total number of raingauges or the total number of the time steps The results ofthe two rainfall events are also shown in Table 2 A largerCV value represents higher variability thus less even rainfalldistribution The CV value of the spatial evenness for EventA is larger than that for Event B which means Event A hashigher variability in space

412 Spatial Distribution of Precipitation In Figure 5 com-pared with CTRL experiment (Figure 5(a)) the forecasts ofprecipitation area and accumulation are increased notice-ably depending on different data assimilation The CTRL

Advances in Meteorology 5

Cum

ulat

ive r

ainf

all (

mm

)25

20

15

10

5

0

Rain

fall

inte

nsity

(mm

3 h

our)

Time (month-day-hour)

0

05

1

15

2

25

3

35

45

4

5

6-18

-16-

18-4

6-18

-76-

18-1

06-

18-1

36-

18-1

66-

18-1

96-

18-2

26-

19-1

6-19

-46-

19-7

6-19

-10

6-19

-13

6-19

-16

6-19

-19

6-19

-22

(a)

Cum

ulat

ive r

ainf

all (

mm

)

0

2

4

6

8

10

12

Rain

fall

inte

nsity

(mm

15

hou

r)

Time (month-day-hour)

005115225335

454

5

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

(b)

Figure 4The time series bars and the cumulative curves of the observed precipitation for the two heavy rainfall events at Laohugou station(a) Event A (b) Event B

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18

(mm)20

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(d)

Figure 5 The forecast precipitation for Event A from the different numerical experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

6 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N∘94 E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(d)

Figure 6 The forecast precipitation for Event B from the different numerical experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

experiment predicts only a small amount of rainfall on thenorthwestern of the study area while in DA-SAT experimentprecipitation area extends to the central and southern regionsof the study area and rainfall accumulation increases sig-nificantly (Figure 5(b)) But in the northwest of the regionthe rainfall accumulation for DA-SAT experiment is similarwith that shown in Figure 5(a) In DA-OBS experiment(Figure 5(c)) precipitation accumulation on the northwest-ern of the region is greater than that in CTRL experimentbut the precipitation area is similar in both experimentsFigure 5(d) shows the result of DA-BOTH experiment inwhich the precipitation area is the largest and accumulationis the biggest compared with the other three experiments

The results for Event B are shown in Figure 6 Theprecipitation area on the northwestern of the study region issignificantly increased by DA-SAT experiment (Figure 6(b))relative to that in CTRL experiment In Figure 6(c)

the precipitation area in DA-OBS experiment is same asthat in Figure 6(a) but the precipitation accumulation isobviously decreased on the southeastern of the regionFigure 6(d) shows the DA-BOTH experiment result whichhas the similar precipitation area and accumulation as shownin DA-SAT experiment

To sum up the data assimilation has important influenceon the precipitation forecast with significantly changes of therainfall area and accumulation It should be noticed that theeffects of the forecast depends on the different rainfall event

The rain-gauge data were used to analyze the changedprecipitation area and accumulation after data assimilationexperiments Figure 7 compares the measured and mod-eled results in the rainfall Event A In CTRL experiment(Figure 7(a)) almost all the triangles lie below the line 119910 = 119909However in DA-SAT experiment the green triangles whichrepresent the observation sites (Qilian Yeniugou Tuole

Advances in Meteorology 7

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(a)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(b)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(c)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)y = x

(d)

Figure 7 Differences between rainfall values measured at the sites and those from the numerical experiments for Event A (a) CTRL (b)DA-SAT (c) DA-OBS (d) DA-BOTH

and Wulan) in the central and southern regions of the studyarea are closer to the line 119910 = 119909 (Figure 7(b)) That is tosay DA-SAT experiment improves the predicted accuracy forthe central and southern regions of the study area throughproviding greater precipitation accumulation Similarly inFigure 7(c) for DA-OBS experiment the red triangles whichrepresent the observation sites (Jiuquan Yumenzheng Anxiand Dunhuang) in northwestern of the region are closer tothe line119910 = 119909 than that is shown inCTRL experiment InDA-BOTH experiment all the red and green triangles are closestto the line 119910 = 119909 (Figure 7(d))

It is noticed that in Figure 8 many more triangles forEvent B lie above the 119910 = 119909 line than Event A Asshown in CTRL experiment (Figure 8(a)) the green triangleswhich represent observation sites (Hezuo Lintao linxia andYuzhong) on the southeastern of the study area are lie aboveand far away from the 119910 = 119909 line After data assimilationthe green triangles in three data assimilation experimentsare closer to the 119910 = 119909 line in Figures 8(b)ndash8(d) Thered triangles represent the observation sites (Jiuquan Tuole

Dingxin and Jinta) in the northwestern of the study areaIn DA-SAT experiment (Figure 8(b)) two red triangles withmore than 10mm of rainfall are closer to the 119910 = 119909 line whilein DA-OBS experiment (Figure 8(c)) other two red triangleswith less than 5mm of rainfall are closer to the 119910 = 119909 linecompared with that in Figure 8(a) In DA-BOTH experiment(Figure 8(d)) the results are similar to that as shown in DA-SAT experiment except the red triangles are closer to the119910 = 119909 line

The above results suggest that data assimilation havepositive effects on the rainfall forecast through changingthe rainfall area and accumulation for the study area Andfor the two events with different evenness in time andspace the results after data assimilation are different ForEvent A data assimilation experiments provide larger rainfallareas and accumulation to improve the predicted accuracyAnd for Event B data assimilation experiments improve theaccuracy on the southeastern of the study area by providingsmaller precipitation accumulation This means that CTRLexperiment underestimates the convective rainfall for Event

8 Advances in Meteorology

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(a)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(b)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(c)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)y = x

(d)

Figure 8 Differences between rainfall values measured at the sites and those from the numerical experiments for Event B (a) CTRL (b)DA-SAT (c) DA-OBS (d) DA-BOTH

A and overestimates the stratiform rainfall of Event B on thesoutheastern of the region

In order to evaluate quantitatively the impact of dataassimilation experiments which have positive effects on therainfall forecast rain-gauge data from 43 sites were usedTable 3 shows the comparison between the measured andmodeled results The CTRL experiments have the lowestprecision of simulation for both Event A and Event B whose1198772 are the smallest and RMSE are the biggest compared with

three data assimilation experiments The greatest improve-ment appears in DA-BOTH experiment for the two eventsFor Event A the 1198772 for DA-BOTH experiment is increasedby 046 and its RMSE is reduced by 47 and for Event Bthe 1198772 for DA-BOTH experiment is increased by 032 andits RMSE is reduced by 06 Additionally it is found that inDA-SAT experiment the improvement is bigger than DA-OBS experimentTheME of rainfall accumulation shows thatfor the entire rainfall range it has positives value for Event Abut has negative value for Event B This means that the fourexperiments produce underestimated rainfall accumulationsfor Event A but overestimated rainfall accumulations for

Event BThat is because in the precipitation ranges of 0ndash5mmand 5ndash10mm almost all the ME values are positive for EventA and all the ME are negative for Event B

In this section for the two events it is found that theforecast precision in DA-SAT experiment is higher than thatin DA-OBS experiment This may be because there are fewstations located on the high-altitude region of northeasternTibetan Plateau DA-BOTH experiment has the biggestprecision compared with the other two data assimilationexperiments which means that the improvement after theassimilation of the satellite radiances combined with surfaceand upper-air meteorological observations is greater thanthat after only assimilating satellite radiances

413 Temporal Variations of Precipitation The greater dif-ference between Event A and Event B is the rainfall dis-tribution in time Figure 9(a) shows the rainfall intensitiesand cumulative curves of the measured and modeled resultsfor Event A at Laohugou station The cumulative curvesof CTRL experiment are lie below and far away from thatof the measured values However during the period from

Advances in Meteorology 9

Table 3 Differences between observation and numerical experiments

Event ID Experiments 1198772 RMSE 0ndash5mm ME

5ndash10mm 10ndash425 (235) 0ndash425 (235)

Event A

CTRL 024 105 minus015 42 216 56DA-SAT 043 81 12 31 172 35DA-OBS 033 92 04 39 189 46DA-BOTH 070 58 minus03 11 95 18

Event B

CTRL 032 55 minus28 minus14 84 minus02DA-SAT 059 50 minus24 minus16 54 minus05DA-OBS 048 54 minus19 minus23 46 minus07DA-BOTH 064 49 minus25 minus13 50 minus06

0

1

2

3

4

5

6

7

8

9

100

5

10

15

20

25

6-18

-36-

18-6

6-18

-96-

18-1

26-

18-1

56-

18-1

86-

18-2

16-

18-2

46-

19-3

6-19

-66-

19-9

6-19

-12

6-19

-15

6-19

-18

6-19

-21

6-19

-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRL

DA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

20

40

60

80

100

120

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)Measured valueCTRLDA-OBS

DA-SATDA-BOTH

Rain

fall

inte

nsity

(mm

15

hou

r)

(b)

Figure 9The rainfall intensities and cumulative curves of the values measured at the sites and the simulated results from the four numericalexperiments (a) Event A (b) Event B

000 to 2400 on July 19 the cumulative curves for DA-SATand DA-SAT experiments are closer to that of the measuredvalues compared with the other two experiments This maybe because the maximum precipitation intensities simulatedby DA-BOTH and DA-SAT experiments are closer to theobserved values from 600 to 900 on July 19

In Figure 9(b) for event B the precipitation intensityin CTRL experiment is significantly larger than the mea-sured values in first 6 hours Therefore the cumulativecurve in the CTRL experiment is far above that of themeasured values due to the inaccurate initial field Afterdata assimilation the rainfall intensities in DA-SAT and DA-BOTH experiments are closer to the observations in first 6hours while the rainfall intensity in DA-OBS experiment ismuch less than the observations As a result the cumulativecurve in DA-BOTH experiment is closest to the curve

of the observed values DA-SAT experiment is next andDA-OBS experiment is farthest It is noteworthy that forEvent B the cumulative curve in DA-SAT experiment iscloser to that of the observed values in comparison withEvent A

The absolute error (AE) has been calculated for bothEventA andEvent B In Figure 10(a) for EventA theAEof thenumerical experiments becomes bigger and bigger with theincrease of rainfall intensity while the values in DA-BOTHandDA-SAT experiments are relatively smaller Itmeans thatfor the convective rainfall like Event A WRF model failsin capturing the whole process of the event However inFigure 10(b) for Event B with even distribution in time theAE of the numerical experiments is relatively constant whilethe values for CTRL and DA-OBS experiments are relativelylarger

10 Advances in Meteorology

0

2

4

6

8

10

12

140

2

4

6

8

10

12

14

06-1

8-3

06-1

8-6

06-1

8-9

06-1

8-12

06-1

8-15

06-1

8-18

06-1

8-21

06-1

8-24

06-1

9-3

06-1

9-6

06-1

9-9

06-1

9-12

06-1

9-15

06-1

9-18

06-1

9-21

06-1

9-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

10

20

30

40

50

60

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

15

hou

r)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(b)

Figure 10TheAEof rainfall intensity between the valuesmeasured at the sites and the simulated results from the four numerical experiments(a) Event A (b) Event B

The results shows that for the temporal variation ofrainfall the improvement in DA-SAT experiment is betterthan that in DA-OBS experiment due to the lack of groundobservational data at high-altitude region And for Event Bthe improvement in DA-SAT is greater than Event A

42 Impacts of Data Assimilation on the Initial Field Fromthe above analysis it is clear that data assimilation exper-iments can improve rainfall forecast The purpose of dataassimilation is to acquire accurate initial fields for numericalmodels To reveal the impact of data assimilation experimentson the initial fields EventA is chosen as an example to analyzethe difference in the initial fields between data assimilationand control experiments The geopotential height and mois-ture flux at 850 hPa are selected for analysis

Figure 11 shows the geopotential height differencesbetween data assimilation and control experimentsCompared with CTRL experiment (Figure 11(a)) thereis no evident change of geopotential height in DA-OBS experiment (Figure 11(b)) or DA-SAT experiment(Figure 11(c)) Figure 8(d) shows the geopotential heightfor DA-BOTH experiment which is much smaller thanthe values from CTRL experiment The region with thelargest changes is mainly located in Qinghai Provinceand the northwest of Gansu Province which is consistentwith the improved forecasts of precipitation area andaccumulation in DA-BOTH experiment As can be seen inFigure 12 after data assimilation the moisture flux changenoticeably in the study area in comparison with the controlexperiment (Figure 12(a)) The moisture flux of DA-SATexperiment is markedly increased in the north of the studyarea (Figure 12(b)) It may bring more precipitation in the

northeast of Qinghai Province In Figure 12(c) there is aclear flow of moisture to the northwestern part of GansuProvince which results in the heavy rain The biggest changeof moisture flux occurs in Figure 12(d) for DA-BOTHexperiment which covers the largest area compared withother numerical experiments and it could give greatestestimated precipitation in northeastern Tibetan Plateau

5 Conclusions

In this study the 3DVar assimilation system is used toimprove the forecasting of heavy rain in spatial and timedistribution in northeastern Tibetan Plateau A control andthree data assimilation experiments were designed Theseexperiments demonstrate that the rain forecast is significantlyimproved after data assimilation through enhancing theaccuracy of the initial field Two heavy rainfall events withdifferent evenness in space and time are used to examine theimprovement for rainfall forecast

For the spatial distribution data assimilation experimentschanged significantly the rainfall area and accumulationfor the study area For Event A the data assimilationexperiments provide larger rainfall areas and accumulationwhile for Event B the data assimilation experiments providesmaller precipitation accumulation on the southeastern ofthe study area Rain-gauge data from 43 sites were usedto evaluate the impact after data assimilation The resultsshows that DA-BOTH experiment has the highest predictionaccuracy followed by the DA-SAT experiment and thenDA-OBS experiment while the CTRL experiment has thelowest accuracy For the temporal variation the forecastresults have great differences between Event A and Event B

Advances in Meteorology 11

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(d)

Figure 11 Geopotential height initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

The satellite radiances for Event B have greater positive effectthan that for Event A For the two events the cumulativecurve in DA-BOTH experiment is closest to the curveof the measured values while in DA-OBS experiment theaccumulation curves are far below to the curve of themeasured values due to the lack of ground observations inthe high-altitude region In conclusion both for the spatialand temporal distribution of rainfall the satellite radianceshave greater effect than surface and upper-air meteorologicalobservations in high-latitude regions of the northeasternedge and the improvement for the assimilation of satelliteradiances together with surface and upper-air meteorologicalobservations is greater than that for assimilating either satel-lite radiances ormeteorology observations Event A is chosento evaluate the impact of data assimilation on the initialfields It is found that better and more detailed informationis added to the initial fields (eg geopotential height andmoisture flux) especially on the northwest of Gansu Province

andQinghai Province where the forecasts of precipitation areimproved significantly

It should be mentioned that many other assimilationtechniques such as 4DVar EnKF and so forth need to betested in this study These approaches have great potentialalthough they currently suffer from unaffordable computercosts Meanwhile we need to analyse the impact of differentparameterization schemes on the WRF model after dataassimilation because in this study the conclusions drawn aresubject to the specific parameterization schemes used here

Appendix

Statistic Calculation

The coefficient of determination (1198772) root mean squaredeviation (RMSE) mean error (ME) and absolute error (AE)

12 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(d)

Figure 12 Moisture flux initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

were used to evaluate the impact of the DA experiments inthis study They are defined as

119877

=

(1119873)sum119873

119894=1(119865119894 minus 119865) (119874119894 minus 119874)

radic(1119873)sum119873

119894=1(119865119894 minus 119865)

2radic(1119873)sum

119873

119894=1(119874119894 minus 119874)

2

RMSE = radic 1119873

119873

sum

119894=1

(119865119894 minus 119874119894)2

ME = 1119873

119873

sum

119894=1

(119874119894 minus 119865119894)

AE = 1119873

119873

sum

119894=1

1003816100381610038161003816119865119894 minus 1198741198941003816100381610038161003816

(A1)

Here 119865 is simulated value 119874 is observed value 119873 is thenumber of sites

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work was supported by the National natural ScienceFoundation of China (41190080 40971290) Strategic LeadingTechnology Special Fund of Chinese Academy of Sciences(XDB03030204) the National Scientific and TechnologicalSupport Project (2013BAB05B03) and the Opening Fund ofKey Laboratory of Land Surface Process and Climate Changein Cold and Arid Regions CAS (LPCC201205)

Advances in Meteorology 13

References

[1] P Chambon S Q Zhang A Y Hou M Zupanski and SCheung ldquoAssessing the impact of pre-GPMmicrowave precipi-tation observations in the GoddardWRF ensemble data assim-ilation systemrdquo Quarterly Journal of the Royal MeteorologicalSociety vol 140 no 681 pp 1219ndash1235 2013

[2] J-H Seo Y H Lee and Y-H Kim ldquoFeature selection forvery short-term heavy rainfall prediction using evolutionarycomputationrdquo Advances in Meteorology vol 2014 Article ID203545 15 pages 2014

[3] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M JEsteban-Parra and Y Castro-Dıez ldquoHigh-resolution projec-tions of mean and extreme precipitation over Spain usingthe WRF model (2070ndash2099 versus 1970ndash1999)rdquo Journal ofGeophysical Research D Atmospheres vol 117 no 12 Article IDD12108 2012

[4] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M J SEsteban-Parra and Y Castro-Dıez ldquoEvaluation of WRF meanand extreme precipitation over Spain present climate (1970ndash99)rdquo Journal of Climate vol 25 no 14 pp 4883ndash4895 2012

[5] R M Cardoso P M M Soares P M A Miranda and MBelo-Pereira ldquoWRF high resolution simulation of Iberianmeanand extreme precipitation climaterdquo International Journal ofClimatology vol 33 no 11 pp 2591ndash2608 2013

[6] L Calvetti and A J Pereira Filho ldquoEnsemble hydrometeoro-logical forecasts using WRF hourly QPF and topmodel for amiddle watershedrdquo Advances in Meteorology vol 2014 ArticleID 484120 12 pages 2014

[7] S Q Zhang M Zupanski A Y Hou X Lin and S H CheungldquoAssimilation of precipitation-affected radiances in a cloud-resolving wrf ensemble data assimilation systemrdquo MonthlyWeather Review vol 141 no 2 pp 754ndash772 2013

[8] J Liu M Bray and D Han ldquoExploring the effect of dataassimilation by WRF-3DVar for numerical rainfall predictionwith different types of storm eventsrdquoHydrological Processes vol27 no 25 pp 3627ndash3640 2013

[9] C-H Xiong L-F Zhang J-P Guan J Peng and B ZhangldquoAnalysis and numerical study of a hybrid BGM-3DVAR dataassimilation scheme using satellite radiance data for heavy rainforecastsrdquo Journal of Hydrodynamics vol 25 no 3 pp 430ndash4392013

[10] D M Barker W Huang Y R Guo and Q N Xiao ldquoA three-dimensional (3DVAR) data assimilation system for use withMM5 implementation and initial resultsrdquo Monthly WeatherReview vol 132 pp 897ndash914 2004

[11] Q Xiao and J Sun ldquoMultiple-radar data assimilation and short-range quantitative precipitation forecasting of a squall lineobserved during IHOP 2002rdquoMonthlyWeather Review vol 135no 10 pp 3381ndash3404 2007

[12] Z Sokol ldquoEffects of an assimilation of radar and satellite dataon a very-short range forecast of heavy convective rainfallsrdquoAtmospheric Research vol 93 no 1ndash3 pp 188ndash206 2009

[13] P Bauer P Lopez A Benedetti D Salmond and E MoreauldquoImplementation of 1D+4D-Var assimilation of precipitation-affected microwave radiances at ECMWF I 1D-Varrdquo QuarterlyJournal of the Royal Meteorological Society vol 132 no 620 pp2277ndash2306 2006

[14] P Bauer P Lopez D Salmond A Benedetti S Saarinen andM Bonazzola ldquoImplementation of 1D+4D-Var assimilation of

precipitation-affected microwave radiances at ECMWF II 4D-Varrdquo Quarterly Journal of the Royal Meteorological Society vol132 no 620 pp 2307ndash2332 2006

[15] P Bauer A J Geer P Lopez and D Salmond ldquoDirect 4D-Var assimilation of all-sky radiances Part I ImplementationrdquoQuarterly Journal of the Royal Meteorological Society vol 136no 652 pp 1868ndash1885 2010

[16] A J Geer P Bauer and P Lopez ldquoDirect 4D-Var assimilationof all-sky radiances Part II assessmentrdquo Quarterly Journal ofthe RoyalMeteorological Society vol 136 no 652 pp 1886ndash19052010

[17] Z Sokol and D Rezacova ldquoAssimilation of radar reflectivityinto the LM COSMOmodel with a high horizontal resolutionrdquoMeteorological Applications vol 13 no 4 pp 317ndash330 2006

[18] Z Sokol ldquoAssimilation of extrapolated radar reflectivity into aNWP model and its impact on a precipitation forecast at highresolutionrdquo Atmospheric Research vol 100 no 2-3 pp 201ndash2122011

[19] HWang J Sun S Fan and X-Y Huang ldquoIndirect assimilationof radar reflectivity with WRF 3D-var and its impact onprediction of four summertime convective eventsrdquo Journal ofApplied Meteorology and Climatology vol 52 no 4 pp 889ndash902 2013

[20] B Macpherson ldquoOperational experience with assimilation ofrainfall data in the Met Office mesoscale modelrdquo Meteorologyand Atmospheric Physics vol 76 no 1-2 pp 3ndash8 2001

[21] K Stephan S Klink and C Schraff ldquoAssimilation of radar-derived rain rates into the convective-scale model COSMO-DEat DWDrdquo Quarterly Journal of the Royal Meteorological Societyvol 134 no 634 pp 1315ndash1326 2008

[22] K Alapaty D S Niyogi F Chen P Pyle A Chandrasekar andN Seaman ldquoDevelopment of the flux-adjusting surface dataassimilation system for mesoscale modelsrdquo Journal of AppliedMeteorology and Climatology vol 47 no 9 pp 2331ndash2350 2008

[23] F H Ruggiero G D Modica and A E Lipton ldquoAssimilationof satellite imager data and surface observations to improveanalysis of circulations forced by cloud shading contrastsrdquoMonthly Weather Review vol 128 no 2 pp 434ndash448 2000

[24] J P Hacker and D Rostkier-Edelstein ldquoPBL state estimationwith surface observations a column model and an ensemblefilterrdquo Monthly Weather Review vol 135 no 8 pp 2958ndash29722007

[25] D J Stensrud N Yussouf D C Dowell and M C ConiglioldquoAssimilating surface data into a mesoscale model ensemblecold pool analyses from spring 2007rdquoAtmospheric Research vol93 no 1ndash3 pp 207ndash220 2009

[26] Q Wan and J Xu ldquoA numerical study of the rainstormcharacteristics of the June 2005 flash flood with WRFGSIdata assimilation system over south-east Chinardquo HydrologicalProcesses vol 25 no 8 pp 1327ndash1341 2011

[27] J Liu M Bray and D Han ldquoA study on WRF radar dataassimilation for hydrological rainfall predictionrdquoHydrology andEarth System Sciences vol 17 no 8 pp 3095ndash3110 2013

[28] K Ide P Courtier M Ghil and A C Lorenc ldquoUnified notationfor data assimilation operational sequential and variationalrdquoJournal of the Meteorological Society of Japan vol 75 no 1 pp181ndash189 1997

[29] A C Lorenc ldquoAnalysis methods for numerical weather predic-tionrdquo Quarterly Journal of the Royal Meteorological Society vol112 no 474 pp 1177ndash1194 1966

14 Advances in Meteorology

[30] D F Parrish and J C Derber ldquoThe nationalmeteorological cen-terrsquos spectral statistical interpolation analysis systemrdquo MonthlyWeather Review vol 120 no 8 pp 1747ndash1763 1992

[31] T B Zhao and C B Fu ldquoPreliminary comparison and analysisbetween ERA-40 NCEP-2 reanalysis and observations overChinardquo Climatic and Environmental Research vol 11 no 1 pp14ndash32 2006

[32] L Ma T Zhang Q Li O W Frauenfeld and D QinldquoEvaluation of ERA-40 NCEP-1 and NCEP-2 reanalysis airtemperatures with ground-based measurements in ChinardquoJournal of Geophysical Research Atmospheres (1984ndash2012) vol13 no D15 1984

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ClimatologyJournal of

EcologyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

EarthquakesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

Mining

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

International Journal of

Geophysics

OceanographyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofPetroleum Engineering

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Atmospheric SciencesInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MineralogyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MeteorologyAdvances in

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geological ResearchJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geology Advances in

2 Advances in Meteorology

WPS domain configuration

46∘N

44∘N

42∘N

40∘N

38∘N

36∘N

34∘N

32∘N

30∘N

90∘E 95∘E 100∘E 105∘E 110∘E

d02

d01

Gansu Province

Qinghai Province

Figure 1 Domains for the ARW-WRF forecasts The outer box is the coarse grid with a resolution of 12 km (d01) the inner box is the nestedgrid (d02) with a resolution of 4 km

the performance of NWP models in spite of the poor initialconditions provided by the global NWP models [12] In thecase of high temporal and spatial resolution satellite datathe European Center for Medium-Range Weather Forecasts(ECMWF) pioneered the direct assimilation of microwaveradiance data affected by precipitation first in a 1 + 4Dvarassimilation approach [13 14] and later in an implementationof all-sky radiance assimilation in the operational 4Dvarsystem [15 16] Additionally many investigations have shownthat rainfall forecasts from theNWPmodels can be improvednoticeably with the assimilation of radar reflectivity [17ndash19]or radar-derived precipitation data [20 21] Nevertheless theremote sensing data need to be validated against ground truth[22] and surface observations have a wealth of informationthat can be used to simulate mesoscale weather phenomena[23] Therefore the assimilation of surface observations in aNWP model is likely to improve the model performance aswell Previous studies have beenmade to improve simulationsof weather parameters by using direct or modeled surfaceobservations These include the assimilation of tempera-ture water vapor mixing ratio and winds [24] and 2mpotential temperature 2m dew point temperature and 10mwind observations [25] into the NWP model to determineplanetary boundary layer (PBL) profiles and to analyse thesurface cold pool respectively The results show a markedimprovement in the model simulations after the assimilationIn the WRF assimilation system previous research hasalso shown that by assimilating only the satellite data theimprovement in precipitation forecasts is not as great as whenthe assimilation of satellite data is combined with that ofsurface observations [26 27]

In this study the WRF-3Dvar system is used to explorethe effect of the assimilation of NCAR surface and upper-air observations and of AMSU-A and AMSU-B microwaveradiance data on forecasts of heavy precipitation on thenortheastern Tibetan Plateau Two heavy rain events thatoccurred in June 2013 were selected in order to evaluate

the improvement for rainfall forecast after data assimilationThe paper is organized as follows Section 2 gives a conciseoverview of the method and experiment design Section 3provides information on the study area and the data Theresults of the data assimilation experiments are presented andevaluated in Section 4 and conclusions are given in Section 5

2 Method and Experimental Design

21 WRF Model Set-Up The numerical data assimilationexperiments in this study are conducted using the AdvancedResearch WRF model Version 35 WRF is a nonhydrostaticprimitive-equation mesoscale meteorological model withadvanced dynamics physics and numerical schemes (detailsof the model is at httpwwwmmmucaredu) As shown inFigure 1 the model domains are two-way nested with 12 km(208 times 182) and 4 km (235 times 182) horizontal spacing Eachdomain has 28 vertical pressure levels with the top levelset at 50 hPa The WRF physical parameterization schemesused in this study include the Purdue Lin microphysicalparameterization Rapid Radiative Transfer Model (RRTM)longwave radiation Dudhia shortwave radiation Monin-Obukhov surface layer Noah land surface Mellor-Yamada-Janjic (MYJ) planetary boundary layer scheme and Grell-Devenyi (GD) cumulus scheme The projection method isLambert

22 3D-Var Data Assimilation Data assimilation is the tech-nique by which observations are combined with a NWPproduct (called the first guess or background field) andtheir respective error statistics to provide an improved esti-mate (the analysis) of the atmospheric (or oceanic) stateVariational (Var) data assimilation achieves this through

Advances in Meteorology 3

the iterative minimization of a prescribed cost (or penalty)function [28]

119869 (119909) =1

2(119909 minus 119909

119887)119879

119861minus1(119909 minus 119909

119887)

+1

2(119910 minus 119910

0)119879

119877minus1(119910 minus 119910

0)

(1)

where 119909 is the analysis to be found that minimizes thecost function 119869(119909) 119909119887 is the first guess of the NWP model1199100 is the assimilated observation and 119910 = 119867(119909) is the

model-derived observation transformed from the analysis 119909by the observation operator 119867 for comparison against 1199100The solution for the cost function given by (1) representsa posteriori maximum likelihood (minimum variance) esti-mate of the true state given the two sources of a priori datathe first guess 119909119887 and the observation 1199100 [29] The fit toindividual observation points is weighted by the estimatesof their errors that is 119861 and 119877 which are the backgrounderror covariance matrix and the observation error covariancematrix respectively

The WRF-3DVar system developed by Barker et al [10]is used in this study in tandem with the WRF model forassimilating the satellite radiance data and the traditionalobservations The performance of the data-assimilation sys-tem largely depends on the plausibility of the backgrounderror covariance (BE) that is thematrix119861 in (1) In this studythe ldquoCV5rdquo background error option is used with the controlvariables of stream function unbalanced temperature unbal-anced potential velocity unbalanced surface pressure andpseudo relative humidity The background error covariancematrix is generated via the National Meteorological Center(NMC) method [30] for our own forecasting domain

23 Experimental Design Four sets of experiments have beenconducted using two domains As shown in Table 1 themodel simulation without data assimilation will be referredto as the control experiment (CTRL) Three assimilationexperiments are designed using different observation param-eters from different data sources In the DA-OBS experimentmeasurements of pressure geopotential height temperaturedewpoint temperature wind direction and speed fromNCARare assimilated while in theDA-SAT experiment onlysatellite radiance data are assimilated Finally in the DA-BOTH experiment both NCAR observations and AMSU-A(B) radiance data are assimilated

3 Study Area and Data

As shown in Figure 2 the northeastern of the TibetanPlateau (94∘391015840ndash103∘271015840E 35∘511015840ndash40∘311015840N) range of elevationbetween 758m and 5725m asl is the head of many inlandrivers which play an important role in the hydrology andagriculture in the downstream arid region A total of 43national observation stations have been used to verify thespatial distribution of the simulated precipitation amongwhich the Wulan station has the highest elevation of 3800mandDunhuang has the lowest altitude of 1137mTheobserved

Table 1 Details of the observed data used in the assimilationexperiments

Experiment name Assimilated dataCTRL No data assimilation

DA-OBS NCAR surface and upper-airobservation

DA-SAT AMSU-A and AMSU-B radiancedata

DA-BOTH NCAR observation andAMSU-A (B) radiance data

precipitation at Laohugou station (elevation 4200m) hasbeen used to evaluate the temporal variations of the simulatedrainfall which is measured by T200B every half hour

The initial and boundary conditions necessary to run theWRF are the ERA-Interim data at 1∘ times 1∘ grid resolutionobtained from the European Centre for Medium-RangeWeather Forecasts (ECMWF) rather than the NCEP-NCARFinal Analysis (FNL) data This is because several studieshave demonstrated that the reliability of ERA-Reanalysis datais higher than the NCEP data in China [31 32] In themodel integration the coordinates of the central point are381∘N and 989∘E WRF-3DVar experiments have been con-ducted with modified initial conditions which were obtainedby assimilating other measured data The assimilated dataincludes NCAR surface and upper-air observations andAMSUA and AMSUB radiance data The surface and upper-air data assimilated in this study are obtained from theldquods3370rdquo in the NCAR archives which contain measure-ments of pressure geopotential height temperature dewpoint temperature wind direction and speed from fixed andmobile landsea stations The data are initially downloadedin PREPBUFR format and can be assimilated directly intoWRFDA The AMSU-A and AMSU-B satellite radiance data(NOAA-1516171819) from theNOAAATOVS instrumentscan be read in WRFDA via CRTM202 which is in BUFRformat

4 Results and Discussion

41 Impacts of Data Assimilation on the Precipitation Forecast

411 Rainfall Event In June 2013 two heavy rain eventsoccurred in the northeast of the Tibetan Plateau Thedurations of the events and the maximummean rainfallaccumulation observed by rain gauge network are shownin Table 2 The two events are of different types accordingto the evenness of rainfall distribution in time and spaceFigure 3 illustrates the spatial distribution of the rainfallaccumulation for the durations of the two events whileFigure 4 presents the time series bars and the cumulativecurves of the observed precipitation for the two events atLaohugou station By comparing the evenness of the rainfalldistribution in time and space in Figures 3 and 4 it can befound that Event A has even rainfall distribution neither inspace nor in time The rainfall distribution of Event B is alsouneven in space but continuous and almost constant in time

4 Advances in Meteorology

Table 2 Durations maximummean rainfall accumulation and spatialtemporal evenness of the heavy rain events

Event ID Start time End time Spatial evenness Temporal evenness Maximum precipitation Mean precipitationA 1862013 000 1962013 2400 146 144 425 67B 762013 000 762013 2400 115 063 235 53

LaohugouRain-gauge stations

River

94∘E

40∘N

38∘N

36∘N

96∘E 98∘E 100∘E 102∘E 104∘E

Figure 2 Location map showing the northeastern edge of the Tibetan Plateau and 44 observation stations

03

0

11

17

25

39

52

71

114

228

425

40∘N

38∘N

36∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(a)

0

04

12

19

24

27

59

81

132

17

235

40∘N

38∘N

36∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(b)

Figure 3 Spatial distribution of the rainfall accumulation for the durations of the two events shown in Thiessen polygons (a) Event A (b)Event B

As for the meteorological characteristics of the storm eventsaccording to the weather analysis charts Event A is likelycaused by very strong local convections and Event B may bea stratiform storm

The rainfall evennessunevenness of the two heavy rainevents can be further verified quantitatively by using thecoefficient of variability (CV)

CV = radic 1119873

119873

sum

119894=1

(119909119894

119909minus 1)

2

(2)

where 119909119894 is the rainfall accumulation of each rain gauge 119894(when calculating the evenness in space) or the average areal

rainfall at each time step 119894 (for the evenness in time) 119909 isthe mean value of 119909119894 and 119873 is the total number of raingauges or the total number of the time steps The results ofthe two rainfall events are also shown in Table 2 A largerCV value represents higher variability thus less even rainfalldistribution The CV value of the spatial evenness for EventA is larger than that for Event B which means Event A hashigher variability in space

412 Spatial Distribution of Precipitation In Figure 5 com-pared with CTRL experiment (Figure 5(a)) the forecasts ofprecipitation area and accumulation are increased notice-ably depending on different data assimilation The CTRL

Advances in Meteorology 5

Cum

ulat

ive r

ainf

all (

mm

)25

20

15

10

5

0

Rain

fall

inte

nsity

(mm

3 h

our)

Time (month-day-hour)

0

05

1

15

2

25

3

35

45

4

5

6-18

-16-

18-4

6-18

-76-

18-1

06-

18-1

36-

18-1

66-

18-1

96-

18-2

26-

19-1

6-19

-46-

19-7

6-19

-10

6-19

-13

6-19

-16

6-19

-19

6-19

-22

(a)

Cum

ulat

ive r

ainf

all (

mm

)

0

2

4

6

8

10

12

Rain

fall

inte

nsity

(mm

15

hou

r)

Time (month-day-hour)

005115225335

454

5

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

(b)

Figure 4The time series bars and the cumulative curves of the observed precipitation for the two heavy rainfall events at Laohugou station(a) Event A (b) Event B

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18

(mm)20

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(d)

Figure 5 The forecast precipitation for Event A from the different numerical experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

6 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N∘94 E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(d)

Figure 6 The forecast precipitation for Event B from the different numerical experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

experiment predicts only a small amount of rainfall on thenorthwestern of the study area while in DA-SAT experimentprecipitation area extends to the central and southern regionsof the study area and rainfall accumulation increases sig-nificantly (Figure 5(b)) But in the northwest of the regionthe rainfall accumulation for DA-SAT experiment is similarwith that shown in Figure 5(a) In DA-OBS experiment(Figure 5(c)) precipitation accumulation on the northwest-ern of the region is greater than that in CTRL experimentbut the precipitation area is similar in both experimentsFigure 5(d) shows the result of DA-BOTH experiment inwhich the precipitation area is the largest and accumulationis the biggest compared with the other three experiments

The results for Event B are shown in Figure 6 Theprecipitation area on the northwestern of the study region issignificantly increased by DA-SAT experiment (Figure 6(b))relative to that in CTRL experiment In Figure 6(c)

the precipitation area in DA-OBS experiment is same asthat in Figure 6(a) but the precipitation accumulation isobviously decreased on the southeastern of the regionFigure 6(d) shows the DA-BOTH experiment result whichhas the similar precipitation area and accumulation as shownin DA-SAT experiment

To sum up the data assimilation has important influenceon the precipitation forecast with significantly changes of therainfall area and accumulation It should be noticed that theeffects of the forecast depends on the different rainfall event

The rain-gauge data were used to analyze the changedprecipitation area and accumulation after data assimilationexperiments Figure 7 compares the measured and mod-eled results in the rainfall Event A In CTRL experiment(Figure 7(a)) almost all the triangles lie below the line 119910 = 119909However in DA-SAT experiment the green triangles whichrepresent the observation sites (Qilian Yeniugou Tuole

Advances in Meteorology 7

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(a)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(b)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(c)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)y = x

(d)

Figure 7 Differences between rainfall values measured at the sites and those from the numerical experiments for Event A (a) CTRL (b)DA-SAT (c) DA-OBS (d) DA-BOTH

and Wulan) in the central and southern regions of the studyarea are closer to the line 119910 = 119909 (Figure 7(b)) That is tosay DA-SAT experiment improves the predicted accuracy forthe central and southern regions of the study area throughproviding greater precipitation accumulation Similarly inFigure 7(c) for DA-OBS experiment the red triangles whichrepresent the observation sites (Jiuquan Yumenzheng Anxiand Dunhuang) in northwestern of the region are closer tothe line119910 = 119909 than that is shown inCTRL experiment InDA-BOTH experiment all the red and green triangles are closestto the line 119910 = 119909 (Figure 7(d))

It is noticed that in Figure 8 many more triangles forEvent B lie above the 119910 = 119909 line than Event A Asshown in CTRL experiment (Figure 8(a)) the green triangleswhich represent observation sites (Hezuo Lintao linxia andYuzhong) on the southeastern of the study area are lie aboveand far away from the 119910 = 119909 line After data assimilationthe green triangles in three data assimilation experimentsare closer to the 119910 = 119909 line in Figures 8(b)ndash8(d) Thered triangles represent the observation sites (Jiuquan Tuole

Dingxin and Jinta) in the northwestern of the study areaIn DA-SAT experiment (Figure 8(b)) two red triangles withmore than 10mm of rainfall are closer to the 119910 = 119909 line whilein DA-OBS experiment (Figure 8(c)) other two red triangleswith less than 5mm of rainfall are closer to the 119910 = 119909 linecompared with that in Figure 8(a) In DA-BOTH experiment(Figure 8(d)) the results are similar to that as shown in DA-SAT experiment except the red triangles are closer to the119910 = 119909 line

The above results suggest that data assimilation havepositive effects on the rainfall forecast through changingthe rainfall area and accumulation for the study area Andfor the two events with different evenness in time andspace the results after data assimilation are different ForEvent A data assimilation experiments provide larger rainfallareas and accumulation to improve the predicted accuracyAnd for Event B data assimilation experiments improve theaccuracy on the southeastern of the study area by providingsmaller precipitation accumulation This means that CTRLexperiment underestimates the convective rainfall for Event

8 Advances in Meteorology

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(a)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(b)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(c)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)y = x

(d)

Figure 8 Differences between rainfall values measured at the sites and those from the numerical experiments for Event B (a) CTRL (b)DA-SAT (c) DA-OBS (d) DA-BOTH

A and overestimates the stratiform rainfall of Event B on thesoutheastern of the region

In order to evaluate quantitatively the impact of dataassimilation experiments which have positive effects on therainfall forecast rain-gauge data from 43 sites were usedTable 3 shows the comparison between the measured andmodeled results The CTRL experiments have the lowestprecision of simulation for both Event A and Event B whose1198772 are the smallest and RMSE are the biggest compared with

three data assimilation experiments The greatest improve-ment appears in DA-BOTH experiment for the two eventsFor Event A the 1198772 for DA-BOTH experiment is increasedby 046 and its RMSE is reduced by 47 and for Event Bthe 1198772 for DA-BOTH experiment is increased by 032 andits RMSE is reduced by 06 Additionally it is found that inDA-SAT experiment the improvement is bigger than DA-OBS experimentTheME of rainfall accumulation shows thatfor the entire rainfall range it has positives value for Event Abut has negative value for Event B This means that the fourexperiments produce underestimated rainfall accumulationsfor Event A but overestimated rainfall accumulations for

Event BThat is because in the precipitation ranges of 0ndash5mmand 5ndash10mm almost all the ME values are positive for EventA and all the ME are negative for Event B

In this section for the two events it is found that theforecast precision in DA-SAT experiment is higher than thatin DA-OBS experiment This may be because there are fewstations located on the high-altitude region of northeasternTibetan Plateau DA-BOTH experiment has the biggestprecision compared with the other two data assimilationexperiments which means that the improvement after theassimilation of the satellite radiances combined with surfaceand upper-air meteorological observations is greater thanthat after only assimilating satellite radiances

413 Temporal Variations of Precipitation The greater dif-ference between Event A and Event B is the rainfall dis-tribution in time Figure 9(a) shows the rainfall intensitiesand cumulative curves of the measured and modeled resultsfor Event A at Laohugou station The cumulative curvesof CTRL experiment are lie below and far away from thatof the measured values However during the period from

Advances in Meteorology 9

Table 3 Differences between observation and numerical experiments

Event ID Experiments 1198772 RMSE 0ndash5mm ME

5ndash10mm 10ndash425 (235) 0ndash425 (235)

Event A

CTRL 024 105 minus015 42 216 56DA-SAT 043 81 12 31 172 35DA-OBS 033 92 04 39 189 46DA-BOTH 070 58 minus03 11 95 18

Event B

CTRL 032 55 minus28 minus14 84 minus02DA-SAT 059 50 minus24 minus16 54 minus05DA-OBS 048 54 minus19 minus23 46 minus07DA-BOTH 064 49 minus25 minus13 50 minus06

0

1

2

3

4

5

6

7

8

9

100

5

10

15

20

25

6-18

-36-

18-6

6-18

-96-

18-1

26-

18-1

56-

18-1

86-

18-2

16-

18-2

46-

19-3

6-19

-66-

19-9

6-19

-12

6-19

-15

6-19

-18

6-19

-21

6-19

-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRL

DA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

20

40

60

80

100

120

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)Measured valueCTRLDA-OBS

DA-SATDA-BOTH

Rain

fall

inte

nsity

(mm

15

hou

r)

(b)

Figure 9The rainfall intensities and cumulative curves of the values measured at the sites and the simulated results from the four numericalexperiments (a) Event A (b) Event B

000 to 2400 on July 19 the cumulative curves for DA-SATand DA-SAT experiments are closer to that of the measuredvalues compared with the other two experiments This maybe because the maximum precipitation intensities simulatedby DA-BOTH and DA-SAT experiments are closer to theobserved values from 600 to 900 on July 19

In Figure 9(b) for event B the precipitation intensityin CTRL experiment is significantly larger than the mea-sured values in first 6 hours Therefore the cumulativecurve in the CTRL experiment is far above that of themeasured values due to the inaccurate initial field Afterdata assimilation the rainfall intensities in DA-SAT and DA-BOTH experiments are closer to the observations in first 6hours while the rainfall intensity in DA-OBS experiment ismuch less than the observations As a result the cumulativecurve in DA-BOTH experiment is closest to the curve

of the observed values DA-SAT experiment is next andDA-OBS experiment is farthest It is noteworthy that forEvent B the cumulative curve in DA-SAT experiment iscloser to that of the observed values in comparison withEvent A

The absolute error (AE) has been calculated for bothEventA andEvent B In Figure 10(a) for EventA theAEof thenumerical experiments becomes bigger and bigger with theincrease of rainfall intensity while the values in DA-BOTHandDA-SAT experiments are relatively smaller Itmeans thatfor the convective rainfall like Event A WRF model failsin capturing the whole process of the event However inFigure 10(b) for Event B with even distribution in time theAE of the numerical experiments is relatively constant whilethe values for CTRL and DA-OBS experiments are relativelylarger

10 Advances in Meteorology

0

2

4

6

8

10

12

140

2

4

6

8

10

12

14

06-1

8-3

06-1

8-6

06-1

8-9

06-1

8-12

06-1

8-15

06-1

8-18

06-1

8-21

06-1

8-24

06-1

9-3

06-1

9-6

06-1

9-9

06-1

9-12

06-1

9-15

06-1

9-18

06-1

9-21

06-1

9-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

10

20

30

40

50

60

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

15

hou

r)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(b)

Figure 10TheAEof rainfall intensity between the valuesmeasured at the sites and the simulated results from the four numerical experiments(a) Event A (b) Event B

The results shows that for the temporal variation ofrainfall the improvement in DA-SAT experiment is betterthan that in DA-OBS experiment due to the lack of groundobservational data at high-altitude region And for Event Bthe improvement in DA-SAT is greater than Event A

42 Impacts of Data Assimilation on the Initial Field Fromthe above analysis it is clear that data assimilation exper-iments can improve rainfall forecast The purpose of dataassimilation is to acquire accurate initial fields for numericalmodels To reveal the impact of data assimilation experimentson the initial fields EventA is chosen as an example to analyzethe difference in the initial fields between data assimilationand control experiments The geopotential height and mois-ture flux at 850 hPa are selected for analysis

Figure 11 shows the geopotential height differencesbetween data assimilation and control experimentsCompared with CTRL experiment (Figure 11(a)) thereis no evident change of geopotential height in DA-OBS experiment (Figure 11(b)) or DA-SAT experiment(Figure 11(c)) Figure 8(d) shows the geopotential heightfor DA-BOTH experiment which is much smaller thanthe values from CTRL experiment The region with thelargest changes is mainly located in Qinghai Provinceand the northwest of Gansu Province which is consistentwith the improved forecasts of precipitation area andaccumulation in DA-BOTH experiment As can be seen inFigure 12 after data assimilation the moisture flux changenoticeably in the study area in comparison with the controlexperiment (Figure 12(a)) The moisture flux of DA-SATexperiment is markedly increased in the north of the studyarea (Figure 12(b)) It may bring more precipitation in the

northeast of Qinghai Province In Figure 12(c) there is aclear flow of moisture to the northwestern part of GansuProvince which results in the heavy rain The biggest changeof moisture flux occurs in Figure 12(d) for DA-BOTHexperiment which covers the largest area compared withother numerical experiments and it could give greatestestimated precipitation in northeastern Tibetan Plateau

5 Conclusions

In this study the 3DVar assimilation system is used toimprove the forecasting of heavy rain in spatial and timedistribution in northeastern Tibetan Plateau A control andthree data assimilation experiments were designed Theseexperiments demonstrate that the rain forecast is significantlyimproved after data assimilation through enhancing theaccuracy of the initial field Two heavy rainfall events withdifferent evenness in space and time are used to examine theimprovement for rainfall forecast

For the spatial distribution data assimilation experimentschanged significantly the rainfall area and accumulationfor the study area For Event A the data assimilationexperiments provide larger rainfall areas and accumulationwhile for Event B the data assimilation experiments providesmaller precipitation accumulation on the southeastern ofthe study area Rain-gauge data from 43 sites were usedto evaluate the impact after data assimilation The resultsshows that DA-BOTH experiment has the highest predictionaccuracy followed by the DA-SAT experiment and thenDA-OBS experiment while the CTRL experiment has thelowest accuracy For the temporal variation the forecastresults have great differences between Event A and Event B

Advances in Meteorology 11

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(d)

Figure 11 Geopotential height initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

The satellite radiances for Event B have greater positive effectthan that for Event A For the two events the cumulativecurve in DA-BOTH experiment is closest to the curveof the measured values while in DA-OBS experiment theaccumulation curves are far below to the curve of themeasured values due to the lack of ground observations inthe high-altitude region In conclusion both for the spatialand temporal distribution of rainfall the satellite radianceshave greater effect than surface and upper-air meteorologicalobservations in high-latitude regions of the northeasternedge and the improvement for the assimilation of satelliteradiances together with surface and upper-air meteorologicalobservations is greater than that for assimilating either satel-lite radiances ormeteorology observations Event A is chosento evaluate the impact of data assimilation on the initialfields It is found that better and more detailed informationis added to the initial fields (eg geopotential height andmoisture flux) especially on the northwest of Gansu Province

andQinghai Province where the forecasts of precipitation areimproved significantly

It should be mentioned that many other assimilationtechniques such as 4DVar EnKF and so forth need to betested in this study These approaches have great potentialalthough they currently suffer from unaffordable computercosts Meanwhile we need to analyse the impact of differentparameterization schemes on the WRF model after dataassimilation because in this study the conclusions drawn aresubject to the specific parameterization schemes used here

Appendix

Statistic Calculation

The coefficient of determination (1198772) root mean squaredeviation (RMSE) mean error (ME) and absolute error (AE)

12 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(d)

Figure 12 Moisture flux initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

were used to evaluate the impact of the DA experiments inthis study They are defined as

119877

=

(1119873)sum119873

119894=1(119865119894 minus 119865) (119874119894 minus 119874)

radic(1119873)sum119873

119894=1(119865119894 minus 119865)

2radic(1119873)sum

119873

119894=1(119874119894 minus 119874)

2

RMSE = radic 1119873

119873

sum

119894=1

(119865119894 minus 119874119894)2

ME = 1119873

119873

sum

119894=1

(119874119894 minus 119865119894)

AE = 1119873

119873

sum

119894=1

1003816100381610038161003816119865119894 minus 1198741198941003816100381610038161003816

(A1)

Here 119865 is simulated value 119874 is observed value 119873 is thenumber of sites

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work was supported by the National natural ScienceFoundation of China (41190080 40971290) Strategic LeadingTechnology Special Fund of Chinese Academy of Sciences(XDB03030204) the National Scientific and TechnologicalSupport Project (2013BAB05B03) and the Opening Fund ofKey Laboratory of Land Surface Process and Climate Changein Cold and Arid Regions CAS (LPCC201205)

Advances in Meteorology 13

References

[1] P Chambon S Q Zhang A Y Hou M Zupanski and SCheung ldquoAssessing the impact of pre-GPMmicrowave precipi-tation observations in the GoddardWRF ensemble data assim-ilation systemrdquo Quarterly Journal of the Royal MeteorologicalSociety vol 140 no 681 pp 1219ndash1235 2013

[2] J-H Seo Y H Lee and Y-H Kim ldquoFeature selection forvery short-term heavy rainfall prediction using evolutionarycomputationrdquo Advances in Meteorology vol 2014 Article ID203545 15 pages 2014

[3] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M JEsteban-Parra and Y Castro-Dıez ldquoHigh-resolution projec-tions of mean and extreme precipitation over Spain usingthe WRF model (2070ndash2099 versus 1970ndash1999)rdquo Journal ofGeophysical Research D Atmospheres vol 117 no 12 Article IDD12108 2012

[4] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M J SEsteban-Parra and Y Castro-Dıez ldquoEvaluation of WRF meanand extreme precipitation over Spain present climate (1970ndash99)rdquo Journal of Climate vol 25 no 14 pp 4883ndash4895 2012

[5] R M Cardoso P M M Soares P M A Miranda and MBelo-Pereira ldquoWRF high resolution simulation of Iberianmeanand extreme precipitation climaterdquo International Journal ofClimatology vol 33 no 11 pp 2591ndash2608 2013

[6] L Calvetti and A J Pereira Filho ldquoEnsemble hydrometeoro-logical forecasts using WRF hourly QPF and topmodel for amiddle watershedrdquo Advances in Meteorology vol 2014 ArticleID 484120 12 pages 2014

[7] S Q Zhang M Zupanski A Y Hou X Lin and S H CheungldquoAssimilation of precipitation-affected radiances in a cloud-resolving wrf ensemble data assimilation systemrdquo MonthlyWeather Review vol 141 no 2 pp 754ndash772 2013

[8] J Liu M Bray and D Han ldquoExploring the effect of dataassimilation by WRF-3DVar for numerical rainfall predictionwith different types of storm eventsrdquoHydrological Processes vol27 no 25 pp 3627ndash3640 2013

[9] C-H Xiong L-F Zhang J-P Guan J Peng and B ZhangldquoAnalysis and numerical study of a hybrid BGM-3DVAR dataassimilation scheme using satellite radiance data for heavy rainforecastsrdquo Journal of Hydrodynamics vol 25 no 3 pp 430ndash4392013

[10] D M Barker W Huang Y R Guo and Q N Xiao ldquoA three-dimensional (3DVAR) data assimilation system for use withMM5 implementation and initial resultsrdquo Monthly WeatherReview vol 132 pp 897ndash914 2004

[11] Q Xiao and J Sun ldquoMultiple-radar data assimilation and short-range quantitative precipitation forecasting of a squall lineobserved during IHOP 2002rdquoMonthlyWeather Review vol 135no 10 pp 3381ndash3404 2007

[12] Z Sokol ldquoEffects of an assimilation of radar and satellite dataon a very-short range forecast of heavy convective rainfallsrdquoAtmospheric Research vol 93 no 1ndash3 pp 188ndash206 2009

[13] P Bauer P Lopez A Benedetti D Salmond and E MoreauldquoImplementation of 1D+4D-Var assimilation of precipitation-affected microwave radiances at ECMWF I 1D-Varrdquo QuarterlyJournal of the Royal Meteorological Society vol 132 no 620 pp2277ndash2306 2006

[14] P Bauer P Lopez D Salmond A Benedetti S Saarinen andM Bonazzola ldquoImplementation of 1D+4D-Var assimilation of

precipitation-affected microwave radiances at ECMWF II 4D-Varrdquo Quarterly Journal of the Royal Meteorological Society vol132 no 620 pp 2307ndash2332 2006

[15] P Bauer A J Geer P Lopez and D Salmond ldquoDirect 4D-Var assimilation of all-sky radiances Part I ImplementationrdquoQuarterly Journal of the Royal Meteorological Society vol 136no 652 pp 1868ndash1885 2010

[16] A J Geer P Bauer and P Lopez ldquoDirect 4D-Var assimilationof all-sky radiances Part II assessmentrdquo Quarterly Journal ofthe RoyalMeteorological Society vol 136 no 652 pp 1886ndash19052010

[17] Z Sokol and D Rezacova ldquoAssimilation of radar reflectivityinto the LM COSMOmodel with a high horizontal resolutionrdquoMeteorological Applications vol 13 no 4 pp 317ndash330 2006

[18] Z Sokol ldquoAssimilation of extrapolated radar reflectivity into aNWP model and its impact on a precipitation forecast at highresolutionrdquo Atmospheric Research vol 100 no 2-3 pp 201ndash2122011

[19] HWang J Sun S Fan and X-Y Huang ldquoIndirect assimilationof radar reflectivity with WRF 3D-var and its impact onprediction of four summertime convective eventsrdquo Journal ofApplied Meteorology and Climatology vol 52 no 4 pp 889ndash902 2013

[20] B Macpherson ldquoOperational experience with assimilation ofrainfall data in the Met Office mesoscale modelrdquo Meteorologyand Atmospheric Physics vol 76 no 1-2 pp 3ndash8 2001

[21] K Stephan S Klink and C Schraff ldquoAssimilation of radar-derived rain rates into the convective-scale model COSMO-DEat DWDrdquo Quarterly Journal of the Royal Meteorological Societyvol 134 no 634 pp 1315ndash1326 2008

[22] K Alapaty D S Niyogi F Chen P Pyle A Chandrasekar andN Seaman ldquoDevelopment of the flux-adjusting surface dataassimilation system for mesoscale modelsrdquo Journal of AppliedMeteorology and Climatology vol 47 no 9 pp 2331ndash2350 2008

[23] F H Ruggiero G D Modica and A E Lipton ldquoAssimilationof satellite imager data and surface observations to improveanalysis of circulations forced by cloud shading contrastsrdquoMonthly Weather Review vol 128 no 2 pp 434ndash448 2000

[24] J P Hacker and D Rostkier-Edelstein ldquoPBL state estimationwith surface observations a column model and an ensemblefilterrdquo Monthly Weather Review vol 135 no 8 pp 2958ndash29722007

[25] D J Stensrud N Yussouf D C Dowell and M C ConiglioldquoAssimilating surface data into a mesoscale model ensemblecold pool analyses from spring 2007rdquoAtmospheric Research vol93 no 1ndash3 pp 207ndash220 2009

[26] Q Wan and J Xu ldquoA numerical study of the rainstormcharacteristics of the June 2005 flash flood with WRFGSIdata assimilation system over south-east Chinardquo HydrologicalProcesses vol 25 no 8 pp 1327ndash1341 2011

[27] J Liu M Bray and D Han ldquoA study on WRF radar dataassimilation for hydrological rainfall predictionrdquoHydrology andEarth System Sciences vol 17 no 8 pp 3095ndash3110 2013

[28] K Ide P Courtier M Ghil and A C Lorenc ldquoUnified notationfor data assimilation operational sequential and variationalrdquoJournal of the Meteorological Society of Japan vol 75 no 1 pp181ndash189 1997

[29] A C Lorenc ldquoAnalysis methods for numerical weather predic-tionrdquo Quarterly Journal of the Royal Meteorological Society vol112 no 474 pp 1177ndash1194 1966

14 Advances in Meteorology

[30] D F Parrish and J C Derber ldquoThe nationalmeteorological cen-terrsquos spectral statistical interpolation analysis systemrdquo MonthlyWeather Review vol 120 no 8 pp 1747ndash1763 1992

[31] T B Zhao and C B Fu ldquoPreliminary comparison and analysisbetween ERA-40 NCEP-2 reanalysis and observations overChinardquo Climatic and Environmental Research vol 11 no 1 pp14ndash32 2006

[32] L Ma T Zhang Q Li O W Frauenfeld and D QinldquoEvaluation of ERA-40 NCEP-1 and NCEP-2 reanalysis airtemperatures with ground-based measurements in ChinardquoJournal of Geophysical Research Atmospheres (1984ndash2012) vol13 no D15 1984

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ClimatologyJournal of

EcologyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

EarthquakesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

Mining

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

International Journal of

Geophysics

OceanographyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofPetroleum Engineering

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Atmospheric SciencesInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MineralogyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MeteorologyAdvances in

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geological ResearchJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geology Advances in

Advances in Meteorology 3

the iterative minimization of a prescribed cost (or penalty)function [28]

119869 (119909) =1

2(119909 minus 119909

119887)119879

119861minus1(119909 minus 119909

119887)

+1

2(119910 minus 119910

0)119879

119877minus1(119910 minus 119910

0)

(1)

where 119909 is the analysis to be found that minimizes thecost function 119869(119909) 119909119887 is the first guess of the NWP model1199100 is the assimilated observation and 119910 = 119867(119909) is the

model-derived observation transformed from the analysis 119909by the observation operator 119867 for comparison against 1199100The solution for the cost function given by (1) representsa posteriori maximum likelihood (minimum variance) esti-mate of the true state given the two sources of a priori datathe first guess 119909119887 and the observation 1199100 [29] The fit toindividual observation points is weighted by the estimatesof their errors that is 119861 and 119877 which are the backgrounderror covariance matrix and the observation error covariancematrix respectively

The WRF-3DVar system developed by Barker et al [10]is used in this study in tandem with the WRF model forassimilating the satellite radiance data and the traditionalobservations The performance of the data-assimilation sys-tem largely depends on the plausibility of the backgrounderror covariance (BE) that is thematrix119861 in (1) In this studythe ldquoCV5rdquo background error option is used with the controlvariables of stream function unbalanced temperature unbal-anced potential velocity unbalanced surface pressure andpseudo relative humidity The background error covariancematrix is generated via the National Meteorological Center(NMC) method [30] for our own forecasting domain

23 Experimental Design Four sets of experiments have beenconducted using two domains As shown in Table 1 themodel simulation without data assimilation will be referredto as the control experiment (CTRL) Three assimilationexperiments are designed using different observation param-eters from different data sources In the DA-OBS experimentmeasurements of pressure geopotential height temperaturedewpoint temperature wind direction and speed fromNCARare assimilated while in theDA-SAT experiment onlysatellite radiance data are assimilated Finally in the DA-BOTH experiment both NCAR observations and AMSU-A(B) radiance data are assimilated

3 Study Area and Data

As shown in Figure 2 the northeastern of the TibetanPlateau (94∘391015840ndash103∘271015840E 35∘511015840ndash40∘311015840N) range of elevationbetween 758m and 5725m asl is the head of many inlandrivers which play an important role in the hydrology andagriculture in the downstream arid region A total of 43national observation stations have been used to verify thespatial distribution of the simulated precipitation amongwhich the Wulan station has the highest elevation of 3800mandDunhuang has the lowest altitude of 1137mTheobserved

Table 1 Details of the observed data used in the assimilationexperiments

Experiment name Assimilated dataCTRL No data assimilation

DA-OBS NCAR surface and upper-airobservation

DA-SAT AMSU-A and AMSU-B radiancedata

DA-BOTH NCAR observation andAMSU-A (B) radiance data

precipitation at Laohugou station (elevation 4200m) hasbeen used to evaluate the temporal variations of the simulatedrainfall which is measured by T200B every half hour

The initial and boundary conditions necessary to run theWRF are the ERA-Interim data at 1∘ times 1∘ grid resolutionobtained from the European Centre for Medium-RangeWeather Forecasts (ECMWF) rather than the NCEP-NCARFinal Analysis (FNL) data This is because several studieshave demonstrated that the reliability of ERA-Reanalysis datais higher than the NCEP data in China [31 32] In themodel integration the coordinates of the central point are381∘N and 989∘E WRF-3DVar experiments have been con-ducted with modified initial conditions which were obtainedby assimilating other measured data The assimilated dataincludes NCAR surface and upper-air observations andAMSUA and AMSUB radiance data The surface and upper-air data assimilated in this study are obtained from theldquods3370rdquo in the NCAR archives which contain measure-ments of pressure geopotential height temperature dewpoint temperature wind direction and speed from fixed andmobile landsea stations The data are initially downloadedin PREPBUFR format and can be assimilated directly intoWRFDA The AMSU-A and AMSU-B satellite radiance data(NOAA-1516171819) from theNOAAATOVS instrumentscan be read in WRFDA via CRTM202 which is in BUFRformat

4 Results and Discussion

41 Impacts of Data Assimilation on the Precipitation Forecast

411 Rainfall Event In June 2013 two heavy rain eventsoccurred in the northeast of the Tibetan Plateau Thedurations of the events and the maximummean rainfallaccumulation observed by rain gauge network are shownin Table 2 The two events are of different types accordingto the evenness of rainfall distribution in time and spaceFigure 3 illustrates the spatial distribution of the rainfallaccumulation for the durations of the two events whileFigure 4 presents the time series bars and the cumulativecurves of the observed precipitation for the two events atLaohugou station By comparing the evenness of the rainfalldistribution in time and space in Figures 3 and 4 it can befound that Event A has even rainfall distribution neither inspace nor in time The rainfall distribution of Event B is alsouneven in space but continuous and almost constant in time

4 Advances in Meteorology

Table 2 Durations maximummean rainfall accumulation and spatialtemporal evenness of the heavy rain events

Event ID Start time End time Spatial evenness Temporal evenness Maximum precipitation Mean precipitationA 1862013 000 1962013 2400 146 144 425 67B 762013 000 762013 2400 115 063 235 53

LaohugouRain-gauge stations

River

94∘E

40∘N

38∘N

36∘N

96∘E 98∘E 100∘E 102∘E 104∘E

Figure 2 Location map showing the northeastern edge of the Tibetan Plateau and 44 observation stations

03

0

11

17

25

39

52

71

114

228

425

40∘N

38∘N

36∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(a)

0

04

12

19

24

27

59

81

132

17

235

40∘N

38∘N

36∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(b)

Figure 3 Spatial distribution of the rainfall accumulation for the durations of the two events shown in Thiessen polygons (a) Event A (b)Event B

As for the meteorological characteristics of the storm eventsaccording to the weather analysis charts Event A is likelycaused by very strong local convections and Event B may bea stratiform storm

The rainfall evennessunevenness of the two heavy rainevents can be further verified quantitatively by using thecoefficient of variability (CV)

CV = radic 1119873

119873

sum

119894=1

(119909119894

119909minus 1)

2

(2)

where 119909119894 is the rainfall accumulation of each rain gauge 119894(when calculating the evenness in space) or the average areal

rainfall at each time step 119894 (for the evenness in time) 119909 isthe mean value of 119909119894 and 119873 is the total number of raingauges or the total number of the time steps The results ofthe two rainfall events are also shown in Table 2 A largerCV value represents higher variability thus less even rainfalldistribution The CV value of the spatial evenness for EventA is larger than that for Event B which means Event A hashigher variability in space

412 Spatial Distribution of Precipitation In Figure 5 com-pared with CTRL experiment (Figure 5(a)) the forecasts ofprecipitation area and accumulation are increased notice-ably depending on different data assimilation The CTRL

Advances in Meteorology 5

Cum

ulat

ive r

ainf

all (

mm

)25

20

15

10

5

0

Rain

fall

inte

nsity

(mm

3 h

our)

Time (month-day-hour)

0

05

1

15

2

25

3

35

45

4

5

6-18

-16-

18-4

6-18

-76-

18-1

06-

18-1

36-

18-1

66-

18-1

96-

18-2

26-

19-1

6-19

-46-

19-7

6-19

-10

6-19

-13

6-19

-16

6-19

-19

6-19

-22

(a)

Cum

ulat

ive r

ainf

all (

mm

)

0

2

4

6

8

10

12

Rain

fall

inte

nsity

(mm

15

hou

r)

Time (month-day-hour)

005115225335

454

5

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

(b)

Figure 4The time series bars and the cumulative curves of the observed precipitation for the two heavy rainfall events at Laohugou station(a) Event A (b) Event B

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18

(mm)20

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(d)

Figure 5 The forecast precipitation for Event A from the different numerical experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

6 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N∘94 E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(d)

Figure 6 The forecast precipitation for Event B from the different numerical experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

experiment predicts only a small amount of rainfall on thenorthwestern of the study area while in DA-SAT experimentprecipitation area extends to the central and southern regionsof the study area and rainfall accumulation increases sig-nificantly (Figure 5(b)) But in the northwest of the regionthe rainfall accumulation for DA-SAT experiment is similarwith that shown in Figure 5(a) In DA-OBS experiment(Figure 5(c)) precipitation accumulation on the northwest-ern of the region is greater than that in CTRL experimentbut the precipitation area is similar in both experimentsFigure 5(d) shows the result of DA-BOTH experiment inwhich the precipitation area is the largest and accumulationis the biggest compared with the other three experiments

The results for Event B are shown in Figure 6 Theprecipitation area on the northwestern of the study region issignificantly increased by DA-SAT experiment (Figure 6(b))relative to that in CTRL experiment In Figure 6(c)

the precipitation area in DA-OBS experiment is same asthat in Figure 6(a) but the precipitation accumulation isobviously decreased on the southeastern of the regionFigure 6(d) shows the DA-BOTH experiment result whichhas the similar precipitation area and accumulation as shownin DA-SAT experiment

To sum up the data assimilation has important influenceon the precipitation forecast with significantly changes of therainfall area and accumulation It should be noticed that theeffects of the forecast depends on the different rainfall event

The rain-gauge data were used to analyze the changedprecipitation area and accumulation after data assimilationexperiments Figure 7 compares the measured and mod-eled results in the rainfall Event A In CTRL experiment(Figure 7(a)) almost all the triangles lie below the line 119910 = 119909However in DA-SAT experiment the green triangles whichrepresent the observation sites (Qilian Yeniugou Tuole

Advances in Meteorology 7

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(a)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(b)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(c)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)y = x

(d)

Figure 7 Differences between rainfall values measured at the sites and those from the numerical experiments for Event A (a) CTRL (b)DA-SAT (c) DA-OBS (d) DA-BOTH

and Wulan) in the central and southern regions of the studyarea are closer to the line 119910 = 119909 (Figure 7(b)) That is tosay DA-SAT experiment improves the predicted accuracy forthe central and southern regions of the study area throughproviding greater precipitation accumulation Similarly inFigure 7(c) for DA-OBS experiment the red triangles whichrepresent the observation sites (Jiuquan Yumenzheng Anxiand Dunhuang) in northwestern of the region are closer tothe line119910 = 119909 than that is shown inCTRL experiment InDA-BOTH experiment all the red and green triangles are closestto the line 119910 = 119909 (Figure 7(d))

It is noticed that in Figure 8 many more triangles forEvent B lie above the 119910 = 119909 line than Event A Asshown in CTRL experiment (Figure 8(a)) the green triangleswhich represent observation sites (Hezuo Lintao linxia andYuzhong) on the southeastern of the study area are lie aboveand far away from the 119910 = 119909 line After data assimilationthe green triangles in three data assimilation experimentsare closer to the 119910 = 119909 line in Figures 8(b)ndash8(d) Thered triangles represent the observation sites (Jiuquan Tuole

Dingxin and Jinta) in the northwestern of the study areaIn DA-SAT experiment (Figure 8(b)) two red triangles withmore than 10mm of rainfall are closer to the 119910 = 119909 line whilein DA-OBS experiment (Figure 8(c)) other two red triangleswith less than 5mm of rainfall are closer to the 119910 = 119909 linecompared with that in Figure 8(a) In DA-BOTH experiment(Figure 8(d)) the results are similar to that as shown in DA-SAT experiment except the red triangles are closer to the119910 = 119909 line

The above results suggest that data assimilation havepositive effects on the rainfall forecast through changingthe rainfall area and accumulation for the study area Andfor the two events with different evenness in time andspace the results after data assimilation are different ForEvent A data assimilation experiments provide larger rainfallareas and accumulation to improve the predicted accuracyAnd for Event B data assimilation experiments improve theaccuracy on the southeastern of the study area by providingsmaller precipitation accumulation This means that CTRLexperiment underestimates the convective rainfall for Event

8 Advances in Meteorology

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(a)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(b)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(c)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)y = x

(d)

Figure 8 Differences between rainfall values measured at the sites and those from the numerical experiments for Event B (a) CTRL (b)DA-SAT (c) DA-OBS (d) DA-BOTH

A and overestimates the stratiform rainfall of Event B on thesoutheastern of the region

In order to evaluate quantitatively the impact of dataassimilation experiments which have positive effects on therainfall forecast rain-gauge data from 43 sites were usedTable 3 shows the comparison between the measured andmodeled results The CTRL experiments have the lowestprecision of simulation for both Event A and Event B whose1198772 are the smallest and RMSE are the biggest compared with

three data assimilation experiments The greatest improve-ment appears in DA-BOTH experiment for the two eventsFor Event A the 1198772 for DA-BOTH experiment is increasedby 046 and its RMSE is reduced by 47 and for Event Bthe 1198772 for DA-BOTH experiment is increased by 032 andits RMSE is reduced by 06 Additionally it is found that inDA-SAT experiment the improvement is bigger than DA-OBS experimentTheME of rainfall accumulation shows thatfor the entire rainfall range it has positives value for Event Abut has negative value for Event B This means that the fourexperiments produce underestimated rainfall accumulationsfor Event A but overestimated rainfall accumulations for

Event BThat is because in the precipitation ranges of 0ndash5mmand 5ndash10mm almost all the ME values are positive for EventA and all the ME are negative for Event B

In this section for the two events it is found that theforecast precision in DA-SAT experiment is higher than thatin DA-OBS experiment This may be because there are fewstations located on the high-altitude region of northeasternTibetan Plateau DA-BOTH experiment has the biggestprecision compared with the other two data assimilationexperiments which means that the improvement after theassimilation of the satellite radiances combined with surfaceand upper-air meteorological observations is greater thanthat after only assimilating satellite radiances

413 Temporal Variations of Precipitation The greater dif-ference between Event A and Event B is the rainfall dis-tribution in time Figure 9(a) shows the rainfall intensitiesand cumulative curves of the measured and modeled resultsfor Event A at Laohugou station The cumulative curvesof CTRL experiment are lie below and far away from thatof the measured values However during the period from

Advances in Meteorology 9

Table 3 Differences between observation and numerical experiments

Event ID Experiments 1198772 RMSE 0ndash5mm ME

5ndash10mm 10ndash425 (235) 0ndash425 (235)

Event A

CTRL 024 105 minus015 42 216 56DA-SAT 043 81 12 31 172 35DA-OBS 033 92 04 39 189 46DA-BOTH 070 58 minus03 11 95 18

Event B

CTRL 032 55 minus28 minus14 84 minus02DA-SAT 059 50 minus24 minus16 54 minus05DA-OBS 048 54 minus19 minus23 46 minus07DA-BOTH 064 49 minus25 minus13 50 minus06

0

1

2

3

4

5

6

7

8

9

100

5

10

15

20

25

6-18

-36-

18-6

6-18

-96-

18-1

26-

18-1

56-

18-1

86-

18-2

16-

18-2

46-

19-3

6-19

-66-

19-9

6-19

-12

6-19

-15

6-19

-18

6-19

-21

6-19

-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRL

DA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

20

40

60

80

100

120

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)Measured valueCTRLDA-OBS

DA-SATDA-BOTH

Rain

fall

inte

nsity

(mm

15

hou

r)

(b)

Figure 9The rainfall intensities and cumulative curves of the values measured at the sites and the simulated results from the four numericalexperiments (a) Event A (b) Event B

000 to 2400 on July 19 the cumulative curves for DA-SATand DA-SAT experiments are closer to that of the measuredvalues compared with the other two experiments This maybe because the maximum precipitation intensities simulatedby DA-BOTH and DA-SAT experiments are closer to theobserved values from 600 to 900 on July 19

In Figure 9(b) for event B the precipitation intensityin CTRL experiment is significantly larger than the mea-sured values in first 6 hours Therefore the cumulativecurve in the CTRL experiment is far above that of themeasured values due to the inaccurate initial field Afterdata assimilation the rainfall intensities in DA-SAT and DA-BOTH experiments are closer to the observations in first 6hours while the rainfall intensity in DA-OBS experiment ismuch less than the observations As a result the cumulativecurve in DA-BOTH experiment is closest to the curve

of the observed values DA-SAT experiment is next andDA-OBS experiment is farthest It is noteworthy that forEvent B the cumulative curve in DA-SAT experiment iscloser to that of the observed values in comparison withEvent A

The absolute error (AE) has been calculated for bothEventA andEvent B In Figure 10(a) for EventA theAEof thenumerical experiments becomes bigger and bigger with theincrease of rainfall intensity while the values in DA-BOTHandDA-SAT experiments are relatively smaller Itmeans thatfor the convective rainfall like Event A WRF model failsin capturing the whole process of the event However inFigure 10(b) for Event B with even distribution in time theAE of the numerical experiments is relatively constant whilethe values for CTRL and DA-OBS experiments are relativelylarger

10 Advances in Meteorology

0

2

4

6

8

10

12

140

2

4

6

8

10

12

14

06-1

8-3

06-1

8-6

06-1

8-9

06-1

8-12

06-1

8-15

06-1

8-18

06-1

8-21

06-1

8-24

06-1

9-3

06-1

9-6

06-1

9-9

06-1

9-12

06-1

9-15

06-1

9-18

06-1

9-21

06-1

9-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

10

20

30

40

50

60

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

15

hou

r)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(b)

Figure 10TheAEof rainfall intensity between the valuesmeasured at the sites and the simulated results from the four numerical experiments(a) Event A (b) Event B

The results shows that for the temporal variation ofrainfall the improvement in DA-SAT experiment is betterthan that in DA-OBS experiment due to the lack of groundobservational data at high-altitude region And for Event Bthe improvement in DA-SAT is greater than Event A

42 Impacts of Data Assimilation on the Initial Field Fromthe above analysis it is clear that data assimilation exper-iments can improve rainfall forecast The purpose of dataassimilation is to acquire accurate initial fields for numericalmodels To reveal the impact of data assimilation experimentson the initial fields EventA is chosen as an example to analyzethe difference in the initial fields between data assimilationand control experiments The geopotential height and mois-ture flux at 850 hPa are selected for analysis

Figure 11 shows the geopotential height differencesbetween data assimilation and control experimentsCompared with CTRL experiment (Figure 11(a)) thereis no evident change of geopotential height in DA-OBS experiment (Figure 11(b)) or DA-SAT experiment(Figure 11(c)) Figure 8(d) shows the geopotential heightfor DA-BOTH experiment which is much smaller thanthe values from CTRL experiment The region with thelargest changes is mainly located in Qinghai Provinceand the northwest of Gansu Province which is consistentwith the improved forecasts of precipitation area andaccumulation in DA-BOTH experiment As can be seen inFigure 12 after data assimilation the moisture flux changenoticeably in the study area in comparison with the controlexperiment (Figure 12(a)) The moisture flux of DA-SATexperiment is markedly increased in the north of the studyarea (Figure 12(b)) It may bring more precipitation in the

northeast of Qinghai Province In Figure 12(c) there is aclear flow of moisture to the northwestern part of GansuProvince which results in the heavy rain The biggest changeof moisture flux occurs in Figure 12(d) for DA-BOTHexperiment which covers the largest area compared withother numerical experiments and it could give greatestestimated precipitation in northeastern Tibetan Plateau

5 Conclusions

In this study the 3DVar assimilation system is used toimprove the forecasting of heavy rain in spatial and timedistribution in northeastern Tibetan Plateau A control andthree data assimilation experiments were designed Theseexperiments demonstrate that the rain forecast is significantlyimproved after data assimilation through enhancing theaccuracy of the initial field Two heavy rainfall events withdifferent evenness in space and time are used to examine theimprovement for rainfall forecast

For the spatial distribution data assimilation experimentschanged significantly the rainfall area and accumulationfor the study area For Event A the data assimilationexperiments provide larger rainfall areas and accumulationwhile for Event B the data assimilation experiments providesmaller precipitation accumulation on the southeastern ofthe study area Rain-gauge data from 43 sites were usedto evaluate the impact after data assimilation The resultsshows that DA-BOTH experiment has the highest predictionaccuracy followed by the DA-SAT experiment and thenDA-OBS experiment while the CTRL experiment has thelowest accuracy For the temporal variation the forecastresults have great differences between Event A and Event B

Advances in Meteorology 11

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(d)

Figure 11 Geopotential height initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

The satellite radiances for Event B have greater positive effectthan that for Event A For the two events the cumulativecurve in DA-BOTH experiment is closest to the curveof the measured values while in DA-OBS experiment theaccumulation curves are far below to the curve of themeasured values due to the lack of ground observations inthe high-altitude region In conclusion both for the spatialand temporal distribution of rainfall the satellite radianceshave greater effect than surface and upper-air meteorologicalobservations in high-latitude regions of the northeasternedge and the improvement for the assimilation of satelliteradiances together with surface and upper-air meteorologicalobservations is greater than that for assimilating either satel-lite radiances ormeteorology observations Event A is chosento evaluate the impact of data assimilation on the initialfields It is found that better and more detailed informationis added to the initial fields (eg geopotential height andmoisture flux) especially on the northwest of Gansu Province

andQinghai Province where the forecasts of precipitation areimproved significantly

It should be mentioned that many other assimilationtechniques such as 4DVar EnKF and so forth need to betested in this study These approaches have great potentialalthough they currently suffer from unaffordable computercosts Meanwhile we need to analyse the impact of differentparameterization schemes on the WRF model after dataassimilation because in this study the conclusions drawn aresubject to the specific parameterization schemes used here

Appendix

Statistic Calculation

The coefficient of determination (1198772) root mean squaredeviation (RMSE) mean error (ME) and absolute error (AE)

12 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(d)

Figure 12 Moisture flux initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

were used to evaluate the impact of the DA experiments inthis study They are defined as

119877

=

(1119873)sum119873

119894=1(119865119894 minus 119865) (119874119894 minus 119874)

radic(1119873)sum119873

119894=1(119865119894 minus 119865)

2radic(1119873)sum

119873

119894=1(119874119894 minus 119874)

2

RMSE = radic 1119873

119873

sum

119894=1

(119865119894 minus 119874119894)2

ME = 1119873

119873

sum

119894=1

(119874119894 minus 119865119894)

AE = 1119873

119873

sum

119894=1

1003816100381610038161003816119865119894 minus 1198741198941003816100381610038161003816

(A1)

Here 119865 is simulated value 119874 is observed value 119873 is thenumber of sites

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work was supported by the National natural ScienceFoundation of China (41190080 40971290) Strategic LeadingTechnology Special Fund of Chinese Academy of Sciences(XDB03030204) the National Scientific and TechnologicalSupport Project (2013BAB05B03) and the Opening Fund ofKey Laboratory of Land Surface Process and Climate Changein Cold and Arid Regions CAS (LPCC201205)

Advances in Meteorology 13

References

[1] P Chambon S Q Zhang A Y Hou M Zupanski and SCheung ldquoAssessing the impact of pre-GPMmicrowave precipi-tation observations in the GoddardWRF ensemble data assim-ilation systemrdquo Quarterly Journal of the Royal MeteorologicalSociety vol 140 no 681 pp 1219ndash1235 2013

[2] J-H Seo Y H Lee and Y-H Kim ldquoFeature selection forvery short-term heavy rainfall prediction using evolutionarycomputationrdquo Advances in Meteorology vol 2014 Article ID203545 15 pages 2014

[3] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M JEsteban-Parra and Y Castro-Dıez ldquoHigh-resolution projec-tions of mean and extreme precipitation over Spain usingthe WRF model (2070ndash2099 versus 1970ndash1999)rdquo Journal ofGeophysical Research D Atmospheres vol 117 no 12 Article IDD12108 2012

[4] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M J SEsteban-Parra and Y Castro-Dıez ldquoEvaluation of WRF meanand extreme precipitation over Spain present climate (1970ndash99)rdquo Journal of Climate vol 25 no 14 pp 4883ndash4895 2012

[5] R M Cardoso P M M Soares P M A Miranda and MBelo-Pereira ldquoWRF high resolution simulation of Iberianmeanand extreme precipitation climaterdquo International Journal ofClimatology vol 33 no 11 pp 2591ndash2608 2013

[6] L Calvetti and A J Pereira Filho ldquoEnsemble hydrometeoro-logical forecasts using WRF hourly QPF and topmodel for amiddle watershedrdquo Advances in Meteorology vol 2014 ArticleID 484120 12 pages 2014

[7] S Q Zhang M Zupanski A Y Hou X Lin and S H CheungldquoAssimilation of precipitation-affected radiances in a cloud-resolving wrf ensemble data assimilation systemrdquo MonthlyWeather Review vol 141 no 2 pp 754ndash772 2013

[8] J Liu M Bray and D Han ldquoExploring the effect of dataassimilation by WRF-3DVar for numerical rainfall predictionwith different types of storm eventsrdquoHydrological Processes vol27 no 25 pp 3627ndash3640 2013

[9] C-H Xiong L-F Zhang J-P Guan J Peng and B ZhangldquoAnalysis and numerical study of a hybrid BGM-3DVAR dataassimilation scheme using satellite radiance data for heavy rainforecastsrdquo Journal of Hydrodynamics vol 25 no 3 pp 430ndash4392013

[10] D M Barker W Huang Y R Guo and Q N Xiao ldquoA three-dimensional (3DVAR) data assimilation system for use withMM5 implementation and initial resultsrdquo Monthly WeatherReview vol 132 pp 897ndash914 2004

[11] Q Xiao and J Sun ldquoMultiple-radar data assimilation and short-range quantitative precipitation forecasting of a squall lineobserved during IHOP 2002rdquoMonthlyWeather Review vol 135no 10 pp 3381ndash3404 2007

[12] Z Sokol ldquoEffects of an assimilation of radar and satellite dataon a very-short range forecast of heavy convective rainfallsrdquoAtmospheric Research vol 93 no 1ndash3 pp 188ndash206 2009

[13] P Bauer P Lopez A Benedetti D Salmond and E MoreauldquoImplementation of 1D+4D-Var assimilation of precipitation-affected microwave radiances at ECMWF I 1D-Varrdquo QuarterlyJournal of the Royal Meteorological Society vol 132 no 620 pp2277ndash2306 2006

[14] P Bauer P Lopez D Salmond A Benedetti S Saarinen andM Bonazzola ldquoImplementation of 1D+4D-Var assimilation of

precipitation-affected microwave radiances at ECMWF II 4D-Varrdquo Quarterly Journal of the Royal Meteorological Society vol132 no 620 pp 2307ndash2332 2006

[15] P Bauer A J Geer P Lopez and D Salmond ldquoDirect 4D-Var assimilation of all-sky radiances Part I ImplementationrdquoQuarterly Journal of the Royal Meteorological Society vol 136no 652 pp 1868ndash1885 2010

[16] A J Geer P Bauer and P Lopez ldquoDirect 4D-Var assimilationof all-sky radiances Part II assessmentrdquo Quarterly Journal ofthe RoyalMeteorological Society vol 136 no 652 pp 1886ndash19052010

[17] Z Sokol and D Rezacova ldquoAssimilation of radar reflectivityinto the LM COSMOmodel with a high horizontal resolutionrdquoMeteorological Applications vol 13 no 4 pp 317ndash330 2006

[18] Z Sokol ldquoAssimilation of extrapolated radar reflectivity into aNWP model and its impact on a precipitation forecast at highresolutionrdquo Atmospheric Research vol 100 no 2-3 pp 201ndash2122011

[19] HWang J Sun S Fan and X-Y Huang ldquoIndirect assimilationof radar reflectivity with WRF 3D-var and its impact onprediction of four summertime convective eventsrdquo Journal ofApplied Meteorology and Climatology vol 52 no 4 pp 889ndash902 2013

[20] B Macpherson ldquoOperational experience with assimilation ofrainfall data in the Met Office mesoscale modelrdquo Meteorologyand Atmospheric Physics vol 76 no 1-2 pp 3ndash8 2001

[21] K Stephan S Klink and C Schraff ldquoAssimilation of radar-derived rain rates into the convective-scale model COSMO-DEat DWDrdquo Quarterly Journal of the Royal Meteorological Societyvol 134 no 634 pp 1315ndash1326 2008

[22] K Alapaty D S Niyogi F Chen P Pyle A Chandrasekar andN Seaman ldquoDevelopment of the flux-adjusting surface dataassimilation system for mesoscale modelsrdquo Journal of AppliedMeteorology and Climatology vol 47 no 9 pp 2331ndash2350 2008

[23] F H Ruggiero G D Modica and A E Lipton ldquoAssimilationof satellite imager data and surface observations to improveanalysis of circulations forced by cloud shading contrastsrdquoMonthly Weather Review vol 128 no 2 pp 434ndash448 2000

[24] J P Hacker and D Rostkier-Edelstein ldquoPBL state estimationwith surface observations a column model and an ensemblefilterrdquo Monthly Weather Review vol 135 no 8 pp 2958ndash29722007

[25] D J Stensrud N Yussouf D C Dowell and M C ConiglioldquoAssimilating surface data into a mesoscale model ensemblecold pool analyses from spring 2007rdquoAtmospheric Research vol93 no 1ndash3 pp 207ndash220 2009

[26] Q Wan and J Xu ldquoA numerical study of the rainstormcharacteristics of the June 2005 flash flood with WRFGSIdata assimilation system over south-east Chinardquo HydrologicalProcesses vol 25 no 8 pp 1327ndash1341 2011

[27] J Liu M Bray and D Han ldquoA study on WRF radar dataassimilation for hydrological rainfall predictionrdquoHydrology andEarth System Sciences vol 17 no 8 pp 3095ndash3110 2013

[28] K Ide P Courtier M Ghil and A C Lorenc ldquoUnified notationfor data assimilation operational sequential and variationalrdquoJournal of the Meteorological Society of Japan vol 75 no 1 pp181ndash189 1997

[29] A C Lorenc ldquoAnalysis methods for numerical weather predic-tionrdquo Quarterly Journal of the Royal Meteorological Society vol112 no 474 pp 1177ndash1194 1966

14 Advances in Meteorology

[30] D F Parrish and J C Derber ldquoThe nationalmeteorological cen-terrsquos spectral statistical interpolation analysis systemrdquo MonthlyWeather Review vol 120 no 8 pp 1747ndash1763 1992

[31] T B Zhao and C B Fu ldquoPreliminary comparison and analysisbetween ERA-40 NCEP-2 reanalysis and observations overChinardquo Climatic and Environmental Research vol 11 no 1 pp14ndash32 2006

[32] L Ma T Zhang Q Li O W Frauenfeld and D QinldquoEvaluation of ERA-40 NCEP-1 and NCEP-2 reanalysis airtemperatures with ground-based measurements in ChinardquoJournal of Geophysical Research Atmospheres (1984ndash2012) vol13 no D15 1984

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ClimatologyJournal of

EcologyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

EarthquakesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

Mining

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

International Journal of

Geophysics

OceanographyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofPetroleum Engineering

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Atmospheric SciencesInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MineralogyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MeteorologyAdvances in

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geological ResearchJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geology Advances in

4 Advances in Meteorology

Table 2 Durations maximummean rainfall accumulation and spatialtemporal evenness of the heavy rain events

Event ID Start time End time Spatial evenness Temporal evenness Maximum precipitation Mean precipitationA 1862013 000 1962013 2400 146 144 425 67B 762013 000 762013 2400 115 063 235 53

LaohugouRain-gauge stations

River

94∘E

40∘N

38∘N

36∘N

96∘E 98∘E 100∘E 102∘E 104∘E

Figure 2 Location map showing the northeastern edge of the Tibetan Plateau and 44 observation stations

03

0

11

17

25

39

52

71

114

228

425

40∘N

38∘N

36∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(a)

0

04

12

19

24

27

59

81

132

17

235

40∘N

38∘N

36∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(b)

Figure 3 Spatial distribution of the rainfall accumulation for the durations of the two events shown in Thiessen polygons (a) Event A (b)Event B

As for the meteorological characteristics of the storm eventsaccording to the weather analysis charts Event A is likelycaused by very strong local convections and Event B may bea stratiform storm

The rainfall evennessunevenness of the two heavy rainevents can be further verified quantitatively by using thecoefficient of variability (CV)

CV = radic 1119873

119873

sum

119894=1

(119909119894

119909minus 1)

2

(2)

where 119909119894 is the rainfall accumulation of each rain gauge 119894(when calculating the evenness in space) or the average areal

rainfall at each time step 119894 (for the evenness in time) 119909 isthe mean value of 119909119894 and 119873 is the total number of raingauges or the total number of the time steps The results ofthe two rainfall events are also shown in Table 2 A largerCV value represents higher variability thus less even rainfalldistribution The CV value of the spatial evenness for EventA is larger than that for Event B which means Event A hashigher variability in space

412 Spatial Distribution of Precipitation In Figure 5 com-pared with CTRL experiment (Figure 5(a)) the forecasts ofprecipitation area and accumulation are increased notice-ably depending on different data assimilation The CTRL

Advances in Meteorology 5

Cum

ulat

ive r

ainf

all (

mm

)25

20

15

10

5

0

Rain

fall

inte

nsity

(mm

3 h

our)

Time (month-day-hour)

0

05

1

15

2

25

3

35

45

4

5

6-18

-16-

18-4

6-18

-76-

18-1

06-

18-1

36-

18-1

66-

18-1

96-

18-2

26-

19-1

6-19

-46-

19-7

6-19

-10

6-19

-13

6-19

-16

6-19

-19

6-19

-22

(a)

Cum

ulat

ive r

ainf

all (

mm

)

0

2

4

6

8

10

12

Rain

fall

inte

nsity

(mm

15

hou

r)

Time (month-day-hour)

005115225335

454

5

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

(b)

Figure 4The time series bars and the cumulative curves of the observed precipitation for the two heavy rainfall events at Laohugou station(a) Event A (b) Event B

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18

(mm)20

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(d)

Figure 5 The forecast precipitation for Event A from the different numerical experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

6 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N∘94 E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(d)

Figure 6 The forecast precipitation for Event B from the different numerical experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

experiment predicts only a small amount of rainfall on thenorthwestern of the study area while in DA-SAT experimentprecipitation area extends to the central and southern regionsof the study area and rainfall accumulation increases sig-nificantly (Figure 5(b)) But in the northwest of the regionthe rainfall accumulation for DA-SAT experiment is similarwith that shown in Figure 5(a) In DA-OBS experiment(Figure 5(c)) precipitation accumulation on the northwest-ern of the region is greater than that in CTRL experimentbut the precipitation area is similar in both experimentsFigure 5(d) shows the result of DA-BOTH experiment inwhich the precipitation area is the largest and accumulationis the biggest compared with the other three experiments

The results for Event B are shown in Figure 6 Theprecipitation area on the northwestern of the study region issignificantly increased by DA-SAT experiment (Figure 6(b))relative to that in CTRL experiment In Figure 6(c)

the precipitation area in DA-OBS experiment is same asthat in Figure 6(a) but the precipitation accumulation isobviously decreased on the southeastern of the regionFigure 6(d) shows the DA-BOTH experiment result whichhas the similar precipitation area and accumulation as shownin DA-SAT experiment

To sum up the data assimilation has important influenceon the precipitation forecast with significantly changes of therainfall area and accumulation It should be noticed that theeffects of the forecast depends on the different rainfall event

The rain-gauge data were used to analyze the changedprecipitation area and accumulation after data assimilationexperiments Figure 7 compares the measured and mod-eled results in the rainfall Event A In CTRL experiment(Figure 7(a)) almost all the triangles lie below the line 119910 = 119909However in DA-SAT experiment the green triangles whichrepresent the observation sites (Qilian Yeniugou Tuole

Advances in Meteorology 7

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(a)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(b)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(c)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)y = x

(d)

Figure 7 Differences between rainfall values measured at the sites and those from the numerical experiments for Event A (a) CTRL (b)DA-SAT (c) DA-OBS (d) DA-BOTH

and Wulan) in the central and southern regions of the studyarea are closer to the line 119910 = 119909 (Figure 7(b)) That is tosay DA-SAT experiment improves the predicted accuracy forthe central and southern regions of the study area throughproviding greater precipitation accumulation Similarly inFigure 7(c) for DA-OBS experiment the red triangles whichrepresent the observation sites (Jiuquan Yumenzheng Anxiand Dunhuang) in northwestern of the region are closer tothe line119910 = 119909 than that is shown inCTRL experiment InDA-BOTH experiment all the red and green triangles are closestto the line 119910 = 119909 (Figure 7(d))

It is noticed that in Figure 8 many more triangles forEvent B lie above the 119910 = 119909 line than Event A Asshown in CTRL experiment (Figure 8(a)) the green triangleswhich represent observation sites (Hezuo Lintao linxia andYuzhong) on the southeastern of the study area are lie aboveand far away from the 119910 = 119909 line After data assimilationthe green triangles in three data assimilation experimentsare closer to the 119910 = 119909 line in Figures 8(b)ndash8(d) Thered triangles represent the observation sites (Jiuquan Tuole

Dingxin and Jinta) in the northwestern of the study areaIn DA-SAT experiment (Figure 8(b)) two red triangles withmore than 10mm of rainfall are closer to the 119910 = 119909 line whilein DA-OBS experiment (Figure 8(c)) other two red triangleswith less than 5mm of rainfall are closer to the 119910 = 119909 linecompared with that in Figure 8(a) In DA-BOTH experiment(Figure 8(d)) the results are similar to that as shown in DA-SAT experiment except the red triangles are closer to the119910 = 119909 line

The above results suggest that data assimilation havepositive effects on the rainfall forecast through changingthe rainfall area and accumulation for the study area Andfor the two events with different evenness in time andspace the results after data assimilation are different ForEvent A data assimilation experiments provide larger rainfallareas and accumulation to improve the predicted accuracyAnd for Event B data assimilation experiments improve theaccuracy on the southeastern of the study area by providingsmaller precipitation accumulation This means that CTRLexperiment underestimates the convective rainfall for Event

8 Advances in Meteorology

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(a)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(b)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(c)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)y = x

(d)

Figure 8 Differences between rainfall values measured at the sites and those from the numerical experiments for Event B (a) CTRL (b)DA-SAT (c) DA-OBS (d) DA-BOTH

A and overestimates the stratiform rainfall of Event B on thesoutheastern of the region

In order to evaluate quantitatively the impact of dataassimilation experiments which have positive effects on therainfall forecast rain-gauge data from 43 sites were usedTable 3 shows the comparison between the measured andmodeled results The CTRL experiments have the lowestprecision of simulation for both Event A and Event B whose1198772 are the smallest and RMSE are the biggest compared with

three data assimilation experiments The greatest improve-ment appears in DA-BOTH experiment for the two eventsFor Event A the 1198772 for DA-BOTH experiment is increasedby 046 and its RMSE is reduced by 47 and for Event Bthe 1198772 for DA-BOTH experiment is increased by 032 andits RMSE is reduced by 06 Additionally it is found that inDA-SAT experiment the improvement is bigger than DA-OBS experimentTheME of rainfall accumulation shows thatfor the entire rainfall range it has positives value for Event Abut has negative value for Event B This means that the fourexperiments produce underestimated rainfall accumulationsfor Event A but overestimated rainfall accumulations for

Event BThat is because in the precipitation ranges of 0ndash5mmand 5ndash10mm almost all the ME values are positive for EventA and all the ME are negative for Event B

In this section for the two events it is found that theforecast precision in DA-SAT experiment is higher than thatin DA-OBS experiment This may be because there are fewstations located on the high-altitude region of northeasternTibetan Plateau DA-BOTH experiment has the biggestprecision compared with the other two data assimilationexperiments which means that the improvement after theassimilation of the satellite radiances combined with surfaceand upper-air meteorological observations is greater thanthat after only assimilating satellite radiances

413 Temporal Variations of Precipitation The greater dif-ference between Event A and Event B is the rainfall dis-tribution in time Figure 9(a) shows the rainfall intensitiesand cumulative curves of the measured and modeled resultsfor Event A at Laohugou station The cumulative curvesof CTRL experiment are lie below and far away from thatof the measured values However during the period from

Advances in Meteorology 9

Table 3 Differences between observation and numerical experiments

Event ID Experiments 1198772 RMSE 0ndash5mm ME

5ndash10mm 10ndash425 (235) 0ndash425 (235)

Event A

CTRL 024 105 minus015 42 216 56DA-SAT 043 81 12 31 172 35DA-OBS 033 92 04 39 189 46DA-BOTH 070 58 minus03 11 95 18

Event B

CTRL 032 55 minus28 minus14 84 minus02DA-SAT 059 50 minus24 minus16 54 minus05DA-OBS 048 54 minus19 minus23 46 minus07DA-BOTH 064 49 minus25 minus13 50 minus06

0

1

2

3

4

5

6

7

8

9

100

5

10

15

20

25

6-18

-36-

18-6

6-18

-96-

18-1

26-

18-1

56-

18-1

86-

18-2

16-

18-2

46-

19-3

6-19

-66-

19-9

6-19

-12

6-19

-15

6-19

-18

6-19

-21

6-19

-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRL

DA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

20

40

60

80

100

120

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)Measured valueCTRLDA-OBS

DA-SATDA-BOTH

Rain

fall

inte

nsity

(mm

15

hou

r)

(b)

Figure 9The rainfall intensities and cumulative curves of the values measured at the sites and the simulated results from the four numericalexperiments (a) Event A (b) Event B

000 to 2400 on July 19 the cumulative curves for DA-SATand DA-SAT experiments are closer to that of the measuredvalues compared with the other two experiments This maybe because the maximum precipitation intensities simulatedby DA-BOTH and DA-SAT experiments are closer to theobserved values from 600 to 900 on July 19

In Figure 9(b) for event B the precipitation intensityin CTRL experiment is significantly larger than the mea-sured values in first 6 hours Therefore the cumulativecurve in the CTRL experiment is far above that of themeasured values due to the inaccurate initial field Afterdata assimilation the rainfall intensities in DA-SAT and DA-BOTH experiments are closer to the observations in first 6hours while the rainfall intensity in DA-OBS experiment ismuch less than the observations As a result the cumulativecurve in DA-BOTH experiment is closest to the curve

of the observed values DA-SAT experiment is next andDA-OBS experiment is farthest It is noteworthy that forEvent B the cumulative curve in DA-SAT experiment iscloser to that of the observed values in comparison withEvent A

The absolute error (AE) has been calculated for bothEventA andEvent B In Figure 10(a) for EventA theAEof thenumerical experiments becomes bigger and bigger with theincrease of rainfall intensity while the values in DA-BOTHandDA-SAT experiments are relatively smaller Itmeans thatfor the convective rainfall like Event A WRF model failsin capturing the whole process of the event However inFigure 10(b) for Event B with even distribution in time theAE of the numerical experiments is relatively constant whilethe values for CTRL and DA-OBS experiments are relativelylarger

10 Advances in Meteorology

0

2

4

6

8

10

12

140

2

4

6

8

10

12

14

06-1

8-3

06-1

8-6

06-1

8-9

06-1

8-12

06-1

8-15

06-1

8-18

06-1

8-21

06-1

8-24

06-1

9-3

06-1

9-6

06-1

9-9

06-1

9-12

06-1

9-15

06-1

9-18

06-1

9-21

06-1

9-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

10

20

30

40

50

60

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

15

hou

r)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(b)

Figure 10TheAEof rainfall intensity between the valuesmeasured at the sites and the simulated results from the four numerical experiments(a) Event A (b) Event B

The results shows that for the temporal variation ofrainfall the improvement in DA-SAT experiment is betterthan that in DA-OBS experiment due to the lack of groundobservational data at high-altitude region And for Event Bthe improvement in DA-SAT is greater than Event A

42 Impacts of Data Assimilation on the Initial Field Fromthe above analysis it is clear that data assimilation exper-iments can improve rainfall forecast The purpose of dataassimilation is to acquire accurate initial fields for numericalmodels To reveal the impact of data assimilation experimentson the initial fields EventA is chosen as an example to analyzethe difference in the initial fields between data assimilationand control experiments The geopotential height and mois-ture flux at 850 hPa are selected for analysis

Figure 11 shows the geopotential height differencesbetween data assimilation and control experimentsCompared with CTRL experiment (Figure 11(a)) thereis no evident change of geopotential height in DA-OBS experiment (Figure 11(b)) or DA-SAT experiment(Figure 11(c)) Figure 8(d) shows the geopotential heightfor DA-BOTH experiment which is much smaller thanthe values from CTRL experiment The region with thelargest changes is mainly located in Qinghai Provinceand the northwest of Gansu Province which is consistentwith the improved forecasts of precipitation area andaccumulation in DA-BOTH experiment As can be seen inFigure 12 after data assimilation the moisture flux changenoticeably in the study area in comparison with the controlexperiment (Figure 12(a)) The moisture flux of DA-SATexperiment is markedly increased in the north of the studyarea (Figure 12(b)) It may bring more precipitation in the

northeast of Qinghai Province In Figure 12(c) there is aclear flow of moisture to the northwestern part of GansuProvince which results in the heavy rain The biggest changeof moisture flux occurs in Figure 12(d) for DA-BOTHexperiment which covers the largest area compared withother numerical experiments and it could give greatestestimated precipitation in northeastern Tibetan Plateau

5 Conclusions

In this study the 3DVar assimilation system is used toimprove the forecasting of heavy rain in spatial and timedistribution in northeastern Tibetan Plateau A control andthree data assimilation experiments were designed Theseexperiments demonstrate that the rain forecast is significantlyimproved after data assimilation through enhancing theaccuracy of the initial field Two heavy rainfall events withdifferent evenness in space and time are used to examine theimprovement for rainfall forecast

For the spatial distribution data assimilation experimentschanged significantly the rainfall area and accumulationfor the study area For Event A the data assimilationexperiments provide larger rainfall areas and accumulationwhile for Event B the data assimilation experiments providesmaller precipitation accumulation on the southeastern ofthe study area Rain-gauge data from 43 sites were usedto evaluate the impact after data assimilation The resultsshows that DA-BOTH experiment has the highest predictionaccuracy followed by the DA-SAT experiment and thenDA-OBS experiment while the CTRL experiment has thelowest accuracy For the temporal variation the forecastresults have great differences between Event A and Event B

Advances in Meteorology 11

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(d)

Figure 11 Geopotential height initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

The satellite radiances for Event B have greater positive effectthan that for Event A For the two events the cumulativecurve in DA-BOTH experiment is closest to the curveof the measured values while in DA-OBS experiment theaccumulation curves are far below to the curve of themeasured values due to the lack of ground observations inthe high-altitude region In conclusion both for the spatialand temporal distribution of rainfall the satellite radianceshave greater effect than surface and upper-air meteorologicalobservations in high-latitude regions of the northeasternedge and the improvement for the assimilation of satelliteradiances together with surface and upper-air meteorologicalobservations is greater than that for assimilating either satel-lite radiances ormeteorology observations Event A is chosento evaluate the impact of data assimilation on the initialfields It is found that better and more detailed informationis added to the initial fields (eg geopotential height andmoisture flux) especially on the northwest of Gansu Province

andQinghai Province where the forecasts of precipitation areimproved significantly

It should be mentioned that many other assimilationtechniques such as 4DVar EnKF and so forth need to betested in this study These approaches have great potentialalthough they currently suffer from unaffordable computercosts Meanwhile we need to analyse the impact of differentparameterization schemes on the WRF model after dataassimilation because in this study the conclusions drawn aresubject to the specific parameterization schemes used here

Appendix

Statistic Calculation

The coefficient of determination (1198772) root mean squaredeviation (RMSE) mean error (ME) and absolute error (AE)

12 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(d)

Figure 12 Moisture flux initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

were used to evaluate the impact of the DA experiments inthis study They are defined as

119877

=

(1119873)sum119873

119894=1(119865119894 minus 119865) (119874119894 minus 119874)

radic(1119873)sum119873

119894=1(119865119894 minus 119865)

2radic(1119873)sum

119873

119894=1(119874119894 minus 119874)

2

RMSE = radic 1119873

119873

sum

119894=1

(119865119894 minus 119874119894)2

ME = 1119873

119873

sum

119894=1

(119874119894 minus 119865119894)

AE = 1119873

119873

sum

119894=1

1003816100381610038161003816119865119894 minus 1198741198941003816100381610038161003816

(A1)

Here 119865 is simulated value 119874 is observed value 119873 is thenumber of sites

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work was supported by the National natural ScienceFoundation of China (41190080 40971290) Strategic LeadingTechnology Special Fund of Chinese Academy of Sciences(XDB03030204) the National Scientific and TechnologicalSupport Project (2013BAB05B03) and the Opening Fund ofKey Laboratory of Land Surface Process and Climate Changein Cold and Arid Regions CAS (LPCC201205)

Advances in Meteorology 13

References

[1] P Chambon S Q Zhang A Y Hou M Zupanski and SCheung ldquoAssessing the impact of pre-GPMmicrowave precipi-tation observations in the GoddardWRF ensemble data assim-ilation systemrdquo Quarterly Journal of the Royal MeteorologicalSociety vol 140 no 681 pp 1219ndash1235 2013

[2] J-H Seo Y H Lee and Y-H Kim ldquoFeature selection forvery short-term heavy rainfall prediction using evolutionarycomputationrdquo Advances in Meteorology vol 2014 Article ID203545 15 pages 2014

[3] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M JEsteban-Parra and Y Castro-Dıez ldquoHigh-resolution projec-tions of mean and extreme precipitation over Spain usingthe WRF model (2070ndash2099 versus 1970ndash1999)rdquo Journal ofGeophysical Research D Atmospheres vol 117 no 12 Article IDD12108 2012

[4] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M J SEsteban-Parra and Y Castro-Dıez ldquoEvaluation of WRF meanand extreme precipitation over Spain present climate (1970ndash99)rdquo Journal of Climate vol 25 no 14 pp 4883ndash4895 2012

[5] R M Cardoso P M M Soares P M A Miranda and MBelo-Pereira ldquoWRF high resolution simulation of Iberianmeanand extreme precipitation climaterdquo International Journal ofClimatology vol 33 no 11 pp 2591ndash2608 2013

[6] L Calvetti and A J Pereira Filho ldquoEnsemble hydrometeoro-logical forecasts using WRF hourly QPF and topmodel for amiddle watershedrdquo Advances in Meteorology vol 2014 ArticleID 484120 12 pages 2014

[7] S Q Zhang M Zupanski A Y Hou X Lin and S H CheungldquoAssimilation of precipitation-affected radiances in a cloud-resolving wrf ensemble data assimilation systemrdquo MonthlyWeather Review vol 141 no 2 pp 754ndash772 2013

[8] J Liu M Bray and D Han ldquoExploring the effect of dataassimilation by WRF-3DVar for numerical rainfall predictionwith different types of storm eventsrdquoHydrological Processes vol27 no 25 pp 3627ndash3640 2013

[9] C-H Xiong L-F Zhang J-P Guan J Peng and B ZhangldquoAnalysis and numerical study of a hybrid BGM-3DVAR dataassimilation scheme using satellite radiance data for heavy rainforecastsrdquo Journal of Hydrodynamics vol 25 no 3 pp 430ndash4392013

[10] D M Barker W Huang Y R Guo and Q N Xiao ldquoA three-dimensional (3DVAR) data assimilation system for use withMM5 implementation and initial resultsrdquo Monthly WeatherReview vol 132 pp 897ndash914 2004

[11] Q Xiao and J Sun ldquoMultiple-radar data assimilation and short-range quantitative precipitation forecasting of a squall lineobserved during IHOP 2002rdquoMonthlyWeather Review vol 135no 10 pp 3381ndash3404 2007

[12] Z Sokol ldquoEffects of an assimilation of radar and satellite dataon a very-short range forecast of heavy convective rainfallsrdquoAtmospheric Research vol 93 no 1ndash3 pp 188ndash206 2009

[13] P Bauer P Lopez A Benedetti D Salmond and E MoreauldquoImplementation of 1D+4D-Var assimilation of precipitation-affected microwave radiances at ECMWF I 1D-Varrdquo QuarterlyJournal of the Royal Meteorological Society vol 132 no 620 pp2277ndash2306 2006

[14] P Bauer P Lopez D Salmond A Benedetti S Saarinen andM Bonazzola ldquoImplementation of 1D+4D-Var assimilation of

precipitation-affected microwave radiances at ECMWF II 4D-Varrdquo Quarterly Journal of the Royal Meteorological Society vol132 no 620 pp 2307ndash2332 2006

[15] P Bauer A J Geer P Lopez and D Salmond ldquoDirect 4D-Var assimilation of all-sky radiances Part I ImplementationrdquoQuarterly Journal of the Royal Meteorological Society vol 136no 652 pp 1868ndash1885 2010

[16] A J Geer P Bauer and P Lopez ldquoDirect 4D-Var assimilationof all-sky radiances Part II assessmentrdquo Quarterly Journal ofthe RoyalMeteorological Society vol 136 no 652 pp 1886ndash19052010

[17] Z Sokol and D Rezacova ldquoAssimilation of radar reflectivityinto the LM COSMOmodel with a high horizontal resolutionrdquoMeteorological Applications vol 13 no 4 pp 317ndash330 2006

[18] Z Sokol ldquoAssimilation of extrapolated radar reflectivity into aNWP model and its impact on a precipitation forecast at highresolutionrdquo Atmospheric Research vol 100 no 2-3 pp 201ndash2122011

[19] HWang J Sun S Fan and X-Y Huang ldquoIndirect assimilationof radar reflectivity with WRF 3D-var and its impact onprediction of four summertime convective eventsrdquo Journal ofApplied Meteorology and Climatology vol 52 no 4 pp 889ndash902 2013

[20] B Macpherson ldquoOperational experience with assimilation ofrainfall data in the Met Office mesoscale modelrdquo Meteorologyand Atmospheric Physics vol 76 no 1-2 pp 3ndash8 2001

[21] K Stephan S Klink and C Schraff ldquoAssimilation of radar-derived rain rates into the convective-scale model COSMO-DEat DWDrdquo Quarterly Journal of the Royal Meteorological Societyvol 134 no 634 pp 1315ndash1326 2008

[22] K Alapaty D S Niyogi F Chen P Pyle A Chandrasekar andN Seaman ldquoDevelopment of the flux-adjusting surface dataassimilation system for mesoscale modelsrdquo Journal of AppliedMeteorology and Climatology vol 47 no 9 pp 2331ndash2350 2008

[23] F H Ruggiero G D Modica and A E Lipton ldquoAssimilationof satellite imager data and surface observations to improveanalysis of circulations forced by cloud shading contrastsrdquoMonthly Weather Review vol 128 no 2 pp 434ndash448 2000

[24] J P Hacker and D Rostkier-Edelstein ldquoPBL state estimationwith surface observations a column model and an ensemblefilterrdquo Monthly Weather Review vol 135 no 8 pp 2958ndash29722007

[25] D J Stensrud N Yussouf D C Dowell and M C ConiglioldquoAssimilating surface data into a mesoscale model ensemblecold pool analyses from spring 2007rdquoAtmospheric Research vol93 no 1ndash3 pp 207ndash220 2009

[26] Q Wan and J Xu ldquoA numerical study of the rainstormcharacteristics of the June 2005 flash flood with WRFGSIdata assimilation system over south-east Chinardquo HydrologicalProcesses vol 25 no 8 pp 1327ndash1341 2011

[27] J Liu M Bray and D Han ldquoA study on WRF radar dataassimilation for hydrological rainfall predictionrdquoHydrology andEarth System Sciences vol 17 no 8 pp 3095ndash3110 2013

[28] K Ide P Courtier M Ghil and A C Lorenc ldquoUnified notationfor data assimilation operational sequential and variationalrdquoJournal of the Meteorological Society of Japan vol 75 no 1 pp181ndash189 1997

[29] A C Lorenc ldquoAnalysis methods for numerical weather predic-tionrdquo Quarterly Journal of the Royal Meteorological Society vol112 no 474 pp 1177ndash1194 1966

14 Advances in Meteorology

[30] D F Parrish and J C Derber ldquoThe nationalmeteorological cen-terrsquos spectral statistical interpolation analysis systemrdquo MonthlyWeather Review vol 120 no 8 pp 1747ndash1763 1992

[31] T B Zhao and C B Fu ldquoPreliminary comparison and analysisbetween ERA-40 NCEP-2 reanalysis and observations overChinardquo Climatic and Environmental Research vol 11 no 1 pp14ndash32 2006

[32] L Ma T Zhang Q Li O W Frauenfeld and D QinldquoEvaluation of ERA-40 NCEP-1 and NCEP-2 reanalysis airtemperatures with ground-based measurements in ChinardquoJournal of Geophysical Research Atmospheres (1984ndash2012) vol13 no D15 1984

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ClimatologyJournal of

EcologyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

EarthquakesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

Mining

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

International Journal of

Geophysics

OceanographyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofPetroleum Engineering

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Atmospheric SciencesInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MineralogyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MeteorologyAdvances in

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geological ResearchJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geology Advances in

Advances in Meteorology 5

Cum

ulat

ive r

ainf

all (

mm

)25

20

15

10

5

0

Rain

fall

inte

nsity

(mm

3 h

our)

Time (month-day-hour)

0

05

1

15

2

25

3

35

45

4

5

6-18

-16-

18-4

6-18

-76-

18-1

06-

18-1

36-

18-1

66-

18-1

96-

18-2

26-

19-1

6-19

-46-

19-7

6-19

-10

6-19

-13

6-19

-16

6-19

-19

6-19

-22

(a)

Cum

ulat

ive r

ainf

all (

mm

)

0

2

4

6

8

10

12

Rain

fall

inte

nsity

(mm

15

hou

r)

Time (month-day-hour)

005115225335

454

5

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

(b)

Figure 4The time series bars and the cumulative curves of the observed precipitation for the two heavy rainfall events at Laohugou station(a) Event A (b) Event B

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18

(mm)20

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(d)

Figure 5 The forecast precipitation for Event A from the different numerical experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

6 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N∘94 E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(d)

Figure 6 The forecast precipitation for Event B from the different numerical experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

experiment predicts only a small amount of rainfall on thenorthwestern of the study area while in DA-SAT experimentprecipitation area extends to the central and southern regionsof the study area and rainfall accumulation increases sig-nificantly (Figure 5(b)) But in the northwest of the regionthe rainfall accumulation for DA-SAT experiment is similarwith that shown in Figure 5(a) In DA-OBS experiment(Figure 5(c)) precipitation accumulation on the northwest-ern of the region is greater than that in CTRL experimentbut the precipitation area is similar in both experimentsFigure 5(d) shows the result of DA-BOTH experiment inwhich the precipitation area is the largest and accumulationis the biggest compared with the other three experiments

The results for Event B are shown in Figure 6 Theprecipitation area on the northwestern of the study region issignificantly increased by DA-SAT experiment (Figure 6(b))relative to that in CTRL experiment In Figure 6(c)

the precipitation area in DA-OBS experiment is same asthat in Figure 6(a) but the precipitation accumulation isobviously decreased on the southeastern of the regionFigure 6(d) shows the DA-BOTH experiment result whichhas the similar precipitation area and accumulation as shownin DA-SAT experiment

To sum up the data assimilation has important influenceon the precipitation forecast with significantly changes of therainfall area and accumulation It should be noticed that theeffects of the forecast depends on the different rainfall event

The rain-gauge data were used to analyze the changedprecipitation area and accumulation after data assimilationexperiments Figure 7 compares the measured and mod-eled results in the rainfall Event A In CTRL experiment(Figure 7(a)) almost all the triangles lie below the line 119910 = 119909However in DA-SAT experiment the green triangles whichrepresent the observation sites (Qilian Yeniugou Tuole

Advances in Meteorology 7

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(a)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(b)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(c)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)y = x

(d)

Figure 7 Differences between rainfall values measured at the sites and those from the numerical experiments for Event A (a) CTRL (b)DA-SAT (c) DA-OBS (d) DA-BOTH

and Wulan) in the central and southern regions of the studyarea are closer to the line 119910 = 119909 (Figure 7(b)) That is tosay DA-SAT experiment improves the predicted accuracy forthe central and southern regions of the study area throughproviding greater precipitation accumulation Similarly inFigure 7(c) for DA-OBS experiment the red triangles whichrepresent the observation sites (Jiuquan Yumenzheng Anxiand Dunhuang) in northwestern of the region are closer tothe line119910 = 119909 than that is shown inCTRL experiment InDA-BOTH experiment all the red and green triangles are closestto the line 119910 = 119909 (Figure 7(d))

It is noticed that in Figure 8 many more triangles forEvent B lie above the 119910 = 119909 line than Event A Asshown in CTRL experiment (Figure 8(a)) the green triangleswhich represent observation sites (Hezuo Lintao linxia andYuzhong) on the southeastern of the study area are lie aboveand far away from the 119910 = 119909 line After data assimilationthe green triangles in three data assimilation experimentsare closer to the 119910 = 119909 line in Figures 8(b)ndash8(d) Thered triangles represent the observation sites (Jiuquan Tuole

Dingxin and Jinta) in the northwestern of the study areaIn DA-SAT experiment (Figure 8(b)) two red triangles withmore than 10mm of rainfall are closer to the 119910 = 119909 line whilein DA-OBS experiment (Figure 8(c)) other two red triangleswith less than 5mm of rainfall are closer to the 119910 = 119909 linecompared with that in Figure 8(a) In DA-BOTH experiment(Figure 8(d)) the results are similar to that as shown in DA-SAT experiment except the red triangles are closer to the119910 = 119909 line

The above results suggest that data assimilation havepositive effects on the rainfall forecast through changingthe rainfall area and accumulation for the study area Andfor the two events with different evenness in time andspace the results after data assimilation are different ForEvent A data assimilation experiments provide larger rainfallareas and accumulation to improve the predicted accuracyAnd for Event B data assimilation experiments improve theaccuracy on the southeastern of the study area by providingsmaller precipitation accumulation This means that CTRLexperiment underestimates the convective rainfall for Event

8 Advances in Meteorology

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(a)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(b)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(c)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)y = x

(d)

Figure 8 Differences between rainfall values measured at the sites and those from the numerical experiments for Event B (a) CTRL (b)DA-SAT (c) DA-OBS (d) DA-BOTH

A and overestimates the stratiform rainfall of Event B on thesoutheastern of the region

In order to evaluate quantitatively the impact of dataassimilation experiments which have positive effects on therainfall forecast rain-gauge data from 43 sites were usedTable 3 shows the comparison between the measured andmodeled results The CTRL experiments have the lowestprecision of simulation for both Event A and Event B whose1198772 are the smallest and RMSE are the biggest compared with

three data assimilation experiments The greatest improve-ment appears in DA-BOTH experiment for the two eventsFor Event A the 1198772 for DA-BOTH experiment is increasedby 046 and its RMSE is reduced by 47 and for Event Bthe 1198772 for DA-BOTH experiment is increased by 032 andits RMSE is reduced by 06 Additionally it is found that inDA-SAT experiment the improvement is bigger than DA-OBS experimentTheME of rainfall accumulation shows thatfor the entire rainfall range it has positives value for Event Abut has negative value for Event B This means that the fourexperiments produce underestimated rainfall accumulationsfor Event A but overestimated rainfall accumulations for

Event BThat is because in the precipitation ranges of 0ndash5mmand 5ndash10mm almost all the ME values are positive for EventA and all the ME are negative for Event B

In this section for the two events it is found that theforecast precision in DA-SAT experiment is higher than thatin DA-OBS experiment This may be because there are fewstations located on the high-altitude region of northeasternTibetan Plateau DA-BOTH experiment has the biggestprecision compared with the other two data assimilationexperiments which means that the improvement after theassimilation of the satellite radiances combined with surfaceand upper-air meteorological observations is greater thanthat after only assimilating satellite radiances

413 Temporal Variations of Precipitation The greater dif-ference between Event A and Event B is the rainfall dis-tribution in time Figure 9(a) shows the rainfall intensitiesand cumulative curves of the measured and modeled resultsfor Event A at Laohugou station The cumulative curvesof CTRL experiment are lie below and far away from thatof the measured values However during the period from

Advances in Meteorology 9

Table 3 Differences between observation and numerical experiments

Event ID Experiments 1198772 RMSE 0ndash5mm ME

5ndash10mm 10ndash425 (235) 0ndash425 (235)

Event A

CTRL 024 105 minus015 42 216 56DA-SAT 043 81 12 31 172 35DA-OBS 033 92 04 39 189 46DA-BOTH 070 58 minus03 11 95 18

Event B

CTRL 032 55 minus28 minus14 84 minus02DA-SAT 059 50 minus24 minus16 54 minus05DA-OBS 048 54 minus19 minus23 46 minus07DA-BOTH 064 49 minus25 minus13 50 minus06

0

1

2

3

4

5

6

7

8

9

100

5

10

15

20

25

6-18

-36-

18-6

6-18

-96-

18-1

26-

18-1

56-

18-1

86-

18-2

16-

18-2

46-

19-3

6-19

-66-

19-9

6-19

-12

6-19

-15

6-19

-18

6-19

-21

6-19

-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRL

DA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

20

40

60

80

100

120

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)Measured valueCTRLDA-OBS

DA-SATDA-BOTH

Rain

fall

inte

nsity

(mm

15

hou

r)

(b)

Figure 9The rainfall intensities and cumulative curves of the values measured at the sites and the simulated results from the four numericalexperiments (a) Event A (b) Event B

000 to 2400 on July 19 the cumulative curves for DA-SATand DA-SAT experiments are closer to that of the measuredvalues compared with the other two experiments This maybe because the maximum precipitation intensities simulatedby DA-BOTH and DA-SAT experiments are closer to theobserved values from 600 to 900 on July 19

In Figure 9(b) for event B the precipitation intensityin CTRL experiment is significantly larger than the mea-sured values in first 6 hours Therefore the cumulativecurve in the CTRL experiment is far above that of themeasured values due to the inaccurate initial field Afterdata assimilation the rainfall intensities in DA-SAT and DA-BOTH experiments are closer to the observations in first 6hours while the rainfall intensity in DA-OBS experiment ismuch less than the observations As a result the cumulativecurve in DA-BOTH experiment is closest to the curve

of the observed values DA-SAT experiment is next andDA-OBS experiment is farthest It is noteworthy that forEvent B the cumulative curve in DA-SAT experiment iscloser to that of the observed values in comparison withEvent A

The absolute error (AE) has been calculated for bothEventA andEvent B In Figure 10(a) for EventA theAEof thenumerical experiments becomes bigger and bigger with theincrease of rainfall intensity while the values in DA-BOTHandDA-SAT experiments are relatively smaller Itmeans thatfor the convective rainfall like Event A WRF model failsin capturing the whole process of the event However inFigure 10(b) for Event B with even distribution in time theAE of the numerical experiments is relatively constant whilethe values for CTRL and DA-OBS experiments are relativelylarger

10 Advances in Meteorology

0

2

4

6

8

10

12

140

2

4

6

8

10

12

14

06-1

8-3

06-1

8-6

06-1

8-9

06-1

8-12

06-1

8-15

06-1

8-18

06-1

8-21

06-1

8-24

06-1

9-3

06-1

9-6

06-1

9-9

06-1

9-12

06-1

9-15

06-1

9-18

06-1

9-21

06-1

9-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

10

20

30

40

50

60

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

15

hou

r)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(b)

Figure 10TheAEof rainfall intensity between the valuesmeasured at the sites and the simulated results from the four numerical experiments(a) Event A (b) Event B

The results shows that for the temporal variation ofrainfall the improvement in DA-SAT experiment is betterthan that in DA-OBS experiment due to the lack of groundobservational data at high-altitude region And for Event Bthe improvement in DA-SAT is greater than Event A

42 Impacts of Data Assimilation on the Initial Field Fromthe above analysis it is clear that data assimilation exper-iments can improve rainfall forecast The purpose of dataassimilation is to acquire accurate initial fields for numericalmodels To reveal the impact of data assimilation experimentson the initial fields EventA is chosen as an example to analyzethe difference in the initial fields between data assimilationand control experiments The geopotential height and mois-ture flux at 850 hPa are selected for analysis

Figure 11 shows the geopotential height differencesbetween data assimilation and control experimentsCompared with CTRL experiment (Figure 11(a)) thereis no evident change of geopotential height in DA-OBS experiment (Figure 11(b)) or DA-SAT experiment(Figure 11(c)) Figure 8(d) shows the geopotential heightfor DA-BOTH experiment which is much smaller thanthe values from CTRL experiment The region with thelargest changes is mainly located in Qinghai Provinceand the northwest of Gansu Province which is consistentwith the improved forecasts of precipitation area andaccumulation in DA-BOTH experiment As can be seen inFigure 12 after data assimilation the moisture flux changenoticeably in the study area in comparison with the controlexperiment (Figure 12(a)) The moisture flux of DA-SATexperiment is markedly increased in the north of the studyarea (Figure 12(b)) It may bring more precipitation in the

northeast of Qinghai Province In Figure 12(c) there is aclear flow of moisture to the northwestern part of GansuProvince which results in the heavy rain The biggest changeof moisture flux occurs in Figure 12(d) for DA-BOTHexperiment which covers the largest area compared withother numerical experiments and it could give greatestestimated precipitation in northeastern Tibetan Plateau

5 Conclusions

In this study the 3DVar assimilation system is used toimprove the forecasting of heavy rain in spatial and timedistribution in northeastern Tibetan Plateau A control andthree data assimilation experiments were designed Theseexperiments demonstrate that the rain forecast is significantlyimproved after data assimilation through enhancing theaccuracy of the initial field Two heavy rainfall events withdifferent evenness in space and time are used to examine theimprovement for rainfall forecast

For the spatial distribution data assimilation experimentschanged significantly the rainfall area and accumulationfor the study area For Event A the data assimilationexperiments provide larger rainfall areas and accumulationwhile for Event B the data assimilation experiments providesmaller precipitation accumulation on the southeastern ofthe study area Rain-gauge data from 43 sites were usedto evaluate the impact after data assimilation The resultsshows that DA-BOTH experiment has the highest predictionaccuracy followed by the DA-SAT experiment and thenDA-OBS experiment while the CTRL experiment has thelowest accuracy For the temporal variation the forecastresults have great differences between Event A and Event B

Advances in Meteorology 11

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(d)

Figure 11 Geopotential height initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

The satellite radiances for Event B have greater positive effectthan that for Event A For the two events the cumulativecurve in DA-BOTH experiment is closest to the curveof the measured values while in DA-OBS experiment theaccumulation curves are far below to the curve of themeasured values due to the lack of ground observations inthe high-altitude region In conclusion both for the spatialand temporal distribution of rainfall the satellite radianceshave greater effect than surface and upper-air meteorologicalobservations in high-latitude regions of the northeasternedge and the improvement for the assimilation of satelliteradiances together with surface and upper-air meteorologicalobservations is greater than that for assimilating either satel-lite radiances ormeteorology observations Event A is chosento evaluate the impact of data assimilation on the initialfields It is found that better and more detailed informationis added to the initial fields (eg geopotential height andmoisture flux) especially on the northwest of Gansu Province

andQinghai Province where the forecasts of precipitation areimproved significantly

It should be mentioned that many other assimilationtechniques such as 4DVar EnKF and so forth need to betested in this study These approaches have great potentialalthough they currently suffer from unaffordable computercosts Meanwhile we need to analyse the impact of differentparameterization schemes on the WRF model after dataassimilation because in this study the conclusions drawn aresubject to the specific parameterization schemes used here

Appendix

Statistic Calculation

The coefficient of determination (1198772) root mean squaredeviation (RMSE) mean error (ME) and absolute error (AE)

12 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(d)

Figure 12 Moisture flux initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

were used to evaluate the impact of the DA experiments inthis study They are defined as

119877

=

(1119873)sum119873

119894=1(119865119894 minus 119865) (119874119894 minus 119874)

radic(1119873)sum119873

119894=1(119865119894 minus 119865)

2radic(1119873)sum

119873

119894=1(119874119894 minus 119874)

2

RMSE = radic 1119873

119873

sum

119894=1

(119865119894 minus 119874119894)2

ME = 1119873

119873

sum

119894=1

(119874119894 minus 119865119894)

AE = 1119873

119873

sum

119894=1

1003816100381610038161003816119865119894 minus 1198741198941003816100381610038161003816

(A1)

Here 119865 is simulated value 119874 is observed value 119873 is thenumber of sites

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work was supported by the National natural ScienceFoundation of China (41190080 40971290) Strategic LeadingTechnology Special Fund of Chinese Academy of Sciences(XDB03030204) the National Scientific and TechnologicalSupport Project (2013BAB05B03) and the Opening Fund ofKey Laboratory of Land Surface Process and Climate Changein Cold and Arid Regions CAS (LPCC201205)

Advances in Meteorology 13

References

[1] P Chambon S Q Zhang A Y Hou M Zupanski and SCheung ldquoAssessing the impact of pre-GPMmicrowave precipi-tation observations in the GoddardWRF ensemble data assim-ilation systemrdquo Quarterly Journal of the Royal MeteorologicalSociety vol 140 no 681 pp 1219ndash1235 2013

[2] J-H Seo Y H Lee and Y-H Kim ldquoFeature selection forvery short-term heavy rainfall prediction using evolutionarycomputationrdquo Advances in Meteorology vol 2014 Article ID203545 15 pages 2014

[3] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M JEsteban-Parra and Y Castro-Dıez ldquoHigh-resolution projec-tions of mean and extreme precipitation over Spain usingthe WRF model (2070ndash2099 versus 1970ndash1999)rdquo Journal ofGeophysical Research D Atmospheres vol 117 no 12 Article IDD12108 2012

[4] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M J SEsteban-Parra and Y Castro-Dıez ldquoEvaluation of WRF meanand extreme precipitation over Spain present climate (1970ndash99)rdquo Journal of Climate vol 25 no 14 pp 4883ndash4895 2012

[5] R M Cardoso P M M Soares P M A Miranda and MBelo-Pereira ldquoWRF high resolution simulation of Iberianmeanand extreme precipitation climaterdquo International Journal ofClimatology vol 33 no 11 pp 2591ndash2608 2013

[6] L Calvetti and A J Pereira Filho ldquoEnsemble hydrometeoro-logical forecasts using WRF hourly QPF and topmodel for amiddle watershedrdquo Advances in Meteorology vol 2014 ArticleID 484120 12 pages 2014

[7] S Q Zhang M Zupanski A Y Hou X Lin and S H CheungldquoAssimilation of precipitation-affected radiances in a cloud-resolving wrf ensemble data assimilation systemrdquo MonthlyWeather Review vol 141 no 2 pp 754ndash772 2013

[8] J Liu M Bray and D Han ldquoExploring the effect of dataassimilation by WRF-3DVar for numerical rainfall predictionwith different types of storm eventsrdquoHydrological Processes vol27 no 25 pp 3627ndash3640 2013

[9] C-H Xiong L-F Zhang J-P Guan J Peng and B ZhangldquoAnalysis and numerical study of a hybrid BGM-3DVAR dataassimilation scheme using satellite radiance data for heavy rainforecastsrdquo Journal of Hydrodynamics vol 25 no 3 pp 430ndash4392013

[10] D M Barker W Huang Y R Guo and Q N Xiao ldquoA three-dimensional (3DVAR) data assimilation system for use withMM5 implementation and initial resultsrdquo Monthly WeatherReview vol 132 pp 897ndash914 2004

[11] Q Xiao and J Sun ldquoMultiple-radar data assimilation and short-range quantitative precipitation forecasting of a squall lineobserved during IHOP 2002rdquoMonthlyWeather Review vol 135no 10 pp 3381ndash3404 2007

[12] Z Sokol ldquoEffects of an assimilation of radar and satellite dataon a very-short range forecast of heavy convective rainfallsrdquoAtmospheric Research vol 93 no 1ndash3 pp 188ndash206 2009

[13] P Bauer P Lopez A Benedetti D Salmond and E MoreauldquoImplementation of 1D+4D-Var assimilation of precipitation-affected microwave radiances at ECMWF I 1D-Varrdquo QuarterlyJournal of the Royal Meteorological Society vol 132 no 620 pp2277ndash2306 2006

[14] P Bauer P Lopez D Salmond A Benedetti S Saarinen andM Bonazzola ldquoImplementation of 1D+4D-Var assimilation of

precipitation-affected microwave radiances at ECMWF II 4D-Varrdquo Quarterly Journal of the Royal Meteorological Society vol132 no 620 pp 2307ndash2332 2006

[15] P Bauer A J Geer P Lopez and D Salmond ldquoDirect 4D-Var assimilation of all-sky radiances Part I ImplementationrdquoQuarterly Journal of the Royal Meteorological Society vol 136no 652 pp 1868ndash1885 2010

[16] A J Geer P Bauer and P Lopez ldquoDirect 4D-Var assimilationof all-sky radiances Part II assessmentrdquo Quarterly Journal ofthe RoyalMeteorological Society vol 136 no 652 pp 1886ndash19052010

[17] Z Sokol and D Rezacova ldquoAssimilation of radar reflectivityinto the LM COSMOmodel with a high horizontal resolutionrdquoMeteorological Applications vol 13 no 4 pp 317ndash330 2006

[18] Z Sokol ldquoAssimilation of extrapolated radar reflectivity into aNWP model and its impact on a precipitation forecast at highresolutionrdquo Atmospheric Research vol 100 no 2-3 pp 201ndash2122011

[19] HWang J Sun S Fan and X-Y Huang ldquoIndirect assimilationof radar reflectivity with WRF 3D-var and its impact onprediction of four summertime convective eventsrdquo Journal ofApplied Meteorology and Climatology vol 52 no 4 pp 889ndash902 2013

[20] B Macpherson ldquoOperational experience with assimilation ofrainfall data in the Met Office mesoscale modelrdquo Meteorologyand Atmospheric Physics vol 76 no 1-2 pp 3ndash8 2001

[21] K Stephan S Klink and C Schraff ldquoAssimilation of radar-derived rain rates into the convective-scale model COSMO-DEat DWDrdquo Quarterly Journal of the Royal Meteorological Societyvol 134 no 634 pp 1315ndash1326 2008

[22] K Alapaty D S Niyogi F Chen P Pyle A Chandrasekar andN Seaman ldquoDevelopment of the flux-adjusting surface dataassimilation system for mesoscale modelsrdquo Journal of AppliedMeteorology and Climatology vol 47 no 9 pp 2331ndash2350 2008

[23] F H Ruggiero G D Modica and A E Lipton ldquoAssimilationof satellite imager data and surface observations to improveanalysis of circulations forced by cloud shading contrastsrdquoMonthly Weather Review vol 128 no 2 pp 434ndash448 2000

[24] J P Hacker and D Rostkier-Edelstein ldquoPBL state estimationwith surface observations a column model and an ensemblefilterrdquo Monthly Weather Review vol 135 no 8 pp 2958ndash29722007

[25] D J Stensrud N Yussouf D C Dowell and M C ConiglioldquoAssimilating surface data into a mesoscale model ensemblecold pool analyses from spring 2007rdquoAtmospheric Research vol93 no 1ndash3 pp 207ndash220 2009

[26] Q Wan and J Xu ldquoA numerical study of the rainstormcharacteristics of the June 2005 flash flood with WRFGSIdata assimilation system over south-east Chinardquo HydrologicalProcesses vol 25 no 8 pp 1327ndash1341 2011

[27] J Liu M Bray and D Han ldquoA study on WRF radar dataassimilation for hydrological rainfall predictionrdquoHydrology andEarth System Sciences vol 17 no 8 pp 3095ndash3110 2013

[28] K Ide P Courtier M Ghil and A C Lorenc ldquoUnified notationfor data assimilation operational sequential and variationalrdquoJournal of the Meteorological Society of Japan vol 75 no 1 pp181ndash189 1997

[29] A C Lorenc ldquoAnalysis methods for numerical weather predic-tionrdquo Quarterly Journal of the Royal Meteorological Society vol112 no 474 pp 1177ndash1194 1966

14 Advances in Meteorology

[30] D F Parrish and J C Derber ldquoThe nationalmeteorological cen-terrsquos spectral statistical interpolation analysis systemrdquo MonthlyWeather Review vol 120 no 8 pp 1747ndash1763 1992

[31] T B Zhao and C B Fu ldquoPreliminary comparison and analysisbetween ERA-40 NCEP-2 reanalysis and observations overChinardquo Climatic and Environmental Research vol 11 no 1 pp14ndash32 2006

[32] L Ma T Zhang Q Li O W Frauenfeld and D QinldquoEvaluation of ERA-40 NCEP-1 and NCEP-2 reanalysis airtemperatures with ground-based measurements in ChinardquoJournal of Geophysical Research Atmospheres (1984ndash2012) vol13 no D15 1984

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ClimatologyJournal of

EcologyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

EarthquakesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

Mining

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

International Journal of

Geophysics

OceanographyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofPetroleum Engineering

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Atmospheric SciencesInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MineralogyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MeteorologyAdvances in

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geological ResearchJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geology Advances in

6 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N∘94 E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

2 4 6 8 10 12 14 16 18 20

(mm)

(d)

Figure 6 The forecast precipitation for Event B from the different numerical experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

experiment predicts only a small amount of rainfall on thenorthwestern of the study area while in DA-SAT experimentprecipitation area extends to the central and southern regionsof the study area and rainfall accumulation increases sig-nificantly (Figure 5(b)) But in the northwest of the regionthe rainfall accumulation for DA-SAT experiment is similarwith that shown in Figure 5(a) In DA-OBS experiment(Figure 5(c)) precipitation accumulation on the northwest-ern of the region is greater than that in CTRL experimentbut the precipitation area is similar in both experimentsFigure 5(d) shows the result of DA-BOTH experiment inwhich the precipitation area is the largest and accumulationis the biggest compared with the other three experiments

The results for Event B are shown in Figure 6 Theprecipitation area on the northwestern of the study region issignificantly increased by DA-SAT experiment (Figure 6(b))relative to that in CTRL experiment In Figure 6(c)

the precipitation area in DA-OBS experiment is same asthat in Figure 6(a) but the precipitation accumulation isobviously decreased on the southeastern of the regionFigure 6(d) shows the DA-BOTH experiment result whichhas the similar precipitation area and accumulation as shownin DA-SAT experiment

To sum up the data assimilation has important influenceon the precipitation forecast with significantly changes of therainfall area and accumulation It should be noticed that theeffects of the forecast depends on the different rainfall event

The rain-gauge data were used to analyze the changedprecipitation area and accumulation after data assimilationexperiments Figure 7 compares the measured and mod-eled results in the rainfall Event A In CTRL experiment(Figure 7(a)) almost all the triangles lie below the line 119910 = 119909However in DA-SAT experiment the green triangles whichrepresent the observation sites (Qilian Yeniugou Tuole

Advances in Meteorology 7

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(a)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(b)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(c)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)y = x

(d)

Figure 7 Differences between rainfall values measured at the sites and those from the numerical experiments for Event A (a) CTRL (b)DA-SAT (c) DA-OBS (d) DA-BOTH

and Wulan) in the central and southern regions of the studyarea are closer to the line 119910 = 119909 (Figure 7(b)) That is tosay DA-SAT experiment improves the predicted accuracy forthe central and southern regions of the study area throughproviding greater precipitation accumulation Similarly inFigure 7(c) for DA-OBS experiment the red triangles whichrepresent the observation sites (Jiuquan Yumenzheng Anxiand Dunhuang) in northwestern of the region are closer tothe line119910 = 119909 than that is shown inCTRL experiment InDA-BOTH experiment all the red and green triangles are closestto the line 119910 = 119909 (Figure 7(d))

It is noticed that in Figure 8 many more triangles forEvent B lie above the 119910 = 119909 line than Event A Asshown in CTRL experiment (Figure 8(a)) the green triangleswhich represent observation sites (Hezuo Lintao linxia andYuzhong) on the southeastern of the study area are lie aboveand far away from the 119910 = 119909 line After data assimilationthe green triangles in three data assimilation experimentsare closer to the 119910 = 119909 line in Figures 8(b)ndash8(d) Thered triangles represent the observation sites (Jiuquan Tuole

Dingxin and Jinta) in the northwestern of the study areaIn DA-SAT experiment (Figure 8(b)) two red triangles withmore than 10mm of rainfall are closer to the 119910 = 119909 line whilein DA-OBS experiment (Figure 8(c)) other two red triangleswith less than 5mm of rainfall are closer to the 119910 = 119909 linecompared with that in Figure 8(a) In DA-BOTH experiment(Figure 8(d)) the results are similar to that as shown in DA-SAT experiment except the red triangles are closer to the119910 = 119909 line

The above results suggest that data assimilation havepositive effects on the rainfall forecast through changingthe rainfall area and accumulation for the study area Andfor the two events with different evenness in time andspace the results after data assimilation are different ForEvent A data assimilation experiments provide larger rainfallareas and accumulation to improve the predicted accuracyAnd for Event B data assimilation experiments improve theaccuracy on the southeastern of the study area by providingsmaller precipitation accumulation This means that CTRLexperiment underestimates the convective rainfall for Event

8 Advances in Meteorology

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(a)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(b)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(c)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)y = x

(d)

Figure 8 Differences between rainfall values measured at the sites and those from the numerical experiments for Event B (a) CTRL (b)DA-SAT (c) DA-OBS (d) DA-BOTH

A and overestimates the stratiform rainfall of Event B on thesoutheastern of the region

In order to evaluate quantitatively the impact of dataassimilation experiments which have positive effects on therainfall forecast rain-gauge data from 43 sites were usedTable 3 shows the comparison between the measured andmodeled results The CTRL experiments have the lowestprecision of simulation for both Event A and Event B whose1198772 are the smallest and RMSE are the biggest compared with

three data assimilation experiments The greatest improve-ment appears in DA-BOTH experiment for the two eventsFor Event A the 1198772 for DA-BOTH experiment is increasedby 046 and its RMSE is reduced by 47 and for Event Bthe 1198772 for DA-BOTH experiment is increased by 032 andits RMSE is reduced by 06 Additionally it is found that inDA-SAT experiment the improvement is bigger than DA-OBS experimentTheME of rainfall accumulation shows thatfor the entire rainfall range it has positives value for Event Abut has negative value for Event B This means that the fourexperiments produce underestimated rainfall accumulationsfor Event A but overestimated rainfall accumulations for

Event BThat is because in the precipitation ranges of 0ndash5mmand 5ndash10mm almost all the ME values are positive for EventA and all the ME are negative for Event B

In this section for the two events it is found that theforecast precision in DA-SAT experiment is higher than thatin DA-OBS experiment This may be because there are fewstations located on the high-altitude region of northeasternTibetan Plateau DA-BOTH experiment has the biggestprecision compared with the other two data assimilationexperiments which means that the improvement after theassimilation of the satellite radiances combined with surfaceand upper-air meteorological observations is greater thanthat after only assimilating satellite radiances

413 Temporal Variations of Precipitation The greater dif-ference between Event A and Event B is the rainfall dis-tribution in time Figure 9(a) shows the rainfall intensitiesand cumulative curves of the measured and modeled resultsfor Event A at Laohugou station The cumulative curvesof CTRL experiment are lie below and far away from thatof the measured values However during the period from

Advances in Meteorology 9

Table 3 Differences between observation and numerical experiments

Event ID Experiments 1198772 RMSE 0ndash5mm ME

5ndash10mm 10ndash425 (235) 0ndash425 (235)

Event A

CTRL 024 105 minus015 42 216 56DA-SAT 043 81 12 31 172 35DA-OBS 033 92 04 39 189 46DA-BOTH 070 58 minus03 11 95 18

Event B

CTRL 032 55 minus28 minus14 84 minus02DA-SAT 059 50 minus24 minus16 54 minus05DA-OBS 048 54 minus19 minus23 46 minus07DA-BOTH 064 49 minus25 minus13 50 minus06

0

1

2

3

4

5

6

7

8

9

100

5

10

15

20

25

6-18

-36-

18-6

6-18

-96-

18-1

26-

18-1

56-

18-1

86-

18-2

16-

18-2

46-

19-3

6-19

-66-

19-9

6-19

-12

6-19

-15

6-19

-18

6-19

-21

6-19

-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRL

DA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

20

40

60

80

100

120

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)Measured valueCTRLDA-OBS

DA-SATDA-BOTH

Rain

fall

inte

nsity

(mm

15

hou

r)

(b)

Figure 9The rainfall intensities and cumulative curves of the values measured at the sites and the simulated results from the four numericalexperiments (a) Event A (b) Event B

000 to 2400 on July 19 the cumulative curves for DA-SATand DA-SAT experiments are closer to that of the measuredvalues compared with the other two experiments This maybe because the maximum precipitation intensities simulatedby DA-BOTH and DA-SAT experiments are closer to theobserved values from 600 to 900 on July 19

In Figure 9(b) for event B the precipitation intensityin CTRL experiment is significantly larger than the mea-sured values in first 6 hours Therefore the cumulativecurve in the CTRL experiment is far above that of themeasured values due to the inaccurate initial field Afterdata assimilation the rainfall intensities in DA-SAT and DA-BOTH experiments are closer to the observations in first 6hours while the rainfall intensity in DA-OBS experiment ismuch less than the observations As a result the cumulativecurve in DA-BOTH experiment is closest to the curve

of the observed values DA-SAT experiment is next andDA-OBS experiment is farthest It is noteworthy that forEvent B the cumulative curve in DA-SAT experiment iscloser to that of the observed values in comparison withEvent A

The absolute error (AE) has been calculated for bothEventA andEvent B In Figure 10(a) for EventA theAEof thenumerical experiments becomes bigger and bigger with theincrease of rainfall intensity while the values in DA-BOTHandDA-SAT experiments are relatively smaller Itmeans thatfor the convective rainfall like Event A WRF model failsin capturing the whole process of the event However inFigure 10(b) for Event B with even distribution in time theAE of the numerical experiments is relatively constant whilethe values for CTRL and DA-OBS experiments are relativelylarger

10 Advances in Meteorology

0

2

4

6

8

10

12

140

2

4

6

8

10

12

14

06-1

8-3

06-1

8-6

06-1

8-9

06-1

8-12

06-1

8-15

06-1

8-18

06-1

8-21

06-1

8-24

06-1

9-3

06-1

9-6

06-1

9-9

06-1

9-12

06-1

9-15

06-1

9-18

06-1

9-21

06-1

9-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

10

20

30

40

50

60

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

15

hou

r)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(b)

Figure 10TheAEof rainfall intensity between the valuesmeasured at the sites and the simulated results from the four numerical experiments(a) Event A (b) Event B

The results shows that for the temporal variation ofrainfall the improvement in DA-SAT experiment is betterthan that in DA-OBS experiment due to the lack of groundobservational data at high-altitude region And for Event Bthe improvement in DA-SAT is greater than Event A

42 Impacts of Data Assimilation on the Initial Field Fromthe above analysis it is clear that data assimilation exper-iments can improve rainfall forecast The purpose of dataassimilation is to acquire accurate initial fields for numericalmodels To reveal the impact of data assimilation experimentson the initial fields EventA is chosen as an example to analyzethe difference in the initial fields between data assimilationand control experiments The geopotential height and mois-ture flux at 850 hPa are selected for analysis

Figure 11 shows the geopotential height differencesbetween data assimilation and control experimentsCompared with CTRL experiment (Figure 11(a)) thereis no evident change of geopotential height in DA-OBS experiment (Figure 11(b)) or DA-SAT experiment(Figure 11(c)) Figure 8(d) shows the geopotential heightfor DA-BOTH experiment which is much smaller thanthe values from CTRL experiment The region with thelargest changes is mainly located in Qinghai Provinceand the northwest of Gansu Province which is consistentwith the improved forecasts of precipitation area andaccumulation in DA-BOTH experiment As can be seen inFigure 12 after data assimilation the moisture flux changenoticeably in the study area in comparison with the controlexperiment (Figure 12(a)) The moisture flux of DA-SATexperiment is markedly increased in the north of the studyarea (Figure 12(b)) It may bring more precipitation in the

northeast of Qinghai Province In Figure 12(c) there is aclear flow of moisture to the northwestern part of GansuProvince which results in the heavy rain The biggest changeof moisture flux occurs in Figure 12(d) for DA-BOTHexperiment which covers the largest area compared withother numerical experiments and it could give greatestestimated precipitation in northeastern Tibetan Plateau

5 Conclusions

In this study the 3DVar assimilation system is used toimprove the forecasting of heavy rain in spatial and timedistribution in northeastern Tibetan Plateau A control andthree data assimilation experiments were designed Theseexperiments demonstrate that the rain forecast is significantlyimproved after data assimilation through enhancing theaccuracy of the initial field Two heavy rainfall events withdifferent evenness in space and time are used to examine theimprovement for rainfall forecast

For the spatial distribution data assimilation experimentschanged significantly the rainfall area and accumulationfor the study area For Event A the data assimilationexperiments provide larger rainfall areas and accumulationwhile for Event B the data assimilation experiments providesmaller precipitation accumulation on the southeastern ofthe study area Rain-gauge data from 43 sites were usedto evaluate the impact after data assimilation The resultsshows that DA-BOTH experiment has the highest predictionaccuracy followed by the DA-SAT experiment and thenDA-OBS experiment while the CTRL experiment has thelowest accuracy For the temporal variation the forecastresults have great differences between Event A and Event B

Advances in Meteorology 11

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(d)

Figure 11 Geopotential height initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

The satellite radiances for Event B have greater positive effectthan that for Event A For the two events the cumulativecurve in DA-BOTH experiment is closest to the curveof the measured values while in DA-OBS experiment theaccumulation curves are far below to the curve of themeasured values due to the lack of ground observations inthe high-altitude region In conclusion both for the spatialand temporal distribution of rainfall the satellite radianceshave greater effect than surface and upper-air meteorologicalobservations in high-latitude regions of the northeasternedge and the improvement for the assimilation of satelliteradiances together with surface and upper-air meteorologicalobservations is greater than that for assimilating either satel-lite radiances ormeteorology observations Event A is chosento evaluate the impact of data assimilation on the initialfields It is found that better and more detailed informationis added to the initial fields (eg geopotential height andmoisture flux) especially on the northwest of Gansu Province

andQinghai Province where the forecasts of precipitation areimproved significantly

It should be mentioned that many other assimilationtechniques such as 4DVar EnKF and so forth need to betested in this study These approaches have great potentialalthough they currently suffer from unaffordable computercosts Meanwhile we need to analyse the impact of differentparameterization schemes on the WRF model after dataassimilation because in this study the conclusions drawn aresubject to the specific parameterization schemes used here

Appendix

Statistic Calculation

The coefficient of determination (1198772) root mean squaredeviation (RMSE) mean error (ME) and absolute error (AE)

12 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(d)

Figure 12 Moisture flux initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

were used to evaluate the impact of the DA experiments inthis study They are defined as

119877

=

(1119873)sum119873

119894=1(119865119894 minus 119865) (119874119894 minus 119874)

radic(1119873)sum119873

119894=1(119865119894 minus 119865)

2radic(1119873)sum

119873

119894=1(119874119894 minus 119874)

2

RMSE = radic 1119873

119873

sum

119894=1

(119865119894 minus 119874119894)2

ME = 1119873

119873

sum

119894=1

(119874119894 minus 119865119894)

AE = 1119873

119873

sum

119894=1

1003816100381610038161003816119865119894 minus 1198741198941003816100381610038161003816

(A1)

Here 119865 is simulated value 119874 is observed value 119873 is thenumber of sites

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work was supported by the National natural ScienceFoundation of China (41190080 40971290) Strategic LeadingTechnology Special Fund of Chinese Academy of Sciences(XDB03030204) the National Scientific and TechnologicalSupport Project (2013BAB05B03) and the Opening Fund ofKey Laboratory of Land Surface Process and Climate Changein Cold and Arid Regions CAS (LPCC201205)

Advances in Meteorology 13

References

[1] P Chambon S Q Zhang A Y Hou M Zupanski and SCheung ldquoAssessing the impact of pre-GPMmicrowave precipi-tation observations in the GoddardWRF ensemble data assim-ilation systemrdquo Quarterly Journal of the Royal MeteorologicalSociety vol 140 no 681 pp 1219ndash1235 2013

[2] J-H Seo Y H Lee and Y-H Kim ldquoFeature selection forvery short-term heavy rainfall prediction using evolutionarycomputationrdquo Advances in Meteorology vol 2014 Article ID203545 15 pages 2014

[3] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M JEsteban-Parra and Y Castro-Dıez ldquoHigh-resolution projec-tions of mean and extreme precipitation over Spain usingthe WRF model (2070ndash2099 versus 1970ndash1999)rdquo Journal ofGeophysical Research D Atmospheres vol 117 no 12 Article IDD12108 2012

[4] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M J SEsteban-Parra and Y Castro-Dıez ldquoEvaluation of WRF meanand extreme precipitation over Spain present climate (1970ndash99)rdquo Journal of Climate vol 25 no 14 pp 4883ndash4895 2012

[5] R M Cardoso P M M Soares P M A Miranda and MBelo-Pereira ldquoWRF high resolution simulation of Iberianmeanand extreme precipitation climaterdquo International Journal ofClimatology vol 33 no 11 pp 2591ndash2608 2013

[6] L Calvetti and A J Pereira Filho ldquoEnsemble hydrometeoro-logical forecasts using WRF hourly QPF and topmodel for amiddle watershedrdquo Advances in Meteorology vol 2014 ArticleID 484120 12 pages 2014

[7] S Q Zhang M Zupanski A Y Hou X Lin and S H CheungldquoAssimilation of precipitation-affected radiances in a cloud-resolving wrf ensemble data assimilation systemrdquo MonthlyWeather Review vol 141 no 2 pp 754ndash772 2013

[8] J Liu M Bray and D Han ldquoExploring the effect of dataassimilation by WRF-3DVar for numerical rainfall predictionwith different types of storm eventsrdquoHydrological Processes vol27 no 25 pp 3627ndash3640 2013

[9] C-H Xiong L-F Zhang J-P Guan J Peng and B ZhangldquoAnalysis and numerical study of a hybrid BGM-3DVAR dataassimilation scheme using satellite radiance data for heavy rainforecastsrdquo Journal of Hydrodynamics vol 25 no 3 pp 430ndash4392013

[10] D M Barker W Huang Y R Guo and Q N Xiao ldquoA three-dimensional (3DVAR) data assimilation system for use withMM5 implementation and initial resultsrdquo Monthly WeatherReview vol 132 pp 897ndash914 2004

[11] Q Xiao and J Sun ldquoMultiple-radar data assimilation and short-range quantitative precipitation forecasting of a squall lineobserved during IHOP 2002rdquoMonthlyWeather Review vol 135no 10 pp 3381ndash3404 2007

[12] Z Sokol ldquoEffects of an assimilation of radar and satellite dataon a very-short range forecast of heavy convective rainfallsrdquoAtmospheric Research vol 93 no 1ndash3 pp 188ndash206 2009

[13] P Bauer P Lopez A Benedetti D Salmond and E MoreauldquoImplementation of 1D+4D-Var assimilation of precipitation-affected microwave radiances at ECMWF I 1D-Varrdquo QuarterlyJournal of the Royal Meteorological Society vol 132 no 620 pp2277ndash2306 2006

[14] P Bauer P Lopez D Salmond A Benedetti S Saarinen andM Bonazzola ldquoImplementation of 1D+4D-Var assimilation of

precipitation-affected microwave radiances at ECMWF II 4D-Varrdquo Quarterly Journal of the Royal Meteorological Society vol132 no 620 pp 2307ndash2332 2006

[15] P Bauer A J Geer P Lopez and D Salmond ldquoDirect 4D-Var assimilation of all-sky radiances Part I ImplementationrdquoQuarterly Journal of the Royal Meteorological Society vol 136no 652 pp 1868ndash1885 2010

[16] A J Geer P Bauer and P Lopez ldquoDirect 4D-Var assimilationof all-sky radiances Part II assessmentrdquo Quarterly Journal ofthe RoyalMeteorological Society vol 136 no 652 pp 1886ndash19052010

[17] Z Sokol and D Rezacova ldquoAssimilation of radar reflectivityinto the LM COSMOmodel with a high horizontal resolutionrdquoMeteorological Applications vol 13 no 4 pp 317ndash330 2006

[18] Z Sokol ldquoAssimilation of extrapolated radar reflectivity into aNWP model and its impact on a precipitation forecast at highresolutionrdquo Atmospheric Research vol 100 no 2-3 pp 201ndash2122011

[19] HWang J Sun S Fan and X-Y Huang ldquoIndirect assimilationof radar reflectivity with WRF 3D-var and its impact onprediction of four summertime convective eventsrdquo Journal ofApplied Meteorology and Climatology vol 52 no 4 pp 889ndash902 2013

[20] B Macpherson ldquoOperational experience with assimilation ofrainfall data in the Met Office mesoscale modelrdquo Meteorologyand Atmospheric Physics vol 76 no 1-2 pp 3ndash8 2001

[21] K Stephan S Klink and C Schraff ldquoAssimilation of radar-derived rain rates into the convective-scale model COSMO-DEat DWDrdquo Quarterly Journal of the Royal Meteorological Societyvol 134 no 634 pp 1315ndash1326 2008

[22] K Alapaty D S Niyogi F Chen P Pyle A Chandrasekar andN Seaman ldquoDevelopment of the flux-adjusting surface dataassimilation system for mesoscale modelsrdquo Journal of AppliedMeteorology and Climatology vol 47 no 9 pp 2331ndash2350 2008

[23] F H Ruggiero G D Modica and A E Lipton ldquoAssimilationof satellite imager data and surface observations to improveanalysis of circulations forced by cloud shading contrastsrdquoMonthly Weather Review vol 128 no 2 pp 434ndash448 2000

[24] J P Hacker and D Rostkier-Edelstein ldquoPBL state estimationwith surface observations a column model and an ensemblefilterrdquo Monthly Weather Review vol 135 no 8 pp 2958ndash29722007

[25] D J Stensrud N Yussouf D C Dowell and M C ConiglioldquoAssimilating surface data into a mesoscale model ensemblecold pool analyses from spring 2007rdquoAtmospheric Research vol93 no 1ndash3 pp 207ndash220 2009

[26] Q Wan and J Xu ldquoA numerical study of the rainstormcharacteristics of the June 2005 flash flood with WRFGSIdata assimilation system over south-east Chinardquo HydrologicalProcesses vol 25 no 8 pp 1327ndash1341 2011

[27] J Liu M Bray and D Han ldquoA study on WRF radar dataassimilation for hydrological rainfall predictionrdquoHydrology andEarth System Sciences vol 17 no 8 pp 3095ndash3110 2013

[28] K Ide P Courtier M Ghil and A C Lorenc ldquoUnified notationfor data assimilation operational sequential and variationalrdquoJournal of the Meteorological Society of Japan vol 75 no 1 pp181ndash189 1997

[29] A C Lorenc ldquoAnalysis methods for numerical weather predic-tionrdquo Quarterly Journal of the Royal Meteorological Society vol112 no 474 pp 1177ndash1194 1966

14 Advances in Meteorology

[30] D F Parrish and J C Derber ldquoThe nationalmeteorological cen-terrsquos spectral statistical interpolation analysis systemrdquo MonthlyWeather Review vol 120 no 8 pp 1747ndash1763 1992

[31] T B Zhao and C B Fu ldquoPreliminary comparison and analysisbetween ERA-40 NCEP-2 reanalysis and observations overChinardquo Climatic and Environmental Research vol 11 no 1 pp14ndash32 2006

[32] L Ma T Zhang Q Li O W Frauenfeld and D QinldquoEvaluation of ERA-40 NCEP-1 and NCEP-2 reanalysis airtemperatures with ground-based measurements in ChinardquoJournal of Geophysical Research Atmospheres (1984ndash2012) vol13 no D15 1984

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ClimatologyJournal of

EcologyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

EarthquakesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

Mining

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

International Journal of

Geophysics

OceanographyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofPetroleum Engineering

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Atmospheric SciencesInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MineralogyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MeteorologyAdvances in

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geological ResearchJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geology Advances in

Advances in Meteorology 7

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(a)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(b)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(c)

0 5 10 15 20 25 30 35 40 450

5

10

15

20

25

30

35

40

45

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)y = x

(d)

Figure 7 Differences between rainfall values measured at the sites and those from the numerical experiments for Event A (a) CTRL (b)DA-SAT (c) DA-OBS (d) DA-BOTH

and Wulan) in the central and southern regions of the studyarea are closer to the line 119910 = 119909 (Figure 7(b)) That is tosay DA-SAT experiment improves the predicted accuracy forthe central and southern regions of the study area throughproviding greater precipitation accumulation Similarly inFigure 7(c) for DA-OBS experiment the red triangles whichrepresent the observation sites (Jiuquan Yumenzheng Anxiand Dunhuang) in northwestern of the region are closer tothe line119910 = 119909 than that is shown inCTRL experiment InDA-BOTH experiment all the red and green triangles are closestto the line 119910 = 119909 (Figure 7(d))

It is noticed that in Figure 8 many more triangles forEvent B lie above the 119910 = 119909 line than Event A Asshown in CTRL experiment (Figure 8(a)) the green triangleswhich represent observation sites (Hezuo Lintao linxia andYuzhong) on the southeastern of the study area are lie aboveand far away from the 119910 = 119909 line After data assimilationthe green triangles in three data assimilation experimentsare closer to the 119910 = 119909 line in Figures 8(b)ndash8(d) Thered triangles represent the observation sites (Jiuquan Tuole

Dingxin and Jinta) in the northwestern of the study areaIn DA-SAT experiment (Figure 8(b)) two red triangles withmore than 10mm of rainfall are closer to the 119910 = 119909 line whilein DA-OBS experiment (Figure 8(c)) other two red triangleswith less than 5mm of rainfall are closer to the 119910 = 119909 linecompared with that in Figure 8(a) In DA-BOTH experiment(Figure 8(d)) the results are similar to that as shown in DA-SAT experiment except the red triangles are closer to the119910 = 119909 line

The above results suggest that data assimilation havepositive effects on the rainfall forecast through changingthe rainfall area and accumulation for the study area Andfor the two events with different evenness in time andspace the results after data assimilation are different ForEvent A data assimilation experiments provide larger rainfallareas and accumulation to improve the predicted accuracyAnd for Event B data assimilation experiments improve theaccuracy on the southeastern of the study area by providingsmaller precipitation accumulation This means that CTRLexperiment underestimates the convective rainfall for Event

8 Advances in Meteorology

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(a)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(b)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(c)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)y = x

(d)

Figure 8 Differences between rainfall values measured at the sites and those from the numerical experiments for Event B (a) CTRL (b)DA-SAT (c) DA-OBS (d) DA-BOTH

A and overestimates the stratiform rainfall of Event B on thesoutheastern of the region

In order to evaluate quantitatively the impact of dataassimilation experiments which have positive effects on therainfall forecast rain-gauge data from 43 sites were usedTable 3 shows the comparison between the measured andmodeled results The CTRL experiments have the lowestprecision of simulation for both Event A and Event B whose1198772 are the smallest and RMSE are the biggest compared with

three data assimilation experiments The greatest improve-ment appears in DA-BOTH experiment for the two eventsFor Event A the 1198772 for DA-BOTH experiment is increasedby 046 and its RMSE is reduced by 47 and for Event Bthe 1198772 for DA-BOTH experiment is increased by 032 andits RMSE is reduced by 06 Additionally it is found that inDA-SAT experiment the improvement is bigger than DA-OBS experimentTheME of rainfall accumulation shows thatfor the entire rainfall range it has positives value for Event Abut has negative value for Event B This means that the fourexperiments produce underestimated rainfall accumulationsfor Event A but overestimated rainfall accumulations for

Event BThat is because in the precipitation ranges of 0ndash5mmand 5ndash10mm almost all the ME values are positive for EventA and all the ME are negative for Event B

In this section for the two events it is found that theforecast precision in DA-SAT experiment is higher than thatin DA-OBS experiment This may be because there are fewstations located on the high-altitude region of northeasternTibetan Plateau DA-BOTH experiment has the biggestprecision compared with the other two data assimilationexperiments which means that the improvement after theassimilation of the satellite radiances combined with surfaceand upper-air meteorological observations is greater thanthat after only assimilating satellite radiances

413 Temporal Variations of Precipitation The greater dif-ference between Event A and Event B is the rainfall dis-tribution in time Figure 9(a) shows the rainfall intensitiesand cumulative curves of the measured and modeled resultsfor Event A at Laohugou station The cumulative curvesof CTRL experiment are lie below and far away from thatof the measured values However during the period from

Advances in Meteorology 9

Table 3 Differences between observation and numerical experiments

Event ID Experiments 1198772 RMSE 0ndash5mm ME

5ndash10mm 10ndash425 (235) 0ndash425 (235)

Event A

CTRL 024 105 minus015 42 216 56DA-SAT 043 81 12 31 172 35DA-OBS 033 92 04 39 189 46DA-BOTH 070 58 minus03 11 95 18

Event B

CTRL 032 55 minus28 minus14 84 minus02DA-SAT 059 50 minus24 minus16 54 minus05DA-OBS 048 54 minus19 minus23 46 minus07DA-BOTH 064 49 minus25 minus13 50 minus06

0

1

2

3

4

5

6

7

8

9

100

5

10

15

20

25

6-18

-36-

18-6

6-18

-96-

18-1

26-

18-1

56-

18-1

86-

18-2

16-

18-2

46-

19-3

6-19

-66-

19-9

6-19

-12

6-19

-15

6-19

-18

6-19

-21

6-19

-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRL

DA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

20

40

60

80

100

120

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)Measured valueCTRLDA-OBS

DA-SATDA-BOTH

Rain

fall

inte

nsity

(mm

15

hou

r)

(b)

Figure 9The rainfall intensities and cumulative curves of the values measured at the sites and the simulated results from the four numericalexperiments (a) Event A (b) Event B

000 to 2400 on July 19 the cumulative curves for DA-SATand DA-SAT experiments are closer to that of the measuredvalues compared with the other two experiments This maybe because the maximum precipitation intensities simulatedby DA-BOTH and DA-SAT experiments are closer to theobserved values from 600 to 900 on July 19

In Figure 9(b) for event B the precipitation intensityin CTRL experiment is significantly larger than the mea-sured values in first 6 hours Therefore the cumulativecurve in the CTRL experiment is far above that of themeasured values due to the inaccurate initial field Afterdata assimilation the rainfall intensities in DA-SAT and DA-BOTH experiments are closer to the observations in first 6hours while the rainfall intensity in DA-OBS experiment ismuch less than the observations As a result the cumulativecurve in DA-BOTH experiment is closest to the curve

of the observed values DA-SAT experiment is next andDA-OBS experiment is farthest It is noteworthy that forEvent B the cumulative curve in DA-SAT experiment iscloser to that of the observed values in comparison withEvent A

The absolute error (AE) has been calculated for bothEventA andEvent B In Figure 10(a) for EventA theAEof thenumerical experiments becomes bigger and bigger with theincrease of rainfall intensity while the values in DA-BOTHandDA-SAT experiments are relatively smaller Itmeans thatfor the convective rainfall like Event A WRF model failsin capturing the whole process of the event However inFigure 10(b) for Event B with even distribution in time theAE of the numerical experiments is relatively constant whilethe values for CTRL and DA-OBS experiments are relativelylarger

10 Advances in Meteorology

0

2

4

6

8

10

12

140

2

4

6

8

10

12

14

06-1

8-3

06-1

8-6

06-1

8-9

06-1

8-12

06-1

8-15

06-1

8-18

06-1

8-21

06-1

8-24

06-1

9-3

06-1

9-6

06-1

9-9

06-1

9-12

06-1

9-15

06-1

9-18

06-1

9-21

06-1

9-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

10

20

30

40

50

60

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

15

hou

r)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(b)

Figure 10TheAEof rainfall intensity between the valuesmeasured at the sites and the simulated results from the four numerical experiments(a) Event A (b) Event B

The results shows that for the temporal variation ofrainfall the improvement in DA-SAT experiment is betterthan that in DA-OBS experiment due to the lack of groundobservational data at high-altitude region And for Event Bthe improvement in DA-SAT is greater than Event A

42 Impacts of Data Assimilation on the Initial Field Fromthe above analysis it is clear that data assimilation exper-iments can improve rainfall forecast The purpose of dataassimilation is to acquire accurate initial fields for numericalmodels To reveal the impact of data assimilation experimentson the initial fields EventA is chosen as an example to analyzethe difference in the initial fields between data assimilationand control experiments The geopotential height and mois-ture flux at 850 hPa are selected for analysis

Figure 11 shows the geopotential height differencesbetween data assimilation and control experimentsCompared with CTRL experiment (Figure 11(a)) thereis no evident change of geopotential height in DA-OBS experiment (Figure 11(b)) or DA-SAT experiment(Figure 11(c)) Figure 8(d) shows the geopotential heightfor DA-BOTH experiment which is much smaller thanthe values from CTRL experiment The region with thelargest changes is mainly located in Qinghai Provinceand the northwest of Gansu Province which is consistentwith the improved forecasts of precipitation area andaccumulation in DA-BOTH experiment As can be seen inFigure 12 after data assimilation the moisture flux changenoticeably in the study area in comparison with the controlexperiment (Figure 12(a)) The moisture flux of DA-SATexperiment is markedly increased in the north of the studyarea (Figure 12(b)) It may bring more precipitation in the

northeast of Qinghai Province In Figure 12(c) there is aclear flow of moisture to the northwestern part of GansuProvince which results in the heavy rain The biggest changeof moisture flux occurs in Figure 12(d) for DA-BOTHexperiment which covers the largest area compared withother numerical experiments and it could give greatestestimated precipitation in northeastern Tibetan Plateau

5 Conclusions

In this study the 3DVar assimilation system is used toimprove the forecasting of heavy rain in spatial and timedistribution in northeastern Tibetan Plateau A control andthree data assimilation experiments were designed Theseexperiments demonstrate that the rain forecast is significantlyimproved after data assimilation through enhancing theaccuracy of the initial field Two heavy rainfall events withdifferent evenness in space and time are used to examine theimprovement for rainfall forecast

For the spatial distribution data assimilation experimentschanged significantly the rainfall area and accumulationfor the study area For Event A the data assimilationexperiments provide larger rainfall areas and accumulationwhile for Event B the data assimilation experiments providesmaller precipitation accumulation on the southeastern ofthe study area Rain-gauge data from 43 sites were usedto evaluate the impact after data assimilation The resultsshows that DA-BOTH experiment has the highest predictionaccuracy followed by the DA-SAT experiment and thenDA-OBS experiment while the CTRL experiment has thelowest accuracy For the temporal variation the forecastresults have great differences between Event A and Event B

Advances in Meteorology 11

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(d)

Figure 11 Geopotential height initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

The satellite radiances for Event B have greater positive effectthan that for Event A For the two events the cumulativecurve in DA-BOTH experiment is closest to the curveof the measured values while in DA-OBS experiment theaccumulation curves are far below to the curve of themeasured values due to the lack of ground observations inthe high-altitude region In conclusion both for the spatialand temporal distribution of rainfall the satellite radianceshave greater effect than surface and upper-air meteorologicalobservations in high-latitude regions of the northeasternedge and the improvement for the assimilation of satelliteradiances together with surface and upper-air meteorologicalobservations is greater than that for assimilating either satel-lite radiances ormeteorology observations Event A is chosento evaluate the impact of data assimilation on the initialfields It is found that better and more detailed informationis added to the initial fields (eg geopotential height andmoisture flux) especially on the northwest of Gansu Province

andQinghai Province where the forecasts of precipitation areimproved significantly

It should be mentioned that many other assimilationtechniques such as 4DVar EnKF and so forth need to betested in this study These approaches have great potentialalthough they currently suffer from unaffordable computercosts Meanwhile we need to analyse the impact of differentparameterization schemes on the WRF model after dataassimilation because in this study the conclusions drawn aresubject to the specific parameterization schemes used here

Appendix

Statistic Calculation

The coefficient of determination (1198772) root mean squaredeviation (RMSE) mean error (ME) and absolute error (AE)

12 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(d)

Figure 12 Moisture flux initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

were used to evaluate the impact of the DA experiments inthis study They are defined as

119877

=

(1119873)sum119873

119894=1(119865119894 minus 119865) (119874119894 minus 119874)

radic(1119873)sum119873

119894=1(119865119894 minus 119865)

2radic(1119873)sum

119873

119894=1(119874119894 minus 119874)

2

RMSE = radic 1119873

119873

sum

119894=1

(119865119894 minus 119874119894)2

ME = 1119873

119873

sum

119894=1

(119874119894 minus 119865119894)

AE = 1119873

119873

sum

119894=1

1003816100381610038161003816119865119894 minus 1198741198941003816100381610038161003816

(A1)

Here 119865 is simulated value 119874 is observed value 119873 is thenumber of sites

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work was supported by the National natural ScienceFoundation of China (41190080 40971290) Strategic LeadingTechnology Special Fund of Chinese Academy of Sciences(XDB03030204) the National Scientific and TechnologicalSupport Project (2013BAB05B03) and the Opening Fund ofKey Laboratory of Land Surface Process and Climate Changein Cold and Arid Regions CAS (LPCC201205)

Advances in Meteorology 13

References

[1] P Chambon S Q Zhang A Y Hou M Zupanski and SCheung ldquoAssessing the impact of pre-GPMmicrowave precipi-tation observations in the GoddardWRF ensemble data assim-ilation systemrdquo Quarterly Journal of the Royal MeteorologicalSociety vol 140 no 681 pp 1219ndash1235 2013

[2] J-H Seo Y H Lee and Y-H Kim ldquoFeature selection forvery short-term heavy rainfall prediction using evolutionarycomputationrdquo Advances in Meteorology vol 2014 Article ID203545 15 pages 2014

[3] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M JEsteban-Parra and Y Castro-Dıez ldquoHigh-resolution projec-tions of mean and extreme precipitation over Spain usingthe WRF model (2070ndash2099 versus 1970ndash1999)rdquo Journal ofGeophysical Research D Atmospheres vol 117 no 12 Article IDD12108 2012

[4] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M J SEsteban-Parra and Y Castro-Dıez ldquoEvaluation of WRF meanand extreme precipitation over Spain present climate (1970ndash99)rdquo Journal of Climate vol 25 no 14 pp 4883ndash4895 2012

[5] R M Cardoso P M M Soares P M A Miranda and MBelo-Pereira ldquoWRF high resolution simulation of Iberianmeanand extreme precipitation climaterdquo International Journal ofClimatology vol 33 no 11 pp 2591ndash2608 2013

[6] L Calvetti and A J Pereira Filho ldquoEnsemble hydrometeoro-logical forecasts using WRF hourly QPF and topmodel for amiddle watershedrdquo Advances in Meteorology vol 2014 ArticleID 484120 12 pages 2014

[7] S Q Zhang M Zupanski A Y Hou X Lin and S H CheungldquoAssimilation of precipitation-affected radiances in a cloud-resolving wrf ensemble data assimilation systemrdquo MonthlyWeather Review vol 141 no 2 pp 754ndash772 2013

[8] J Liu M Bray and D Han ldquoExploring the effect of dataassimilation by WRF-3DVar for numerical rainfall predictionwith different types of storm eventsrdquoHydrological Processes vol27 no 25 pp 3627ndash3640 2013

[9] C-H Xiong L-F Zhang J-P Guan J Peng and B ZhangldquoAnalysis and numerical study of a hybrid BGM-3DVAR dataassimilation scheme using satellite radiance data for heavy rainforecastsrdquo Journal of Hydrodynamics vol 25 no 3 pp 430ndash4392013

[10] D M Barker W Huang Y R Guo and Q N Xiao ldquoA three-dimensional (3DVAR) data assimilation system for use withMM5 implementation and initial resultsrdquo Monthly WeatherReview vol 132 pp 897ndash914 2004

[11] Q Xiao and J Sun ldquoMultiple-radar data assimilation and short-range quantitative precipitation forecasting of a squall lineobserved during IHOP 2002rdquoMonthlyWeather Review vol 135no 10 pp 3381ndash3404 2007

[12] Z Sokol ldquoEffects of an assimilation of radar and satellite dataon a very-short range forecast of heavy convective rainfallsrdquoAtmospheric Research vol 93 no 1ndash3 pp 188ndash206 2009

[13] P Bauer P Lopez A Benedetti D Salmond and E MoreauldquoImplementation of 1D+4D-Var assimilation of precipitation-affected microwave radiances at ECMWF I 1D-Varrdquo QuarterlyJournal of the Royal Meteorological Society vol 132 no 620 pp2277ndash2306 2006

[14] P Bauer P Lopez D Salmond A Benedetti S Saarinen andM Bonazzola ldquoImplementation of 1D+4D-Var assimilation of

precipitation-affected microwave radiances at ECMWF II 4D-Varrdquo Quarterly Journal of the Royal Meteorological Society vol132 no 620 pp 2307ndash2332 2006

[15] P Bauer A J Geer P Lopez and D Salmond ldquoDirect 4D-Var assimilation of all-sky radiances Part I ImplementationrdquoQuarterly Journal of the Royal Meteorological Society vol 136no 652 pp 1868ndash1885 2010

[16] A J Geer P Bauer and P Lopez ldquoDirect 4D-Var assimilationof all-sky radiances Part II assessmentrdquo Quarterly Journal ofthe RoyalMeteorological Society vol 136 no 652 pp 1886ndash19052010

[17] Z Sokol and D Rezacova ldquoAssimilation of radar reflectivityinto the LM COSMOmodel with a high horizontal resolutionrdquoMeteorological Applications vol 13 no 4 pp 317ndash330 2006

[18] Z Sokol ldquoAssimilation of extrapolated radar reflectivity into aNWP model and its impact on a precipitation forecast at highresolutionrdquo Atmospheric Research vol 100 no 2-3 pp 201ndash2122011

[19] HWang J Sun S Fan and X-Y Huang ldquoIndirect assimilationof radar reflectivity with WRF 3D-var and its impact onprediction of four summertime convective eventsrdquo Journal ofApplied Meteorology and Climatology vol 52 no 4 pp 889ndash902 2013

[20] B Macpherson ldquoOperational experience with assimilation ofrainfall data in the Met Office mesoscale modelrdquo Meteorologyand Atmospheric Physics vol 76 no 1-2 pp 3ndash8 2001

[21] K Stephan S Klink and C Schraff ldquoAssimilation of radar-derived rain rates into the convective-scale model COSMO-DEat DWDrdquo Quarterly Journal of the Royal Meteorological Societyvol 134 no 634 pp 1315ndash1326 2008

[22] K Alapaty D S Niyogi F Chen P Pyle A Chandrasekar andN Seaman ldquoDevelopment of the flux-adjusting surface dataassimilation system for mesoscale modelsrdquo Journal of AppliedMeteorology and Climatology vol 47 no 9 pp 2331ndash2350 2008

[23] F H Ruggiero G D Modica and A E Lipton ldquoAssimilationof satellite imager data and surface observations to improveanalysis of circulations forced by cloud shading contrastsrdquoMonthly Weather Review vol 128 no 2 pp 434ndash448 2000

[24] J P Hacker and D Rostkier-Edelstein ldquoPBL state estimationwith surface observations a column model and an ensemblefilterrdquo Monthly Weather Review vol 135 no 8 pp 2958ndash29722007

[25] D J Stensrud N Yussouf D C Dowell and M C ConiglioldquoAssimilating surface data into a mesoscale model ensemblecold pool analyses from spring 2007rdquoAtmospheric Research vol93 no 1ndash3 pp 207ndash220 2009

[26] Q Wan and J Xu ldquoA numerical study of the rainstormcharacteristics of the June 2005 flash flood with WRFGSIdata assimilation system over south-east Chinardquo HydrologicalProcesses vol 25 no 8 pp 1327ndash1341 2011

[27] J Liu M Bray and D Han ldquoA study on WRF radar dataassimilation for hydrological rainfall predictionrdquoHydrology andEarth System Sciences vol 17 no 8 pp 3095ndash3110 2013

[28] K Ide P Courtier M Ghil and A C Lorenc ldquoUnified notationfor data assimilation operational sequential and variationalrdquoJournal of the Meteorological Society of Japan vol 75 no 1 pp181ndash189 1997

[29] A C Lorenc ldquoAnalysis methods for numerical weather predic-tionrdquo Quarterly Journal of the Royal Meteorological Society vol112 no 474 pp 1177ndash1194 1966

14 Advances in Meteorology

[30] D F Parrish and J C Derber ldquoThe nationalmeteorological cen-terrsquos spectral statistical interpolation analysis systemrdquo MonthlyWeather Review vol 120 no 8 pp 1747ndash1763 1992

[31] T B Zhao and C B Fu ldquoPreliminary comparison and analysisbetween ERA-40 NCEP-2 reanalysis and observations overChinardquo Climatic and Environmental Research vol 11 no 1 pp14ndash32 2006

[32] L Ma T Zhang Q Li O W Frauenfeld and D QinldquoEvaluation of ERA-40 NCEP-1 and NCEP-2 reanalysis airtemperatures with ground-based measurements in ChinardquoJournal of Geophysical Research Atmospheres (1984ndash2012) vol13 no D15 1984

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ClimatologyJournal of

EcologyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

EarthquakesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

Mining

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

International Journal of

Geophysics

OceanographyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofPetroleum Engineering

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Atmospheric SciencesInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MineralogyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MeteorologyAdvances in

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geological ResearchJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geology Advances in

8 Advances in Meteorology

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(a)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(b)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)

y = x

(c)

0 5 10 15 20 250

5

10

15

20

25

Measured precipitation (mm)

Sim

ulat

ed p

reci

pita

tion

(mm

)y = x

(d)

Figure 8 Differences between rainfall values measured at the sites and those from the numerical experiments for Event B (a) CTRL (b)DA-SAT (c) DA-OBS (d) DA-BOTH

A and overestimates the stratiform rainfall of Event B on thesoutheastern of the region

In order to evaluate quantitatively the impact of dataassimilation experiments which have positive effects on therainfall forecast rain-gauge data from 43 sites were usedTable 3 shows the comparison between the measured andmodeled results The CTRL experiments have the lowestprecision of simulation for both Event A and Event B whose1198772 are the smallest and RMSE are the biggest compared with

three data assimilation experiments The greatest improve-ment appears in DA-BOTH experiment for the two eventsFor Event A the 1198772 for DA-BOTH experiment is increasedby 046 and its RMSE is reduced by 47 and for Event Bthe 1198772 for DA-BOTH experiment is increased by 032 andits RMSE is reduced by 06 Additionally it is found that inDA-SAT experiment the improvement is bigger than DA-OBS experimentTheME of rainfall accumulation shows thatfor the entire rainfall range it has positives value for Event Abut has negative value for Event B This means that the fourexperiments produce underestimated rainfall accumulationsfor Event A but overestimated rainfall accumulations for

Event BThat is because in the precipitation ranges of 0ndash5mmand 5ndash10mm almost all the ME values are positive for EventA and all the ME are negative for Event B

In this section for the two events it is found that theforecast precision in DA-SAT experiment is higher than thatin DA-OBS experiment This may be because there are fewstations located on the high-altitude region of northeasternTibetan Plateau DA-BOTH experiment has the biggestprecision compared with the other two data assimilationexperiments which means that the improvement after theassimilation of the satellite radiances combined with surfaceand upper-air meteorological observations is greater thanthat after only assimilating satellite radiances

413 Temporal Variations of Precipitation The greater dif-ference between Event A and Event B is the rainfall dis-tribution in time Figure 9(a) shows the rainfall intensitiesand cumulative curves of the measured and modeled resultsfor Event A at Laohugou station The cumulative curvesof CTRL experiment are lie below and far away from thatof the measured values However during the period from

Advances in Meteorology 9

Table 3 Differences between observation and numerical experiments

Event ID Experiments 1198772 RMSE 0ndash5mm ME

5ndash10mm 10ndash425 (235) 0ndash425 (235)

Event A

CTRL 024 105 minus015 42 216 56DA-SAT 043 81 12 31 172 35DA-OBS 033 92 04 39 189 46DA-BOTH 070 58 minus03 11 95 18

Event B

CTRL 032 55 minus28 minus14 84 minus02DA-SAT 059 50 minus24 minus16 54 minus05DA-OBS 048 54 minus19 minus23 46 minus07DA-BOTH 064 49 minus25 minus13 50 minus06

0

1

2

3

4

5

6

7

8

9

100

5

10

15

20

25

6-18

-36-

18-6

6-18

-96-

18-1

26-

18-1

56-

18-1

86-

18-2

16-

18-2

46-

19-3

6-19

-66-

19-9

6-19

-12

6-19

-15

6-19

-18

6-19

-21

6-19

-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRL

DA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

20

40

60

80

100

120

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)Measured valueCTRLDA-OBS

DA-SATDA-BOTH

Rain

fall

inte

nsity

(mm

15

hou

r)

(b)

Figure 9The rainfall intensities and cumulative curves of the values measured at the sites and the simulated results from the four numericalexperiments (a) Event A (b) Event B

000 to 2400 on July 19 the cumulative curves for DA-SATand DA-SAT experiments are closer to that of the measuredvalues compared with the other two experiments This maybe because the maximum precipitation intensities simulatedby DA-BOTH and DA-SAT experiments are closer to theobserved values from 600 to 900 on July 19

In Figure 9(b) for event B the precipitation intensityin CTRL experiment is significantly larger than the mea-sured values in first 6 hours Therefore the cumulativecurve in the CTRL experiment is far above that of themeasured values due to the inaccurate initial field Afterdata assimilation the rainfall intensities in DA-SAT and DA-BOTH experiments are closer to the observations in first 6hours while the rainfall intensity in DA-OBS experiment ismuch less than the observations As a result the cumulativecurve in DA-BOTH experiment is closest to the curve

of the observed values DA-SAT experiment is next andDA-OBS experiment is farthest It is noteworthy that forEvent B the cumulative curve in DA-SAT experiment iscloser to that of the observed values in comparison withEvent A

The absolute error (AE) has been calculated for bothEventA andEvent B In Figure 10(a) for EventA theAEof thenumerical experiments becomes bigger and bigger with theincrease of rainfall intensity while the values in DA-BOTHandDA-SAT experiments are relatively smaller Itmeans thatfor the convective rainfall like Event A WRF model failsin capturing the whole process of the event However inFigure 10(b) for Event B with even distribution in time theAE of the numerical experiments is relatively constant whilethe values for CTRL and DA-OBS experiments are relativelylarger

10 Advances in Meteorology

0

2

4

6

8

10

12

140

2

4

6

8

10

12

14

06-1

8-3

06-1

8-6

06-1

8-9

06-1

8-12

06-1

8-15

06-1

8-18

06-1

8-21

06-1

8-24

06-1

9-3

06-1

9-6

06-1

9-9

06-1

9-12

06-1

9-15

06-1

9-18

06-1

9-21

06-1

9-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

10

20

30

40

50

60

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

15

hou

r)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(b)

Figure 10TheAEof rainfall intensity between the valuesmeasured at the sites and the simulated results from the four numerical experiments(a) Event A (b) Event B

The results shows that for the temporal variation ofrainfall the improvement in DA-SAT experiment is betterthan that in DA-OBS experiment due to the lack of groundobservational data at high-altitude region And for Event Bthe improvement in DA-SAT is greater than Event A

42 Impacts of Data Assimilation on the Initial Field Fromthe above analysis it is clear that data assimilation exper-iments can improve rainfall forecast The purpose of dataassimilation is to acquire accurate initial fields for numericalmodels To reveal the impact of data assimilation experimentson the initial fields EventA is chosen as an example to analyzethe difference in the initial fields between data assimilationand control experiments The geopotential height and mois-ture flux at 850 hPa are selected for analysis

Figure 11 shows the geopotential height differencesbetween data assimilation and control experimentsCompared with CTRL experiment (Figure 11(a)) thereis no evident change of geopotential height in DA-OBS experiment (Figure 11(b)) or DA-SAT experiment(Figure 11(c)) Figure 8(d) shows the geopotential heightfor DA-BOTH experiment which is much smaller thanthe values from CTRL experiment The region with thelargest changes is mainly located in Qinghai Provinceand the northwest of Gansu Province which is consistentwith the improved forecasts of precipitation area andaccumulation in DA-BOTH experiment As can be seen inFigure 12 after data assimilation the moisture flux changenoticeably in the study area in comparison with the controlexperiment (Figure 12(a)) The moisture flux of DA-SATexperiment is markedly increased in the north of the studyarea (Figure 12(b)) It may bring more precipitation in the

northeast of Qinghai Province In Figure 12(c) there is aclear flow of moisture to the northwestern part of GansuProvince which results in the heavy rain The biggest changeof moisture flux occurs in Figure 12(d) for DA-BOTHexperiment which covers the largest area compared withother numerical experiments and it could give greatestestimated precipitation in northeastern Tibetan Plateau

5 Conclusions

In this study the 3DVar assimilation system is used toimprove the forecasting of heavy rain in spatial and timedistribution in northeastern Tibetan Plateau A control andthree data assimilation experiments were designed Theseexperiments demonstrate that the rain forecast is significantlyimproved after data assimilation through enhancing theaccuracy of the initial field Two heavy rainfall events withdifferent evenness in space and time are used to examine theimprovement for rainfall forecast

For the spatial distribution data assimilation experimentschanged significantly the rainfall area and accumulationfor the study area For Event A the data assimilationexperiments provide larger rainfall areas and accumulationwhile for Event B the data assimilation experiments providesmaller precipitation accumulation on the southeastern ofthe study area Rain-gauge data from 43 sites were usedto evaluate the impact after data assimilation The resultsshows that DA-BOTH experiment has the highest predictionaccuracy followed by the DA-SAT experiment and thenDA-OBS experiment while the CTRL experiment has thelowest accuracy For the temporal variation the forecastresults have great differences between Event A and Event B

Advances in Meteorology 11

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(d)

Figure 11 Geopotential height initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

The satellite radiances for Event B have greater positive effectthan that for Event A For the two events the cumulativecurve in DA-BOTH experiment is closest to the curveof the measured values while in DA-OBS experiment theaccumulation curves are far below to the curve of themeasured values due to the lack of ground observations inthe high-altitude region In conclusion both for the spatialand temporal distribution of rainfall the satellite radianceshave greater effect than surface and upper-air meteorologicalobservations in high-latitude regions of the northeasternedge and the improvement for the assimilation of satelliteradiances together with surface and upper-air meteorologicalobservations is greater than that for assimilating either satel-lite radiances ormeteorology observations Event A is chosento evaluate the impact of data assimilation on the initialfields It is found that better and more detailed informationis added to the initial fields (eg geopotential height andmoisture flux) especially on the northwest of Gansu Province

andQinghai Province where the forecasts of precipitation areimproved significantly

It should be mentioned that many other assimilationtechniques such as 4DVar EnKF and so forth need to betested in this study These approaches have great potentialalthough they currently suffer from unaffordable computercosts Meanwhile we need to analyse the impact of differentparameterization schemes on the WRF model after dataassimilation because in this study the conclusions drawn aresubject to the specific parameterization schemes used here

Appendix

Statistic Calculation

The coefficient of determination (1198772) root mean squaredeviation (RMSE) mean error (ME) and absolute error (AE)

12 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(d)

Figure 12 Moisture flux initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

were used to evaluate the impact of the DA experiments inthis study They are defined as

119877

=

(1119873)sum119873

119894=1(119865119894 minus 119865) (119874119894 minus 119874)

radic(1119873)sum119873

119894=1(119865119894 minus 119865)

2radic(1119873)sum

119873

119894=1(119874119894 minus 119874)

2

RMSE = radic 1119873

119873

sum

119894=1

(119865119894 minus 119874119894)2

ME = 1119873

119873

sum

119894=1

(119874119894 minus 119865119894)

AE = 1119873

119873

sum

119894=1

1003816100381610038161003816119865119894 minus 1198741198941003816100381610038161003816

(A1)

Here 119865 is simulated value 119874 is observed value 119873 is thenumber of sites

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work was supported by the National natural ScienceFoundation of China (41190080 40971290) Strategic LeadingTechnology Special Fund of Chinese Academy of Sciences(XDB03030204) the National Scientific and TechnologicalSupport Project (2013BAB05B03) and the Opening Fund ofKey Laboratory of Land Surface Process and Climate Changein Cold and Arid Regions CAS (LPCC201205)

Advances in Meteorology 13

References

[1] P Chambon S Q Zhang A Y Hou M Zupanski and SCheung ldquoAssessing the impact of pre-GPMmicrowave precipi-tation observations in the GoddardWRF ensemble data assim-ilation systemrdquo Quarterly Journal of the Royal MeteorologicalSociety vol 140 no 681 pp 1219ndash1235 2013

[2] J-H Seo Y H Lee and Y-H Kim ldquoFeature selection forvery short-term heavy rainfall prediction using evolutionarycomputationrdquo Advances in Meteorology vol 2014 Article ID203545 15 pages 2014

[3] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M JEsteban-Parra and Y Castro-Dıez ldquoHigh-resolution projec-tions of mean and extreme precipitation over Spain usingthe WRF model (2070ndash2099 versus 1970ndash1999)rdquo Journal ofGeophysical Research D Atmospheres vol 117 no 12 Article IDD12108 2012

[4] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M J SEsteban-Parra and Y Castro-Dıez ldquoEvaluation of WRF meanand extreme precipitation over Spain present climate (1970ndash99)rdquo Journal of Climate vol 25 no 14 pp 4883ndash4895 2012

[5] R M Cardoso P M M Soares P M A Miranda and MBelo-Pereira ldquoWRF high resolution simulation of Iberianmeanand extreme precipitation climaterdquo International Journal ofClimatology vol 33 no 11 pp 2591ndash2608 2013

[6] L Calvetti and A J Pereira Filho ldquoEnsemble hydrometeoro-logical forecasts using WRF hourly QPF and topmodel for amiddle watershedrdquo Advances in Meteorology vol 2014 ArticleID 484120 12 pages 2014

[7] S Q Zhang M Zupanski A Y Hou X Lin and S H CheungldquoAssimilation of precipitation-affected radiances in a cloud-resolving wrf ensemble data assimilation systemrdquo MonthlyWeather Review vol 141 no 2 pp 754ndash772 2013

[8] J Liu M Bray and D Han ldquoExploring the effect of dataassimilation by WRF-3DVar for numerical rainfall predictionwith different types of storm eventsrdquoHydrological Processes vol27 no 25 pp 3627ndash3640 2013

[9] C-H Xiong L-F Zhang J-P Guan J Peng and B ZhangldquoAnalysis and numerical study of a hybrid BGM-3DVAR dataassimilation scheme using satellite radiance data for heavy rainforecastsrdquo Journal of Hydrodynamics vol 25 no 3 pp 430ndash4392013

[10] D M Barker W Huang Y R Guo and Q N Xiao ldquoA three-dimensional (3DVAR) data assimilation system for use withMM5 implementation and initial resultsrdquo Monthly WeatherReview vol 132 pp 897ndash914 2004

[11] Q Xiao and J Sun ldquoMultiple-radar data assimilation and short-range quantitative precipitation forecasting of a squall lineobserved during IHOP 2002rdquoMonthlyWeather Review vol 135no 10 pp 3381ndash3404 2007

[12] Z Sokol ldquoEffects of an assimilation of radar and satellite dataon a very-short range forecast of heavy convective rainfallsrdquoAtmospheric Research vol 93 no 1ndash3 pp 188ndash206 2009

[13] P Bauer P Lopez A Benedetti D Salmond and E MoreauldquoImplementation of 1D+4D-Var assimilation of precipitation-affected microwave radiances at ECMWF I 1D-Varrdquo QuarterlyJournal of the Royal Meteorological Society vol 132 no 620 pp2277ndash2306 2006

[14] P Bauer P Lopez D Salmond A Benedetti S Saarinen andM Bonazzola ldquoImplementation of 1D+4D-Var assimilation of

precipitation-affected microwave radiances at ECMWF II 4D-Varrdquo Quarterly Journal of the Royal Meteorological Society vol132 no 620 pp 2307ndash2332 2006

[15] P Bauer A J Geer P Lopez and D Salmond ldquoDirect 4D-Var assimilation of all-sky radiances Part I ImplementationrdquoQuarterly Journal of the Royal Meteorological Society vol 136no 652 pp 1868ndash1885 2010

[16] A J Geer P Bauer and P Lopez ldquoDirect 4D-Var assimilationof all-sky radiances Part II assessmentrdquo Quarterly Journal ofthe RoyalMeteorological Society vol 136 no 652 pp 1886ndash19052010

[17] Z Sokol and D Rezacova ldquoAssimilation of radar reflectivityinto the LM COSMOmodel with a high horizontal resolutionrdquoMeteorological Applications vol 13 no 4 pp 317ndash330 2006

[18] Z Sokol ldquoAssimilation of extrapolated radar reflectivity into aNWP model and its impact on a precipitation forecast at highresolutionrdquo Atmospheric Research vol 100 no 2-3 pp 201ndash2122011

[19] HWang J Sun S Fan and X-Y Huang ldquoIndirect assimilationof radar reflectivity with WRF 3D-var and its impact onprediction of four summertime convective eventsrdquo Journal ofApplied Meteorology and Climatology vol 52 no 4 pp 889ndash902 2013

[20] B Macpherson ldquoOperational experience with assimilation ofrainfall data in the Met Office mesoscale modelrdquo Meteorologyand Atmospheric Physics vol 76 no 1-2 pp 3ndash8 2001

[21] K Stephan S Klink and C Schraff ldquoAssimilation of radar-derived rain rates into the convective-scale model COSMO-DEat DWDrdquo Quarterly Journal of the Royal Meteorological Societyvol 134 no 634 pp 1315ndash1326 2008

[22] K Alapaty D S Niyogi F Chen P Pyle A Chandrasekar andN Seaman ldquoDevelopment of the flux-adjusting surface dataassimilation system for mesoscale modelsrdquo Journal of AppliedMeteorology and Climatology vol 47 no 9 pp 2331ndash2350 2008

[23] F H Ruggiero G D Modica and A E Lipton ldquoAssimilationof satellite imager data and surface observations to improveanalysis of circulations forced by cloud shading contrastsrdquoMonthly Weather Review vol 128 no 2 pp 434ndash448 2000

[24] J P Hacker and D Rostkier-Edelstein ldquoPBL state estimationwith surface observations a column model and an ensemblefilterrdquo Monthly Weather Review vol 135 no 8 pp 2958ndash29722007

[25] D J Stensrud N Yussouf D C Dowell and M C ConiglioldquoAssimilating surface data into a mesoscale model ensemblecold pool analyses from spring 2007rdquoAtmospheric Research vol93 no 1ndash3 pp 207ndash220 2009

[26] Q Wan and J Xu ldquoA numerical study of the rainstormcharacteristics of the June 2005 flash flood with WRFGSIdata assimilation system over south-east Chinardquo HydrologicalProcesses vol 25 no 8 pp 1327ndash1341 2011

[27] J Liu M Bray and D Han ldquoA study on WRF radar dataassimilation for hydrological rainfall predictionrdquoHydrology andEarth System Sciences vol 17 no 8 pp 3095ndash3110 2013

[28] K Ide P Courtier M Ghil and A C Lorenc ldquoUnified notationfor data assimilation operational sequential and variationalrdquoJournal of the Meteorological Society of Japan vol 75 no 1 pp181ndash189 1997

[29] A C Lorenc ldquoAnalysis methods for numerical weather predic-tionrdquo Quarterly Journal of the Royal Meteorological Society vol112 no 474 pp 1177ndash1194 1966

14 Advances in Meteorology

[30] D F Parrish and J C Derber ldquoThe nationalmeteorological cen-terrsquos spectral statistical interpolation analysis systemrdquo MonthlyWeather Review vol 120 no 8 pp 1747ndash1763 1992

[31] T B Zhao and C B Fu ldquoPreliminary comparison and analysisbetween ERA-40 NCEP-2 reanalysis and observations overChinardquo Climatic and Environmental Research vol 11 no 1 pp14ndash32 2006

[32] L Ma T Zhang Q Li O W Frauenfeld and D QinldquoEvaluation of ERA-40 NCEP-1 and NCEP-2 reanalysis airtemperatures with ground-based measurements in ChinardquoJournal of Geophysical Research Atmospheres (1984ndash2012) vol13 no D15 1984

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ClimatologyJournal of

EcologyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

EarthquakesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

Mining

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

International Journal of

Geophysics

OceanographyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofPetroleum Engineering

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Atmospheric SciencesInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MineralogyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MeteorologyAdvances in

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geological ResearchJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geology Advances in

Advances in Meteorology 9

Table 3 Differences between observation and numerical experiments

Event ID Experiments 1198772 RMSE 0ndash5mm ME

5ndash10mm 10ndash425 (235) 0ndash425 (235)

Event A

CTRL 024 105 minus015 42 216 56DA-SAT 043 81 12 31 172 35DA-OBS 033 92 04 39 189 46DA-BOTH 070 58 minus03 11 95 18

Event B

CTRL 032 55 minus28 minus14 84 minus02DA-SAT 059 50 minus24 minus16 54 minus05DA-OBS 048 54 minus19 minus23 46 minus07DA-BOTH 064 49 minus25 minus13 50 minus06

0

1

2

3

4

5

6

7

8

9

100

5

10

15

20

25

6-18

-36-

18-6

6-18

-96-

18-1

26-

18-1

56-

18-1

86-

18-2

16-

18-2

46-

19-3

6-19

-66-

19-9

6-19

-12

6-19

-15

6-19

-18

6-19

-21

6-19

-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRL

DA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

20

40

60

80

100

120

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Cum

ulat

ive r

ainf

all (

mm

)

Time (month-day-hour)Measured valueCTRLDA-OBS

DA-SATDA-BOTH

Rain

fall

inte

nsity

(mm

15

hou

r)

(b)

Figure 9The rainfall intensities and cumulative curves of the values measured at the sites and the simulated results from the four numericalexperiments (a) Event A (b) Event B

000 to 2400 on July 19 the cumulative curves for DA-SATand DA-SAT experiments are closer to that of the measuredvalues compared with the other two experiments This maybe because the maximum precipitation intensities simulatedby DA-BOTH and DA-SAT experiments are closer to theobserved values from 600 to 900 on July 19

In Figure 9(b) for event B the precipitation intensityin CTRL experiment is significantly larger than the mea-sured values in first 6 hours Therefore the cumulativecurve in the CTRL experiment is far above that of themeasured values due to the inaccurate initial field Afterdata assimilation the rainfall intensities in DA-SAT and DA-BOTH experiments are closer to the observations in first 6hours while the rainfall intensity in DA-OBS experiment ismuch less than the observations As a result the cumulativecurve in DA-BOTH experiment is closest to the curve

of the observed values DA-SAT experiment is next andDA-OBS experiment is farthest It is noteworthy that forEvent B the cumulative curve in DA-SAT experiment iscloser to that of the observed values in comparison withEvent A

The absolute error (AE) has been calculated for bothEventA andEvent B In Figure 10(a) for EventA theAEof thenumerical experiments becomes bigger and bigger with theincrease of rainfall intensity while the values in DA-BOTHandDA-SAT experiments are relatively smaller Itmeans thatfor the convective rainfall like Event A WRF model failsin capturing the whole process of the event However inFigure 10(b) for Event B with even distribution in time theAE of the numerical experiments is relatively constant whilethe values for CTRL and DA-OBS experiments are relativelylarger

10 Advances in Meteorology

0

2

4

6

8

10

12

140

2

4

6

8

10

12

14

06-1

8-3

06-1

8-6

06-1

8-9

06-1

8-12

06-1

8-15

06-1

8-18

06-1

8-21

06-1

8-24

06-1

9-3

06-1

9-6

06-1

9-9

06-1

9-12

06-1

9-15

06-1

9-18

06-1

9-21

06-1

9-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

10

20

30

40

50

60

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

15

hou

r)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(b)

Figure 10TheAEof rainfall intensity between the valuesmeasured at the sites and the simulated results from the four numerical experiments(a) Event A (b) Event B

The results shows that for the temporal variation ofrainfall the improvement in DA-SAT experiment is betterthan that in DA-OBS experiment due to the lack of groundobservational data at high-altitude region And for Event Bthe improvement in DA-SAT is greater than Event A

42 Impacts of Data Assimilation on the Initial Field Fromthe above analysis it is clear that data assimilation exper-iments can improve rainfall forecast The purpose of dataassimilation is to acquire accurate initial fields for numericalmodels To reveal the impact of data assimilation experimentson the initial fields EventA is chosen as an example to analyzethe difference in the initial fields between data assimilationand control experiments The geopotential height and mois-ture flux at 850 hPa are selected for analysis

Figure 11 shows the geopotential height differencesbetween data assimilation and control experimentsCompared with CTRL experiment (Figure 11(a)) thereis no evident change of geopotential height in DA-OBS experiment (Figure 11(b)) or DA-SAT experiment(Figure 11(c)) Figure 8(d) shows the geopotential heightfor DA-BOTH experiment which is much smaller thanthe values from CTRL experiment The region with thelargest changes is mainly located in Qinghai Provinceand the northwest of Gansu Province which is consistentwith the improved forecasts of precipitation area andaccumulation in DA-BOTH experiment As can be seen inFigure 12 after data assimilation the moisture flux changenoticeably in the study area in comparison with the controlexperiment (Figure 12(a)) The moisture flux of DA-SATexperiment is markedly increased in the north of the studyarea (Figure 12(b)) It may bring more precipitation in the

northeast of Qinghai Province In Figure 12(c) there is aclear flow of moisture to the northwestern part of GansuProvince which results in the heavy rain The biggest changeof moisture flux occurs in Figure 12(d) for DA-BOTHexperiment which covers the largest area compared withother numerical experiments and it could give greatestestimated precipitation in northeastern Tibetan Plateau

5 Conclusions

In this study the 3DVar assimilation system is used toimprove the forecasting of heavy rain in spatial and timedistribution in northeastern Tibetan Plateau A control andthree data assimilation experiments were designed Theseexperiments demonstrate that the rain forecast is significantlyimproved after data assimilation through enhancing theaccuracy of the initial field Two heavy rainfall events withdifferent evenness in space and time are used to examine theimprovement for rainfall forecast

For the spatial distribution data assimilation experimentschanged significantly the rainfall area and accumulationfor the study area For Event A the data assimilationexperiments provide larger rainfall areas and accumulationwhile for Event B the data assimilation experiments providesmaller precipitation accumulation on the southeastern ofthe study area Rain-gauge data from 43 sites were usedto evaluate the impact after data assimilation The resultsshows that DA-BOTH experiment has the highest predictionaccuracy followed by the DA-SAT experiment and thenDA-OBS experiment while the CTRL experiment has thelowest accuracy For the temporal variation the forecastresults have great differences between Event A and Event B

Advances in Meteorology 11

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(d)

Figure 11 Geopotential height initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

The satellite radiances for Event B have greater positive effectthan that for Event A For the two events the cumulativecurve in DA-BOTH experiment is closest to the curveof the measured values while in DA-OBS experiment theaccumulation curves are far below to the curve of themeasured values due to the lack of ground observations inthe high-altitude region In conclusion both for the spatialand temporal distribution of rainfall the satellite radianceshave greater effect than surface and upper-air meteorologicalobservations in high-latitude regions of the northeasternedge and the improvement for the assimilation of satelliteradiances together with surface and upper-air meteorologicalobservations is greater than that for assimilating either satel-lite radiances ormeteorology observations Event A is chosento evaluate the impact of data assimilation on the initialfields It is found that better and more detailed informationis added to the initial fields (eg geopotential height andmoisture flux) especially on the northwest of Gansu Province

andQinghai Province where the forecasts of precipitation areimproved significantly

It should be mentioned that many other assimilationtechniques such as 4DVar EnKF and so forth need to betested in this study These approaches have great potentialalthough they currently suffer from unaffordable computercosts Meanwhile we need to analyse the impact of differentparameterization schemes on the WRF model after dataassimilation because in this study the conclusions drawn aresubject to the specific parameterization schemes used here

Appendix

Statistic Calculation

The coefficient of determination (1198772) root mean squaredeviation (RMSE) mean error (ME) and absolute error (AE)

12 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(d)

Figure 12 Moisture flux initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

were used to evaluate the impact of the DA experiments inthis study They are defined as

119877

=

(1119873)sum119873

119894=1(119865119894 minus 119865) (119874119894 minus 119874)

radic(1119873)sum119873

119894=1(119865119894 minus 119865)

2radic(1119873)sum

119873

119894=1(119874119894 minus 119874)

2

RMSE = radic 1119873

119873

sum

119894=1

(119865119894 minus 119874119894)2

ME = 1119873

119873

sum

119894=1

(119874119894 minus 119865119894)

AE = 1119873

119873

sum

119894=1

1003816100381610038161003816119865119894 minus 1198741198941003816100381610038161003816

(A1)

Here 119865 is simulated value 119874 is observed value 119873 is thenumber of sites

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work was supported by the National natural ScienceFoundation of China (41190080 40971290) Strategic LeadingTechnology Special Fund of Chinese Academy of Sciences(XDB03030204) the National Scientific and TechnologicalSupport Project (2013BAB05B03) and the Opening Fund ofKey Laboratory of Land Surface Process and Climate Changein Cold and Arid Regions CAS (LPCC201205)

Advances in Meteorology 13

References

[1] P Chambon S Q Zhang A Y Hou M Zupanski and SCheung ldquoAssessing the impact of pre-GPMmicrowave precipi-tation observations in the GoddardWRF ensemble data assim-ilation systemrdquo Quarterly Journal of the Royal MeteorologicalSociety vol 140 no 681 pp 1219ndash1235 2013

[2] J-H Seo Y H Lee and Y-H Kim ldquoFeature selection forvery short-term heavy rainfall prediction using evolutionarycomputationrdquo Advances in Meteorology vol 2014 Article ID203545 15 pages 2014

[3] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M JEsteban-Parra and Y Castro-Dıez ldquoHigh-resolution projec-tions of mean and extreme precipitation over Spain usingthe WRF model (2070ndash2099 versus 1970ndash1999)rdquo Journal ofGeophysical Research D Atmospheres vol 117 no 12 Article IDD12108 2012

[4] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M J SEsteban-Parra and Y Castro-Dıez ldquoEvaluation of WRF meanand extreme precipitation over Spain present climate (1970ndash99)rdquo Journal of Climate vol 25 no 14 pp 4883ndash4895 2012

[5] R M Cardoso P M M Soares P M A Miranda and MBelo-Pereira ldquoWRF high resolution simulation of Iberianmeanand extreme precipitation climaterdquo International Journal ofClimatology vol 33 no 11 pp 2591ndash2608 2013

[6] L Calvetti and A J Pereira Filho ldquoEnsemble hydrometeoro-logical forecasts using WRF hourly QPF and topmodel for amiddle watershedrdquo Advances in Meteorology vol 2014 ArticleID 484120 12 pages 2014

[7] S Q Zhang M Zupanski A Y Hou X Lin and S H CheungldquoAssimilation of precipitation-affected radiances in a cloud-resolving wrf ensemble data assimilation systemrdquo MonthlyWeather Review vol 141 no 2 pp 754ndash772 2013

[8] J Liu M Bray and D Han ldquoExploring the effect of dataassimilation by WRF-3DVar for numerical rainfall predictionwith different types of storm eventsrdquoHydrological Processes vol27 no 25 pp 3627ndash3640 2013

[9] C-H Xiong L-F Zhang J-P Guan J Peng and B ZhangldquoAnalysis and numerical study of a hybrid BGM-3DVAR dataassimilation scheme using satellite radiance data for heavy rainforecastsrdquo Journal of Hydrodynamics vol 25 no 3 pp 430ndash4392013

[10] D M Barker W Huang Y R Guo and Q N Xiao ldquoA three-dimensional (3DVAR) data assimilation system for use withMM5 implementation and initial resultsrdquo Monthly WeatherReview vol 132 pp 897ndash914 2004

[11] Q Xiao and J Sun ldquoMultiple-radar data assimilation and short-range quantitative precipitation forecasting of a squall lineobserved during IHOP 2002rdquoMonthlyWeather Review vol 135no 10 pp 3381ndash3404 2007

[12] Z Sokol ldquoEffects of an assimilation of radar and satellite dataon a very-short range forecast of heavy convective rainfallsrdquoAtmospheric Research vol 93 no 1ndash3 pp 188ndash206 2009

[13] P Bauer P Lopez A Benedetti D Salmond and E MoreauldquoImplementation of 1D+4D-Var assimilation of precipitation-affected microwave radiances at ECMWF I 1D-Varrdquo QuarterlyJournal of the Royal Meteorological Society vol 132 no 620 pp2277ndash2306 2006

[14] P Bauer P Lopez D Salmond A Benedetti S Saarinen andM Bonazzola ldquoImplementation of 1D+4D-Var assimilation of

precipitation-affected microwave radiances at ECMWF II 4D-Varrdquo Quarterly Journal of the Royal Meteorological Society vol132 no 620 pp 2307ndash2332 2006

[15] P Bauer A J Geer P Lopez and D Salmond ldquoDirect 4D-Var assimilation of all-sky radiances Part I ImplementationrdquoQuarterly Journal of the Royal Meteorological Society vol 136no 652 pp 1868ndash1885 2010

[16] A J Geer P Bauer and P Lopez ldquoDirect 4D-Var assimilationof all-sky radiances Part II assessmentrdquo Quarterly Journal ofthe RoyalMeteorological Society vol 136 no 652 pp 1886ndash19052010

[17] Z Sokol and D Rezacova ldquoAssimilation of radar reflectivityinto the LM COSMOmodel with a high horizontal resolutionrdquoMeteorological Applications vol 13 no 4 pp 317ndash330 2006

[18] Z Sokol ldquoAssimilation of extrapolated radar reflectivity into aNWP model and its impact on a precipitation forecast at highresolutionrdquo Atmospheric Research vol 100 no 2-3 pp 201ndash2122011

[19] HWang J Sun S Fan and X-Y Huang ldquoIndirect assimilationof radar reflectivity with WRF 3D-var and its impact onprediction of four summertime convective eventsrdquo Journal ofApplied Meteorology and Climatology vol 52 no 4 pp 889ndash902 2013

[20] B Macpherson ldquoOperational experience with assimilation ofrainfall data in the Met Office mesoscale modelrdquo Meteorologyand Atmospheric Physics vol 76 no 1-2 pp 3ndash8 2001

[21] K Stephan S Klink and C Schraff ldquoAssimilation of radar-derived rain rates into the convective-scale model COSMO-DEat DWDrdquo Quarterly Journal of the Royal Meteorological Societyvol 134 no 634 pp 1315ndash1326 2008

[22] K Alapaty D S Niyogi F Chen P Pyle A Chandrasekar andN Seaman ldquoDevelopment of the flux-adjusting surface dataassimilation system for mesoscale modelsrdquo Journal of AppliedMeteorology and Climatology vol 47 no 9 pp 2331ndash2350 2008

[23] F H Ruggiero G D Modica and A E Lipton ldquoAssimilationof satellite imager data and surface observations to improveanalysis of circulations forced by cloud shading contrastsrdquoMonthly Weather Review vol 128 no 2 pp 434ndash448 2000

[24] J P Hacker and D Rostkier-Edelstein ldquoPBL state estimationwith surface observations a column model and an ensemblefilterrdquo Monthly Weather Review vol 135 no 8 pp 2958ndash29722007

[25] D J Stensrud N Yussouf D C Dowell and M C ConiglioldquoAssimilating surface data into a mesoscale model ensemblecold pool analyses from spring 2007rdquoAtmospheric Research vol93 no 1ndash3 pp 207ndash220 2009

[26] Q Wan and J Xu ldquoA numerical study of the rainstormcharacteristics of the June 2005 flash flood with WRFGSIdata assimilation system over south-east Chinardquo HydrologicalProcesses vol 25 no 8 pp 1327ndash1341 2011

[27] J Liu M Bray and D Han ldquoA study on WRF radar dataassimilation for hydrological rainfall predictionrdquoHydrology andEarth System Sciences vol 17 no 8 pp 3095ndash3110 2013

[28] K Ide P Courtier M Ghil and A C Lorenc ldquoUnified notationfor data assimilation operational sequential and variationalrdquoJournal of the Meteorological Society of Japan vol 75 no 1 pp181ndash189 1997

[29] A C Lorenc ldquoAnalysis methods for numerical weather predic-tionrdquo Quarterly Journal of the Royal Meteorological Society vol112 no 474 pp 1177ndash1194 1966

14 Advances in Meteorology

[30] D F Parrish and J C Derber ldquoThe nationalmeteorological cen-terrsquos spectral statistical interpolation analysis systemrdquo MonthlyWeather Review vol 120 no 8 pp 1747ndash1763 1992

[31] T B Zhao and C B Fu ldquoPreliminary comparison and analysisbetween ERA-40 NCEP-2 reanalysis and observations overChinardquo Climatic and Environmental Research vol 11 no 1 pp14ndash32 2006

[32] L Ma T Zhang Q Li O W Frauenfeld and D QinldquoEvaluation of ERA-40 NCEP-1 and NCEP-2 reanalysis airtemperatures with ground-based measurements in ChinardquoJournal of Geophysical Research Atmospheres (1984ndash2012) vol13 no D15 1984

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ClimatologyJournal of

EcologyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

EarthquakesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

Mining

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

International Journal of

Geophysics

OceanographyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofPetroleum Engineering

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Atmospheric SciencesInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MineralogyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MeteorologyAdvances in

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geological ResearchJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geology Advances in

10 Advances in Meteorology

0

2

4

6

8

10

12

140

2

4

6

8

10

12

14

06-1

8-3

06-1

8-6

06-1

8-9

06-1

8-12

06-1

8-15

06-1

8-18

06-1

8-21

06-1

8-24

06-1

9-3

06-1

9-6

06-1

9-9

06-1

9-12

06-1

9-15

06-1

9-18

06-1

9-21

06-1

9-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

3 h

our)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(a)

00

10

20

30

40

50

60

70

8000

10

20

30

40

50

60

6-7-

15

6-7-

36-

7-4

56-

7-6

6-7-

75

6-7-

96-

7-10

56-

7-12

6-7-

135

6-7-

156-

7-16

56-

7-18

6-7-

195

6-7-

216-

7-22

56-

7-24

Abso

lute

erro

r (m

m)

Time (month-day-hour)

Rain

fall

inte

nsity

(mm

15

hou

r)

Measured valueCTRLDA-OBS

DA-SATDA-BOTH

(b)

Figure 10TheAEof rainfall intensity between the valuesmeasured at the sites and the simulated results from the four numerical experiments(a) Event A (b) Event B

The results shows that for the temporal variation ofrainfall the improvement in DA-SAT experiment is betterthan that in DA-OBS experiment due to the lack of groundobservational data at high-altitude region And for Event Bthe improvement in DA-SAT is greater than Event A

42 Impacts of Data Assimilation on the Initial Field Fromthe above analysis it is clear that data assimilation exper-iments can improve rainfall forecast The purpose of dataassimilation is to acquire accurate initial fields for numericalmodels To reveal the impact of data assimilation experimentson the initial fields EventA is chosen as an example to analyzethe difference in the initial fields between data assimilationand control experiments The geopotential height and mois-ture flux at 850 hPa are selected for analysis

Figure 11 shows the geopotential height differencesbetween data assimilation and control experimentsCompared with CTRL experiment (Figure 11(a)) thereis no evident change of geopotential height in DA-OBS experiment (Figure 11(b)) or DA-SAT experiment(Figure 11(c)) Figure 8(d) shows the geopotential heightfor DA-BOTH experiment which is much smaller thanthe values from CTRL experiment The region with thelargest changes is mainly located in Qinghai Provinceand the northwest of Gansu Province which is consistentwith the improved forecasts of precipitation area andaccumulation in DA-BOTH experiment As can be seen inFigure 12 after data assimilation the moisture flux changenoticeably in the study area in comparison with the controlexperiment (Figure 12(a)) The moisture flux of DA-SATexperiment is markedly increased in the north of the studyarea (Figure 12(b)) It may bring more precipitation in the

northeast of Qinghai Province In Figure 12(c) there is aclear flow of moisture to the northwestern part of GansuProvince which results in the heavy rain The biggest changeof moisture flux occurs in Figure 12(d) for DA-BOTHexperiment which covers the largest area compared withother numerical experiments and it could give greatestestimated precipitation in northeastern Tibetan Plateau

5 Conclusions

In this study the 3DVar assimilation system is used toimprove the forecasting of heavy rain in spatial and timedistribution in northeastern Tibetan Plateau A control andthree data assimilation experiments were designed Theseexperiments demonstrate that the rain forecast is significantlyimproved after data assimilation through enhancing theaccuracy of the initial field Two heavy rainfall events withdifferent evenness in space and time are used to examine theimprovement for rainfall forecast

For the spatial distribution data assimilation experimentschanged significantly the rainfall area and accumulationfor the study area For Event A the data assimilationexperiments provide larger rainfall areas and accumulationwhile for Event B the data assimilation experiments providesmaller precipitation accumulation on the southeastern ofthe study area Rain-gauge data from 43 sites were usedto evaluate the impact after data assimilation The resultsshows that DA-BOTH experiment has the highest predictionaccuracy followed by the DA-SAT experiment and thenDA-OBS experiment while the CTRL experiment has thelowest accuracy For the temporal variation the forecastresults have great differences between Event A and Event B

Advances in Meteorology 11

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(d)

Figure 11 Geopotential height initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

The satellite radiances for Event B have greater positive effectthan that for Event A For the two events the cumulativecurve in DA-BOTH experiment is closest to the curveof the measured values while in DA-OBS experiment theaccumulation curves are far below to the curve of themeasured values due to the lack of ground observations inthe high-altitude region In conclusion both for the spatialand temporal distribution of rainfall the satellite radianceshave greater effect than surface and upper-air meteorologicalobservations in high-latitude regions of the northeasternedge and the improvement for the assimilation of satelliteradiances together with surface and upper-air meteorologicalobservations is greater than that for assimilating either satel-lite radiances ormeteorology observations Event A is chosento evaluate the impact of data assimilation on the initialfields It is found that better and more detailed informationis added to the initial fields (eg geopotential height andmoisture flux) especially on the northwest of Gansu Province

andQinghai Province where the forecasts of precipitation areimproved significantly

It should be mentioned that many other assimilationtechniques such as 4DVar EnKF and so forth need to betested in this study These approaches have great potentialalthough they currently suffer from unaffordable computercosts Meanwhile we need to analyse the impact of differentparameterization schemes on the WRF model after dataassimilation because in this study the conclusions drawn aresubject to the specific parameterization schemes used here

Appendix

Statistic Calculation

The coefficient of determination (1198772) root mean squaredeviation (RMSE) mean error (ME) and absolute error (AE)

12 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(d)

Figure 12 Moisture flux initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

were used to evaluate the impact of the DA experiments inthis study They are defined as

119877

=

(1119873)sum119873

119894=1(119865119894 minus 119865) (119874119894 minus 119874)

radic(1119873)sum119873

119894=1(119865119894 minus 119865)

2radic(1119873)sum

119873

119894=1(119874119894 minus 119874)

2

RMSE = radic 1119873

119873

sum

119894=1

(119865119894 minus 119874119894)2

ME = 1119873

119873

sum

119894=1

(119874119894 minus 119865119894)

AE = 1119873

119873

sum

119894=1

1003816100381610038161003816119865119894 minus 1198741198941003816100381610038161003816

(A1)

Here 119865 is simulated value 119874 is observed value 119873 is thenumber of sites

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work was supported by the National natural ScienceFoundation of China (41190080 40971290) Strategic LeadingTechnology Special Fund of Chinese Academy of Sciences(XDB03030204) the National Scientific and TechnologicalSupport Project (2013BAB05B03) and the Opening Fund ofKey Laboratory of Land Surface Process and Climate Changein Cold and Arid Regions CAS (LPCC201205)

Advances in Meteorology 13

References

[1] P Chambon S Q Zhang A Y Hou M Zupanski and SCheung ldquoAssessing the impact of pre-GPMmicrowave precipi-tation observations in the GoddardWRF ensemble data assim-ilation systemrdquo Quarterly Journal of the Royal MeteorologicalSociety vol 140 no 681 pp 1219ndash1235 2013

[2] J-H Seo Y H Lee and Y-H Kim ldquoFeature selection forvery short-term heavy rainfall prediction using evolutionarycomputationrdquo Advances in Meteorology vol 2014 Article ID203545 15 pages 2014

[3] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M JEsteban-Parra and Y Castro-Dıez ldquoHigh-resolution projec-tions of mean and extreme precipitation over Spain usingthe WRF model (2070ndash2099 versus 1970ndash1999)rdquo Journal ofGeophysical Research D Atmospheres vol 117 no 12 Article IDD12108 2012

[4] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M J SEsteban-Parra and Y Castro-Dıez ldquoEvaluation of WRF meanand extreme precipitation over Spain present climate (1970ndash99)rdquo Journal of Climate vol 25 no 14 pp 4883ndash4895 2012

[5] R M Cardoso P M M Soares P M A Miranda and MBelo-Pereira ldquoWRF high resolution simulation of Iberianmeanand extreme precipitation climaterdquo International Journal ofClimatology vol 33 no 11 pp 2591ndash2608 2013

[6] L Calvetti and A J Pereira Filho ldquoEnsemble hydrometeoro-logical forecasts using WRF hourly QPF and topmodel for amiddle watershedrdquo Advances in Meteorology vol 2014 ArticleID 484120 12 pages 2014

[7] S Q Zhang M Zupanski A Y Hou X Lin and S H CheungldquoAssimilation of precipitation-affected radiances in a cloud-resolving wrf ensemble data assimilation systemrdquo MonthlyWeather Review vol 141 no 2 pp 754ndash772 2013

[8] J Liu M Bray and D Han ldquoExploring the effect of dataassimilation by WRF-3DVar for numerical rainfall predictionwith different types of storm eventsrdquoHydrological Processes vol27 no 25 pp 3627ndash3640 2013

[9] C-H Xiong L-F Zhang J-P Guan J Peng and B ZhangldquoAnalysis and numerical study of a hybrid BGM-3DVAR dataassimilation scheme using satellite radiance data for heavy rainforecastsrdquo Journal of Hydrodynamics vol 25 no 3 pp 430ndash4392013

[10] D M Barker W Huang Y R Guo and Q N Xiao ldquoA three-dimensional (3DVAR) data assimilation system for use withMM5 implementation and initial resultsrdquo Monthly WeatherReview vol 132 pp 897ndash914 2004

[11] Q Xiao and J Sun ldquoMultiple-radar data assimilation and short-range quantitative precipitation forecasting of a squall lineobserved during IHOP 2002rdquoMonthlyWeather Review vol 135no 10 pp 3381ndash3404 2007

[12] Z Sokol ldquoEffects of an assimilation of radar and satellite dataon a very-short range forecast of heavy convective rainfallsrdquoAtmospheric Research vol 93 no 1ndash3 pp 188ndash206 2009

[13] P Bauer P Lopez A Benedetti D Salmond and E MoreauldquoImplementation of 1D+4D-Var assimilation of precipitation-affected microwave radiances at ECMWF I 1D-Varrdquo QuarterlyJournal of the Royal Meteorological Society vol 132 no 620 pp2277ndash2306 2006

[14] P Bauer P Lopez D Salmond A Benedetti S Saarinen andM Bonazzola ldquoImplementation of 1D+4D-Var assimilation of

precipitation-affected microwave radiances at ECMWF II 4D-Varrdquo Quarterly Journal of the Royal Meteorological Society vol132 no 620 pp 2307ndash2332 2006

[15] P Bauer A J Geer P Lopez and D Salmond ldquoDirect 4D-Var assimilation of all-sky radiances Part I ImplementationrdquoQuarterly Journal of the Royal Meteorological Society vol 136no 652 pp 1868ndash1885 2010

[16] A J Geer P Bauer and P Lopez ldquoDirect 4D-Var assimilationof all-sky radiances Part II assessmentrdquo Quarterly Journal ofthe RoyalMeteorological Society vol 136 no 652 pp 1886ndash19052010

[17] Z Sokol and D Rezacova ldquoAssimilation of radar reflectivityinto the LM COSMOmodel with a high horizontal resolutionrdquoMeteorological Applications vol 13 no 4 pp 317ndash330 2006

[18] Z Sokol ldquoAssimilation of extrapolated radar reflectivity into aNWP model and its impact on a precipitation forecast at highresolutionrdquo Atmospheric Research vol 100 no 2-3 pp 201ndash2122011

[19] HWang J Sun S Fan and X-Y Huang ldquoIndirect assimilationof radar reflectivity with WRF 3D-var and its impact onprediction of four summertime convective eventsrdquo Journal ofApplied Meteorology and Climatology vol 52 no 4 pp 889ndash902 2013

[20] B Macpherson ldquoOperational experience with assimilation ofrainfall data in the Met Office mesoscale modelrdquo Meteorologyand Atmospheric Physics vol 76 no 1-2 pp 3ndash8 2001

[21] K Stephan S Klink and C Schraff ldquoAssimilation of radar-derived rain rates into the convective-scale model COSMO-DEat DWDrdquo Quarterly Journal of the Royal Meteorological Societyvol 134 no 634 pp 1315ndash1326 2008

[22] K Alapaty D S Niyogi F Chen P Pyle A Chandrasekar andN Seaman ldquoDevelopment of the flux-adjusting surface dataassimilation system for mesoscale modelsrdquo Journal of AppliedMeteorology and Climatology vol 47 no 9 pp 2331ndash2350 2008

[23] F H Ruggiero G D Modica and A E Lipton ldquoAssimilationof satellite imager data and surface observations to improveanalysis of circulations forced by cloud shading contrastsrdquoMonthly Weather Review vol 128 no 2 pp 434ndash448 2000

[24] J P Hacker and D Rostkier-Edelstein ldquoPBL state estimationwith surface observations a column model and an ensemblefilterrdquo Monthly Weather Review vol 135 no 8 pp 2958ndash29722007

[25] D J Stensrud N Yussouf D C Dowell and M C ConiglioldquoAssimilating surface data into a mesoscale model ensemblecold pool analyses from spring 2007rdquoAtmospheric Research vol93 no 1ndash3 pp 207ndash220 2009

[26] Q Wan and J Xu ldquoA numerical study of the rainstormcharacteristics of the June 2005 flash flood with WRFGSIdata assimilation system over south-east Chinardquo HydrologicalProcesses vol 25 no 8 pp 1327ndash1341 2011

[27] J Liu M Bray and D Han ldquoA study on WRF radar dataassimilation for hydrological rainfall predictionrdquoHydrology andEarth System Sciences vol 17 no 8 pp 3095ndash3110 2013

[28] K Ide P Courtier M Ghil and A C Lorenc ldquoUnified notationfor data assimilation operational sequential and variationalrdquoJournal of the Meteorological Society of Japan vol 75 no 1 pp181ndash189 1997

[29] A C Lorenc ldquoAnalysis methods for numerical weather predic-tionrdquo Quarterly Journal of the Royal Meteorological Society vol112 no 474 pp 1177ndash1194 1966

14 Advances in Meteorology

[30] D F Parrish and J C Derber ldquoThe nationalmeteorological cen-terrsquos spectral statistical interpolation analysis systemrdquo MonthlyWeather Review vol 120 no 8 pp 1747ndash1763 1992

[31] T B Zhao and C B Fu ldquoPreliminary comparison and analysisbetween ERA-40 NCEP-2 reanalysis and observations overChinardquo Climatic and Environmental Research vol 11 no 1 pp14ndash32 2006

[32] L Ma T Zhang Q Li O W Frauenfeld and D QinldquoEvaluation of ERA-40 NCEP-1 and NCEP-2 reanalysis airtemperatures with ground-based measurements in ChinardquoJournal of Geophysical Research Atmospheres (1984ndash2012) vol13 no D15 1984

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ClimatologyJournal of

EcologyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

EarthquakesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

Mining

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

International Journal of

Geophysics

OceanographyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofPetroleum Engineering

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Atmospheric SciencesInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MineralogyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MeteorologyAdvances in

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geological ResearchJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geology Advances in

Advances in Meteorology 11

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N94∘E 96∘E 98∘E 100∘E 102∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E

10000 15000 20000 25000 30000 35000 40000 45000

(m2s2)

(d)

Figure 11 Geopotential height initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

The satellite radiances for Event B have greater positive effectthan that for Event A For the two events the cumulativecurve in DA-BOTH experiment is closest to the curveof the measured values while in DA-OBS experiment theaccumulation curves are far below to the curve of themeasured values due to the lack of ground observations inthe high-altitude region In conclusion both for the spatialand temporal distribution of rainfall the satellite radianceshave greater effect than surface and upper-air meteorologicalobservations in high-latitude regions of the northeasternedge and the improvement for the assimilation of satelliteradiances together with surface and upper-air meteorologicalobservations is greater than that for assimilating either satel-lite radiances ormeteorology observations Event A is chosento evaluate the impact of data assimilation on the initialfields It is found that better and more detailed informationis added to the initial fields (eg geopotential height andmoisture flux) especially on the northwest of Gansu Province

andQinghai Province where the forecasts of precipitation areimproved significantly

It should be mentioned that many other assimilationtechniques such as 4DVar EnKF and so forth need to betested in this study These approaches have great potentialalthough they currently suffer from unaffordable computercosts Meanwhile we need to analyse the impact of differentparameterization schemes on the WRF model after dataassimilation because in this study the conclusions drawn aresubject to the specific parameterization schemes used here

Appendix

Statistic Calculation

The coefficient of determination (1198772) root mean squaredeviation (RMSE) mean error (ME) and absolute error (AE)

12 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(d)

Figure 12 Moisture flux initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

were used to evaluate the impact of the DA experiments inthis study They are defined as

119877

=

(1119873)sum119873

119894=1(119865119894 minus 119865) (119874119894 minus 119874)

radic(1119873)sum119873

119894=1(119865119894 minus 119865)

2radic(1119873)sum

119873

119894=1(119874119894 minus 119874)

2

RMSE = radic 1119873

119873

sum

119894=1

(119865119894 minus 119874119894)2

ME = 1119873

119873

sum

119894=1

(119874119894 minus 119865119894)

AE = 1119873

119873

sum

119894=1

1003816100381610038161003816119865119894 minus 1198741198941003816100381610038161003816

(A1)

Here 119865 is simulated value 119874 is observed value 119873 is thenumber of sites

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work was supported by the National natural ScienceFoundation of China (41190080 40971290) Strategic LeadingTechnology Special Fund of Chinese Academy of Sciences(XDB03030204) the National Scientific and TechnologicalSupport Project (2013BAB05B03) and the Opening Fund ofKey Laboratory of Land Surface Process and Climate Changein Cold and Arid Regions CAS (LPCC201205)

Advances in Meteorology 13

References

[1] P Chambon S Q Zhang A Y Hou M Zupanski and SCheung ldquoAssessing the impact of pre-GPMmicrowave precipi-tation observations in the GoddardWRF ensemble data assim-ilation systemrdquo Quarterly Journal of the Royal MeteorologicalSociety vol 140 no 681 pp 1219ndash1235 2013

[2] J-H Seo Y H Lee and Y-H Kim ldquoFeature selection forvery short-term heavy rainfall prediction using evolutionarycomputationrdquo Advances in Meteorology vol 2014 Article ID203545 15 pages 2014

[3] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M JEsteban-Parra and Y Castro-Dıez ldquoHigh-resolution projec-tions of mean and extreme precipitation over Spain usingthe WRF model (2070ndash2099 versus 1970ndash1999)rdquo Journal ofGeophysical Research D Atmospheres vol 117 no 12 Article IDD12108 2012

[4] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M J SEsteban-Parra and Y Castro-Dıez ldquoEvaluation of WRF meanand extreme precipitation over Spain present climate (1970ndash99)rdquo Journal of Climate vol 25 no 14 pp 4883ndash4895 2012

[5] R M Cardoso P M M Soares P M A Miranda and MBelo-Pereira ldquoWRF high resolution simulation of Iberianmeanand extreme precipitation climaterdquo International Journal ofClimatology vol 33 no 11 pp 2591ndash2608 2013

[6] L Calvetti and A J Pereira Filho ldquoEnsemble hydrometeoro-logical forecasts using WRF hourly QPF and topmodel for amiddle watershedrdquo Advances in Meteorology vol 2014 ArticleID 484120 12 pages 2014

[7] S Q Zhang M Zupanski A Y Hou X Lin and S H CheungldquoAssimilation of precipitation-affected radiances in a cloud-resolving wrf ensemble data assimilation systemrdquo MonthlyWeather Review vol 141 no 2 pp 754ndash772 2013

[8] J Liu M Bray and D Han ldquoExploring the effect of dataassimilation by WRF-3DVar for numerical rainfall predictionwith different types of storm eventsrdquoHydrological Processes vol27 no 25 pp 3627ndash3640 2013

[9] C-H Xiong L-F Zhang J-P Guan J Peng and B ZhangldquoAnalysis and numerical study of a hybrid BGM-3DVAR dataassimilation scheme using satellite radiance data for heavy rainforecastsrdquo Journal of Hydrodynamics vol 25 no 3 pp 430ndash4392013

[10] D M Barker W Huang Y R Guo and Q N Xiao ldquoA three-dimensional (3DVAR) data assimilation system for use withMM5 implementation and initial resultsrdquo Monthly WeatherReview vol 132 pp 897ndash914 2004

[11] Q Xiao and J Sun ldquoMultiple-radar data assimilation and short-range quantitative precipitation forecasting of a squall lineobserved during IHOP 2002rdquoMonthlyWeather Review vol 135no 10 pp 3381ndash3404 2007

[12] Z Sokol ldquoEffects of an assimilation of radar and satellite dataon a very-short range forecast of heavy convective rainfallsrdquoAtmospheric Research vol 93 no 1ndash3 pp 188ndash206 2009

[13] P Bauer P Lopez A Benedetti D Salmond and E MoreauldquoImplementation of 1D+4D-Var assimilation of precipitation-affected microwave radiances at ECMWF I 1D-Varrdquo QuarterlyJournal of the Royal Meteorological Society vol 132 no 620 pp2277ndash2306 2006

[14] P Bauer P Lopez D Salmond A Benedetti S Saarinen andM Bonazzola ldquoImplementation of 1D+4D-Var assimilation of

precipitation-affected microwave radiances at ECMWF II 4D-Varrdquo Quarterly Journal of the Royal Meteorological Society vol132 no 620 pp 2307ndash2332 2006

[15] P Bauer A J Geer P Lopez and D Salmond ldquoDirect 4D-Var assimilation of all-sky radiances Part I ImplementationrdquoQuarterly Journal of the Royal Meteorological Society vol 136no 652 pp 1868ndash1885 2010

[16] A J Geer P Bauer and P Lopez ldquoDirect 4D-Var assimilationof all-sky radiances Part II assessmentrdquo Quarterly Journal ofthe RoyalMeteorological Society vol 136 no 652 pp 1886ndash19052010

[17] Z Sokol and D Rezacova ldquoAssimilation of radar reflectivityinto the LM COSMOmodel with a high horizontal resolutionrdquoMeteorological Applications vol 13 no 4 pp 317ndash330 2006

[18] Z Sokol ldquoAssimilation of extrapolated radar reflectivity into aNWP model and its impact on a precipitation forecast at highresolutionrdquo Atmospheric Research vol 100 no 2-3 pp 201ndash2122011

[19] HWang J Sun S Fan and X-Y Huang ldquoIndirect assimilationof radar reflectivity with WRF 3D-var and its impact onprediction of four summertime convective eventsrdquo Journal ofApplied Meteorology and Climatology vol 52 no 4 pp 889ndash902 2013

[20] B Macpherson ldquoOperational experience with assimilation ofrainfall data in the Met Office mesoscale modelrdquo Meteorologyand Atmospheric Physics vol 76 no 1-2 pp 3ndash8 2001

[21] K Stephan S Klink and C Schraff ldquoAssimilation of radar-derived rain rates into the convective-scale model COSMO-DEat DWDrdquo Quarterly Journal of the Royal Meteorological Societyvol 134 no 634 pp 1315ndash1326 2008

[22] K Alapaty D S Niyogi F Chen P Pyle A Chandrasekar andN Seaman ldquoDevelopment of the flux-adjusting surface dataassimilation system for mesoscale modelsrdquo Journal of AppliedMeteorology and Climatology vol 47 no 9 pp 2331ndash2350 2008

[23] F H Ruggiero G D Modica and A E Lipton ldquoAssimilationof satellite imager data and surface observations to improveanalysis of circulations forced by cloud shading contrastsrdquoMonthly Weather Review vol 128 no 2 pp 434ndash448 2000

[24] J P Hacker and D Rostkier-Edelstein ldquoPBL state estimationwith surface observations a column model and an ensemblefilterrdquo Monthly Weather Review vol 135 no 8 pp 2958ndash29722007

[25] D J Stensrud N Yussouf D C Dowell and M C ConiglioldquoAssimilating surface data into a mesoscale model ensemblecold pool analyses from spring 2007rdquoAtmospheric Research vol93 no 1ndash3 pp 207ndash220 2009

[26] Q Wan and J Xu ldquoA numerical study of the rainstormcharacteristics of the June 2005 flash flood with WRFGSIdata assimilation system over south-east Chinardquo HydrologicalProcesses vol 25 no 8 pp 1327ndash1341 2011

[27] J Liu M Bray and D Han ldquoA study on WRF radar dataassimilation for hydrological rainfall predictionrdquoHydrology andEarth System Sciences vol 17 no 8 pp 3095ndash3110 2013

[28] K Ide P Courtier M Ghil and A C Lorenc ldquoUnified notationfor data assimilation operational sequential and variationalrdquoJournal of the Meteorological Society of Japan vol 75 no 1 pp181ndash189 1997

[29] A C Lorenc ldquoAnalysis methods for numerical weather predic-tionrdquo Quarterly Journal of the Royal Meteorological Society vol112 no 474 pp 1177ndash1194 1966

14 Advances in Meteorology

[30] D F Parrish and J C Derber ldquoThe nationalmeteorological cen-terrsquos spectral statistical interpolation analysis systemrdquo MonthlyWeather Review vol 120 no 8 pp 1747ndash1763 1992

[31] T B Zhao and C B Fu ldquoPreliminary comparison and analysisbetween ERA-40 NCEP-2 reanalysis and observations overChinardquo Climatic and Environmental Research vol 11 no 1 pp14ndash32 2006

[32] L Ma T Zhang Q Li O W Frauenfeld and D QinldquoEvaluation of ERA-40 NCEP-1 and NCEP-2 reanalysis airtemperatures with ground-based measurements in ChinardquoJournal of Geophysical Research Atmospheres (1984ndash2012) vol13 no D15 1984

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ClimatologyJournal of

EcologyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

EarthquakesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

Mining

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

International Journal of

Geophysics

OceanographyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofPetroleum Engineering

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Atmospheric SciencesInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MineralogyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MeteorologyAdvances in

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geological ResearchJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geology Advances in

12 Advances in Meteorology

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(a)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

(b)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(c)

41∘N

40∘N

39∘N

38∘N

37∘N

36∘N

35∘N

94∘E 96∘E 98∘E 100∘E 102∘E 104∘E

1 2 3 4 5 6 7 8 9 10 11 12 13 14

(kgmmiddots)

(d)

Figure 12 Moisture flux initial field for the CTRL and DA experiments (a) CTRL (b) DA-SAT (c) DA-OBS (d) DA-BOTH

were used to evaluate the impact of the DA experiments inthis study They are defined as

119877

=

(1119873)sum119873

119894=1(119865119894 minus 119865) (119874119894 minus 119874)

radic(1119873)sum119873

119894=1(119865119894 minus 119865)

2radic(1119873)sum

119873

119894=1(119874119894 minus 119874)

2

RMSE = radic 1119873

119873

sum

119894=1

(119865119894 minus 119874119894)2

ME = 1119873

119873

sum

119894=1

(119874119894 minus 119865119894)

AE = 1119873

119873

sum

119894=1

1003816100381610038161003816119865119894 minus 1198741198941003816100381610038161003816

(A1)

Here 119865 is simulated value 119874 is observed value 119873 is thenumber of sites

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work was supported by the National natural ScienceFoundation of China (41190080 40971290) Strategic LeadingTechnology Special Fund of Chinese Academy of Sciences(XDB03030204) the National Scientific and TechnologicalSupport Project (2013BAB05B03) and the Opening Fund ofKey Laboratory of Land Surface Process and Climate Changein Cold and Arid Regions CAS (LPCC201205)

Advances in Meteorology 13

References

[1] P Chambon S Q Zhang A Y Hou M Zupanski and SCheung ldquoAssessing the impact of pre-GPMmicrowave precipi-tation observations in the GoddardWRF ensemble data assim-ilation systemrdquo Quarterly Journal of the Royal MeteorologicalSociety vol 140 no 681 pp 1219ndash1235 2013

[2] J-H Seo Y H Lee and Y-H Kim ldquoFeature selection forvery short-term heavy rainfall prediction using evolutionarycomputationrdquo Advances in Meteorology vol 2014 Article ID203545 15 pages 2014

[3] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M JEsteban-Parra and Y Castro-Dıez ldquoHigh-resolution projec-tions of mean and extreme precipitation over Spain usingthe WRF model (2070ndash2099 versus 1970ndash1999)rdquo Journal ofGeophysical Research D Atmospheres vol 117 no 12 Article IDD12108 2012

[4] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M J SEsteban-Parra and Y Castro-Dıez ldquoEvaluation of WRF meanand extreme precipitation over Spain present climate (1970ndash99)rdquo Journal of Climate vol 25 no 14 pp 4883ndash4895 2012

[5] R M Cardoso P M M Soares P M A Miranda and MBelo-Pereira ldquoWRF high resolution simulation of Iberianmeanand extreme precipitation climaterdquo International Journal ofClimatology vol 33 no 11 pp 2591ndash2608 2013

[6] L Calvetti and A J Pereira Filho ldquoEnsemble hydrometeoro-logical forecasts using WRF hourly QPF and topmodel for amiddle watershedrdquo Advances in Meteorology vol 2014 ArticleID 484120 12 pages 2014

[7] S Q Zhang M Zupanski A Y Hou X Lin and S H CheungldquoAssimilation of precipitation-affected radiances in a cloud-resolving wrf ensemble data assimilation systemrdquo MonthlyWeather Review vol 141 no 2 pp 754ndash772 2013

[8] J Liu M Bray and D Han ldquoExploring the effect of dataassimilation by WRF-3DVar for numerical rainfall predictionwith different types of storm eventsrdquoHydrological Processes vol27 no 25 pp 3627ndash3640 2013

[9] C-H Xiong L-F Zhang J-P Guan J Peng and B ZhangldquoAnalysis and numerical study of a hybrid BGM-3DVAR dataassimilation scheme using satellite radiance data for heavy rainforecastsrdquo Journal of Hydrodynamics vol 25 no 3 pp 430ndash4392013

[10] D M Barker W Huang Y R Guo and Q N Xiao ldquoA three-dimensional (3DVAR) data assimilation system for use withMM5 implementation and initial resultsrdquo Monthly WeatherReview vol 132 pp 897ndash914 2004

[11] Q Xiao and J Sun ldquoMultiple-radar data assimilation and short-range quantitative precipitation forecasting of a squall lineobserved during IHOP 2002rdquoMonthlyWeather Review vol 135no 10 pp 3381ndash3404 2007

[12] Z Sokol ldquoEffects of an assimilation of radar and satellite dataon a very-short range forecast of heavy convective rainfallsrdquoAtmospheric Research vol 93 no 1ndash3 pp 188ndash206 2009

[13] P Bauer P Lopez A Benedetti D Salmond and E MoreauldquoImplementation of 1D+4D-Var assimilation of precipitation-affected microwave radiances at ECMWF I 1D-Varrdquo QuarterlyJournal of the Royal Meteorological Society vol 132 no 620 pp2277ndash2306 2006

[14] P Bauer P Lopez D Salmond A Benedetti S Saarinen andM Bonazzola ldquoImplementation of 1D+4D-Var assimilation of

precipitation-affected microwave radiances at ECMWF II 4D-Varrdquo Quarterly Journal of the Royal Meteorological Society vol132 no 620 pp 2307ndash2332 2006

[15] P Bauer A J Geer P Lopez and D Salmond ldquoDirect 4D-Var assimilation of all-sky radiances Part I ImplementationrdquoQuarterly Journal of the Royal Meteorological Society vol 136no 652 pp 1868ndash1885 2010

[16] A J Geer P Bauer and P Lopez ldquoDirect 4D-Var assimilationof all-sky radiances Part II assessmentrdquo Quarterly Journal ofthe RoyalMeteorological Society vol 136 no 652 pp 1886ndash19052010

[17] Z Sokol and D Rezacova ldquoAssimilation of radar reflectivityinto the LM COSMOmodel with a high horizontal resolutionrdquoMeteorological Applications vol 13 no 4 pp 317ndash330 2006

[18] Z Sokol ldquoAssimilation of extrapolated radar reflectivity into aNWP model and its impact on a precipitation forecast at highresolutionrdquo Atmospheric Research vol 100 no 2-3 pp 201ndash2122011

[19] HWang J Sun S Fan and X-Y Huang ldquoIndirect assimilationof radar reflectivity with WRF 3D-var and its impact onprediction of four summertime convective eventsrdquo Journal ofApplied Meteorology and Climatology vol 52 no 4 pp 889ndash902 2013

[20] B Macpherson ldquoOperational experience with assimilation ofrainfall data in the Met Office mesoscale modelrdquo Meteorologyand Atmospheric Physics vol 76 no 1-2 pp 3ndash8 2001

[21] K Stephan S Klink and C Schraff ldquoAssimilation of radar-derived rain rates into the convective-scale model COSMO-DEat DWDrdquo Quarterly Journal of the Royal Meteorological Societyvol 134 no 634 pp 1315ndash1326 2008

[22] K Alapaty D S Niyogi F Chen P Pyle A Chandrasekar andN Seaman ldquoDevelopment of the flux-adjusting surface dataassimilation system for mesoscale modelsrdquo Journal of AppliedMeteorology and Climatology vol 47 no 9 pp 2331ndash2350 2008

[23] F H Ruggiero G D Modica and A E Lipton ldquoAssimilationof satellite imager data and surface observations to improveanalysis of circulations forced by cloud shading contrastsrdquoMonthly Weather Review vol 128 no 2 pp 434ndash448 2000

[24] J P Hacker and D Rostkier-Edelstein ldquoPBL state estimationwith surface observations a column model and an ensemblefilterrdquo Monthly Weather Review vol 135 no 8 pp 2958ndash29722007

[25] D J Stensrud N Yussouf D C Dowell and M C ConiglioldquoAssimilating surface data into a mesoscale model ensemblecold pool analyses from spring 2007rdquoAtmospheric Research vol93 no 1ndash3 pp 207ndash220 2009

[26] Q Wan and J Xu ldquoA numerical study of the rainstormcharacteristics of the June 2005 flash flood with WRFGSIdata assimilation system over south-east Chinardquo HydrologicalProcesses vol 25 no 8 pp 1327ndash1341 2011

[27] J Liu M Bray and D Han ldquoA study on WRF radar dataassimilation for hydrological rainfall predictionrdquoHydrology andEarth System Sciences vol 17 no 8 pp 3095ndash3110 2013

[28] K Ide P Courtier M Ghil and A C Lorenc ldquoUnified notationfor data assimilation operational sequential and variationalrdquoJournal of the Meteorological Society of Japan vol 75 no 1 pp181ndash189 1997

[29] A C Lorenc ldquoAnalysis methods for numerical weather predic-tionrdquo Quarterly Journal of the Royal Meteorological Society vol112 no 474 pp 1177ndash1194 1966

14 Advances in Meteorology

[30] D F Parrish and J C Derber ldquoThe nationalmeteorological cen-terrsquos spectral statistical interpolation analysis systemrdquo MonthlyWeather Review vol 120 no 8 pp 1747ndash1763 1992

[31] T B Zhao and C B Fu ldquoPreliminary comparison and analysisbetween ERA-40 NCEP-2 reanalysis and observations overChinardquo Climatic and Environmental Research vol 11 no 1 pp14ndash32 2006

[32] L Ma T Zhang Q Li O W Frauenfeld and D QinldquoEvaluation of ERA-40 NCEP-1 and NCEP-2 reanalysis airtemperatures with ground-based measurements in ChinardquoJournal of Geophysical Research Atmospheres (1984ndash2012) vol13 no D15 1984

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ClimatologyJournal of

EcologyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

EarthquakesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

Mining

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

International Journal of

Geophysics

OceanographyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofPetroleum Engineering

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Atmospheric SciencesInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MineralogyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MeteorologyAdvances in

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geological ResearchJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geology Advances in

Advances in Meteorology 13

References

[1] P Chambon S Q Zhang A Y Hou M Zupanski and SCheung ldquoAssessing the impact of pre-GPMmicrowave precipi-tation observations in the GoddardWRF ensemble data assim-ilation systemrdquo Quarterly Journal of the Royal MeteorologicalSociety vol 140 no 681 pp 1219ndash1235 2013

[2] J-H Seo Y H Lee and Y-H Kim ldquoFeature selection forvery short-term heavy rainfall prediction using evolutionarycomputationrdquo Advances in Meteorology vol 2014 Article ID203545 15 pages 2014

[3] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M JEsteban-Parra and Y Castro-Dıez ldquoHigh-resolution projec-tions of mean and extreme precipitation over Spain usingthe WRF model (2070ndash2099 versus 1970ndash1999)rdquo Journal ofGeophysical Research D Atmospheres vol 117 no 12 Article IDD12108 2012

[4] D Argueso J M Hidalgo-Munoz S R Gamiz-Fortis M J SEsteban-Parra and Y Castro-Dıez ldquoEvaluation of WRF meanand extreme precipitation over Spain present climate (1970ndash99)rdquo Journal of Climate vol 25 no 14 pp 4883ndash4895 2012

[5] R M Cardoso P M M Soares P M A Miranda and MBelo-Pereira ldquoWRF high resolution simulation of Iberianmeanand extreme precipitation climaterdquo International Journal ofClimatology vol 33 no 11 pp 2591ndash2608 2013

[6] L Calvetti and A J Pereira Filho ldquoEnsemble hydrometeoro-logical forecasts using WRF hourly QPF and topmodel for amiddle watershedrdquo Advances in Meteorology vol 2014 ArticleID 484120 12 pages 2014

[7] S Q Zhang M Zupanski A Y Hou X Lin and S H CheungldquoAssimilation of precipitation-affected radiances in a cloud-resolving wrf ensemble data assimilation systemrdquo MonthlyWeather Review vol 141 no 2 pp 754ndash772 2013

[8] J Liu M Bray and D Han ldquoExploring the effect of dataassimilation by WRF-3DVar for numerical rainfall predictionwith different types of storm eventsrdquoHydrological Processes vol27 no 25 pp 3627ndash3640 2013

[9] C-H Xiong L-F Zhang J-P Guan J Peng and B ZhangldquoAnalysis and numerical study of a hybrid BGM-3DVAR dataassimilation scheme using satellite radiance data for heavy rainforecastsrdquo Journal of Hydrodynamics vol 25 no 3 pp 430ndash4392013

[10] D M Barker W Huang Y R Guo and Q N Xiao ldquoA three-dimensional (3DVAR) data assimilation system for use withMM5 implementation and initial resultsrdquo Monthly WeatherReview vol 132 pp 897ndash914 2004

[11] Q Xiao and J Sun ldquoMultiple-radar data assimilation and short-range quantitative precipitation forecasting of a squall lineobserved during IHOP 2002rdquoMonthlyWeather Review vol 135no 10 pp 3381ndash3404 2007

[12] Z Sokol ldquoEffects of an assimilation of radar and satellite dataon a very-short range forecast of heavy convective rainfallsrdquoAtmospheric Research vol 93 no 1ndash3 pp 188ndash206 2009

[13] P Bauer P Lopez A Benedetti D Salmond and E MoreauldquoImplementation of 1D+4D-Var assimilation of precipitation-affected microwave radiances at ECMWF I 1D-Varrdquo QuarterlyJournal of the Royal Meteorological Society vol 132 no 620 pp2277ndash2306 2006

[14] P Bauer P Lopez D Salmond A Benedetti S Saarinen andM Bonazzola ldquoImplementation of 1D+4D-Var assimilation of

precipitation-affected microwave radiances at ECMWF II 4D-Varrdquo Quarterly Journal of the Royal Meteorological Society vol132 no 620 pp 2307ndash2332 2006

[15] P Bauer A J Geer P Lopez and D Salmond ldquoDirect 4D-Var assimilation of all-sky radiances Part I ImplementationrdquoQuarterly Journal of the Royal Meteorological Society vol 136no 652 pp 1868ndash1885 2010

[16] A J Geer P Bauer and P Lopez ldquoDirect 4D-Var assimilationof all-sky radiances Part II assessmentrdquo Quarterly Journal ofthe RoyalMeteorological Society vol 136 no 652 pp 1886ndash19052010

[17] Z Sokol and D Rezacova ldquoAssimilation of radar reflectivityinto the LM COSMOmodel with a high horizontal resolutionrdquoMeteorological Applications vol 13 no 4 pp 317ndash330 2006

[18] Z Sokol ldquoAssimilation of extrapolated radar reflectivity into aNWP model and its impact on a precipitation forecast at highresolutionrdquo Atmospheric Research vol 100 no 2-3 pp 201ndash2122011

[19] HWang J Sun S Fan and X-Y Huang ldquoIndirect assimilationof radar reflectivity with WRF 3D-var and its impact onprediction of four summertime convective eventsrdquo Journal ofApplied Meteorology and Climatology vol 52 no 4 pp 889ndash902 2013

[20] B Macpherson ldquoOperational experience with assimilation ofrainfall data in the Met Office mesoscale modelrdquo Meteorologyand Atmospheric Physics vol 76 no 1-2 pp 3ndash8 2001

[21] K Stephan S Klink and C Schraff ldquoAssimilation of radar-derived rain rates into the convective-scale model COSMO-DEat DWDrdquo Quarterly Journal of the Royal Meteorological Societyvol 134 no 634 pp 1315ndash1326 2008

[22] K Alapaty D S Niyogi F Chen P Pyle A Chandrasekar andN Seaman ldquoDevelopment of the flux-adjusting surface dataassimilation system for mesoscale modelsrdquo Journal of AppliedMeteorology and Climatology vol 47 no 9 pp 2331ndash2350 2008

[23] F H Ruggiero G D Modica and A E Lipton ldquoAssimilationof satellite imager data and surface observations to improveanalysis of circulations forced by cloud shading contrastsrdquoMonthly Weather Review vol 128 no 2 pp 434ndash448 2000

[24] J P Hacker and D Rostkier-Edelstein ldquoPBL state estimationwith surface observations a column model and an ensemblefilterrdquo Monthly Weather Review vol 135 no 8 pp 2958ndash29722007

[25] D J Stensrud N Yussouf D C Dowell and M C ConiglioldquoAssimilating surface data into a mesoscale model ensemblecold pool analyses from spring 2007rdquoAtmospheric Research vol93 no 1ndash3 pp 207ndash220 2009

[26] Q Wan and J Xu ldquoA numerical study of the rainstormcharacteristics of the June 2005 flash flood with WRFGSIdata assimilation system over south-east Chinardquo HydrologicalProcesses vol 25 no 8 pp 1327ndash1341 2011

[27] J Liu M Bray and D Han ldquoA study on WRF radar dataassimilation for hydrological rainfall predictionrdquoHydrology andEarth System Sciences vol 17 no 8 pp 3095ndash3110 2013

[28] K Ide P Courtier M Ghil and A C Lorenc ldquoUnified notationfor data assimilation operational sequential and variationalrdquoJournal of the Meteorological Society of Japan vol 75 no 1 pp181ndash189 1997

[29] A C Lorenc ldquoAnalysis methods for numerical weather predic-tionrdquo Quarterly Journal of the Royal Meteorological Society vol112 no 474 pp 1177ndash1194 1966

14 Advances in Meteorology

[30] D F Parrish and J C Derber ldquoThe nationalmeteorological cen-terrsquos spectral statistical interpolation analysis systemrdquo MonthlyWeather Review vol 120 no 8 pp 1747ndash1763 1992

[31] T B Zhao and C B Fu ldquoPreliminary comparison and analysisbetween ERA-40 NCEP-2 reanalysis and observations overChinardquo Climatic and Environmental Research vol 11 no 1 pp14ndash32 2006

[32] L Ma T Zhang Q Li O W Frauenfeld and D QinldquoEvaluation of ERA-40 NCEP-1 and NCEP-2 reanalysis airtemperatures with ground-based measurements in ChinardquoJournal of Geophysical Research Atmospheres (1984ndash2012) vol13 no D15 1984

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ClimatologyJournal of

EcologyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

EarthquakesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

Mining

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

International Journal of

Geophysics

OceanographyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofPetroleum Engineering

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Atmospheric SciencesInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MineralogyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MeteorologyAdvances in

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geological ResearchJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geology Advances in

14 Advances in Meteorology

[30] D F Parrish and J C Derber ldquoThe nationalmeteorological cen-terrsquos spectral statistical interpolation analysis systemrdquo MonthlyWeather Review vol 120 no 8 pp 1747ndash1763 1992

[31] T B Zhao and C B Fu ldquoPreliminary comparison and analysisbetween ERA-40 NCEP-2 reanalysis and observations overChinardquo Climatic and Environmental Research vol 11 no 1 pp14ndash32 2006

[32] L Ma T Zhang Q Li O W Frauenfeld and D QinldquoEvaluation of ERA-40 NCEP-1 and NCEP-2 reanalysis airtemperatures with ground-based measurements in ChinardquoJournal of Geophysical Research Atmospheres (1984ndash2012) vol13 no D15 1984

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ClimatologyJournal of

EcologyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

EarthquakesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

Mining

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

International Journal of

Geophysics

OceanographyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofPetroleum Engineering

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Atmospheric SciencesInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MineralogyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MeteorologyAdvances in

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geological ResearchJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geology Advances in

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ClimatologyJournal of

EcologyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

EarthquakesJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom

Applied ampEnvironmentalSoil Science

Volume 2014

Mining

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

International Journal of

Geophysics

OceanographyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of Computational Environmental SciencesHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal ofPetroleum Engineering

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

GeochemistryHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Atmospheric SciencesInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OceanographyHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MineralogyInternational Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MeteorologyAdvances in

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Paleontology JournalHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ScientificaHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geological ResearchJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Geology Advances in