14
IRI/ARCS Regional Applications Project J. Roads, S. Chen ECPC D. Lettenmaier, E. Salathe, E. Miles Univ. Washington H. Juang, J. Han, NCEP S. Cocke, T. Larow, FSU J. -H. Qian, S. Zebiak, Andrew Robertson, IRI Work plan: 7/1/02-6/30/05 1. Background Previously, the ARCS/IRI Regional modeling consortium had developed a cooperative research plan concentrated upon a comparison of regional simulations forced by global analyses over Brazil (Roads et al. 2002). While this project set the stage for developing corresponding regional forecasts, there are still a number of unanswered questions about regional downscaling including: (1) Is the spatial character of the interannual variability improved [at the regional scale]? (2) Are the details provided by the higher resolution predictable, or are they just 'noise' on top of the regional signal? (3) Is probabilistic information changed/improved (i.e. tighter and/or more reliable PDF) regionally and/or locally? (4) Is temporal character of variability improved? (i.e. precipitation frequency, frequency of extreme events and wet/dry spells). The original goal was to continue the comparison for Brazil and to begin to use the global models used for seasonal prediction for the boundary conditions instead of reanalysis. However, in discussions with OGP and IRI, it became apparent that the proposed regional comparison project over Brazil did not have as high a priority as one that would develop additional applications for specific regions. Our regional consortium took these OGP suggestions quite seriously and decided to drop its effort to focus on a specific region (Brazil) and to instead try to help develop regional applications for other regions, while at the same time attempting to answer some of the original downscaling questions. Leaders have been identified for each application project who agree to not only evaluate their own forecasts but also forecasts from other participants. Regional issues that this consortium will address including: (1) seasonal forecast and global change hydrologic predictions over the US West as well as the US Southwest; (2) crop forecasting over the southeast; (3) fire danger forecasts over the US; (4) regional applications over Brazil. It should be noted that not every participant intends to contribute equally to every application project. It should also be noted that not every application project currently being undertaken by the ARCS is included in this effort. Nonetheless, the proposed application projects provide a beginning regional focus and additional coordination and application efforts may become entrained later, depending upon future funding and interest. It should be noted that not all applications intend to use regional models and some will utilize instead only the global forecasts along with empirical or statistical downscaling appropriate for a particular application model. In that regard, it should be noted that this statistical downscaling is likely to be successful so long as the application input is restricted to standard variables like temperature and precipitation. However, some of the applications require additional variables, such as surface wind and humidity, whose long-term observations are unavailable. Therefore it may ultimately prove advantageous to use the comprehensive output from the regional models.

IRI/ARCS Regional Applications Project J. Roads, S. Chen

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: IRI/ARCS Regional Applications Project J. Roads, S. Chen

IRI/ARCS Regional Applications ProjectJ. Roads, S. Chen ECPCD. Lettenmaier, E. Salathe, E. Miles Univ. WashingtonH. Juang, J. Han, NCEPS. Cocke, T. Larow, FSUJ. -H. Qian, S. Zebiak, Andrew Robertson, IRI

Work plan: 7/1/02-6/30/05

1. BackgroundPreviously, the ARCS/IRI Regional modeling consortium had developed a cooperative researchplan concentrated upon a comparison of regional simulations forced by global analyses overBrazil (Roads et al. 2002). While this project set the stage for developing corresponding regionalforecasts, there are still a number of unanswered questions about regional downscalingincluding: (1) Is the spatial character of the interannual variability improved [at the regionalscale]? (2) Are the details provided by the higher resolution predictable, or are they just 'noise' ontop of the regional signal? (3) Is probabilistic information changed/improved (i.e. tighter and/ormore reliable PDF) regionally and/or locally? (4) Is temporal character of variability improved?(i.e. precipitation frequency, frequency of extreme events and wet/dry spells).

The original goal was to continue the comparison for Brazil and to begin to use the globalmodels used for seasonal prediction for the boundary conditions instead of reanalysis. However,in discussions with OGP and IRI, it became apparent that the proposed regional comparisonproject over Brazil did not have as high a priority as one that would develop additionalapplications for specific regions. Our regional consortium took these OGP suggestions quiteseriously and decided to drop its effort to focus on a specific region (Brazil) and to instead try tohelp develop regional applications for other regions, while at the same time attempting to answersome of the original downscaling questions. Leaders have been identified for each applicationproject who agree to not only evaluate their own forecasts but also forecasts from otherparticipants.

Regional issues that this consortium will address including: (1) seasonal forecast and globalchange hydrologic predictions over the US West as well as the US Southwest; (2) cropforecasting over the southeast; (3) fire danger forecasts over the US; (4) regional applicationsover Brazil. It should be noted that not every participant intends to contribute equally to everyapplication project. It should also be noted that not every application project currently beingundertaken by the ARCS is included in this effort. Nonetheless, the proposed application projectsprovide a beginning regional focus and additional coordination and application efforts maybecome entrained later, depending upon future funding and interest.

It should be noted that not all applications intend to use regional models and some will utilizeinstead only the global forecasts along with empirical or statistical downscaling appropriate for aparticular application model. In that regard, it should be noted that this statistical downscaling islikely to be successful so long as the application input is restricted to standard variables liketemperature and precipitation. However, some of the applications require additional variables,such as surface wind and humidity, whose long-term observations are unavailable. Therefore itmay ultimately prove advantageous to use the comprehensive output from the regional models.

Page 2: IRI/ARCS Regional Applications Project J. Roads, S. Chen

2

However, one might also simply use a mixture of global model output and judiciously downscaleonly precipitation or temperature. Again the focus is on developing the regional application asopposed to developing a rigorous downscaling effort.

In the following sections, we describe the participating models as well as the various identifiedapplications.

2. Models 2.1 Global Models NCEP’s operational global seasonal forecast model will be one of the models examined.Operational 7-month forecasts from a 20 member ensemble along with a rolling climatologyfrom 21 (the corresponding month but for each of the previous 21 years) 10 member ensembles.The challenge with this model will be to provide appropriate data for various regionalapplications as well as to drive various regional models. The IRI has also archived ECHAM4.5 (T42, L18) data for 50+ years of simulation for a 24member ensemble and 30+ years of retrospective forecasts using persisted SSTA for a 12-member ensemble.. ECPC’s global 16 week forecasts, made every weekend beginning 00 UTC, Sept. 27, 1997, alsoprovides a large ensemble for driving similar regional forecasts. In fact, regional forecasts arebeing made for several regions in order to develop useful regional applications as well as toassess regional forecast skill. The ECPC global model is currently an earlier version of the NCEPglobal model (the NCEP/NCAR reanalysis model), but there are plans to upgrade the model to aversion corresponding to the current operational seasonal forecast model. Given the bandwidthand storage limitations, we believe it is important to run the global model locally as well as toaccess the current operational archives. The FSU global model is a spectral primitive equation model typically run T63 (1.875 deg.) orhigher resolution and 27 vertical sigma levels. The global model is coupled to several oceanmodels, including the HYCOM model and a version of the Max Planck HOPE model. The globalmodel shares the same physical parameterizations as the FSU Nested Regional Spectral Model, Another global model to be used for some of the regional downscaling experiments is the NASASeasonal to Interannual Prediction Project (NSIPP, See Suarez et al.) model. Limited simulationsare available from this global model to drive various regional models and Roads et al. (2002)used this global model for comparison to regional models. 2.2 Regional ModelsECPC's regional spectral model (RSM) was originally developed at NCEP (Juang et al. 1997)and has been used for many our regional modeling efforts (e.g. Chen et al., 1999; Roads andChen, 2000). The RSM has the same vertical structure (sigma coordinates) and physicalparameterizations as NCEP's global spectral model (GSM) used for the NCEP/NCAR andNCEP-DOE reanalysis (Kalnay et al. 1996; Kanamitsu et al. 2002). Base fields may be eitherglobal or regional model output or analysis fields. Perturbation fields are represented by doubleFourier trigonometric functions, which are relaxed to zero at lateral boundaries.

Page 3: IRI/ARCS Regional Applications Project J. Roads, S. Chen

3

A new CVS version of the RSM has been developed to closely emulate the new SeasonalForecast Model at NCEP and to also be used for global change experiments (Han and Roads2001). The new CVS version will provide a possible methodology for the community to shareand test new GSM parameterizations being developed at NCEP in various regions. New physicsare also being developed for the global and regional models and will include a prognostic cloudscheme (which will replace the large scale precipitation) with cloud interation in radiationpackage (the cloud tuning for radiation in the previous versions is gone). Angular momentun andmoisture mixings are being implemented to avoid the false production of hurricanes. The surfacevegitation is being changed to a hetearogeneous type with 13 kinds. The gravity wave drag willhave multi-direction effects depending on wind direction.

The FSU RSM is similar to the NCEP RSM. The FSUNRSM also includes a number of user-selectable parameterizations, including 3 radiation schemes and 6 deep convection schemes. Forthe previous intercomparison run, the FSUNRSM was run with 27 vertical sigma levels, CCM3.6 radiation and the Zhang-McFarlane deep convection scheme. While the BATS and SSiB landsurface schemes are currently being implemented, a simplified land surface scheme which had24 land use (vegetation) categories (based on USGS data) and 3 soil temperature layers was usedhere. Soil moisture and albedo are based on vegetation type and season.

The IRI is using a version of the NCEP RSM for some regional climate modeling studies, but forthis comparison, the IRI will be using RegCM2, developed by Giorgi, et al. (1990). RegCM2is agridpoint model and for this comparison had 14-layers, CCM3 radiation, MM5-Grell cumulusparameterization, Holtstag PBL, and BATS land-surface package. Various regional models(RegCM2, RSM, etc) driven by the hindcast/forecast (1997-present) will be tested over themodel domain. 2.3 Application Models A number of groups will be using the variable infiltration capacity (VIC) hydrologic model formaking streamflow and surface hydrology forecasts. VIC is described in detail by Liang et al.(1994) and at http://hydro.washington.edu. As shown in various land surface comparisons (E.g.Pitman et al. 1999), VIC not only provides a useful macroscale hydrologic budget it alsocompares well with other models and observations in small scale regions. VIC balances bothenergy and water over a grid mesh, in this application at a 1/8-degree resolution, using a 3-hourlytime step. At the 1/8-degree resolution, the model represents about 23,000 computational gridcells within the Mississippi River Basin. The VIC model computes the vertical energy andmoisture fluxes in a grid cell based on a specification at each grid cell of soil properties andvegetation coverage. VIC includes the representation of subgrid variability in soil infiltrationcapacity, and a mosaic of vegetation classes in any grid cell. Drainage between the soil layers(three were used in this application) is entirely gravity-driven, and the unsaturated hydraulicconductivity is a function of the degree of saturation of the soil, with base flow produced fromthe lowest soil layer using the nonlinear ARNO formulation (Todini, 1996). To account forsubgrid variability in infiltration, the VIC model uses a variable infiltration capacity schemebased on Zhao et al. (1980). This scheme uses a spatial probability distribution to characterizeavailable infiltration capacity as a function of the relative saturated area of the grid cell.Precipitation in excess of the available infiltration capacity forms surface runoff.

Page 4: IRI/ARCS Regional Applications Project J. Roads, S. Chen

4

One of the crop models that will be used to simulate maize yield is the CERES-Maize simulationmodel (Ritchie et al., 1998). This model is a dynamic process based crop model that simulatesplant response to soil, weather, water stress and management practices. The model calculatesdevelopment, growth and partitioning processes on a daily basis, beginning with planting andending at harvest maturity. The model can be used to assess the impacts of weather (Mavromatisand Jones, 1998; Mearns et al., 1999) and management decisions (Jones et al, 2000; Hansen etal., 2001). Preliminary results for selected locations in Florida are encouraging and show greaterskill of the FSU regional model to predict Maize yields when compared to the FSU global model(Jagtap et al 2002). The National Fire Danger Rating System (NFDRS, Deeming et al. 1977), which has served theUS fire management community for over 20 years, and is used to monitor daily fluctuations offire danger across broad geographic areas, prior to fire occurrence will be used to forecastpotential fire danger. Fire managers currently only determine daily readiness levels based on thisinformation but have expressed an interest in looking at long-range forecasts of suchinformation. Development of high-resolution information may also be useful to the FARSITEsystem, which predicts rate of spread based upon local winds. Both systems integrate the effectsof fuels, topography, and weather information to fire spread and intensity. At their core is theRothermel fire spread model, which simulate the quasi-steady state forward rate of spread of theflaming front at the head of the fire. The NFDRS, as well as the Hawaii Division of Foresty andWildlife, also uses the Keetch-Byram Drought Index (KBDI) to evaluate drought effects on firepotential. The KBDI simulates the counterbalancing effects of precipitation andevapotranspiration on moisture content in the first 20 cm of the soil. The KBDI requires high-resolution precipitation data. In addition, soil moisture from the regional models can be utilizedfor comparison to the KBDI.

4. Experimental Applications4.1 Experimental West-wide ensemble hydrologic prediction systemWood et al. (2002) demonstrated a hydrologic forecasting method that uses monthly to seasonalensemble climate forecasts as input to a macroscale hydrology model that in turn producesensembles of runoff, soil moisture, snow pack and streamflow. The method has been appliedover the U.S. East Coast during the summer 2000drought, and to the Columbia River basin during the2000-2001 Pacific Northwest drought (seewww.ce.washington.edu/pub/HYDRO/aww/west_fcast/west_fcast.htm). Weare working to expand the prior experimentalapplications to the entire western U.S. (domain at right),on an operational basis.

The current forecast approach uses monthly NCEP/CMBGlobal Spectral Model (GSM) climate forecasts. Theforecasts comprise seven-month lead monthly surfaceprecipitation and temperature ensembles (20 member) atT62 spatial resolution. CMB also produces a rollingclimatology: 21 10-member GSM ensembles run for thesame monthly forecast period, but using observed sea

Page 5: IRI/ARCS Regional Applications Project J. Roads, S. Chen

5

surface temperatures for each of the years 1979-99. Using the climatology as statistical contextfor the forecasts, we bias-correct and downscale the ensembles to 1/4 or 1/8 degree horizontalresolution. Then, disaggregated to a daily time step, the ensembles drive the VIC hydrologymodel (Liang et al., 1994), producing streamflow forecasts, which serve as inflows to reservoirmanagement models.

The goals of the forecast expansion and development project are as follows:• Make operational the hydrologic forecasting approach over the western U.S. domain shown

above.• Move from aperiodic experimental forecasts to quasi-operational forecast products.• Calibrate streamflow forecast points throughout the domain, and identify potentially

associated uses and users.• Develop climatological datasets at different scales pertaining to the bias-correction and

downscaling steps, and for forecast verification.• Use the CMB retrospective ensemble climatology to assess streamflow forecast accuracy, as

compared with “naturalized” stream flows at control points within the domain.• For routine implementation over the large domain, automate various processing steps through

enhancements to existing software.• Identify and evaluate potential real time data sources for use in hydrology model

initialization (spin-up), a critical factor for forecast accuracy.• Standardize ongoing retrospective efforts to verify recent forecasts and diagnose prediction

accuracy and identify sources of error.• Enhance our existing web site for disseminating forecasts and forecast retrospective

evaluation results, and will forge links to interested operating agencies• Investigate approaches to making the forecast results available to these communities other

than water management.

Forecast variables: precipitationLength of forecast: 7 months or longerTime period for forecasts: 1979-presentFrequency of forecasts: once a month (20 member ensemble)Domain:Resolution (200 kms)

4.2 Southwest Climate PredictionsECPC is developing a corresponding experimental US Southwest hydrologic prediction system,using the Regional Spectral Model (RSM) and the Variable Infiltration Capacity (VIC) macro-scale hydrologic land surface model, along with a routing model to simulate streamflow forspecific basins (Rio Grande and Colorado River). In particular, daily US observations ofprecipitation (25 kms) from the Univ. of Arizona Precipitation Estimation precipitation productsand RSM daily forecasts of temperature (max and min), wind speed, and solar radiation are usedto continuously force water and energy balance versions of the VIC at daily time scales. VICdaily, weekly, and monthly forecasts are also made from the routine ECPC RSM forecasts forthe US (50 kms) and Southwest forecasts (25 kms). As will be shown, there are noticeabledifferences between the streamflow (and other variables) forecasts and continuous simulationsdue to various biases in the forecast variables (especially precipitation). As long as these forecastbiases are empirically corrected, useful predictions can be made. Streamflow predictions can also

Page 6: IRI/ARCS Regional Applications Project J. Roads, S. Chen

6

be empirically corrected a posteriori. Daily Land Data Assimilation System (LDAS) productsfrom NCEP are also being used to initialize and force the VIC model but present additionalassimilation problems, such as how to incorporate the LDAS soil moisture, snow waterequivalent and other state variables.

Forecast variables: Precipitation (mm)Daily maximum temperature (C)Daily minimum temperature (C)Sub-daily air temperature (C)Wind speed (m/s)Air pressure (kpa)Atmospheric vapor pressure (kpa)Atmospheric density (kg/m^3)Downwart Shortwave rad. (w/m^2)

Length of forecast: 3 months or longerTime period for forecasts: 00 UTC Sept. 27,1997-presentFrequency of forecasts: once a week,Domain: 25.04752 - 42.25518 N, 241.78625 - 261.25453 EResolution (200-25 kms)

Page 7: IRI/ARCS Regional Applications Project J. Roads, S. Chen

7

Eventually we would like to expand our region of interest beyond U.S. Southwest to the entireUS. We are also working with scientists from Taiwan in integrating a very similar hydrologicprediction system for streamflow seasonal forecasts. The global/ atmospheric forecasts fromECPC drive regional forecasts and a conceptual lumped water balance model (Tong and Haith1995).

4.3 Climate Change Impacts in the Pacific NorthwestThere are a number of disciplinary process models of the Pacific Northwest involving climate,hydrology, Puget Sound circulation, and Puget Sound ecosystem dynamics, which are wellconstructed and stand alone, but which have not been used systematically to address issues ofclimate change. To enable these models to use the growing knowledge of climate change andvariability to assess impacts, a dedicated effort will be required to produce climate-changescenarios and input parameters that address the specific modeling needs.

We propose to tailor the various climate products and model outputs currently underdevelopment in the Climate Impacts Group and CDEP ARC so that existing models in the usercommunity can be coordinated by simulating responses to a common set of standard climate-change scenarios. The models we are considering require input datasets that simulate theperturbed atmospheric and hydrologic conditions of a changed climate. We are at an appropriatestage in the development of climate modeling and regional downscaling to produce acomprehensive collection of scenarios and provide the various data at appropriate resolution toforce these models. Specific regional process models that we propose to focus on in developingthe dataset are: 1) The Washington State Department of Ecology's South Puget Sound Model(SPASM). 2) Lifecycle model of Coho. 3) Pacific Northwest Estuary model. 3) Californiacurrent ecosystem model.

These regional process models are currently used to diagnose systems under present climateconditions and are run with observed present-day input parameters, such as meteorologicalconditions or river flows. In the Climate Impacts Group, we have developed tools to performdownscaling of long global climate simulations using statistical methods and have a mesoscalemodel (mm5) available for physical downscaling of shorter simulations (Widmann et al. 2001,Wood et al, 2001). We also have the hydrological modeling tools to simulate the correspondingstream flows. The resulting data may be used to perturb or replace the meteorological forcing inthe process models and assess their response to climate change.

4.4 Crop ForecastingFSU is interested in assessing the skill of both the global and regional models for use indownscaling to crop models. One measure of the skill will be the crop yields determined by thecrop models and verified against the observed yields for selected locations in Florida andGeorgia. The timing of the growth stages of the crops (i.e., germination, flowering etc.) will alsobe examined between the global and regional models to assess the sensitivity of the downscaling.

Page 8: IRI/ARCS Regional Applications Project J. Roads, S. Chen

8

Forecast variables (input): tmax, tmin, precipitation, net surface solar radiationLength of forecast: 6 monthsTime period of forecasts: March thru August (spring growing season) November thru March (winter growing season) 1997 to presentDomain: (14.8N,100W; 42.5N,67.5W)Resolution: 200-20 km

4.4 Fire danger forecastsFSU is interested in assessing the skill of both the global and regional models for use indownscaling to crop models. One measure of the skill will be the crop yields determined by thecrop models and verified against the observed yields for selected locations in Florida andGeorgia. The timing of the growth stages of the crops (i.e., germination, flowering etc.) will alsobe examined between the global and regional models to assess the sensitivity of the downscaling.

One of the crop models that will be used to simulate maize yield is the CERES-Maize simulationmodel (Ritchie et al., 1998). This model is a dynamic process based crop model that simulatesplant response to soil, weather, water stress and management practices. The model calculatesdevelopment, growth and partitioning processes on a daily basis, beginning with planting andending at harvest maturity. The model can be used to assess the impacts of weather (Mavromatisand Jones, 1998; Mearns et al., 1999) and management decisions (Jones et al, 2000; Hansen etal., 2001). Preliminary results for selected locations in Florida are encouraging and show greaterskill of the FSU regional model to predict Maize yields when compared to the FSU global model(Jagtap et al 2002).

Page 9: IRI/ARCS Regional Applications Project J. Roads, S. Chen

9

ECPC is developing an experimental US fire danger prediction system using the RegionalSpectral Model (RSM) and the National Fire Danger Rating System described above. Our goalis to assess the meteorological forecast skill of the basic input variables, utilize the operationalanalyses to update these variables for the initial state, and to eventually assess the skill of thesefire potential forecasts by comparison to fire occurrence and size data. Eventually we intend tomove beyond the US boundaries to other regions by using satellite information as well asweather information to drive fire danger models (Burgan et al. 1998). For example, we havestarted to build ECPC long-range global/regional forecasting system into the fire-danger ratingsystem of Heilongjiang Meteorological Bureau, China. We are also working with the USFS toprovide necessary input data for their seasonal forecasts for specific regions. Their currentmethodology relies on statistical relationships between forecast 500 mb heights and fire dangervariables and thus we will also be providing ensemble forecasts of the 500 mb heights todetermine their influence upon ensemble seasonal forecasts of seasonal fire danger. Forecast variables: 2M T Temperature at 2:00 LST (Fahrenheit) 2M H Relative Humidity at 2:00 LST (percent) PPT 24 hour precipitation (inches) CLD Cloud amount at 2:00 (percent) WS Windspeed (miles/hour) precipitation duration fudged = 24hrpcp / 3.81 Max T, Min T for preceding 24 hours Max H, Min H for preceding 24 hours Precipitation Duration during preceding 24 hour

500 mb heightsLength of forecast: 4 monthsTime period for forecasts: 00 UTC Sept. 27,1997-presentFrequency of forecasts: once a week. (Every Sat. at 00UTC).Domain: 21.13881 - 51.41708 N, 229.24575 - 294.10550 E Resolution: (200-25 kms)

Page 10: IRI/ARCS Regional Applications Project J. Roads, S. Chen

10

4.5 Brazil The IRI's regional model domain will cover both Northeast Brazil and the La Plata Basin ofSouth America. These two targeted regions are chosen due to their potentialy high climatepredictability and practical applications. The regional climate models (RCM) will be driven bythe IRI and NCEP GCM hindcasts/forcasts. Various regional models (RegCM2, RSM) driven bythe global hindcast/forcasts will be tested over the model domain. The initial simulation andprediction period will be from 1997 to present. Regional climate predictability, seasonalvariability, high-resolution spatial and temporal statistics, and systematic errors of the regionalmodels over the targeted domains will be examined in the hindcast/forcast mode. The best GCM-RCM model configuration will be used to produce ensemble seasonal forcasts for subsequentapplications in Northeast Brazil and the Plata Basin. An ARCS downscaling activity in the PlataBasin could also provide an important contribution to a major focus of the CLIVAR Variabilityof the American Monsoon Systems (VAMOS) project, which is planned for the Plata Basin. ThisVAMOS project (PLATIN) will focus on the variability and predictability of hydroclimate over

Page 11: IRI/ARCS Regional Applications Project J. Roads, S. Chen

11

the Plata Basin, and is interested in fostering linkages with applications. The IRI integrated climate prediction and application project over the Ceara State in theNortheast Brazil will evaluate climate impacts in agriculture, hydrology, and industry, andattempt to reduce vulnerability to recurrent droughts. The ongoing studies at the IRI andFUNCEME (an institute in Ceara, Brazil) for this region include the RSM ensemble downscalingand prediction over a domain tested for the best simulation over the Ceara State, and theapplications by using statistical hydrology models, economic models (such as crop-pricemodels), etc. The ARCs downscaling study will further provide additional information and toolsto reinforce the application researches over this region. Firstly, in an enlarged model domaincovering both the Ceara as well as the Plata Basin (discussed below), RegCM2 will be run toprovide a multi-model ensemble simulation and forcast. Secondly, RSM will also be run in theenlarged domain by NECP and/or IRI over the enlarged domain. Special attention will be paid onimproving the simulation of the intensity and positioning of the Atlantic ITCZ, which is crucialfor the precipitation forcast in the Northeast Brazil. The application value of the ARCsdownscaling results will be studied by taking advantage of the existing tools envolved in theCeara project. The full-set of the IRI's regional model results will be archived in the IRI MassStorage System for the application model users. The IRI will further collaborate with other ARCs to apply the VIC model (UW) in the PlataRiver basin in southeastern South America, using downscaled meteorological variables(principally daily Tmax, Tmin and rainfall) derived from GCM and regional-model simulationssuch as RegCM2. The study will compare this physically-based approach with various statisticalmethods for hydrological simulation and forecast, and quantify the strengths and weaknesses ofeach approach. IRI also plans to leverage and build upon existing contacts with water managersand agronomists in the region, including: Alexandre Guetter (Curitiba, Brazil) and WalterBaethgen and Ricardo Romero (Montevideo, Uruguay), together with an existing applicationsproject in Argentina involving crop modeling for farm cooperatives, established throughagronomists at the University of Florida, but which also involves several scientists at IRI.Several crop models are available, and these generally require daily values of Tmax, Tmin andrainfall. Forecast variables: precipitation (mm/day) daily maiximum temperature (C) daily minimum temperature (C) sub-daily air temperature (C) wind speed (m/s) surface shortwave radiation (w/m^2) Length of forecasts: 7 months Time period of forecasts: 1997-present Frequency of forecasts: once 3 months Domain: 90W-21W, 41S-15N Resolution 50km. 5. Summary The previous highly focused ARCS/IRI regional modeling intercomparison project centered overBrazil has now evolved toward a larger project with less of a regional focus than the previous

Page 12: IRI/ARCS Regional Applications Project J. Roads, S. Chen

12

project but with the promise of eventually developing a greater number of regional applicationsin several additional regions. There are certainly a number of overlapping efforts. Most of theprojects intend to examine and incorporate the NCEP and IRI global forecast products as part oftheir regional applications. Several of the US based projects also intend to examine andincorporate the NCEP regional forecast products. In addition to statistical downscaling, severalefforts will also use physical downscaling to drive the regional applications (NCEP, IRI, ECPC,and FSU RSM). Some of the hydrologic efforts also use the VIC model Some of the crop modelsdeveloped for the US Southeast are also li similar to crop models developed over Brazil. A key to this coordinated effort will be the development of two types of reduced data sets. Forexample, for some purposes, all that will be needed will be monthly mean forecasts oftemperature, precipitation and the 500 mb height. Other application efforts will require 8 x dailyvalues for precipitation, surface air temperature, surface humidity, wind speed, downward solar,downward infrared, tmax, tmin, rhmax, rhmin, cloudiness at 14:00 LST. The challenge will be to adequately coordinate such a project among the diverse groups andinterests. Below, we describe a 3 year work plan along with the appropriate leaders of eachmodeling or application project. The identified leaders will be in charge of developing theappropriate community data sets and making contact with other groups interested in participatingin the individual efforts. Year 1Hydrologic models will be set up for the US West, US Southwest, Brazil.Crop models will be set up for the US Southeast.Fire Danger models will be set up for the USVarious application models will be set up for BrazilModel input for each of these applications will be developed from individual efforts and willmake use of more comprehensive global and regional forecasts being developed at NCEP andIRI. Year 2 Develop protocol to distribute forecast data through individual projects. Skill will begin to be assessed for the various applications and regions Additional model input from the broader community will be collected and developed foradditional application. Year 3Finalize model skill assessments for both meteorological input and applications.Submit papers for publication, and begin to develop additional regional applications

Project LeadersNCEP Global and Regional model forecasts, H. Juang, J. HanIRI Global model forecasts, J. Qian, S. Zebiak, A. RobertsonBrazil Applications, J. Qian, S. Zebiak, A. RobertsonUS West Hydrology, D. LettenmaierUS Southwest Regional Climate, J. Roads, J. ChenUS Northwest Hydrology, E. Salathe, E. Miles

Page 13: IRI/ARCS Regional Applications Project J. Roads, S. Chen

13

S.E. US Crop Forecasting, S. Cocke, T. LarowFire Danger Forecasts, J. Roads, S. Chen

ReferencesBurgan, R. E., R.W. Klaver, J.M. Klaver, 1998: Fuel models and fire potential from satellite and

surface observations. Intl. J. Wildland Fire 8, 159-170.Chen, S. -C., J.O. Roads, H. -M. H. Juang, M. Kanamitsu, 1999: Global to regional simulation of

California's wintertime precipitation. J. Geophys. Res., 104(24), 31517-31532,Cocke, S. D. and T. E. LaRow, 2000: Seasonal Predictions using a Regional Spectral Model

Embedded within a Coupled Ocean Atmosphere Model. Mon. Wea. Rev., 128, 689-708.Deeming, John E., Robert E. Burgan, Jack D. Cohen. The National Fire-Danger Rating System –

1978. Ogden, UT: United States Department of Agriculture, Forest Service,Intermountain Forest and Range Experiment Station, General Technical Report INT-39.1977: 66p.

Fujioka, F.M., 1990: The art of long-range fire weather forecasting. Proceedings, InternationalSymposium on Fire and the Environment, March 1990. USDA For. Serv. Gen. Tech.Rep. SE-69. 219--223.

Jagtap, S. S., J. W. Jones, D. E. Hanley, T. E. LaRow, S. D. Cocke and J. J. O'Brien 2002: Issuesand challenges: Integrating regional spectral model climate forecasts with crop models.(In Preparation).

Juang, H. and M. Kanamitsu, 1994: The NMC nested regional spectral model. Mon. Wea. Rev.,

122, 3-26.

Juang, H., S. Hong, and M. Kanamitsu 1997: The NMC nested regional spectral model. Anupdate. Bull. Amer. Meteor. Soc., 78, 2125-2143.

Han, J. and J. Roads, 2002: US Climate Sensitivity Simulated with the NCEP Regional SpectralModel. Climate Change (submitted)

Hansen, J.W., J.W. Jones, A. Irmak, and F. Royce, 2001: El Nino-Southern Oscillation impactson crop production in the Southern United States. Impacts of El Nino and ClimateVariability on Agriculture. C. Rosenzweig, K.J. Boote, S. Hollinger, A. Iglesias and J.Phillips (Eds.), ASA Special Publication No. 63, American Society of Agronomy,Madison, Wisconsin, USA, 57-78.

Jones,J.W., J.W. Hansen, F.S. Royce, and C.D. Messina, 2000: Potential benefits of climateforecasting to agriculture. Agric. Ecosys. Env., 82, 169-184.

Kalnay, E., M. Kanamitsu, R. Kistler, W. Collins, D. Deaven, L. Gandin, M. Iredell, S. Saha, G.White, J. Woolen, Y. Zhu, A. Leetma, R. Reynolds, M. Chelliah, W. Ebisuzaki, W.Higgins, J. Janowiak, K.C. Mo, R. Jenne, and D. Joseph, 1996: The NCEP/NCAR 40-Year Reanalysis Project, Bulletin of the American Meteorological Society, 77:437-471.

Kanamitsu, M., W. Ebisuzaki, J. Woollen, S. -K. Yang, J. Hnilo, M. Fiorino, and J. Potter, 2002:NCEP/DOE AMIP-II REANALYSIS (R-2). Bulletin Amer. Meteor. Soc. (In press)

Liang, X. , D.P. Lettenmaier, E.F. Wood, and S.J. Burges "A Simple Hydrologically BasedModel of Land and Energy Fluxes for General Circulation Models," Journal ofGeophysical Research, 99(D7), 14, 415-14, 1994.

Mavromatis, T. and P.D. Jones, 1998: Comparison of climate change scenario constructionmethodologies for impact assessment studies. Agric. For. Meteorol., 91, 51-67.

Page 14: IRI/ARCS Regional Applications Project J. Roads, S. Chen

14

Mearns, L.O., T. Mavromatis, E. Tsvetsinkaya, C. Hays, W. Easterling, 1999: Comparativeresponses of EPIC and CERES crop model to high and low spatial resolution climatechange scenarios. J. Geophys. Res., 40 (D6), 6623-6646.

Pitman, A. J., A. Henderson-Sellers, C. E. Desborough, Z.-L. Yang, F. Abramopoulos, A.Boone, R. E. Dickinson, N. Gedney, R. Koster, E. Kowalczyk, D. Lettenmaier, X. Liang,J.-F. Mahfouf, J. Noilhan, J. Polcher, W. Qu, A. Robock, C. Rosenzweig, C. A.Schlosser, A. B. Shmakin, J. Smith, M. Suarez, D. Verseghy, P. Wetzel, E. Wood, Y.Xue, 1999: Key results and implications from Phase 1(c) of the Project forIntercomparison of Land-surface Parametrization Schemes, Climate Dynamics, 15, 673-684.

Ritchie, J.T., U. Singh, D.C. Godwin, and W.T. Bowen, 1998: Cereal growth, development, andyield. Understanding Options for Agricultural Production. G.Y. Tsuji, G. Hoogenboom,P.K. Thornton, (Eds.), Kluwer Academic Publisher, Dordrecht, The Netherlands, 79-98.

Roads, J.O. and S-C. Chen, 2000: Surface Water and Energy Budgets in the NCEP RegionalSpectral Model. JGR-Atmospheres. 105, 29, 539-29550.

Roads, J., S.-C. Chen, M. Kanamitsu, 2002. US Regional Climate Simulations and SeasonalForecasts. Journal of Geophysical Research-Atmospheres (submitted).

Roads, J., S. Chen, L. Druyan, M. Fulakeza, S. Cocke, T. Larrow, J. Qian, 2002: The IRI/ARCSregional model comparison project (in preparation)

Suarez, M. J. and L. L. Takacs, 1995. Documentation of the ARIES/GEOS Dynamical Core:Version 2, NASA Technical Memorandum no. 104606, v. 5., 53pp.

Todini, E., 1996: The ARNO rainfall-runoff model, J. Hydrol., 175, 339-382.Tung, C.P., and D.A. Haith, 1995: Global warming effect on New York stream flows. Journal of

Water Resource Planning and Management, 21, 216-225.Widmann, M., C. S. Bretherton, and E. P. Salathe, 2001: Statistical precipitation downscaling

over the Northwestern United States using numerically simulated precipitation as apredictor", submitted to Journal of Climate.http://www.atmos.washington.edu/~salathe/papers/downscale/

Wood, A.W., Maurer, E.P., Kumar, A. and D.P. Lettenmaier, 2002. “Long range experimentalhydrologic forecasting for the eastern U.S.”, in press, Journal of Geophysical Research.(www.ce.washington.edu/pub/HYDRO/aww/east_fcast/aww_jgr2002.pdf)

Zhao, R.-J., L.-R. Fang, X.-R. Liu, and Q.-S. Zhang, 1980: The Xinanjiang model, InHydrological Forecasting Proceedings, Oxford Symposium, IASH 129:351-356.