EBQUJWF 0CTFSWBUJPO (VJEBODF QQMJFE UP 5ZQIPPO 3VTB … · (2007) have compared various adaptive observation strategies for Atlantic hurricanes. To prepare for the THORPEX-Pacific

ASIA-PACIFIC JOURNAL OF ATMOSPHERIC SCIENCES, 44, 3, 2008, p. 297-312

1. Introduction

Two kinds of uncertainties may be involved in nu-merical weather prediction (NWP). One stems from an imperfect model, insufficient knowledge or im-perfect parameterization of the physical processes involved, and imperfect numerical solution of the governing partial differential equations. The other is initial condition (i.e., analysis) uncertainty, which is caused by imperfect realization of the current state of the atmosphere. The initial condition is produced by a data assimilation system that combines ob-servations and prior information (typically a pre-

vious numerical forecast) while taking into account the uncertainties associated with each. Both types of uncertainties may degrade NWP forecasts by ampli-fying during the forward model integration. Although the initial condition uncertainties may evolve as a major part of forecast error for short-term weather forecasts, such uncertainties may be reduced by enhancing the type and number of observations. In the past, meteorological observational sites usu-ally were near human residences to monitor weather conditions for living or industrial purposes. More re-cently, more weather observations have become available in previously data-void areas.

Adaptive (or targeted) observation guidance (or strategies) recently have been developed to reveal re-gions where more observations would improve weather forecasts. That is, the goal of adaptive (or tar-geted) observations is to decrease the forecast error by placing observations in regions where additional

Adaptive Observation Guidance Applied to Typhoon Rusa:

Implications for THORPEX-PARC 2008

Hyun Mee Kim1, Byoung-Joo Jung

1, Yeon-Hee Kim

2 and Hee-Sang Lee

2

1Atmospheric Predictability and Data Assimilation Laboratory, Department of Atmospheric Sciences, Yonsei University, Seoul, Republic of Korea

2Forecast Research Team, National Institute of Meteorological Research, Seoul, Republic of Korea(Manuscript received 11 June 2008; in final form 8 August 2008)

Abstract In this study, total energy (TE) singular vectors (SVs) based on the fifth generation Pennsylvania State University/National

Center for Atmospheric Research (PSU/NCAR) Mesoscale Model (MM5) and its tangent linear and adjoint models with a Lanczos algorithm are applied to Typhoon Rusa. The structures of SVs are evaluated to understand the sensitivity of forecasts with respect to the initial conditions, and thence to suggest the sensitive regions in terms of adaptive observations. In addition, the implications of applying this adaptive observation strategy for the THORPEX- Pacific Asian Regional Campaign (T-PARC) are discussed. Sensitive regions identified by TESVs are located horizontally in inflow regions along the edge of the subtropical high and the mid-latitude trough. Sensitive regions are located vertically in the mid-tropo-sphere with several secondary peaks throughout the troposphere. In contrast to the upward energy propagation mechanism of SV development in extratropical cyclones, this does not occur for Typhoon Rusa. Vertically-confined, upshear-tilted SV structures under the mid-latitude trough are noticed for the temperature component of SV. The results of this study are quite consistent with recent studies on targeted observation strategies of tropical cyclones, and demonstrate that the TESVs can capture the signal of the environmental features affecting the evolution of Typhoon Rusa. The TESV guidance shown in this study will be calculated and provided for real-time field experiments during the T-PARC.

Key words: Adaptive observations, adaptive observation guidance, sensitive regions, THORPEX, T-PARC, singular vectors, tropical cyclone

Corresponding Author: : Hyun Mee Kim, Department of Atmospheric Sciences, Yonsei University, Shinchon-dong134, Seodaemun-ku, Seoul, 120-749, Republic of Korea.Phone : +82-2-2123-5683, Fax : +82-2-365-5163E-mail: [email protected]

298 ASIA-PACIFIC JOURNAL OF ATMOSPHERIC SCIENCES

observations are expected to improve a forecast of interest. These regions may be considered "sensitive" in the sense that changes to the initial conditions in these regions are expected to have a larger effect on a particular measure of forecast skill than changes in other regions (Kim et al., 2004).

Adaptive observation strategies may be roughly categorized as uncertainty-based and dynamics-based strategies. Uncertainty-based adaptive observation strategies use estimates of analysis and forecast er-rors (e.g., Bishop and Toth, 1999; Bishop et al., 2001; Hamill and Snyder, 2002; Aberson and Etherton, 2006; Majumdar et al., 2006). The Ensemble Transform Kalman Filter (ETKF) is the typical uncertainty- based adaptive observation strategy. Dynamics- based adaptive observation strategies use the dynamics of the flow (Bergot, 1999; Bergot et al., 1999; Pu and Kalnay, 1999; Buizza and Montani, 1999; Gelaro et al., 1999; Langland et al., 1999; Montani et al., 1999; Peng and Reynolds, 2005, 2006; Majumdar, 2006; Kim and Jung, 2006; Wu et al., 2007). Adjoint sensi-tivity and singular vectors (SVs) are typical of the dy-namics-based adaptive observation strategy. Adjoint sensitivity is the gradient of some forecast measure with respect to the model control variables (e.g., Errico, 1997) or to the observations (Baker and Daley, 2000). Given an initial norm, a final norm, a basic state, and a forecast model, SVs are those per-turbations that amplify the most rapidly over a speci-fied time period. Because SVs grow rapidly, they have been used for adaptive observations by identify-ing regions with large sensitivity to small perturba-tions (e.g., Palmer et al., 1998).

Sensitive regions for adaptive observations change depending on weather events, forecast regions of in-terest, and adaptive observation strategies. Until re-cently, most of the studies that have used adaptive ob-servation strategies to identify sensitive regions for adaptive observations have focused on extratropical cyclogenesis over the northeast Pacific (e.g., North Pacific Experiment (NORPEX); Winter Storm Reconnaissance Program) or Atlantic (i.e., Fronts and Atlantic Storm Track Experiments (FASTEX); Atlantic THORPEX1) Observing System Test (Atlantic-TOST)). Recently, adaptive observation

strategies have been applied for tropical cyclone (TC) evolution. Aberson (2003) conducted sensi-tivity analysis study on tropical cyclones in the Atlantic. Majumdar et al. (2006) and Reynolds et al. (2007) have compared various adaptive observation strategies for Atlantic hurricanes.

To prepare for the THORPEX-Pacific Regional Campaign (T-PARC), which is planned for the west-ern North Pacific in 2008 to improve typhoon fore-casts, some studies have addressed sensitive regions for Pacific typhoons using adaptive observation guidance. Peng and Reynolds (2006) used SVs to study the dynamics of TC evolution in both the Atlantic and the Pacific during the 2003 summer season. Kim and Jung (2006) investigated sensitive regions for Typhoon Rusa using adjoint-based fore-cast sensitivities. Wu et al. (2007) utilized an ad-joint-derived sensitivity steering vector (ADSSV) technique to detect sensitive regions for adaptive ob-servations of several typhoons.

Typhoon events that affect Korea are closely re-lated to the scientific issues of typhoon track targeting and recurvature considered in T-PARC. Therefore, to prepare for T-PARC, dynamics-based adaptive ob-servation strategies (i.e., adjoint-based forecast sen-sitivities and SVs) using the fifth-generation Pennsylvania State University/National Center for Atmospheric Research (PSU/NCAR) Mesoscale Model (MM5) and its tangent linear and adjoint mod-els have been applied to several typhoons. The focus of this study is Typhoon Rusa, which caused sig-nificant societal and economic damage in Korea.

In this study, adaptive observation guidance using SVs is applied to Typhoon Rusa to understand fore-cast sensitivity with respect to the initial conditions, and thence to suggest the sensitive regions in terms of adaptive observations. Section 2 describes the mathematical formulations used to calculate total en-ergy (TE) norm SVs (TESVs). Case study descrip-

1) The Observing system Research and Predictability Experiment (THORPEX) is a WMO/WWRP program from 2003 to 2012 to accelerate improvements in theaccuracy of 1 to 14-day high impact weather forecasts for the benefit of the society and economy (WMO, 2004).

31 August 2008 Hyun Mee Kim et al. 299

tion and configurations of the model and SV calcu-lations are presented in section 3, and the SV sensitiv-ities for Typhoon Rusa and linearity evaluation are presented in section 4. In section 5, implications for T-PARC are presented. Section 6 contains a summary and discussion.

2. Total energy norm SVs

The calculation of SVs essentially involves select-ing the projection coefficients subject to the con-straints that the initial disturbance has unit amplitude in a specified norm and evolves to have maximum amplitude in a specified norm after some finite opti-mization time, optt τ= . In this study, the initial and fi-nal norms are the dry TE defined by Zou et al. (1997) as

22 2 2 2

d, ,

1E2 Nx y

gu v wσ

′ ′ ′ ′= + + + θ θ ∫∫∫

(1)

221 '

s

p dydxdc

+ σ ρ ,

where dE is the dry TE in a non-hydrostatic model, u′, v′, w′ are the zonal, meridional, vertical wind per-turbations, ′θ is the potential temperature perturba-tion, p′ is the pressure perturbation, N , θ , ρ , sc are Brunt-Väisälä frequency, potential temperature, density, speed of sound at the reference level, and x , y, σ denote zonal, meridional, and vertical coor-dinates, respectively.

The constrained optimization problem seeks to maximize the Rayleigh quotient, 2λ (the amplifica-tion factor),

2λ′ ′

=′ ′

PMx (0),CPMx (0)x (0),Cx (0) ,

(2)

at the time optt τ= , where the inner product is denoted by 〈,〉, P is a local projection operator that makes the state vector be zero outside a given domain2) (Buizza, 1994b), M is the tangent linear model

(TLM) of the nonlinear model, and C is the dry TE norm. In (2), the state vector at the initial time is as-sumed to evolve linearly. By defining the local pro-jection operator, the amplitude of the state vector with norm C at the optimization time is maximized over a specific region. The maximum of this ratio is real-ized when ′x (0) is the leading SV of the TLM M for the C norm, that is, ′x (0) satisfies

2λ′ ′T TM P CPMx (0) = Cx (0) . (3)

The generalized eigenvalue problem in (3) can be re-duced to an ordinary eigenvalue problem by multi-plying both sides of (3) by the inverse of the square root of C. Then a Lanczos type algorithm (e.g., Ehrendorfer and Errico, 1995; Kim, 2003; Kim et al., 2004) is used to solve for ′x (0) in (3).

3. Case study and model description

a. Case study

In addition to a record-breaking daily rainfall amount of 870.5 mm on the east coast of the Korean peninsula on 31 August 2002, Typhoon Rusa caused

Fig. 1. The RSMC Tokyo-Typhoon Center best track (-+- symbols) and the 24-h MM5 forecast track (-●-) of Typhoon Rusa. Each symbol is plotted at 6 hour intervals.

2) In this study, the regions of non-zero projection operatorare referred to as the verification region.


Fig. 2. 200-500 hPa layer-average PV (shading interval of 1 PVU) and mean sea-level pressure (MSLP) (contour interval of 4 hPa) of analyses at (a) 1200 UTC 30, (b) 0000 UTC 31, (c) 1200 UTC 31 August 2002, and of forecasts at (d) 1200 UTC 30, (e) 0000 UTC 31, and (f) 1200 UTC 31 August 2002. The box in (f) denotes the geographic region for maximizing the dry TE at 24 h, which is referred to as the verification region.


many casualties and extensive property damage in Korea. After its formation on 23 August 2002, Typhoon Rusa moved northwestward and recurved northeastward along the edge of the subtropical high after 1200 UTC 30 August (Fig. 1a). Rusa approached the southern part of the Korean peninsula at 0600 UTC 31 August. The MM5 model is integrated for 24 hours from 1200 UTC 30 to 1200 UTC 31 August 2002 (Fig. 1). Even though the predicted motion was slightly faster than observed, the predicted path gen-erally simulates the observed track well.

Analyses of 200 – 500 hPa layer-average potential vorticity (PV) from 1200 UTC 30 to 1200 UTC 31 August 2002 are shown in Fig. 2. At 1200 UTC 30 August, a large PV reservoir was north of the Korean peninsula, and isolated regions of large PV were over the center and northwest of Rusa (Fig. 2a). At 0000 UTC 31 August, the isolated large PV region north-west of Rusa merged with the large PV over the TC center as Rusa approached higher latitudes (Fig. 2b). At 1200 UTC 31 August, the large PV near the storm center merged with the PV reservoir north of the Korean peninsula (Fig. 2c). The fifth-generation Pennsylvania State University/National Center for Atmospheric Research (PSU/NCAR) Mesoscale Model (MM5) forecasts (Figs. 2d, e, and f) have sim-ilar overall characteristics as the analysis fields, which implies that the predicted fields generally sim-ulate the analysis fields well.

b. Model and SV configurations

To calculate SVs, this study uses the MM5 adjoint modeling system (Zou et al., 1997) and a Lanczos algorithm. The model is centered at 36˚ N and 123˚ E with a 100 km horizontal grid spacing in a 50 x 50 domain and 14 evenly spaced sigma levels in the ver-tical from the surface to 50 hPa. The model initial and lateral boundary conditions are from the National Centers for Environmental Prediction (NCEP) final analysis (FNL; 1o x 1o global grid). The Optimum Interpolation Sea Surface Temperature (OISST) ver-sion 2 (Reynolds et al., 2002) is used for the lower boundary condition over the ocean. Physical param-eterizations used for the nonlinear basic state in-

tegrations include the Grell convective scheme, a bulk aerodynamic formulation of the planetary boundary layer, a simple radiational cooling scheme, horizontal and vertical diffusion, dry convective ad-justment, and explicit treatment of cloud water, rain, snow, and ice. The same physical parameterizations are used in the TLM and adjoint model integrations, except the effect of moisture is neglected. The initial time of the SV calculation is 1200 UTC 30 August 2002, the same as that of the MM5 model integration, and the optimization time interval is 24 hours.

4. Characteristics of TESVs

a. Growth rates and structures of SVs

The amplification factors of individual SVs are shown in Fig. 3. The relative contributions of the leading five SVs to energy of the total 20 SVs are 51.8 %, which implies the leading five SVs explain more than half of the total energy.

Horizontal structures of each SV3) component on 800 hPa are shown in Fig. 4. A large positive (nega-tive) zonal wind component of SV is located around 400 km (1000 km) southeast of the storm center (Fig.

Fig. 3. Amplification factors (ordinate) for the first twenty SVs (abscissa) with 24 hour optimization time.

3) All the SVs from Fig. 4 to Fig. 7 are for the leading SV.In addition, SV without specification refers to the SV at theinitial time.


4a). Regions of large zonal wind components of SVs correspond well with inflow regions toward the storm center, as mentioned by Peng and Reynolds (2006) and Kim and Jung (2008a). The meridional wind component of the SVs (Fig. 4b) has two large positive sensitivity regions around 1000 km east and 700 km west of the storm center. The larger magni-tude sensitivity region to the east (i.e., in the right semi circle of the storm center) is also associated with the inflow toward the TC in the western quadrant of the subtropical high (Fig. 4b). Although the temperature component of the SVs also has maxima in the right semi circle of the storm, it is part of an overall elon-

gated structure from northeast China to the East China Sea (Fig. 4c).

Vertical structures of each component of the SV along the solid lines in Fig. 4 are shown in Fig. 5. The vertical structure of the zonal wind component of the SV is shown superposed on the PV in Fig. 5a. The PV column larger than 1 PVU through most of the troposphere is located near the TC center, with anoth-er larger PV region northeast of the Korean peninsula that corresponds to the mid-latitude trough in Fig. 2. The zonal wind component of the SV has large pos-itive and negative sensitivities in the lower tropo-sphere south of the TC center, which is associated

Fig. 4. Leading SVs on 800 hPa at the analysis time: (a) zonal wind component of SV (contour interval of 1 m s-1), (b) meridional wind component of SV (contour interval of 1 m s-1), and (c) temperature component of SV (contour interval of 0.5 K). △ indicates the center of Typhoon Rusa.


with the large positive and negative sensitivities southeast of the TC center in Fig. 4a. For the meri-dional wind component of the SV (Fig. 5b), max-imum sensitivities have vertically confined struc-tures that are west and east of the TC center in the mid-troposphere, with the larger magnitudes east of the TC center. The positive sign of the meridional wind component of the SV perturbation is consistent with inflow toward the TC, which likely leads to growth of dry TE in the verification region at the final time. By contrast, the temperature component of the

SV (Fig. 4c) has an upshear-tilted structure under the upper trough, which is a typical SV structure for ex-tratropical cyclones (e.g., Kleist and Morgan, 2005).

Horizontal structures of each component of the 24-h evolved SV on 350 hPa are shown in Fig. 6. A large positive (negative) zonal wind SV component is over (500 km north of) the storm center (Fig. 6a). The vertical cross-section along the line DD’ through the TC center indicates that the zonal wind SV com-ponent has a maximum in the lower part of the TC center (Fig. 7a). The meridional wind component of

Fig. 5. Vertical cross-sections along the (a) line AA’ indicated in Fig. 4a of PV (shaded, interval of 1 PVU) superposed on the zonal wind component of SV (contour interval of 2 m s-1), (b) line BB’ in Fig. 4b of PV (shaded, interval of 1 PVU) superposed on the meridional wind component of SV (contour interval of 1 m s-1), and (c) line CC’ in Fig. 4c of PV (shaded, interval of 1 PVU) superposed on the temperature component of SV (contour interval of 0.5 K). The ordinate represents the vertical level (hPa) and the abscissas in (a), (b), and (c) denote the line AA’, BB’, and CC’ in Fig. 4, respectively. The numbers in the abscissas represent latitude and longitude of A, A’, B, B’, C, and C’.


the evolved SV has large positive sensitivities north-east (i.e., right semi circle) of the TC center (Fig. 6b). The vertical cross-section of the meridional wind component of the SV along the line EE’ indicates that the largest magnitude of the SV is now in the upper troposphere (Fig. 7b). The SV temperature compo-nent has its maximum over the TC center (Fig. 6c). The vertical cross-section of the temperature compo-nent of the SV along the line FF’ has a large temper-ature perturbation in the upper troposphere (Fig. 7c).

The PV is vertical at this time and the upper and lower boundary temperature perturbations have increased, similar to extratropical cyclones (Fig. 7c).

b. Energy-weighted SVs

The vertically-integrated, energy-weighted SV is used to represent general features of the sensitivity of Typhoon Rusa. The energy-weighted SV is calcu-lated to combine all the SVs for different model varia-

Fig. 6. Leading SVs on 350 hPa at 24 h: (a) zonal wind component of SV (contour interval of 20 m s-1), (b) meridional wind component of SV (contour interval of 20 m s-1), and (c) temperature component of SV (contour interval of 5 K).


bles with different units into a single SV field with the unit of energy (J kg-1). The vertically-integrated, energy-weighted 1st, 2nd, 3rd SV and MSLP at the ini-tial and final times are shown in Fig. 8. For the leading SV, the initial SV has large sensitivities of elongated half-circled structures from northeast China to the East China Sea (Fig. 8a). The sensitive regions are closely associated with the mid-latitude trough and the region between the typhoon and the subtropical ridge to the east. The 2nd SV has structures very sim-ilar to the leading SV except larger mid-latitude sen-

sitivities (Fig. 8c), and the 3rd SV has major sensitiv-ities north of the Korean peninsula (Fig. 8e). The 1st and 2nd evolved SVs after 24 h have large sensitivities over the TC center (Figs. 8b and d), and 600 km north of the TC center for the 3rd SV (Fig. 8f). These varia-tions in the evolved SVs imply that the TC forecasts may be sensitive to the choice of the initial conditions, and the sensitivities of the forecast to the initial con-ditions can be indicated by SVs. Adaptive ob-servations in the sensitive regions indicated by SVs may reduce the initial condition uncertainties in

Fig. 7. Vertical cross-sections along the (a) line DD’ indicated in Fig. 6a of PV (shaded, interval of 1 PVU) superposed on the zonal wind component of SV (contour interval of 30 m s-1), (b) line EE’ in Fig. 6b of PV (shaded, interval of 1 PVU) superposed on the meridional wind component of SV (contour interval of 20 m s-1), and (c) line FF’ in Fig. 6c of PV (shaded, interval of 1 PVU) superposed on the temperature component of SV (contour interval of 5 K). The ordinate represents the vertical level (hPa) and the abscissas in (a), (b), and (c) denote the line DD’, EE’, and FF’ in Fig. 6, respectively. The numbers in the abscissas represent latitude and longitude of D, D’, E, E’, F, and F’.


Fig. 8. Vertically-integrated, energy-weighted SV (J kg-1, shaded, interval varies) and MSLP (solid, contour interval of 4 hPa) at (a), (c), (e) 0h and (b), (d), (f) 24 h for 1st, 2nd, and 3rd SV, respectively. The boxes in (b), (d), and (f) denote the verification region.


those sensitive regions, and lead the better forecasts. The largest contributions to the vertically-in-

tegrated, energy-weighted leading SV in Fig. 8a are from two layers, corresponding to the lower and mid-dle parts of the atmosphere (Fig. 9). The en-ergy-weighted SVs in the lower part (Fig. 9a) are asso-ciated with the sensitivity region in Fig. 8e to the north of the typhoon. The energy-weighted SV in the middle part (Fig. 9b) is associated with the mid-latitude trough and the region between the typhoon and the subtropical high to the east (Fig. 9b). In conjunction with Fig. 8, large sensitivities in the mid-latitudes are located in the lower and middle part of the troposphere, which is con-sistent with the results in Kim and Jung (2008a).

Vertical profiles of the 1st, 2nd, and 3rd en-ergy-weighted SVs are shown in Fig. 10. The max-imum of the leading SV for both the total energy and kinetic energy at the initial time (Fig. 10a) is located near the mid-troposphere with several secondary peaks near lower troposphere, mid- to upper- tropo-sphere, and at the upper boundary. By contrast, the leading SV for the potential energy has a smaller con-tribution and a more uniform structure (also true for the 2nd and 3rd SV). The leading SV at the final time (Fig. 10d) has increased in amplitude by at least an order of magnitude and has maxima at the lower and

upper troposphere. Major and minor peaks of the 2nd SV for both the total and kinetic energy at the initial (Fig. 10b) and final (Fig. 10e) times are similar to those of the leading SV at the corresponding times, except for a relatively smaller peak in the upper tropo-sphere at the final time. The maxima of the 3rd SV for the total and kinetic energy at the initial (Fig. 10c) and final (Fig. 10f) times are located in the lower tropo-sphere and the 3rd SV for the potential energy also has a maximum in the lower troposphere.

The composites of the vertical profiles of the 1st to 5th energy-weighted SVs for the total, kinetic, and po-tential energy (Fig. 11) have similar characteristics to those of the two leading SVs in Fig. 10. While the max-imum of the composite SV at the initial time (Fig. 11a) is near the mid-troposphere, with secondary peaks near the lower troposphere and the upper boundary, the maximum of the composite SV at the final time (Fig. 11b) is located in the lower troposphere, with a secondary peak at the upper troposphere. In contrast to the vertical SV profiles for extratropical cyclones, the kinetic energy (KE) of the initial SV is dominant except at upper boundary. In addition, upward energy propagation during SV evolution in extratropical cy-clones (e.g., Buizza and Palmer, 1995; Badger and Hoskins, 2001; Morgan, 2001; Kim and Morgan,

Fig. 9. Vertically-integrated, energy-weighted leading SV (J kg-1, shaded, shading interval varies) for (a) lower part (900 hPa to sea level) superposed on 950 hPa temperature field (contour interval of 5 K) and (b) middle part (300 hPa to 700 hPa) superposed on 500 hPa geopotential height (contour interval of 50 m). The box in (b) denotes the verification region.


2002) is not noticed for Rusa, which is similar to the results in Kim and Jung (2008a).

c. Validation of linearity

The linear growth assumption of SVs is verified by examining the ratio (Zou et al., 1997) and sim-ilarity index (Buizza, 1994a) of 24-h linearly and nonlinearly evolved perturbation magnitudes and structures, respectively. For nonlinear evolution, the scaled leading SV that has maximum magnitudes of 4 m s-1 for u′ or v′, or 2 K for T ′ (e.g., Errico and Reader, 1999) is used as a finite perturbation because the linearity of the finite perturbation has more prac-tical interests. The finite perturbation is obtained by multiplying a fixed number to the leading SV so that

the maximum magnitude of T ′ is 2 K and that of u′and v′ is below 4 m s-1. The ratios for u′, v′, and T ′ compo-nents are 0.885, 0.993, and 0.784, respectively, and the similarity index is around 50 %, which implies similar magnitudes and structures between linear and nonlinear evolutions. The linearity can be further verified by comparing individual structures of the linearly and nonlinearly evolved SVs in the ver-ification region (Fig. 12). The nonlinearly and line-arly evolved zonal wind components of the leading SV at 850 hPa (Figs. 12b and c) are quite similar.

5. Implications for T-PARC

As mentioned in the Introduction, T-PARC will occur in the western North Pacific during 1 August

Fig. 10. Vertical distributions of the energy-weighted SVs (J kg-1, TE; solid, kinetic energy; dashed, potential energy; dotted) at 0 h for (a) 1st SV, (b) 2nd SV, and (c) 3rd SV and at 24 h in the verification region for (d) 1st SV, (e) 2nd SV, and (f) 3rd SV. Notice the different magnitudes along the abscissas of the evolved SVs after 24 h. The ordinate represents the vertical level (σ) and the abscissa denotes the energy-weighted SVs (J kg-1).


Fig. 11. Vertical distributions of the energy-weighted SVs (J kg-1, TE; solid, kinetic energy; dashed, potential energy; dotted): (a) composite of 1st SV to 5th SV at 0 h and (b) composite of 1st SV to 5th SV at 24 h. The ordinate represents the vertical level (σ) and the abscissa denotes the energy-weighted SVs (J kg-1).

Fig. 12. Zonal wind components of (a) initial leading SV (m s-1), (b) linearly evolved leading SV, and (c) nonlinearly evolved leading SV. The boxes in (b) and (c) denote the verification region.


to 6 October 2008 (WMO, 2008). Various organ-izations have developed real-time strategies to sug-gest possible target regions for adaptive observations to improve tropical cyclone track forecasts. These in-clude TESV strategy from the Japan Meteorological Administration (JMA) (Yamaguchi et al., 2008), MM5 TESV strategy from Yonsei University (YSU) (Kim and Jung, 2008a, b), Ensemble Tansform Kalman Filter (ETKF) by the University of Miami (Majumdar, 2006), analysis of ensemble deep-layer mean (DLM) wind variance by NOAA (Aberson, 2003), TESV strategy by the Naval Research Laboratory (NRL) (Peng and Reynolds, 2006), and TESV strategy by the European Center for Medium- range Weather Forecasts (ECMWF) (Buizza et al., 2007). A few more real-time adaptive observation strategies are planned for the T-PARC period. These sensitivity products for adaptive observations will be collected by JMA, Earth Observing Laboratory (EOL) of National Center for Atmospheric Research (NCAR), and PREVention, Information and Early Warning (PREVIEW) of ECMWF, and used to de-cide on the target regions for adaptive observations of typhoons.

Adjoint-based sensitivities (i.e., adjoint-based forecast sensitivities and SVs) using MM5 and its tangent linear and adjoint models have been applied to high-impact weather events (e.g., typhoon, heavy rainfall and snowfall events associated with extra-tropical cyclones, sand and dust storm, etc.) in Korea (e.g., Kim and Jung, 2006; Kim et al., 2008; Kim and Jung, 2008a, b). Adjoint-based sensitivity products as presented in this article may contribute to T-PARC by suggesting sensitive regions for adaptive ob-servations of typhoons. One caution is that the input data for the real-time SV calculation for T-PARC are forecast fields instead of analyses as in this study, due to the preparation time necessary for real-time adap-tive observations as explained in Kim et al. (2004) and Majumdar et al. (2006). In addition to applica-tions to field experiments such as T-PARC, these strategies can be used to determine permanent or semi-permanent observational sites to improve fore-casts and predictability of high-impact weather events in Korea.

6. Summary and discussion

In this study, SVs are applied to Typhoon Rusa as an adaptive observation strategy to understand the sensitivity of the forecast with respect to the initial conditions, and thence to suggest the sensitive re-gions in terms of adaptive observations. Because Typhoon Rusa is nearing recurvature during the peri-od of this study, sensitive regions by SVs are located in the mid-latitude trough regions as well as the in-flow regions along the subtropical high, which is sim-ilar to the results in Peng and Reynolds (2006) and Kim and Jung (2008a).

Both the SVs in this study and the adjoint sensitiv-ities in Kim and Jung (2006) indicate that the inflow regions in the right semi circle of the storm are sensi-tive regions for Rusa. However, the sensitive regions in the mid-latitude trough predicted by SVs are not indicated by the adjoint sensitivities. This difference may be attributed to the different configurations of physical processes used for SV and adjoint sensi-tivity calculations (Kim and Jung, 2008b). In the present study, moist basic state and dry linear in-tegrations are used for SV calculations to detect large-scale influences on the evolution of Typhoon Rusa, such as the mid-latitude trough. In contrast, the adjoint sensitivities in Kim and Jung (2006) used moist basic state and linear integrations so that large sensitivities are concentrated near the storm center. Differences in the vertical distributions of the en-ergy-weighted SV in this study and adjoint sensitiv-ities also may be explained by different config-urations of physical processes for sensitivity calculations. Moist processes used in adjoint sensi-tivity calculations make the sensitivity maximum close to the lower boundary (i.e., close to the source of moisture) as indicated in Kim and Jung (2008b).

Some implications of this adaptive observation strategy for T-PARC and other applications have also been discussed. The results of this study are quite con-sistent with recent studies on targeted observation strategies of tropical cyclones and demonstrate that the TESVs can capture the signal of the environ-mental features affecting the evolution of Typhoon Rusa. The TESV guidance shown in this study will


be calculated and provided for real-time field experi-ments during T-PARC.

Acknowledgments. This study was supported by the principal project "Korea Enhanced Observing Period (KEOP)" of the National Institute of Meteorological Research of Korea Meteorological Administration, the Korea Meteorological Administration Research and Development Program under Grant CATER 2006-2102, and Yonsei University Research Fund of 2005. The authors wish to thank two anony-mous reviewers for their very helpful comments.

REFERENCES

Aberson, S. D., 2003: Targeted observations to improve operational tropical cyclone forecast guidance. Mon. Wea. Rev., 131, 1613‐1628.

_____, and B. J. Etherton, 2006: Targeting and data assim-ilation studies during Hurricane Humberto. J. Atmos. Sci., 62, 175‐186.

Badger, J., and B. J. Hoskins, 2001: Simple initial value problems and mechanisms for baroclinic growth. J. Atmos. Sci., 58, 38‐49.

Baker, N. L., and R. Daley, 2000: Observation and back-ground adjoint sensitivity in the adaptive observation‐targeting problem. Quart. J. Roy. Meteor. Soc., 126, 1431‐1454.

Bergot, T., 1999: Adaptive observations during FASTEX: A systematic survey of upstream flights. Quart. J. R. Meteor. Soc., 125, 3271‐3298.

_____, G. Hello, A. Joly, and S. Malardel, 1999: Adaptive observations: A feasibility study. Mon. Wea. Rev., 127, 743‐765.

Bishop, C. H., and Z. Toth, 1999: Ensemble transformation and adaptive observations. J. Atmos. Sci., 56, 1748‐1765.

_____, B. J. Etherton, and S. J. Majumdar, 2001: Adaptive sampling with the ensemble transform Kalman filter. Part I: Theoretical aspects. Mon. Wea. Rev., 129, 420‐436.

Buizza, R., 1994a: Sensitivity of optimal unstable structures. Quart. J. Roy. Meteor. Soc., 120, 429‐451.

_____, 1994b: Localization of optimal perturbations using a projection operator. Quart. J. Roy. Meteor. Soc., 120, 1647‐1681.

_____, and T. N. Palmer, 1995: The singular‐vector struc-ture of the atmospheric global circulation. J. Atmos. Sci., 52, 1434‐1456.

_____, and A. Montani, 1999: Targeting observations us-ing singular vectors. J. Atmos. Sci., 56, 2965‐2985.

_____, C. Cardinali, G. Kelly, and J.‐N. Thepaut, 2007: The value of observations. II: The value of observations lo-cated in singular‐vector‐based target areas. J. Atmos. Sci., 56, 2965‐2985.

Ehrendorfer, M., and R. M. Errico, 1995: Mesoscale pre-dictability and the spectrum of optimal perturbations. J. Atmos. Sci., 52, 3475‐3500.

Errico, R. M., 1997: What is an adjoint model? Bull. Amer. Meteor. Soc., 78, 2577‐2591.

_____, and K. D. Raeder, 1999: An examination of the ac-curacy of the linearization of a mesoscale model with moist physics. Quart. J. Roy. Meteor. Soc., 125, 169‐195.

Gelaro, R., R. H. Langland, G. D. Rohaly, and T. E. Rosmond, 1999: An assessment of the singular‐vector approach to targeted observing using the FASTEX dataset. Quart. J. Roy. Meteor. Soc., 125, 3299‐3327.

Hamill, T. M., and C. Snyder, 2002: Using improved back-ground‐error covariances from an ensemble Kalman fil-ter for adaptive observations. Mon. Wea. Rev., 130, 1552‐1572.

Kim, H. M., 2003: A computation of adjoint‐based sensi-tivities in a quasigeostrophic model. Korean J. Atmos. Sci., 6, 71‐83.

_____, and M. C. Morgan, 2002: Dependence of singular vector structure and evolution on the choice of norm. J. Atmos. Sci., 59, 3099‐3116.

_____, and B.‐J. Jung, 2006: Adjoint‐based forecast sensi-tivities of Typhoon Rusa. Geophys. Res. Lett., 33, L21813, doi:10.1029/2006GL027289.

_____, and _____, 2008a: Singular vector structure and evolution of a recurving tropical cyclone. Mon. Wea. Rev., doi:10.1175/2008MWR2643.1.

_____, and _____, 2008b: Influence of moist physics and norms on singular vectors for a tropical cyclone. sub-mitted to Mon. Wea. Rev.

_____, M. C. Morgan, and R. E. Morss, 2004: Evolution of analysis error and adjoint‐based sensitivities: Implications for adaptive observations. J. Atmos. Sci., 61, 795‐812.

_____, J. K. Kay, and B.‐J. Jung, 2008: Application of ad-joint‐based forecast sensitivities to Asian dust transport events in Korea. Water Air Soil Pollut., doi:10.1007/ s11270-008-9750-8.

Kleist, D. T., and M. C. Morgan, 2005: Interpretation of the structure and evolution of adjoint‐derived forecast sensitivity gradients. Mon. Wea. Rev., 133, 466‐484.

Langland, R. H., R. Gelaro, G. D. Rohaly, and M. A.


Shapiro, 1999: Targeted observations in FASTEX: Adjoint based targeting procedures and data impact ex-periments in IOP17 and IOP18. Quart. J. Roy. Meteor. Soc., 125, 3241‐3270.

Majumdar, S. J., S. D. Aberson, C. H. Bishop, R. Buizza, M. S. Peng, and C. A. Reynolds, 2006: A comparison of adaptive observing guidance for Atlantic tropical cyclones. Mon. Wea. Rev., 134, 2354‐2372.

Montani, A., A. J. Thorpe, R. Buizza, and P. Unden, 1999: Forecast skill of the ECMWF model using targeted ob-servations during FASTEX. Quart. J. Roy. Meteor. Soc., 125, 3219‐3240.

Morgan, M. C., 2001: A potential vorticity and wave activ-ity diagnosis of optimal perturbation evolution. J. Atmos. Sci., 58, 2518‐2544.

Palmer, T. N., R. Gelaro, J. Barkmeijer, and R. Buizza, 1998: Singular vectors, metrics and adaptive observations. J. Atmos. Sci., 55, 633‐653.

Peng, M. S., and C. A. Reynolds, 2005: Double trouble for typhoon forecasters. Geophys. Res. Lett., 32, L02810, doi:10.1029/2004GL021680.

_____, and _____, 2006: Sensitivity of tropical cyclone forecasts as revealed by singular vectors. J. Atmos. Sci., 63, 2508‐2528.

Pu, Z.‐X., and E. Kalnay, 1999: Targeting observations with the quasi‐inverse linear and adjoint NCEP global

models: Performance during FASTEX. Quart. J. R. Meteor. Soc., 125, 3329‐3337.

Reynolds, C. A., M. S. Peng, S. J. Majumdar, S. D. Aberson, C. H. Bishop, and R. Buizza, 2007: Interpretation of adaptive observing guidance for Atlantic tropical cyclones. Mon. Wea. Rev., 135, 4006‐4029.

Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate., 15, 1609‐1625.

WMO, 2004: THORPEX: A Global Atmospheric Research Programme. [Available online from http://www.wmo. int/thorpex/].

_____, 2008: THORPEX Progress Report. [Available on-line from http://www.wmo.int/thorpex/].

Wu, C. C., J.‐H. Chen, P.‐H. Lin, and K.‐H. Chou, 2007: Targeted observations of tropical cyclone movement based on the adjoint‐derived sensitivity steering vector. J. Atmos. Sci., 64, 2611‐2626.

Yamaguchi, M., T. Iriguchi, T. Nakazawa, and C.‐C. Wu, 2008: An observing system experiments for Typhoon CONSON (2004) using a singular vector method and DOTSTAR data. submitted to Mon. Wea. Rev.

Zou, X., F. Vandenberghe, M. Pondeca, and Y.‐H. Kuo, 1997: Introduction to adjoint techniques and the MM5 adjoint modeling system. NCAR Tech. Note NCAR/TN‐ 435STR, 110 pp.

Documents

EBQUJWF 0CTFSWBUJPO (VJEBODF QQMJFE UP 5ZQIPPO 3VTB … · (2007) have compared various adaptive observation strategies for Atlantic hurricanes. To prepare for the THORPEX-Pacific