mesoscale numerical weather predictionmesoscale models, mostly through case studies or model...

9
Mesoscale Numerical Weather Prediction Ying-Hwa Kuo National Center for Atmospheric Research Boulder, Colorado, U.S.A. 1. Introduction Mesoscale models, with grid resolution higher than synoptic and global models, and with advanced physical parameterizations, have been an important tool for meteorological research over the past twenty years. The research applications of mesoscale models, mostly through case studies or model sensitivity experiments in the 1980s, provided us with important physical insights into mesoscale weather systems such as the mesoscale environment of severe storms (Anthes et al 1982), tropical cyclones (Chang 1982), mesoscale convective systems (Zhang and Fritsch 1986), extratropical cyclones (Kuo and Reed 1988), and heavy rainfall events (Lee and Hong 1989). These mesoscale model applications typically use grid resolution ranging from 20 km to 100 km, generally within the range of grid resolution that does not violate the hydrostatic assumption. Because of the computational limitations, high-resolution mesoscale models were rarely used for real-time numerical weather prediction (NWP), and with a limited number of case studies, it is hard to know the performance and systematic behavior of these high-resolution models in day-to-day weather prediction. Since the early 1990s several important changes took place in mesoscale modeling. First is the introduction of nonhydrostatic dynamics into mesoscale models (e.g., Dudhia 1993). Without the restrictions of the hydrostatic assumption, nonhydrostatic mesoscale models can be run at cloud resolving resolutions (~1 km). This greatly increases the range of scientific problems to which the models can be applied. For example, at such resolution, mesoscale models can explicitly simulate convection and its interaction with the larger scale weather systems in a realistic setting (Kuo and Wang 1996). Another important change is the de-centralization of regional NWP efforts Corresponding author address: Ying-Hwa Kuo, Mesoscale Microscale Meteorology, National Center for Atmospheric Research, P. O. Box 3000 Boulder, CO 80307. E-mail: [email protected] (Mass and Kuo 1998). Real-time weather prediction used to be the sole privilege of a handful of major operational centers simply because of the tremendous amount of human and computational resources required to develop and operate a weather prediction model. This situation began to change in the early 1990s. Currently, there are dozens of groups across the United States that are performing real-time prediction on a regular basis, using a number of different models. For the efforts involving the use of MM5 model alone, there are over 20 real-time sites in the United States (Fig. 1). As pointed out by Mass and Kuo (1998), this transition occurred because of three important factors: (1) the availability of high-performance workstations (or even personal computers) at affordable prices, (2) the sharing of mesoscale models (such as MM5, RAMS, and ARPS) and model components (i.e., physical parameterizations) within the community, and (3) the real-time accessibility of analysis and forecast grid point data from the operational runs at the National Centers for Environmental Prediction (NCEP). Many of the regional NWP efforts have demonstrated significant skills over the operational models (Colle et al. 1999; Davis et al. 1999). Fig. 1. Real-time MM5 sites in the United States and other countries. (Courtesy of J. Bresch) Stimulated by the success of mesoscale NWP efforts, many operational centers around the world are also actively pursuing the development and operational use of nonhydrostatic mesoscale 122

Upload: others

Post on 22-Feb-2020

3 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Mesoscale Numerical Weather Predictionmesoscale models, mostly through case studies or model sensitivity experiments in the 1980s, provided us with important physical insights into

Mesoscale Numerical Weather Prediction

Ying-Hwa Kuo

National Center for Atmospheric Research Boulder, Colorado, U.S.A.

1. Introduction Mesoscale models, with grid resolution higher than synoptic and global models, and with advanced physical parameterizations, have been an important tool for meteorological research over the past twenty years. The research applications of mesoscale models, mostly through case studies or model sensitivity experiments in the 1980s, provided us with important physical insights into mesoscale weather systems such as the mesoscale environment of severe storms (Anthes et al 1982), tropical cyclones (Chang 1982), mesoscale convective systems (Zhang and Fritsch 1986), extratropical cyclones (Kuo and Reed 1988), and heavy rainfall events (Lee and Hong 1989). These mesoscale model applications typically use grid resolution ranging from 20 km to 100 km, generally within the range of grid resolution that does not violate the hydrostatic assumption. Because of the computational limitations, high-resolution mesoscale models were rarely used for real-time numerical weather prediction (NWP), and with a limited number of case studies, it is hard to know the performance and systematic behavior of these high-resolution models in day-to-day weather prediction. Since the early 1990s several important changes took place in mesoscale modeling. First is the introduction of nonhydrostatic dynamics into mesoscale models (e.g., Dudhia 1993). Without the restrictions of the hydrostatic assumption, nonhydrostatic mesoscale models can be run at cloud resolving resolutions (~1 km). This greatly increases the range of scientific problems to which the models can be applied. For example, at such resolution, mesoscale models can explicitly simulate convection and its interaction with the larger scale weather systems in a realistic setting (Kuo and Wang 1996). Another important change is the de-centralization of regional NWP efforts Corresponding author address: Ying-Hwa Kuo, Mesoscale Microscale Meteorology, National Center for Atmospheric Research, P. O. Box 3000 Boulder, CO 80307. E-mail: [email protected]

(Mass and Kuo 1998). Real-time weather prediction used to be the sole privilege of a handful of major operational centers simply because of the tremendous amount of human and computational resources required to develop and operate a weather prediction model. This situation began to change in the early 1990s. Currently, there are dozens of groups across the United States that are performing real-time prediction on a regular basis, using a number of different models. For the efforts involving the use of MM5 model alone, there are over 20 real-time sites in the United States (Fig. 1). As pointed out by Mass and Kuo (1998), this transition occurred because of three important factors: (1) the availability of high-performance workstations (or even personal computers) at affordable prices, (2) the sharing of mesoscale models (such as MM5, RAMS, and ARPS) and model components (i.e., physical parameterizations) within the community, and (3) the real-time accessibility of analysis and forecast grid point data from the operational runs at the National Centers for Environmental Prediction (NCEP). Many of the regional NWP efforts have demonstrated significant skills over the operational models (Colle et al. 1999; Davis et al. 1999).

Fig. 1. Real-time MM5 sites in the United States

and other countries. (Courtesy of J. Bresch) Stimulated by the success of mesoscale NWP efforts, many operational centers around the world are also actively pursuing the development and operational use of nonhydrostatic mesoscale

122

Page 2: Mesoscale Numerical Weather Predictionmesoscale models, mostly through case studies or model sensitivity experiments in the 1980s, provided us with important physical insights into

models. We are at a point in the history of meteorology where mesoscale modeling is experiencing the most rapid growth. In this paper, I will review the current status of mesoscale NWP and the lessons learned from these efforts. I will also discuss issues that are facing the field, and provide suggestions for future research. 2. Selected examples of mesoscale NWP in the U.S. The use of a mesoscale model for real-time numerical weather prediction was first attempted at The Pennsylvania State University (Warner and Seaman 1990) using the PSU/NCAR MM4 model. This was soon followed by similar efforts at Colorado State University (Cotton et al. 1994), the

University of Utah (Horel and Gibson 1994), the University of Washington, and the Kennedy Space Flight Center (Manobianco et al. 1996), using the Regional Atmospheric Modeling System (RAMS), the Utah Local Area Model (ULAM), and the Mesoscale Atmospheric Simulation System (MASS) model. The field of mesoscale NWP has evolved rapidly over the past fifteen years. As of today, there are over three dozen groups in the United States alone that are running mesoscale models for regional prediction at least once a day. With a large number of real-time mesoscale NWP efforts, it is not possible to perform a comprehensive review of all activities. In this section, we will provide a few examples of real-time mesoscale forecast systems (see Table 1) as a way of illustrating the current state of the art.

Table 1. Selected examples of real-time/operational mesoscale numerical weather prediction efforts

Institution Modeling system

Resolution(km)

Vert. Levels

Domain Period (h)

Initializa-tion

Start Times (UTC)

NCEP ETA 12 60 North America

60/84 EDAS3 00/06/12/18

NCEP NMM 8 60 Regional 48 ETA 00/06/12/18

MMM/ NCAR

WRF 10 35 CONUS 48 ETA 0000/1200

FNMOC1 COAMP 81/27 45/15/5

30 North America

24-72 NOGAPS + DA

0000/ 1200

University Washington

MM5 36/12/4 37 Pacific Northwest

72 Cold start – GFS/ETA

0000/ 1200

University Wisconsin

NMS (Tripoli)

60/30 35 CONUS Midwest

48 ETA 0000/ 1200

NCSU MASS 45/15 25 U.S./N.C.

36 Cold start – Early ETA

0000

University Oklahoma

ARPS 27/ 9/3 50 CONUS/ S. Plains

48/12 ETA + obs + radar

0000/ 1200

MMM/ NCAR

MM5 90/30/10/ 3.3

29 Antarctic 72/36 GFS + obs 0000/ 1200

Kennedy Space C.

RAMS 60/15/5/1.25

36 Florida 24 ETA 12-h FCST + obs

0000 1200

CAA4 Taiwan

MM5 45/15/5 31 Taiwan/ E. Asia

24-72 3DVAR 3-h update cycle

ATEC2 MM5 30/10/3.3/1.1

31 Utah 24 ETA + FDDA

0000/ 1200

-- 1. FNMOC -- Fleet Numerical Meteorology and Oceanography Center 2. ATEC -- Army Test and Evaluation Center 3. EDAS – ETA data analysis system 4. CAA – Civil Aeronautics Administration

123

Page 3: Mesoscale Numerical Weather Predictionmesoscale models, mostly through case studies or model sensitivity experiments in the 1980s, provided us with important physical insights into

The real-time forecast system at the University of Washington (UW) was first established in 1995 (http://www.atmos.washington.edu/mm5rt/). This system initially operated on a single processor DEC workstation, running a single domain MM5 at 27-km resolution over the west coast. In 1996, with the upgrade to a 14-processor SUN Ultra Enterprise 4000 computer the system was moved to a 36/12km domain configuration. In 1997 with another upgrade of processors, the UW system was augmented with a 4-km domain over western Washington (WA). Two years later, with the addition of a Compaq ES-40 the 4-km domain was enlarged to include all of western WA. Moreover, four additional ensemble runs were added. In 2000, this was augmented by a 30-processor SUN E6500 workstation. Two additional Linux PC clusters (with 20 and 32 processors) were added by 2002. With enhanced computing power, UW is now performing two sets (with 17 members and 8 members, respectively) of mesoscale ensemble forecasting (Grimit and Eckel 2003). The ensemble members vary in initial conditions, boundary conditions and physics options. Their results show that the ensemble mean possesses better skill than individual members, and the ensemble spread provides good estimates of probability forecast and forecasts of model skills (Grimit and Mass 2002).

The current domain for the UW system is shown in Fig. 2. Another example is the real-time local forecast system with a 1-km grid developed by NCAR for the Army Test and Evaluation Command (ATEC) at the Dugway Proving Ground (DPG) in west-central Utah. This system was developed in support of the U.S. Army’s forecasting of local circulations driven by complex terrain and other variations in land-surface characteristics (Davis et al. 1999). The 1-km resolution was achieved by employing four levels of nesting, with grid resolution varying from 30 km to 1.1 km (see Table 1). Such a high horizontal resolution is required to properly account for the nature of the terrain and land-surface variations, which are responsible for the development of local circulations over the region. To our knowledge, this resolution is finer than that being used in any current real-time or operational weather prediction system (except for the system developed at the Kennedy Space Center, Case et al 2002). The model configuration for the ATEC system is shown in Fig. 3. For detailed information of this site please refer to: http://140.196.88.2/cgi-bin/model/rtfdda_ugui

Fig. 2. Domain of the University of Washington real-time forecast system. The two boxes indicate the 12-km and 4-km domains, respectively. (From Colle et al. 2000)

124

Page 4: Mesoscale Numerical Weather Predictionmesoscale models, mostly through case studies or model sensitivity experiments in the 1980s, provided us with important physical insights into

Fig. 3. Configuration of the real-time MM5 forecast system of US Army Test and Evaluation Center at the

Dugway Proving Ground in Utah. (From Davis et al. 1999) The use of high-resolution mesoscale models for real-time NWP requires a tremendous amount of computational resources. For example, to increase the grid resolution (both horizontally and vertically) of a model by a factor of two would require a factor of 16 increase in computational resources. One might ask: “What have we learned from these high-resolution mesoscale NWP efforts? Do higher-resolution mesoscale models really produce better forecasts that are statistically significant?” To properly answer these questions requires careful statistical evaluation of a large number of forecasts (Davis and Carr 2000). Despite the blossoming of mesoscale NWP efforts, careful statistical evaluation and verification of large numbers of real-time predictions have been rare. Recently, several studies have been devoted to such efforts (Colle et al. 1999; Davis et al. 1999; Colle and Mass 2000; Colle et al. 2002; Hou et al. 2001; Case 2002). The results of these verification studies have produced a number of interesting conclusions:

(i) High resolution mesoscale models demonstrated considerable skill in predicting local circulations driven by local topography and land-surface variations (Davis et al. 1999) Such forecasts were often missed or not resolved by coarse resolution operational models.

(ii) A noticeable improvement in the cold-season precipitation forecast in mountainous regions was found in almost all skill scores as the model horizontal resolution was increased from 36 km to 12 km (Colle et al. 1999). However, subsequent improvement was not found for model resolution increasing from 12 km to 4 km, except for heavy precipitation events (Fig. 4). Further analysis shows that the model tended to overpredict precipitation on the windward slopes and underpredict in the lee of major barriers (Colle et al. 2000). Careful case studies and sensitivity tests indicate that further improvements in microphysical schemes are required in order to more accurately predict precipitation.

125

Page 5: Mesoscale Numerical Weather Predictionmesoscale models, mostly through case studies or model sensitivity experiments in the 1980s, provided us with important physical insights into

(iii) The skill of the model in mesoscale prediction is strongly affected by the quality of its large-scale forecasts (which are, in turn, strongly affected by the initial and boundary conditions provided by the larger scale models). For example, Colle et al. (2000) showed that the model precipitation forecast skill increased considerably if less accurate large-scale forecasts for the verification period were eliminated.

(iv) Ensemble forecast experiments at UW have indicated that the ensemble mean often possesses higher forecast skills than individual members of the ensemble (Grmit and Mass 2002). Also, the ensemble spread provides useful prediction of model forecast skills (Fig. 5). Results from SAMEX (Storm and Mesoscale Ensemble Experiment) indicated that an ensemble of multiple forecast systems possesses considerable skill (Hou et al. 2001). Moreover, variations to model physical parameterizations, as well as the use of boundary condition perturbations consistent with those applied to the interior initial state, are important for regional ensemble forecasting (Stensrud et al. 2000).

Fig. 4. The 24-h bias scores for the 36-, 12-, and 4-

km domains from 1 Jan 1998 through 15 Mar 1998 and 1 Oct 1998 through 8 Mar 1999. Note, there is a large jump between 1.9 and 3.0 inch in precipitation threshold. (From Colle et al. 2000)

3. Outlook for the future With the continued advance in computer technology, we can expect that more computational resources will be available to real-time and

operational mesoscale NWP. This would allow us to run mesoscale weather prediction models at increasing grid resolutions. But, will we see continued improvement in the model skill as the grid resolution is progressively increased? The answer is probably a cautious “yes”. As already shown by the verification studies performed by the UW group, higher grid resolution does not necessarily produce more accurate precipitation forecasts. Model grid resolution is not the only factor (and perhaps, not even the most important factor) that affects the accuracy of mesoscale NWP. If we are to continue to make progress, we need to devote considerable effort to, at least, the following areas: a. Improvement in model physics As the grid resolution of a model is increased, a mesoscale model becomes progressively sensitive to internal/external forcings, which are represented by the model’s physical parameterizations. At present, almost all the model physical parameterizations require some level of improvement. This includes cloud microphysics, PBL physics, atmospheric radiation, land-surface physics, land-use, and soil moisture treatment. For example, Pielke et al (1997) showed that cloud-scale prediction is highly sensitive to the specification of surface land use. Ayotte et al. (1996) showed that almost all existing PBL models have considerable deficiency in modeling the turbulence transport and entrainment at the top of the PBL. Recent cloud-resolving mesoscale model simulations have demonstrated great sensitivity to cloud microphysics (Weisman 2003, personal communications). b. Better use of observations It is well known that the traditional synoptic-scale radiosonde network is grossly inadequate for mesoscale model initialization. To improve the quality of the initial condition for a mesoscale model, we need to take innovative approaches (such as three-dimensional and four-dimensional variational data assimilation – 3DVAR/4DVAR) to assimilating non-traditional observations, such as wind profilers, ground-based GPS water vapor measurements, surface observations, surface rainfall data, satellite radiance, water vapor winds, GPS/MET radio occultation, cloud-track winds, and Doppler radars, into the model. Guo et al. (2000) showed that the use of wind profiler, surface

126

Page 6: Mesoscale Numerical Weather Predictionmesoscale models, mostly through case studies or model sensitivity experiments in the 1980s, provided us with important physical insights into

Fig. 5. Averaged ensemble mean MAE (mean absolute error) for surface wind direction prediction. The

figure shows that the ensemble mean has considerably lower MAE when there is a small spread among the ensemble members, and the MAE is large when the spread is high. (From Grimit and Mass 2002)

observations, surface rainfall and ground-based GPS data improved the prediction of a mesoscale convective system. Sun and Crook (2001) demonstrated considerable success in storm scale prediction with 4DVAR assimilation of single Doppler radar data. c. Ensemble forecasting on the mesoscale Given the considerable uncertainties in the model initial condition and in the model itself, it is desirable to attempt the ensemble forecasting on the mesoscale. Recently, Wandishin et al. (2001) experimented with an ensemble set consisting of five members from the NCEP’s Regional Spectral Model and 10 members from the 80-km ETA model. They showed that ensemble configurations with as few as five members could significantly outperform the higher-resolution 29-km Meso ETA model in precipitation forecasts. Clearly, a significant amount of work is still required in designing and testing an optimal set of ensemble members. Also, more work is required to perform ensemble forecasting at the true mesoscale resolution (below 10 km). d. Real-time verification of mesoscale prediction Careful model verification is extremely important for us to understand the performance of a model and to identify areas for future model development. There are two major challenges to the verification of mesoscale numerical weather prediction. First, there is a lack of mesoscale observations for mesoscale model verification. Secondly, traditional verification methods, based on standard observations (soundings and surface), or based on the instantaneous comparison of analyzed and predicted fields, may not yield useful information.

Davis and Carr (2000) have provided a number of interesting suggestions for mesoscale model verification. e. The WRF model Through a collaborative partnership, principally among NCAR, NOAA, DoD, OU/CAPS, FAA, and universities, the mesoscale meteorological community is developing a next-generation mesoscale model, known as the Weather Research and Forecasting (WRF) model. The goal of the WRF project is to develop an advanced mesoscale forecast and data assimilation system, and to accelerate research advances into operations. The WRF model integrates the fully compressible nonhydrostatic equations of motion. It uses a two level, 3rd order Runge-Kutta (RK3) split-explicit time integration scheme (Wicker and Skamarock 2002), and a 5th order upwind advection scheme. For details on the WRF project, please refer to the web site: http://www.wrf-model.org/. With advanced numerics, the WRF model exhibits low dispersion errors and allows a larger time step, compared to either lower order Runge-Kutta schemes or the popular leapfrog time integration schemes. As a result, the model is capable of capturing small-scale details in its precipitation forecasts. As an example, we show in Figure 6 the 3-h accumulated precipitation from a 4-km observational analysis (based on radar and rain gage), and the WRF and operational NCEP models from 12 UTC 4 June 2002 runs, valid at 1800 UTC. It is interesting to note that the 22-km WRF model possesses more small-scale precipitation details than NCEP models at 8-km or 12-km resolutions.

127

Page 7: Mesoscale Numerical Weather Predictionmesoscale models, mostly through case studies or model sensitivity experiments in the 1980s, provided us with important physical insights into

Fig. 6. Three-hour accumulated precipitation from a 4-km observational analysis (based on radar and rain

gage), 10-km WRF, 22-km WRF, and 8-km NMM (National Mesoscale Model, based on meso-ETA model), and the 12-km ETA. (From Baldwin and Wandishin 2002).

128

Fig. 7. Spectral analysis of variance of 3-h accumulated precipitation for 15-18 UTC 4 June 2002 (From Baldwin and Wandishin 2002).

The corresponding spectral analysis (Fig. 7) shows that the 10- and 22-km WRF model forecasts maintain the variance in the precipitation field down to at least four times the horizontal grid spacing. By comparison, the variance for the operational ETA models drop off much sooner, at greater than ten times the grid spacing (Baldwin and Wandishin 2002). The WRF model provides a focal point and common modeling framework for the modeling community to collaborate on the development effort. With continued improvements in model physics, data assimilation techniques, ensemble forecasting and model verification, we can expect continued advances in mesoscale numerical weather prediction.

Page 8: Mesoscale Numerical Weather Predictionmesoscale models, mostly through case studies or model sensitivity experiments in the 1980s, provided us with important physical insights into

129

Acknowledgement: The author thanks the following people for their generous contributions of information, figures, comments, discussions, and statistics: Mark Albright, Jim Bresch, Brian Colle, Chris Davis, Geoff DiMego, Kevin Droegemeier, Jimy Dudhia, Eric Grimit, Rich Hodur, Joe Klemp, Kevin Manning, John Manobianco, Cliff Mass, Roger Pielke, Ananthakrishna Sarma, Bill Skamarock, Jenny Sun, Wei Wang, Tom Warner, Ted Yamada, and Sayuri Yamada. References Anthes, R. A., Y.-H. Kuo, S. G. Benjamin, Y.-F.

Li, 1982: The evolution of the mesoscale environment of severe local storms: Preliminary modeling results. Mon. Wea. Rev., 110, 1187-1213.

Ayotte, K. W., P. P. Sullivan, A. Andren, S. C. Doney, A. A. M. Holtslag, W. G. Large, J. C. McWilliams, C.-H. Moeng, M. J. Otte, J. J. Tribbia, and J. C. Wygaard, 1996: An evaluation of neutral and convective planetary boundary parameterization relative to large eddy simulations. Boudary-Layer Meteorology, 79, 131-175.

Baldwin, M., and M. Wandishin, 2002: Determining the resolved spatial scales of ETA model precipitation forecasts. Preprints, AMS 15th Conf. on Numerical Weather Prediction, 12-16 Aug. 2002, San Antonia, Texas, 85-88.

Case, J. L., J. Manobianco, A. V. Dianic, M. M. Wheeler, D. E. Harms, C. R. Parks, 2002, Verification of high-resolution RAMS forecasts over East-Central Florida during the 1999 and 2000 summer months. Wea. Forecasting, 17, 1133-1151.

Chang, S. W., 1982: The orographic effects induced by an island mountain range on propagating tropical cyclones. Mon. Wea. Rev., 110, 1255-1270.

Colle, B. A., K. J. Westrick, C. F. Mass, 1999: Evaluation of MM5 and Eta-10 precipitation forecasts over the Pacific Northwest during the cold season. Wea. Forecasting, 14, 137-154.

Colle, B. A., C. F. Mass, 2000: The 5-9 February 1996 flooding event over the Pacific Northwest: Sensitivity studies and evaluation of the MM5 precipitation forecasts. Mon. Wea. Rev., 128, 593-617.

Colle, B. A., C. F. Mass, K. J. Westrick, 2000: MM5 precipitation verification over the

Pacific Northwest during the 1997-99 cool seasons. Wea. Forecasting, 15, 730-744.

Cotton, W. R., G. Thompson, and P. W. Mielke Jr., 1994: Real-time mesoscale prediction on workstations. Bull. Amer. Meteor. Soc., 75, 349-362.

Davis, C. and F. Carr, 2000: Summary of the 1998 Workshop on Mesoscale Model Verification. Bull. Amer. Meteor. Soc., 81, 809-819.

Davis, C., T. Warner, E. Astling, and J. Bowers, 2000: Development and application of an operational, relocatable, mesogamma-scale weather analysis and forecasting system. Tellus, 51A, 710-727.

Dudhia, J., 1993: A nonhydrostatic version of the Penn State-NCAR mesoscale model: Validation tests and simulation of an Atlantic cyclone and cold front. Mon. Wea. Rev., 121, 1493-1513.

Grimit, E. P. and F. A. Eckel, 2003: A basic guide to understanding short-range ensemble forecasts at the University of Washington; [Available from http://www.atmos.washington.edu/~emm5rt/]

Grimit, E. P. and C. F. Mass, 2002: Initial results of a mesoscale short-range ensemble forecasting system over the Pacific Northwest. Wea. Forecasting, 17, 192-205

Guo, Y.-R., Y.-H. Kuo, J. Dudhia, D. Parsons, and C. Rocken, 2000: Four-dimensional variational data assimilation of heterogeneous mesoscale observations for a strong convective case. Mon. Wea. Rev., 128, 619-643.

Horel, J. D., and C. V. Gibson, 1994: Analysis and simulation of a winter storm over Utah. Wea. Forecasting, 9, 479-507.

Hou, D., E. Kalnay, K. K. Drgemeier, 2001: Objective verification of the SAMEX ’98 ensemble forecasts. Mon. Wea. Rev., 129, 73-91.

Kuo, Y.-H., and R. J. Reed, 1988: Numerical simulation of an explosively deepening cyclone in the Eastern Pacific. Mon. Wea. Rev., 116, 2081-2105.

Kuo, Y.-H., and W. Wang, 1996: Simulation of a prefrontal rainband observed in TAMEX IOP 13. Preprints, Seventh Conference on Mesoscale Processes. Reading, U.K., 9-13 September 1996, 335-338.

Lee, D.-K., and S.-Y. Hong, 1989: Numerical experiments of the heavy rainfall event that occurred over Korea during 1-3

Page 9: Mesoscale Numerical Weather Predictionmesoscale models, mostly through case studies or model sensitivity experiments in the 1980s, provided us with important physical insights into

September 1984. J. of Korean Meteor. Soc., 25, 233-260.

Manobianco, J., J. W. Zack, and G. E. Taylor, 1996: Workstation-based real-time mesoscale modeling designed for weather support to operations at the Kennedy Space Center and Cape Canaveral Air Station. Bull. Amer. Meteor. Soc., 77, 653-672.

Mass, C., and Y.-H. Kuo, 1998: Regional real-time numerical weather prediction: Current status and future potential. Bull. Amer. Meteor. Soc., 79, 253-263.

Pielke, R. A., T. L. Lee, J. H., Copeland, J. L. Eastman, C. L. Ziegler, and C. A. Finley, 1997: Use of USGS-provided data to improve weather and climate simulations. Ecological Applications, 7, 3-21.

Stensrud, D. J., J.-W. Bao, and T. T. Warner, 2000: Using initial condition and model physics perturbations in short-range ensemble simulations of mesoscale convective systems. Mon. Wea. Rev., 128, 2077-2107.

Sun, J., and N. A. Crook, 2001: Real-time low-level wind and temperature analysis using WSR-88D data. Wea. Forecasting, 16, 117-132.

Warner, T. W., and N. L. Seaman, 1990: A real-time mesoscale numerical weather prediction system used for research, teaching, and public service at The Pennsylvania State University. Bull. Amer. Meteor. Soc., 71, 792-805.

Wandishin, M. S., S. L. Mullen, D. J. Stensrud, H. E. Brooks, 2001: Evaluation of a short-range multi-model ensemble system. Mon. Wea. Rev. 129, 729-747.

Wicker, L. J., and W. C. Skamarock, 2002: Time splitting methods for elastic models using forward time schemes. Mon. Wea. Rev., 130, 2088-2097.

Zhang, D.-L., and J. M. Fritsch, 1986: A case study of the sensitivity of numerical simulation of mesoscale convective systems to varying initial conditions. Mon. Wea. Rev., 114, 2418-2431.

130