E.
Numerical Weather Prediction 1. Background Numerical weather prediction at RAP involves developing, testing, and implementing operational forecasting systems for new areas of the world. This entails better understanding local dynamical processes through the use of special data, development of improved representations of physical-processes in the models, and objectively verifying the skill of the models at predicting local meteorological processes. The meteorological modeling systems are often coupled with other models such as transport and dispersion models and sound-propagation models. An example of a project is one in which operational mesogamma-scale numerical forecast systems are being developed for and installed at test ranges of the U. S. Army Test and Evaluation Command. The systems, called the 4DWX systems, have been installed at Dugway Proving Ground, Utah; White Sands Missile Range, New Mexico; Aberdeen Test Center, Maryland; Yuma Proving Ground, Arizona; and Cold Regions Test Center, Alaska. The forecast systems include data ingest and quality control modules, display systems for data and model forecasts, and the Penn State/NCAR MM5 modeling system that has been adapted for range requirements. Scientific studies that have been conducted while developing and testing these systems include Davis et al. (1999), Warner and Hsu (2000) and Rife et al. (2002). References Davis, C., T. Warner, J. Bowers, and E. Astling, 1999: Development and application of an operational, relocatable, mesogamma-scale weather analysis and forecasting system. Tellus, 51A, 710-727. Warner, T. T. and H. M. Hsu, 2000: Nested-model simulation of moist convection: The impact of coarse-grid parameterized convection on fine-grid resolved convection through lateral-boundary-condition effects. Mon. Wea. Rev., 128, 2211-2231. Rife, D. L., T. T. Warner, F. Chen, and E. A. Astling, 2002: Mechanisms for diurnal boundary-layer circulations in the Great Basin Desert. Mon. Wea. Rev., 130, 921-938. 2. Modeling system development for the Army Cold Regions Test Center, Alaska The newest operational ATEC modeling system to be developed is for the Army's Cold Regions Test Center in Greely, Alaska. This location has especially interesting meteorology because it is located on the north slope of the Alaska Range, and experiences katabatic wind flow for much of the winter. The system of four nested grids is shown in Figure 1 where the smallest grid has a grid increment of 1.1 km. Twelve-hour forecasts are produced every three hours. Even though this modeling system is being used at near polar latitudes, the objective skill statistics are comparable to the statistics for the ATEC operational modeling systems being used in mid-latitudes. To illustrate the climatology of the low-level wind field, Figure 2 illustrates the wind-direction frequency and the temperature associated with the different directions during the winter months. The katabatic flow from the mountains to the south is obvious, as is the fact that the temperatures are higher when the wind is downslope. This work is being performed by Rong Sheu and Tom Warner.
Figure 1. Grids used for the operational modeling system designed for the Army Cold Regions Test Center, Greely, Alaska.
Figure 2. The near-surface wind and temperature climatology for Fort Greely, Alaska. The length of the solid lines indicate the frequency of the wind from each direction, while the length of the cone indicates the average speed. The color of the cone shows the average temperature (C) when the wind is from each direction. [TOP] 3. Modeling support for airborne toxic-hazard prediction at the Salt Lake City Winter Olympics There are many potential homeland-defense applications for rapidly deployable mesoscale models. One application is to provide meteorological input to transport and dispersion models in order to respond to, or plan for, the use of weapons of mass destruction (WMD) that release radiological, biological, or chemical material into the atmosphere. The ATEC modeling system was used during the Salt Lake City Winter Olympics to provide mesoscale wind fields to emergency-response organizations charged with assessing the possible impact of the release of airborne hazardous material. Figure 3 shows the model grid system used for the Salt Lake City area, with the finest grid having 1.3 km resolution. There was much complex terrain in the area that imposed fine structures on the wind field, thus justifying the use of such high resolution. The meteorological model was coupled to the SCIPUFF/HPAC transport and dispersion model, and users had a graphical interface that allowed the convenient running of different release scenarios. Figure 5 shows a 2-h forecast of the streamlines of the low-level wind in the valleys, and the simulated dosage and concentration one hour after the release of a gas from the shore of the Great Salt Lake, north of the city. Figure 4 shows a perspective view of the gas after it has spread southward through the valley. Upward motion resulting from the daytime heating over the Wasatch Mountains to the east of the Salt Lake Valley transported the gas on the east side of the valley to higher elevations. This work was performed by Yubao Liu, Laurie Carson, Becky Ruttenberg, Daran Rife, Carter Borst, Steve Webb, Doug Lindholm, Rong Sheu, Hank Fisher, David Hahn, Hsiao-Ming Hsu, Simon Low-Nam, Scott Swerdlin and Tom Warner.
Figure 3. Computational grids employed for Salt Lake City Olympics counter-terrorism support. The grid increments were 36, 12, 4, and 1.33 km, from the outer to the inner grids, respectively. The expanded inner grid shows topography, with the darkness of the shading proportional to elevation, the shore of the Great Salt Lake (heavy line), and Olympic event venues (number and letter codes).
Figure 4. The Salt Lake Valley, showing streamlines (yellow) of the forecast 1300 LT 10 December 2001 low-level winds, and the concentration (green) and dosage (blue) of a neutrally buoyant gas released one hour earlier near the shore of the Great Salt Lake. Higher terrain elevations are shown in gray-white shades, and the streamlines end where the plotting level merges with the terrain. The SCIPUFF model was used to calculate the transport and dispersion of the plume.
Figure 5. A perspective view of the Salt Lake Valley from the south, showing streamlines (yellow) of the forecast 1300 LT 10 December 2001 low-level winds, and the concentration (green, blue) of a neutrally buoyant gas released one hour earlier near the shore of the Great Salt Lake. Higher terrain elevations are shown in gray-white shades, and the streamlines end where the plotting level merges with the terrain. The SCIPUFF model was used to calculate the transport and dispersion of the plume. The daytime upslope flow over the Wasatch Mountains to the east of the valley is elevating the plume. [TOP]
Improved verification procedures for high-resolution meteorological models need to be developed. One reason is that high horizontal resolution is employed for the explicit purpose of resolving fine-scale meteorological features. But, if the features are misplaced slightly by the model, objective skill statistics such as root-mean-square-error (RMSE) indicate a poorer-quality forecast than if no feature had been forecast at all. The schematic in Figure 6 illustrates this concept with an example of a wind jet, where the model forecasts the jet to be offset relative to the position of the observed jet. A forecast with no jet will have a smaller RMSE than will a forecast with an offset jet.
Figure 6. Schematic showing the "penalty" that results from the production of a sharper forecast, when the forecast feature is displaced from its true position.
Consider this problem in the context of the forecasts for the Salt Lake City Winter Olympics discussed earlier. Figure 7 shows the Mean Absolute Error (MAE) of the MM5-forecast wind direction for different forecast lead times, and for different times of the day. Also shown are the MAE scores for three NWS models, all with considerably coarser horizontal resolution. The 4DWX and MM5 systems initiate forecasts every 3h, for 6-h time periods. Thus the verification statistics are plotted as a continuous line for the MM5 and RUC models. For Eta and the GFM, output was only available for forecasts initialized at 00 UTC and 12 UTC, so 6-h forecasts appear as only two points. For all forecast lead times, and for all times of the time day, the 4DWX system's skill was better than those of the NWS models. However, given the one to two orders-of-magnitude difference in the horizontal resolution, between MM5 and the other models, the improvement in skill is surprisingly modest. In contrast to these objective statistics, Figs. 8 - 11 provide a subjective impression of the forecast skill for the area of the Salt Lake Valley. For locations not close to the mountain slopes, the early morning winds are oriented down the valley toward the lake and the evening winds are oriented up the valley. The observations near the mountains show nocturnal drainage and afternoon upslope flow. The 1700 LT MM5-model 10-m AGL wind-direction climatology, for the locations of the observations, is shown in Figure 10, where a definite up-valley flow is evident. The station closest to the Wasatch Mountains in the east shows upslope flow. For the Eta model, Figure 11 shows the valley winds to have little directional preference, and to be from the wrong direction for this time in some cases. The paradox being addressed in this model-verification effort is related to why the subjective superiority of MM5 is not reflected to a greater degree in the statistics. This work is being performed by Daran Rife and Tom Warner.
Figure 7. The mean absolute 10-m wind-direction error for the models indicated, at the locations of observations on the inner grid (shown in Fig. 1) covering the Salt Lake Valley.
Figure 8. Wind-direction climatology, for 10 m AGL, at 0500 LT for an approximately 90-day period surrounding the Salt Lake City 2002 Winter Olympics.
Figure 9. Wind-direction climatology, for 10 m AGL, at 1700 LT for an approximately 90-day period surrounding the Salt Lake City 2002 Winter Olympics.
Figure 10. MM5 12-h forecast, wind-direction climatology, for 10 m AGL, at 1700 LT for an approximately 90-day period surrounding the Salt Lake City 2002 Winter Olympics.
Figure 11. NWS Eta model, 12-h forecast, wind-direction climatology, for 10 m AGL, at 1700 LT for an approximately 90-day period surrounding the Salt Lake City 2002 Winter Olympics. [TOP] 5. Modeling support for a Navy missile test in the Pacific A tropical application of the 4DWX system resulted when the White Sands Missile Range was tasked with providing meteorological support for a missile launch from the Pacific Missile Range Facility (PMRF) on Kauai, Hawaii. For the test, White Sands meteorologists forecasted weather conditions at the launch location and for downrange areas. Because the 4DWX model products are available through a web-based interface, it was possible for forecasters at White Sands and PMRF staff to simultaneously view the model-forecast products. Figure 12 shows an example of the web interface, with a display of the forecast low-level winds in the area of the island of Kauai. Substantial terrain and thermal influences are apparent. This work was performed by Yubao Liu, Laurie Carson, Becky Ruttenberg, Daran Rife, Scott Swerdlin and Tom Warner.
Figure 12. The web interface to an MM5 forecast of the low-level winds and specific humidity around the island of Kauai, produced in support of a missile test at the Pacific Missile Range facility on the island. [TOP] 6. Modeling support for ground-operations in Afghanistan Military surface operations can benefit from high-resolution operational weather forecasts in a variety of ways, including the use of precipitation forecasts for soil trafficability estimation, cloud and fog forecasts for assessing the usefulness of electro-optical weapons systems, and wind forecasts for conducting precision air drops and for calculating the transport of hazardous material and obscurants released on the battlefield. One example of the use of the 4DWX model in a rapidly evolving situation was its employment in support of ground operations in Afghanistan in late 2001. The National Ground Intelligence Center (NGIC) required high-resolution meteorological forecasts for different operational areas in order to assess the consequences of the possible release of hazardous material. For this application, the meteorological model had an interface to a DOD transport and dispersion model - the Second-order Closure Integrated PUFF (SCIPUFF) model that is part of the Hazard Prediction and Assessment Capability (HPAC) system. Figure 13 shows the model grid configuration when the system was used for support of operations in the central part of the country, and Figure 14 is an example of a national precipitation forecast. During the period of model support, the grid needed to be moved on short notice. The grid-spacing was 3.3 km for the highest-resolution domain. Feedback from NGIC indicated that the very high resolution forecast winds played an important role in their consequence assessment for hazardous material releases. This work was performed by Yubao Liu, Laurie Carson, Becky Ruttenberg, Daran Rife, Scott Swerdlin and Tom Warner.
Figure 13. Afghan-grids. Nested system of four computational grids, where the grid increments were 3.3, 10, 30, and 90 km, from the inner to outer grids, respectively. Terrain elevation is color coded. Operational forecasts from the grid covering Afghanistan were used for overall guidance about the country's weather, while the inner grid was used to provide high-resolution meteorological data to a transport and dispersion model.
Figure 14. Afghan-precip. Forecast precipitation rates in millimeters per hour from the 10 km grid-increment grid shown in Fig. 2, from a forecast during November 1991. [TOP] 7. Modeling the near-surface flow over the Southern California Bight To assess the impact of the wind stress from high-resolution atmospheric mesoscale models on the coastal circulation calculated from ocean numerical models, a set of 3-month (March - May 1999) simulations using COAMPS (Coupled Ocean/Atmosphere Mesoscale Prediction System) was conducted. A triply nested grid configuration, with grid increments of 81, 27 and 9 km, was located over the Southern California Bight (SCB). The average of the 10-m vertical vorticity over the three-month period is shown in Figure 15 for the innermost domain (grid increment of 9 km). A broad area of cyclonic vorticity over SCB and a narrow band of cyclonic vorticity north of Point Conception were apparent. Similarly, vorticity bands were also found to the north of Monterey Bay and to the south of San Diego along the coast. The near-surface vorticity indicates the existence of the wind-stress curl. Local upwelling (downwelling) is induced if the wind stress curl is cyclonic (anti-cyclonic). Furthermore, the band-shaped cyclonic wind-stress curl along the coast can enhance the alongshore upwelling. The cyclonic vorticity over SCB also indicates the existence of the Catalina eddies during this period. Under proper synoptic conditions, the lee troughing can induce cyclonic vorticity over SCB but the generation of such an eddy was usually in the lee of the coastal mountains immediately to the north of the Santa Barbara Channel. Vorticity over the land was complicated by the interaction of low-level flow over the local topography. This work was performed by Hsiao-Ming Hsu.
Figure 15. The relative vorticity, based on a COAMPS model climatology, at about 10 m AGL.
8. Modeling support for wildfire management in Colorado Anticipation of changes in low-level winds and relative humidity are important to the effective management of wild-land fires, and may be critical to the safety of firefighters. Thus, wildfire managers try to obtain the best meteorological guidance that is available. For guidance about large-scale weather conditions, the operational models of the NWS are sufficient. However, the fact that wildfires often occur in mountainous terrain means that local orographic winds may be as, or more, important than synoptic-scale factors. Thus, higher-resolution models are needed in order to resolve these important local effects. Rapid redeployability is needed because, as the active area of a fire moves, or as new fires develop, the model forecast grid must be moved regularly in order to apply the computational resources where they are needed. The concept was tested during the summer of 2002, when the forecast system was used by NCAR to support firefighters for two fires in Colorado - The Hayman and Missionary Ridge fires. One of the fire managers stated that the front-line firefighters are using (the model output) directly for their planning and tactical ops and that the model output was very important to them. This work was performed by Yubao Liu, Laurie Carson, Becky Ruttenberg, Daran Rife, Scott Swerdlin and Tom Warner.
[TOP]
|
|||||||||||||||
