MM5 Forecasting System

The MM5 model was provided a daily real-time forecast of weather conditions over the UAE and the Clark-Hall model was used as a tool to understand, in detail, the mechanisms for precipitation formation in specific case studies.MM5

Realtime Forecasts

Prior to 2001, only very coarse (50-110 km and greater) resolution general forecast models were run for the UAE and surrounding region. These global models, such as the ECMWF, the United Kingdom Meteorological Office Bracknell model, and the U.S. National Center for Environmental Prediction (NCEP) Medium Range Forecast (MRF) model capture very large, continental-scale meteorological features. A modeling system with factors of ten greater resolution than the global models was necessary to study cloud development and local meteorology of the region.

The word "mesoscale" describes meteorological motions that have a temporal and horizontal spatial scale smaller than continental ("synoptic scale") systems but are larger than individual clouds. The horizontal scale ranges from a few to several hundred kilometers. Time scales range from one to twelve hours. Mesoscale systems can either be caused by geography (for example, sea and land breezes, mountain-valley winds, or effects due to air being forced to flow over terrain features) or by a larger synoptic system (for example, convective bands, mesoscale convective clusters, circulations within fronts).

The PSU/NCAR MM5 is a limited-area non-hydrostatic, terrain-following sigma coordinate model designed to simulate or predict mesoscale or regional-scale atmospheric circulation (vs. large-scale "synoptic" or global models, or smaller-scale cloud-scale models), using gridpoints 1-100 km apart (vs. 40-100 km for a global model). Thus, it can be used to study synoptic monsoons, hurricanes, and cyclones, but also mesoscale fronts, land-sea breezes, mountain-induced flows, and urban heat islands, for example. It has been developed at PSU and NCAR as a community mesoscale model and is constantly being improved by contributions from users at several universities and government laboratories.

The complete MM5 model documentation, including the executables, is available on the web at
http://www.mmm.ucar.edu/mm5/mm5-home.html

The MM5 was implemented on a 14-node Linux PC cluster at NCAR that was purchased through an agreement with DWRS. The system was configured to perform 36-hour forecasts operationally over a region containing the UAE and surrounding areas daily for a three-month period during the UAE field program effort.

The model has continued to run continuously since then and was once again used as a forecasting tool for the two field campaigns in 2002. It has also become a basic tool for the weather forecast offices in Abu Dhabi and Dubai.

Input Data and Grid Configuration

Since MM5 is a regional, limited-area model, it requires initial conditions as well as lateral boundary conditions to run. The initial background state and boundary conditions at later times are supplied by gridded large-scale (i.e., global) model forecast data that cover the modeled region for the entire time the model is integrated. Forecast data from either the 00Z or 12Z AVN global model has provided the atmospheric conditions as well as the 1° sea surface temperature (SST) data. The differential temperature between this sea surface temperature in the Arabian Gulf and the land temperature accounts for the strength and onset of the daily sea breeze.

The three nested domains are shown to the left. The outer (grid 1), middle (grid 2), and inner (grid 3) domains have grid spacing of 30 km, 10 km, and 3.3 km, and are composed of (94 x 94), (94 x 91), and (163 x 124) horizontal grid points, respectively. The nested grids each have 32 vertical levels and are two-way interactive during the simulation, thus they feed information back and forth between the different levels.

Physics Parameter Settings and Modifications

There are many choices for schemes that parameterize physical processes with varying levels of complexity. These physical processes include:

  • Cumulus parameterization schemes that generalize cumulus cloud formation when the model is using resolutions of 10 km or greater by treating the mass, heat, and momentum transport effects of convection when these clouds could not be resolved individually. The Grell cumulus parameterization has been selected for the outer and middle domains, which is based on the rate of destabilization and a simple single-cloud scheme with updraft/downdraft fluxes and compensating motion determining the heating/moistening profile.
  • Planetary boundary layer (PBL) schemes that represent the sensible and latent heat fluxes from the earth's surface that cannot be resolved explicitly. The MRF PBL scheme was selected for the forecast configuration.
  • Cloud physics schemes that treat condensation, formation of rain and possibly ice, the interactions of atmospheric particles, and their sedimentation as precipitation. The Dudhia simple ice scheme was chosen for its computational efficiency. Condensate above the freezing level (0 °C) is assumed to be ice and below it is assumed to be liquid with immediate melting as it falls through this level. Although this simplifies microphysical processes, it allows the model to be run in much less time than the more complex schemes that might be appropriate at finer grid spacing.
  • Radiation schemes that represent atmospheric long-wave and short-wave radiation interaction with clouds, air, and the earth's surface. The cloud radiation scheme was selected for this application. It is sophisticated enough to account for long wave and short wave radiation interactions with cloud and clear air.
  • Ground temperature schemes that represent the changing surface soil temperature and perhaps moisture conditions in response to atmospheric radiation. A five-layer soil model was chosen. This model predicts the soil temperature in 1, 2, 4, 8, and 16-inch deep layers with a fixed substrate below based on the vertical diffusion of heat. It resolves the diurnal temperature variation in the soil, and, along with the soil moisture, accounts for the rapidness of diurnal temperature variability near the surface.

Automated Process

Daily scripts drive processes that fetch the AVN gridded forecast atmospheric data for the 12Z simulation performed at NCEP (we started using a 00Z forecast but switched to 12Z during the Summer 2001), the sea surface temperature, and WMO GTS soundings, and brings them to the forecast cluster at NCAR. These data were integrated for the "best guess" of the state of the atmosphere at the time of model initialization, and interpolated onto the MM5 model grid. Then, a 36-hour forecast is run with the MM5 model.

  • Graphics to analyze the forecast were generated for every 3 hourly period, and transferred to the NCAR/RAP web site for the UAE project. The forecast products were 3 hourly plots of:
  • Surface (temperature (T)/wind, pressure/wind, water vapor mixing ratio, equivalent potential temperature, planetary boundary layer depth, 3 hourly precipitation, and total precipitation)
  • Upper Air (850 mb, 750 mb, 500 mb - geopotential height (H)/T/wind and relative humidity/wind, 500 and 300 mb - H/vorticity)
  • Surface Fluxes of heat and moisture, and soil temperatures
  • Soundings at selected locations (AUH, DXB, ALN)

Site forecasts of surface conditions at 65 sites in and near the UAE are also provided with the daily forecasts. These forecasts are given as hourly data in a text table format.

Find out results of preliminary forecast assessment

Read about Clark-Hall high-resolution modeling studies