24-10-2012, 05:07 PM
Statistical Studies of Mesoscale Forecast Models MM5 and WRF
ABSTRACT
Two mesoscale weather forecasting models--the Mesoscale Model Version 5 (MM5) and the Weather Research and
Forecast (WRF)--were statistically evaluated over two different geographical areas, Utah and western Texas. Using the 40-km
Eta forecast data as input data, forecast calculations of both the models were carried out and the results were compared with
surface observation data. Both models tended to overforecast temperature and dew-point temperature, although the correlation
coefficients between forecast and observations were fairly high. The statistical parameters for MM5 were slightly better than
those for the WRF. For both MM5 and WRF, statistical parameters for wind vector components are inferior to those of
temperature and dew-point temperature. The influences of different input data on the MM5 forecast fields were studied using
the 40-km Eta and the Global Forecast System (GFS). For all surface meteorological parameters, the MM5 with the inputs
from the 40-km Eta performed better than the MM5 with the GFS. The WRF forecasting over western Texas produced better
statistical results than those over Utah, probably due to simpler terrain in western Texas as compared to Utah.