Mesoscale Modeling and Four-Dimensional Data Assimilation in Areas of Highly Complex Terrain

Abstract

A multiscale four-dimensional data assimilation (FDDA) technique, based on Newtonian relaxation, is incorporated into a mesoscale model and evaluated using meteorological and tracer data collected during the Atmospheric Studies in Complex Terrain (ASCOT) field experiment in the winter of 1991. The mesoscale model is used to predict the synoptically driven flows and small-scale circulations influenced by terrain along the Front Range in Colorado in the vicinity of Rocky Flats Plant for four nocturnal periods during the ASCOT field experiment. Data assimilation is used to create dynamically consistent analysis fields based on the mesoscale forecasts and the special asynoptic data taken during this experiment. Observations from towers, minisodars, airsondes, tethersondes, rawinsondes, and profilers near the Rocky Flats Plant, as well as observations from surface stations throughout Colorado, are incorporated into the high-resolution analysis fields. The wind and turbulence quantities produced by the mesoscale model are then used to determine the dispersion of tracer released from the Rocky Flats Plant for each evening. A subjective and statistical evaluation of the meteorological and dispersion results is performed to examine the effect of FDDA on the nocturnal circulations and tracer transport. The mesoscale model is able to qualitatively predict the mesobeta-scale drainage flows from the Front Range into the South Platte River basin; however, the largest wind forecast errors occurred in a region immediately adjacent to the foothills. As expected, the current FDDA technique reduced the overall errors in the atmospheric and dispersion calculation, while the model generated realistic small-scale circulations not resolved by the data. Still, the model did not capture the shallow surface drainage flows just east of the Rocky Flats Plant for two of the evenings during the field experiment. When the model was initialized with the high-resolution analysis fields generated by FDDA and left to forecast on its own, little improvement in the forecasts were seen two hours after the initialization time. This may be due to the fact that only the observed horizontal wind components were assimilated into the analyses generated by FDDA; assimilation of temperature observations was not included in this study.