A Multiscale Four-Dimensional Data Assimilation System Applied in the San Joaquin Valley during SARMAP. Part I: Modeling Design and Basic Performance Characteristics

Abstract

This paper presents results of numerical simulations made with a high-resolution multiscale four-dimensional data assimilation system applied over California during two episodes associated with high ozone concentrations in the San Joaquin Valley. The model used here is the nonhydrostatic Pennsylvania State University?National Center for Atmospheric Research Mesoscale Model (MM5). The focus of the paper is the objective validation of the regional (mesoalpha scale) meteorological results. The mulliscale data assimilation approach produces highly reliable simulations of the wind, temperature, mixed-layer depth, and moisture, each of which is vital to air quality modeling and a host of other mesoscale applications. The significance of this research is threefold. First, it is the first evaluation of this multiscale assimilation system in strongly heated summertime conditions and with comparatively fine grid resolution (4-km inner mesh). Second, the assimilation system has been extended so that temperature soundings can be used to effectively reduce model errors for the simulated mixed-layer depth (which is crucial for correctly simulating boundary layer mixing and air chemistry processes). Third, by withholding half of the special data for use in model verification, it is shown that assimilation of observations at the mesoscale is, indeed, effective. Numerical errors are reduced over the intervening regions between the sites where data are assimilated. By establishing interobservation accuracy, we demonstrate that the data-assimilating model produces spatially consistent solutions without serious distortion of the active dynamical processes. In other words, the model and the observations are each able to contribute to the final numerical solution in a way that reduces error growth and does not disrupt the intervariable consistency among the primitive variable fields.