Development of a nudging-based four-dimensional data assimilation capability for the Weather Research and Forecast (WRF) model

The Community Multiscale Air Quality (CMAQ) Model is EPA's principal numerical air-quality modeling tool for scientific and policy analyses for tropospheric ozone, fine particulate matter, acid deposition and visibility. This model derives its meteorological input from the Penn State/NCAR Mesoscale Model Version 5 (MM5), itself a community model for weather forecasting and analysis. This model is often used in a retrospective mode for dynamic analysis, where four-dimensional data assimilation (FDDA) via Newtonian nudging is applied throughout a model simulation for air quality applications. An FDDA-assisted model simulation covering a historical air-quality episode is then used as input to CMAQ and other air quality models to study the episode.

Nudging FDDA effectively controls model error growth while the model simultaneously produces detailed mesoscale structures which may not be resolved by the observational data. The nudging is used to relax the model solution towards the observations continuously at each time step by adding to the prognostic equations artificial tendency terms based on the difference between the two states. These terms are relatively small compared to the other terms in the model equations, and the result is a dynamically consistent product superior to that obtained with the model or the observations applied separately. The assimilation can be accomplished by nudging the model solution towards gridded analyses based on observations (analysis nudging) or directly towards the individual observations (observation nudging). The result is a time-continuous high quality meteorological data set for air-quality applications.

The nudging FDDA approach, including both the analysis nudging and observation nudging, was originally developed by Penn State (Stauffer and Seaman 1990, 1994; Stauffer et al. 1991) with support from the EPA. Now that the new Weather Research and Forecast (WRF) model is beginning to be used by the weather forecasting community, the FDDA technique is again needed within this model if it is to be used by the air quality community, and EPA is again providing some funding to Penn State to begin development of the nudging FDDA capability for the WRF.

Initially, analysis nudging will be implemented in the Eulerian Mass (Advanced Research) dynamic core of WRF. Similar to the MM5 implementation, the WRF nudging follows the approach described by Stauffer and Seaman (1990; 1994) and Stauffer et al. (1991), but uses the WRF state variables and is updated to provide additional user flexibilities. An overview of the WRF FDDA design plan will be presented along with an updated status report and some results comparing the current FDDA within WRF to that within the MM5. The WRF FDDA code development is intended to be provided to the user community when completed.