A nonlinear least squares four-dimensional variational data assimilation system for PM2.5 forecasts (NASM): Description and preliminary evaluation

Abstract

Air quality is a vital concern globally, especially in China. To improve fine particulate matter (PM2.5) forecasts, a nonlinear least squares four-dimensional variational (NLS-4DVar) data assimilation system was established and applied into the Weather Research and Forecasting model coupled with the “offline” Community Multiscale Air Quality (WRF-CMAQ) model. By assimilating hourly surface PM2.5 observations, the optimal initial conditions (ICs) of the state variable were solved iteratively with the NLS-4DVar method, which uses a Gauss–Newton iterative scheme to handle nonlinearity without a tangent linear or adjoint model, thereby rendering the aerosol assimilation process fast and simple. Observing system simulation experiments (OSSEs) were designed from 10 to 16 November 2018 to evaluate the effectiveness of the NLS-4DVar data assimilation system for PM2.5 forecasts (NASM) assimilation system. The results derived from the OSSEs indicated that the NASM system could effectively assimilate multi-time PM2.5 observations, reduce uncertainty in surface initial PM2.5 concentrations, and thus improve the accuracy of predictions.