Regularization method of assimilating Doppler radar data and its influence on precipitation forecast

Abstract

A new regularization method is proposed to directly assimilate Doppler radar data into mesoscale numerical weather forecast based on the traditional 3DVAR. For seeking the minimum module solution of the Yo=H(X) with bias δ, the regularization method is adopted and leads to a new cost function. A group of experiments were designed to study the case of a locally strong rainstorm occurred in Beijing area on August 14, 2008. The L-curve principle is used to determine the optimal regularization parameter and the result is α=0.1. Numerical results demonstrate that both regularization method and 3DVAR scheme can efficiently assimilate the Doppler radar data, and that the improved initial condition can alleviate the spin-up phenomenon and improve the nowcasting precipitation forecast. However, when an optimal regularization parameter is choser, better improvement, more accurate precipitation forecast and higher TS score are expected.