Cloud-dependent Background Field Error Covariance And Its Application To Data Assimilation

Reasonable assimilation of multi-source observations(e.g.,radar observations)is of great importance to improve the initial conditions of the model and provide more accurate forecasts.Isotropic and homogeneous background field error covariance is applied in traditional variational assimilation systems,ignoring the weather-system-dependence of the background error covariance,while introducing ensemble flow-dependent background field error covariance in the variational framework requires additional ensemble forecasts.In order to introduce more reasonable background error covariance in variational assimilation,this paper proposes a "cloud-dependent" background field error covariance and a assimilation scheme via cloud-dependent background error covariance by introducing background cloud information,which is applied to the assimilation of radar and other multi-source observations,and conducts a series of studies.1)Based on the background cloud classification technique,the background error covariance including hydrometeor control variables was generated in cloudy and clear sky regions,and the background error covariance characteristics of cloudy regions were analyzed in detail.The analysis show that the background error covariance of cloudy regions has greater background error,the difference of hydrometeor control variables between cloudy and clear sky regions is more obvious compared with the conventional control variables.There is also a stronger correlation between hydrometeor and other control variables.2)This paper proposes an assimilation scheme based on cloud-dependent background error covariance,which achieves real-time assimilation of radar and other multi-source observations in cloudy and clear sky regions in a variational framework.Using the background cloud information,the cloud index is calculated and the cloud-dependent background error covariance is generated,so that the background error covariance of each control variable during assimilation varies dynamically with the cloud index,and the larger the cloud index is,the larger the background error covariance is,and vice versa.A series of single observation tests with conventional and radar observations show that the cloud-dependent background field error covariance can dynamically adjust the background field error of each grid point in real time,and the analysis increment vary with the evolution of the cloud system,with obvious anisotropy and dependence of cloud and rain characteristics.3)To further investigate the impact of the "cloud-dependent" background error covariance on forecasting,this paper conducts cycling assimilation and forecasting experiments based on the cloud-dependent background error covariance of radar data for 11 consecutive days during the Meiyu period and conducts a diagnostic analysis of a severe convective rainstorm process.The results show that the cloud-dependent background error covariance can steadily improve the precipitation forecasting capability,and the improvement of the heavy precipitation is more obvious.The application of the cloud-dependent assimilation scheme has a positive impact on radar assimilation,improving the forecasts of dynamic,thermal,water vapor and hydrometeors fields and thus the accuracy of precipitation forecasts.4)The cloud-dependent background error covariance is applied to the CMA-BJ system for operational test.Through the comparative analysis of the differences in the background error covariance characteristics in cloudy and clear sky regions,the single observation tests,and the cycling assimilation and forecasting experiments of two extreme rainfall cases: "8.16 Haidian heavy rainfall" and "7.20 Zhengzhou heavy rainfall".It is preliminarily verified that the clouddependent scheme has a positive impact on the precipitation forecast of the CMA-BJ system.