Research On Direct Variational Assimilation Of Cloud-affected Satellite Microwave Observations

Nowadays, satellite data are widely used in numerical weather prediction systems and have significantly improved the accuracy of numerical weather prediction. However, one of the key reasons why more than 75 percent satellite data are removed in the stage of quality control by the data assimilation system, is that these data are affected by clouds and precipitation. At present, most operational data assimilation systems only digest the clear satellite data, while the radiances in cloudy and rainy areas (especially in typhoon and storm areas) cannot pass the data assimilation systems and are directly deleted. This condition severely restricts the application of satellite data. To explore the assimilation method of cloud-affected satellite microwave observations, a direct variational assimilation scheme is proposed in this paper. In order to make up the deficiency of the hydrometeor information in the first guess field the retrieved profiles of cloud and rain by one-dimensional variational method are combined with atmospheric temperature and humidity parameters as the initial input fields of the radiative transfer model employed by the data assimilation system. The purpose of assimilating the cloud-affected microwave satellite data is reached by starting the scattering module of the radiative transfer model.A new radio-frequency interference (RFI) identification algorithm, namely the modified principle component analysis (MPCA) method was developped using the observations of microwave imager to identify the RFI mixed within ice or snow underlying surface. Comparing the MPCA method with other methods, such as spectral difference method, PCA method, NPCA method and DPCA method, it was found that the new MPCA method with wide range of adaptability and high computing efficiency laid a good foundation for the follow-up application of microwave imager measurements.The 1D-Var algorithm was utilized to retrieve the hydrometeor parameters, such as vertical profiles of cloud liquid water content, ice water content and rain water content based on the RFI corrected observations of microwave imager. All the retrieved products were assessed taking Halong Typhoon,2014 as example. These hydrometeor profiles were complemented as the input to the radiative transform model (observation operators), in which cloud absorption and scattering effect was considered. The retrieved hydrometeor profiles were valiated using the observed reflectivities of CPR (Cloud Profiling Radar) uploaded on CloudSat satellite. The comparison results showed that the areas with higher radar reflectivity matched well with the cloud water and liquid precipitaion regions at high level. In another word, the precision of hydrometeor retrievals from the 1D-Var algorithm were very high. By comparing the biases between the simulations and the satellite observations in the clear-sky and cloudy cases, we found that providing the cloud parameters as auxiliary parameters and activating the scattering module of radiative transfer model can effectively improve the simulations over the typhoon periphery system, greatly reduce the deviations between the simulations and the observations, and increase the amount of satellite data which can be assimilated into the numerical weather prediction system.Taking the retrieved profiles of cloud, rain and atmospheric temperature and humidity profiles together as the first guess of the three-dimensional variational data assimilation system (GSI), the brightness temperatures of AMSU-A in cloudy areas were directly assimilated. Taking typhoon "Halong" occurred in August 2014 as example and three parallel assimilation experiments were designed. The control experiment assimilated conventional data only, the experiment 1 assimilated the conventional data and AMSU-A data in clear sky, and the experiment 2 which is based on the direct assimilation scheme proposed by this paper assimilated the conventional data and all-weather AMSU-A data. By comparing the analysis fields of these three experiments we found that the utilization efficiency of satellite microwave data increased and the temperature, humidity and wind fields at large scale are improved by the experiment 2, especially for the adjustment of the typhoon warm core at high altitude. The enhancement (diminution) of typhoon warm core will enhance (weaken) cyclone by deepening (reducing) the gradient with surrounding environment field during the integrating process of WRF model. Although the initial typhoon location was not close to the observed one, the path forcast error decreased gradually as the increasing of follow-up assimilation circulation and prediction time.