Generalized Variational Merging of Multi-source Precipitation Data Based on the Non-Gaussian Model

摘要

Different from other domestic and foreign research in which the optimum interpolation (Ol) merging algorithm is commonly used, this paper constructs the non-Gaussian model for generalized variational precipitation data merging research based on the non-Gaussianity of precipitation data. For CMORPH data correction, the probability density function ( PDF) matching method is adopted, during which the GAMMA function fitting is utilized, and the generalized variational merging based on non-Gaussian model is used to merge corrected CMORPH precipitation data and station ground observation precipitation data. Meanwhile, we carry out an experiment on CMORPH precipitation data correction and the merging of multisource precipitation data based on non-Gaussian model. By measuring the structural similarity between the merged field and the reference field, we get a merging method that can better retain useful"outliers" which represent weather phenomena. The experimental results accord with our expectations.