Effective Assimilation of Global Precipitation: Simulation Experiments

Past attempts to assimilate precipitation by nudging or variational methods have succeeded 1 in forcing the model precipitation to be close to the observed values. However, the model 2 forecasts tend to lose their additional skill after few forecast hours. In this study, a local 3 ensemble transform Kalman filter (LETKF) is used to effectively assimilate precipitation by 4 allowing ensemble members with better precipitation to receive higher weights. In addition, two 5 other changes in the precipitation assimilation process are proposed to solve the problems related 6 to the non-Gaussianity of the precipitation variable: a) transform the precipitation variable into a 7 Gaussian distribution based on its climatological distribution, and b) only assimilate precipitation 8 at the location where at least some ensemble members have positive precipitation. Unlike most 9 current approaches, both positive and zero rain observations are assimilated effectively. 11 model, a simplified but realistic general circulation model. When the global precipitation is 12 assimilated in addition to rawinsonde observations, both the analyses and the medium range 13 forecasts are significantly improved as compared to only having rawinsonde observations. The 14 improvement is much reduced when only modifying the moisture field by precipitation 15 observations with the same approach. The effect of precipitation assimilation on the analyses is 16 retained on the medium-range forecasts, and is larger in the Southern Hemisphere than that in the 17 Northern Hemisphere because the NH analyses are already accurate by the denser rawinsonde 18 stations. Both the Gaussian transformation and the new observation selection criterion are shown 19 to be beneficial to the precipitation assimilation especially in the case of large observation errors. 21 improves the LETKF analysis. The new approach could be used in the assimilation of other non-22 Gaussian observations.