In this study, we investigate an efficient strategy to directly assimilate satellite precipitation observations into a cloud-resolving-model. The assimilation is based on the nudging of latent heating estimates from satellite microwave observations. Assimilation formulations possibly more effective than nudging exist but they are significantly more intensive from the computational standpoint. Examples of such formulations are the 4-D variational assimilation and the ensemble Kalman filter. In small scale hydrologic applications, it is preferable to use the computational power to better resolve small scale cloud processes than to better fit a simpler model to observations. The cloud resolving model used in the study is the Advanced Regional Prediction System (ARPS) developed at the University of Oklahoma. The satellite microwave passive latent heating are provided by a Bayesian algorithm previously developed by the authors. The assimilation consists of adjusting the sources/sinks in the thermodynamic equations to be consistent with the satellite microwave estimates.

The methodology is applied first to ground radar latent heating estimates. Then satellite observations are synthesized from the radar observations and the assimilation methodology is tested with satellite like latent heating estimates. The impact of both temporal sampling and instantaneous latent heating estimation errors on the short term quantitative precipitation forecasting is investigated. The methodology is applied to two storms that caused flash floods in central and south Europe.