This paper is on the impact of forecasts for a landfalling hurricane ( Karl of 2010) from the use of rain rate initialization at radar resolution within a mesoscale cloud resolving model. The concept of rain rate initialization for cloud resolving non hydrostatic models is made by extending the previous work on physical initialization for global models at high resolutions. In the present study the inner core of the hurricane is resolved at a horizontal resolution of 1.33 kms. A ground based radar (the Alvarado radar of Mexico) was able to provide radar reflectivity data sets as Hurricane Karl was slowly approaching the Mexican coast. This oceanic traverse when the storm was arriving, was of the order of 24 hours. That provided, through validated Z-R relationship, the ain rates translated from the radar reflectivity data sets of the Alvarado radar. Those rain rates were used for rain rate initialization (called physical initialization) within a WRF/ARW model. This is a doubly nested mesoscale model with various options for the physical parameterizations. The inner nest at 1.33 km mesh has explicit clouds. Comparisons of short range forecasts of landfall of hurricane Karl , with and without the rainrate initialization shows major promise of the initialization. An animation of rain rate initialization shows that it is possible to initialize the rains, as inferred from the Alvarado radar , to a spatial correlation of 0.8 stating from a value of 0.1. The forecast skills, inferred from the equitable threat scores, and verified against Radar and TRMM based rains show much larger skills from the rain rate initialized starts. This paper will show forecast validations of extreme events, i.e. very heavy rains over Mexico from these forecasts.