Floods are most frequent among all the natural disasters resulting in loss of lives and property. A combination of meteorological and hydrological extremes usually triggers floods. Prediction of timing, location and intensity of the rainfall and resulting flood inundation extent, well in advance, helps society for better preparedness. Impact of floods is higher in urban areas owing to dense population and economic activities. From a hydrologic perspective, the rainfall-runoff response in urban areas is rapid and highly sensitive due to the presence of high spatial heterogeneity in land use / land cover. Moreover, urbanization has the potential to influence the local micro-climate due to modifications in the surface roughness, temperature, atmospheric boundary layer and moisture availability. Despite this, limited attention is paid towards its impact on space-time distribution of the rainfall, especially in the context of extreme events. The current study uses a model cascading framework incorporating a Numerical Weather Prediction (NWP) model to generate high resolution ensemble precipitation forecasts, and a quasi-distributed hydrological model for runoff estimation. The main focus is to provide reliable rainfall forecasts at convection permitting scales employing detailed urban physical parameterizations together with assimilating high-resolution satellite, and weather radar data. Under-dispersion, a widely encountered problem in convective ensemble precipitation forecasts would be addressed by designing the techniques that consider spatial metrics of the rain fields. The generated ensemble high resolution precipitation forecast will be integrated with a hydrological model to produce probabilistic stream flow forecasts. Initial findings indicate that there is a variability characterized by the configurations of NWP model in high-resolution weather forecasting and thus, judicious selection of NWP configuration is vital for improving forecast accuracy.