Spatial-temporal model of rainfall calibrated by radar data

Small- and medium-scale catchments (1-100 km2) are generally characterized by high spatial and temporal requirements, due to their rapid reactions to rainfall inputs at fine spatial and temporal scales. However, the natural variability of rainfall fields may lead to large uncertainty and bias in rain estimation at those scales. For this reason, flood risk assessment and management (e.g. in urban areas) could greatly benefit from stochastic spatial-temporal modelling of precipitation. The calibration of the Gaussian displacements spatial-temporal rainfall model (GDSTM), i.e. a spatial analog of a point-process model, is presented. GDSTM assumes that rainfall is realized as a sequence of storms, each consisting of a number of cells with clustering in space and time. Both storms and cells are characterized by their centers, durations, and areal extent. The model is calibrated using data collected by the weather radar Polar 55C in Rome, over an area of 100×100 km2, with the radar located at the center. The parameters are estimated with the Hansen method, using data with a resolution of 2×2 km2space-time.