Impacts of the QPE integration using Poisson's Equation for Radar Reflectivity, MPE Eumetsat and Rain Gauge Network on a subtropical basin in Brazil

High resolution Quantitative Precipitation Estimation (QPE) from 4 S-Band weather radars and Eumetsat MPE data was combined with rain gauge network to guide hydrological simulations over Paranapanema Basin in Southern Brazil. The Paranapanema basin has 105,000 km2 covering north of Parana State and South of São Paulo State. For hydrological purposes the area was divided in 22 watersheds according to runoff and terrain elevation. While rain gauges provide accurate measurement at a point, remote sensing helps to retrieve the spatial rainfall pattern. An algorithm to blend these direct and remote measurements with the solution of Poisson's equation has been used to integrate radar and satellite hourly QPE with the rain gauge network over the watersheds. This technique, called Siprec, was based on 1 km x 1 km resolution matrices of radar mosaic, 4km x 4km satellite MPE and 155 rain gauges. Inter-comparisons and cross-validation tests were been conducted with Siprec, radar and satellite QPE for 18 rain gauges that did not participate in the solution, distributed with varying distance from other gauges used in the integration. The statistical results indicated that Siprec decreased the RMSE from 4.2 mm h-1 (radar-only) and 6.4 (satellite-only) to 2.4 mm h-1. Furthermore, the integration increased the correlation coefficient to 0.83 while radar and satellite got 0.59 mm h-1 and 0.41mm h-1, respectively. Most of the errors could be explained by the difference found in the events with precipitation above 10 mm h-1, most likely due the large spatial variability, typical of deep convection. The solution of the Poisson equation acts directly on the data received at a certain time, converging the amplitude to the raingauge values and keeping the spatial distribution of the radar or satellite measurement without a priori adjustments. This is an important advantage in operational environmental because it does not require frequently parallel processing to update the weights like other schemes. The QPE integrated has been used on the Soil Moisture Accounting Procedure (SMAP) hydrological model to simulate the streamflow over 11 watersheds. The results indicate the simulations that use the integrated data overestimates the streamflow during all period when compared with the rain gauge only adjustment. At the streamflow peaks the simulations are very sensible to the radar/satellite QPE overestimations, because it covers all the watersheds. The streamflow simulations produced by SMAP with Siprec QPE designs more realistic results when the watershed has some rain gauges and the radar estimates were unbiased. However, if there are no rain gauges inside the watershed, the simulations depend on the radar or satellite data and any noise or random errors could disenable the streamflow forecasts. The results indicate the technique was successful for watershed that contain a rain gauges network and radar estimates which the equilibrium between the measures will provide results more realistic.