Automatic estimation of rainfall fields for hydrological applications: Blending radar and rain gauge data in real time

Weather radar and rain gauges are complementary sensors in the estimation of rainfall fields. A number of proposed techniques to combine their measurements looking for an improved rainfall field may be found in the literature. Most of them aim to join the quantitative skill of rain gauges with the good spatial description observed by the radar using geostatistical estimators. However, none of these techniques has been considered the reference one in real time applications, mainly because of the difficulties to adapt them to the operational requirements: in these conditions, the definition of valid spatial variability models (correlograms) has not been, up to now, solved in a satisfactory way to avoid manual analysis or a priori simplified assumptions. In this paper, an automatic technique to compute rainfall fields blending radar and rain gauges, which is based on kriging with external drift, is described and its performance is evaluated. The main interest of this new methodology is that it can be fully implemented in a real time framework as the correlograms are automatically computed at each time step applying a fast approach based on the FFT. Cross validation analysis shows that this technique produces accurate estimations of rainfall. At the same time, the spatial structure of the estimated rainfall fields is close to the patterns observed in the radar images. The performance of the proposed methodology is evaluated in several case studies measured in Catalunya (Spain).