Using satellite rain estimates to update heavy rainfall warnings in Hungary

As part of a project funded by the United States - Hungarian Science and Technology Joint Fund, research has been conducted to improve the forecasting of extreme rainfall events and to increase the timeliness of heavy rainfall warnings in Hungary. In particular, we have implemented and tested an automated satellite rain estimation technique wherein METEOSAT observations are used daily to provide precipitation fields over Hungary and neighboring countries. These estimates have the added potential of providing first-guess fields and updates for heavy rainfall prediction techniques (in particular, Probable Maximum Precipitation, or PMP, methods) that rely on soundings made at 12 h synoptic intervals.

For this project, it was necessary to adapt the Griffith-Woodley satellite rain estimation technique to make it applicable to rainfall in Hungary. These adaptations included the translation of GOES digital counts into their METEOSAT equivalents and a revised and streamlined version of the life history method used in the original technique. A third novel element of these revisions involved the use of model-generated pseudo-soundings available every three hours from the Aladin model in place of soundings from European synoptic stations. This change allows Model Adjustment Factors (MAFS) used by the rain estimation technique to be updated every 3 hours at any location where extreme rains are expected.

The performance of the rain estimation technique was carefully examined for three heavy rainfall case studies in the summer of 1998; one of these case studies will be presented here. For this case, the satellite estimates routinely overestimated the rain measured by gauges in the mean, but underestimated maximum rainfall at the location of the event. The overestimation appeared to be a result of decreased resolution of digitalized imagery used between the digitized images available at that time only every three hours. Underestimation at the location of rain gages in extreme rainfalls was at least partially a result of the bilinear interpolation of the satellite data to gauge locations. In fact, individual pixel rainfalls in the vicinity of the gauges could be quite close to the gauge observation. Variations in the MAFs were not strongly related to gauge-satellite differences, and the use of Aladin model-generated pseudo-soundings in the computation of the MAFs produced reasonable results.