Quantitative precipitation estimation in the Alps: where do we stand?

Abstract

Dipartimento di Elettronica, Politecnico di Torino, ItalyAbstract. Quantitative precipitation estimation in a moun-tainous region is of high practical use. Think of hydrogeo-logical risk management, hydroelectric power, road mainte-nance, and tourism. Yet, neither gauge networks nor weatherradars give a perfect answer. Gauge observations in a moun-tainous region such as the Swiss Alps suffer from severeerrors because of wind, snow-drift and spatial variability.Radar scientists and engineers, on the other hand, are facedwith severe beam shielding, strong ground clutter, and dif-ficult operating conditions. A sophisticated elimination ofground clutter, and automatic robust algorithms to correct forshielding and profile effects are a must if one needs quanti-tative estimates of precipitation amounts in a routinely man-ner. At MeteoSwiss many efforts went into the optimisationof hardware stability and data processing of the radar net-work in order to obtain quantitative precipitation estimatesthat meet both the meteorologist’s and the hydrologist’s re-quirements. Recent modifications in the correction schemesof the operational radar product RAIN, the best radar esti-mate of surface precipitation over Switzerland, resulted in asignificant reduction of bias, scatter and false alarms.1 IntroductionMeteoSwiss has more than fourty years of experience withthe operation of weather radars in a mountainous region.Many efforts went into the optimisation of hardware sta-bility (Cavalli, 1998) and data processing of the radar net-work (Joss and Waldvogel, 1990; Joss et al., 1998) in or-der to obtain quantitative precipitation estimates that meetboth the meteorologist’s and the hydrologist’s requirements.One of the results of these efforts is the MeteoSwiss opera-tional radar product RAIN, a two-dimensional map with thebest radar estimate of surface precipitation over Switzerland.The main challenge in a mountainous region is the elimina-Correspondence to: U. Germann(urs.germann@meteoswiss.ch)tion of ground clutter (Lee et al., 1995; Germann and Joss,1999b, 2004), and the correction for shielding and profileeffects (Joss and Lee, 1995; Germann, 1998; Germann andJoss, 1999a; Vignal et al., 2000; Pellarin et al., 2002; Ger-mann and Joss, 2002). In Switzerland about two thirds of allradar measurements are made in the snow.Although of minor relative importance in a mountain-ous context also attenuation (Delrieu et al., 1991; Germann,1999), and variations in the drop size distribution (Joss andZawadzki, 1997; Smith and Joss, 1997; Lee and Zawadzki,2003) introduce significant uncertainties in deriving precipi-tation rates from radar reflectivity.Recent modifications in the correction schemes of theradar product RAIN (Germann and Joss, 2002; Germannet al., 2003; Germann and Joss, 2004) resulted in a significantreduction of bias, scatter and false alarms (Bolliger et al.,2004, and this paper).On the use of weighted multiple regression analysis to cor-rect for residual errors and as a quality check see Gabellaet al. (2004), this conference.2 OutlineThis paper, first, briefly describes recent modifications inradar data processing for quantitative estimates of surfaceprecipitation in Switzerland, and, second, presents an objec-tive evaluation based on measurements of 3 radars, 58 auto-matic gauges and 7 years of data. For details the reader isreferred to the literature.3 Recent improvementsThis is a list of the major recent modifications in radar dataprocessing for quantitative estimates of surface precipitationin Switzerland.