Improving satellite-based quantification of extreme precipitation events with long return period

Extreme precipitation causes large damage and replenish freshwater storage in arid regions. Quantitative information on extremes with low yearly exceedance probability is crucial for risk and water resources management, design of hydraulic structures and early-warning systems. Rain gauges offer relatively long and homogeneous records, but don’t sample uniformly the Earth's surface, leaving vast areas completely ungauged. Satellite observations could help overcoming this limit, but suffer from estimation errors, which may propagate to the estimated extreme quantiles. In this work, rain gauge and satellite data for two different regions, Israel and a portion of south-eastern Austria, are used to derive extreme quantiles associated to low yearly exceedance probability using the novel "Simplified Metastatistical Extreme Value" framework. Differences between satellite and rain gauge estimates are analyzed using a specifically developed error model with the aim of understanding how the current approaches could be refined and improved. The results show that satellite based estimates of extreme quantiles are in good agreement with the rain gauges over Austria, while a slight overestimation is detected over Israel. The developed error model allowed us to predict errors in the estimated quantiles based on errors in the parameters of the statistical model. Correlations between the first two moments of the events' distribution and the statistical model parameters provide the bases for including prior information on satellite estimation errors in the estimation of extreme quantiles.