Assessment of the rainfall forecasts from extrapolation-based INCA nowcasting and AROME forecasts in Austria

Abstract

<p>As one of the essential inputs, accurate precipitation forecasts play a significant role in having a reliable flood warning system. Due to the variability of precipitation in time and space, forecast models need to use the latest observational data as input. However, the accumulation of noise and the imbalance in analyses can decrease the performance of forecasts (Seity et al., 2011). The convection-permitting numerical limited area model AROME (Applications of Research to Operations at Mesoscale) is a non-hydrostatic convective-scale limited-area model with 2.5 km resolution. The three-hourly update cycle AROME has been replaced by the hourly update cycle since 2019 in Austria.</p><p>The aim of this study is to assess the operational AROME and INCA (Integrated Nowcasting through Comprehensive Analysis) forecasts with the three-hourly and one-hourly update cycles for three events over Austria. We use the object-oriented Structure-Amplitude-Location (SAL) approach to evaluate the performance of precipitation forecasts. The SAL approach is based on a minimum threshold method to detect precipitation objects for the forecast verification (Wernli et al., 2009). Here, we use a multi-threshold approach to define precipitation objects using the thunderstorm detection and tracking algorithm (Feldmann et al., 2021). Three heavy precipitation events with different characteristics are selected to investigate the performance of AROME and INCA using the SAL metric. The result of this study helps to understand the role of the update frequency in forecasting.</p>