Forecasts of extreme precipitation in the western Balkans in May 2014: model skill and sensitivity to the vertical co-ordinate

Abstract

This study examines the simulation of the catastrophic precipitating event that occurred on 14–16 May 2014 over the western Balkans, using two regional models, the Eta model and the Weather Research and Forecasting Non-hydrostatic Mesoscale Model (WRF-NMM). This case was an extreme precipitation event with very large precipitation totals exceeding 200 mm in 72 h. The Eta model was run in two variants, switched to use the eta co-ordinate with sloping steps and switched to use the sigma co-ordinate. The aim of this study was to analyse the ability of the regional model to forecast heavy precipitation successfully as well as the impact of the vertical co-ordinate on the forecasting of extreme precipitation events based on realistic model simulation. Experiments show the ability of regional models to reproduce extreme accumulations for 2 days of heavy precipitation with bias adjusted equitable threat score ranging from 0.3 to 0.5 for precipitation greater than the analysed threshold values (25–55 mm day−1), whereby the Eta model exhibits a slight advantage over the WRF-NMM and the run of the Eta model switched to the eta co-ordinate has the smallest frequency bias. For larger observed accumulations exceeding 50 mm day−1 all forecasts tend to be more biased. The root mean square precipitation difference varies insignificantly between the three forecasts. The results of the experiment indicate that, in the case of the extreme precipitation considered in this study, scores of the forecasts of the Eta model for 10 m wind speed and 2 m temperature show a small dependence on the vertical co-ordinate and in favour of the run with the eta co-ordinate.