Case study of the extreme precipitation events in the south-western part of Sweden, August 2014

Abstract: In August 2014, parts of south-western Sweden were flooded due to intense precipitation. It caused large material damages and problems in the infrastructure. Since the sea surface temperature (SST) was higher than normal during the month, it was speculated that it was one of the reasons for the heavy precipitation. In this study, the Weather Research and Forecast (WRF) model was used to simulate two of these extreme precipitation events. Nested domains with 25, 5 and 1 km resolutions were configured. In order to find the best model setup for this study, three different microphysics schemes: WRF singlemoment 3-class (WSM3), Thompson and Morrison, and two different cumulus schemes: Kain-Fritsch and Grell-3 were tested. The results were compared with observations with focus on evaluating the amounts, location and timing of the simulated precipitation. Furthermore, the sensitivity of precipitation to SST was investigated by changing the SST in the entire domain by +2 and -2ºC, compared to a reference simulation. The results showed that the WRF model was able to reproduce the maximum precipitation amounts, but not in the correct location. The timing of the precipitation was captured fairly well in the beginning of the simulations. Increased uncertainties with time indicate the importance of choosing a proper initialization time. Grell-3 cumulus scheme performed better than Kain-Fritsch, which often underestimated the precipitation. This was most pronounced for 5 and 1 km resolutions. The differences may be related to a better adaptation to finer grids in the Grell-3 scheme, e.g. by allowing subsidence effects to be spread also in the surrounding grid columns. It was also found that using cumulus scheme at 5 km was generally preferable, since the resolution was not high enough for treating the small scale convection explicitly. The results were not that sensitive to choice of microphysics scheme. This indicates that it, in some cases, may be enough to use the relatively simple WSM3 scheme that only represent precipitation as rain or snow, instead of the more sophisticated Thompson and Morrison schemes which describe also graupel and hail. Probably, the precipitation during the two chosen days did not involve hail or graupel, why this were not needed to be represented by the schemes. The simulated precipitation was highly sensitive to changes in SST. A 2ºC increase in SST led to an increase of precipitation of about 30-50 % in the most affected areas, which indicate that the relatively high SST in August, in combination with the present weather situation, may have contributed to the extreme precipitation amounts.