Assessment of the WRF model in simulating a catastrophic flash flood

Abstract

The present study examines the ability of the forecast (WRF) model to reproduce a heavy rainfall flash-flood event that hit the urban area of Skopje City, on August 6, 2016. A series of numerical experiments were carried out to evaluate the model’s performance in the simulation of this catastrophic event, which caused great material damage and the loss of 23 human lives. The simulations with the triple-nested WRF-ARW runs as well as the experiment using WRF-NMM dynamic core with the initial data of FNL GDAS showed better skills in a more precise qualitative and quantitative assessment of the total 24-h accumulated precipitation, the location and the relative intensities of rainfall. Explicit treatment of convection without parameterization significantly improves forecast accuracy and reduces forecast errors. The verification results, using standard tests, showed the model’s ability to reproduce the occurred flood. The correlation coefficient is higher for runs with explicit cumulus convection and 4 km resolution with the Yonsei PBL scheme and Thomson microphysics with aerosol climatology. In addition to the influence of the thermodynamic characteristics of the atmosphere, orographic forcing on the development of a strong mesosystem is of great importance for the intensification of convective cells and the production of large amounts of precipitation.