Deficiencies in quantitative precipitation forecasts: sensitivity studies using the COSMO model

Abstract

The quantitative precipitation forecast (QPF) of the COSMO model, like of other models, reveals some deficiencies. The aim of this study is to investigate which physical and numerical schemes have the strongest impact on QPF and, thus, have the highest potential for improving QPF. Test cases are selected that are meant to reflect typical forecast errors in different countries. The 13 test cases fall into two main groups: overestimation of stratiform precipitation (6 cases) and underestimation of convective precipitation (5 cases). 22 sensitivity experiments predominantly regarding numerical and physical schemes are performed. The area averaged 24 h precipitation sums are evaluated. The results show that the strongest impact on QPF is caused by changes of the initial atmospheric humidity and by using the Kain-Fritsch/Bechtold convection scheme instead of the Tiedtke scheme. Both sensitivity experiments change the area averaged precipitation in the range of 30-35 %. This clearly shows that improved simulation of atmospheric water vapour is of utmost importance to achieve better precipitation forecasts. Significant changes are also caused by using the Runge-Kutta time integration scheme instead of the Leapfrog scheme, by applying a modified warm rain and snow physics scheme or a modified Tiedtke convection scheme. The fore-mentioned changes result in differences of area averaged precipitation of roughly 20 %. Only for Greek test cases, which all have a strong influence from the sea, the heat and moisture exchange between surface and atmosphere is of great importance and can cause changes of up to 20 %.