Comparison of precipitation nowcasting by extrapolation and statistical-advection methods

Two models for nowcasting of 1-h, 2-h and 3-h precipitation in the warm part of the year were evaluated. The first model was based on the extrapolation of observed radar reflectivity (COTREC-IPA) and the second one combined the extrapolation with the application of a statistical model (SAMR). The accuracy of the model forecasts was evaluated on independent data using the standard measures of root-mean-squared-error, absolute error, bias and correlation coefficient as well as by spatial verification methods Fractions Skill Score and SAL technique. The results show that SAMR yields slightly better forecasts during the afternoon period. On the other hand very small or no improvement is realized at night and in the very early morning. COTREC-IPA and SAMR forecast a very similar horizontal structure of precipitation patterns but the model forecasts differ in values. SAMR, similarly as COTREC-IPA, is not able to develop new storms or significantly intensify already existing storms. This is caused by a large uncertainty regarding future development. On the other hand, the SAMR model can reliably predict decreases in precipitation intensity.

► Two extrapolation models nowcasting precipitation were compared. ► The first model was based on the extrapolation of observed radar reflectivity. ► The second one applied a statistical model to the extrapolated reflectivity. ► The statistical model improved afternoon forecasts. ► No impact was realized at night and in the morning.