Objectbased verification of tropical precipitation forecasts during the YOTC period

The accurate representation of tropical convection still poses a major challenge to global forecasting systems. However, in particular an improved representation of the physical processes in the tropics would have a positive feedback on the medium-range predictability in the extratropics. Therefore, the investigation of forecast error growth and the identification of systematic forecast errors in the tropics is of great interest. This study aims to investigate the extent of the forecast error growth and to identify systematical errors in position and structure of tropical convection in the ECWMF Integrated Forecast System during the YOTC period. Operational 5-day forecast runs are verified against precipitation estimates of the Tropical Rainfall Measuring Mission (TRMM) using two kinds of verification methods, an object-based and an field based approach. In contrast to classical skill scores, these approaches allow to verify the structural characteristics of the precipitation and enable identification of spatial displacement errors.

Our investigations suggest that the growth of the displacement error makes a substantial contribution to the overall error growth with lead time. However, error growth shows a strong regional variability, with showing a maximum over coastal regions like the Eastern Pacific of the Maritime Continent. Additionally, a significant diurnal cycle of the displacement error is identified, indicating an inaccurate representation of the diurnal cycle of the convection itself. Furthermore, in some parts of the tropics, the intensity error is largest during the first 24 hrs and weakens over the course of the forecast, which hints on the existence of a kind of spin-up problem.