Understanding Sources of Tropical Rainfall Forecast Errors

Despite decades of research on the roles of moist convective processes in large-scale tropical dynamics, tropical quantitative precipitation forecasts (QPF) skill in operational models is still deficient, even at short lead times. Here we start by assessing global QPF skill of the NCEP Global Forecast System (GFS) and ECMWF Integrated Forecast System (IFS) models during April, 2015-March, 2016. Results reveal that in general initial conditions are reasonably well estimated in both forecast systems, as indicated by relatively good QPF scores for the 6-12 hour forecasts. However, overall, the tropical QPF in both systems was not considered useful by typical metrics much beyond 6 days, although there is some conditional skill, as seen by the increased skill of especially the IFS during a Madden-Julian Oscillation (MJO) event of December 2015. To quantify the relationship between QPF skill and the model's ability to simulate the MJO and convectively coupled equatorial waves, we calculate the space-time spectrum of tropical rainfall using forecast rainfall time series. Is is shown that while tropical variability is too weak overall in both models, the IFS is much superior in allowing tropical waves to propagate for longer lead times. As shown in previous studies, the results presented here suggest that the differences in cumulus parameterizations used in each system, in particular, how moist convective processes are coupled to the large scale flow through these physical parameterizations, appear to be the main source of QPF errors.