Performance evaluation of sub-daily ensemble precipitation forecasts

Nowadays, major advances have been made in meteorological forecasts. For instance, ensemble forecast systems have been developed to quantify prediction uncertainty. In this research, sub-daily ensemble precipitation forecasts of five THORPEX Interactive Grand Global Ensemble (TIGGE) models from 2014 to 2018 were evaluated in 10 major basins located in north and west Iran. Furthermore, Bayesian model averaging (BMA) was used to combine five prediction models and construct a grand ensemble. The results indicate that the models had the best performance in the Karun and Western Border basins to the southwest and west, with average performance in the Sefidrood basin in the north. In terms of the prediction of precipitation depth, the European Centre for Medium-Range Weather Forecasts (ECMWF) and the UK Met Office (UKMO) models, and in terms of prediction of precipitation occurrence and non-occurrence, the National Centers for Environmental Prediction (NCEP) model, performed best. Overall, the Japan Meteorological Agency (JMA) and the China Meteorological Administration (CMA) models acquired medium scores. The BMA technique greatly improved the probability forecasts, reducing uncertainties in the numerical models. Moreover, the models' forecasts were weaker with a 6 hr lead time compared with those with 24 hr, which may be attributable to the inaccurate detection of the initiation time of precipitation by the models. In addition, the performance of the UKMO (ECMWF) models with increasing basin elevation increased (decreased), while all models better forecasted precipitation in wet years/seasons than they did in dry years/seasons. Overall, the evaluations showed that the ECMWF, UKMO and NCEP models performed well in the majority of the northern and western basins of Iran.