Assessing improved CPTEC probabilistic forecasts on medium-range timescale

The present study assesses the medium-range probabilistic forecast issued using an improved version of the Centre for Weather Forecasting and Climate Studies (CPTEC) Ensemble Prediction System (EPS). This version includes improvements made to the method used to select the fastest growing modes (the empirical orthogonal functions (EOF)-based method), and the bias correction. Results demonstrate that this improved CPTEC EPS surpasses the skills of the current operational version. An assessment was conducted using the framework of the THORPEX Interactive Grand Global Ensemble (TIGGE), comparing three configurations of the CPTEC EPS to the EPS from the National Centre for Environment Prediction (NCEP) and the Korean Meteorological Agency (KMA). The score adopted for metrics was the continuous ranked probability score (CRPS), calculated for absolute temperature at the 850hPa level (T850), and sea level pressure (SLP). The best improvements were found for T850 over the Northern and Southern Hemispheres, with a general increase of 9days in the useful forecast horizon. SLP augments are more modest, with useful performance extension of 2days in the Northern Hemisphere and 3days in the Southern Hemisphere. All the EPSs over the Southern Hemisphere showed a general decrease in performance when compared with their respective counterparts in the Northern Hemisphere, both for T850 and SLP. Over the Tropics, SLP is better predicted than T850 by most EPSs, but the NCEP EPS presents the best performance over this region with similar scores for the two variables.