Evaluation of Deterministic Wind Speed Forecasting Output of Two Ensemble Post-Processing Methods

In this study, deterministic forecasts of 10-meter wind speed for the next 24, 48 and 72 hours have been produced and analyzed over Iran using BMA and EMOS methods for post-processing of raw outputs of the ensemble systems. The main purpose of this article is to compare deterministic forecasts based on these two methods with each raw ensemble members and the mean of the raw ensemble members. The applied ensemble system consists of eight members with different boundary layer schemes in the Weather Research and Forecasting (WRF) model. Other physical schemes remained the same in the ensemble members. For each ensemble member, the 24, 48 and 72-hour forecasts of 10-meter wind speed have ben conducted over Iran, with a horizontal resolution of 21 km. The Global Forecast System (GFS) is used for initial and boundary conditions of forecasts starting at 1200 UTC for each case. Observational data of 31 synoptic meteorological stations located in provincial capitals have been used for model evaluation, in which model outputs are interpolated to the locations of these stations by a bilinear method. The WRF model is run from 1 March to 31 August 2017, but the results from 11 April to 31 August 2017 are considered as the spin-up period. Indeed, after careful examination of the forecast errors using different spin-up periods, the first 30 days of the simulation are considered as the spin-up for both BMA and EMOS methods. Verification is performed by different methods (accuracy: PC, TS and OR; reliability and resolution: FAR, POFD and POD; skill: CSS, HSS, PSS, GSS and Q; statistical errors: RMSE and MAE) for 10-meter wind speed thresholds less than 3 m/s and more than 5, 10 and 15 m/s for both methods for all forecast lead times. Results indicate significant improvements in accuracy scores (300%), reliability and resolution scores (220%), skill scores (340%). Statistical error scores are also reduced by 24%. Furthermore, applying verification for different climatic regions of Iran (cold, semi-arid, hot-dry, hot-humid and moderate-rainy climate) indicates that in all climatic regions, the best performance in terms of RMSE is for BMA and EMOS methods, with the average reduction of error by 21% and 23% ,respectively. Particularly, in hot and humid climates,  these two methods better improve predictions, and hence, are more promisingas they reduce the error by 44% and 46%, respectively.