Forecasting of monthly precipitation based on ensemble approach using CFSv2-WRF model over Iran(case study: October 2019 to April 2020)

Skill assessment of global weather forecasting systems in different regions and time scales can not only improve the performance of the models based on the use of appropriate parameters for modeling or methods for post-processing, but also increase our understanding about performance of forecasting models in regions and different time scales. Accurate precipitation forecasts can play an important role in water resources management as well as reduce damages caused by heavy rainfall.    In this study, we assess the use of the Weather Research and Forecasting (WRF) model to downscale NCEP Climate Forecast System Version 2 (CFSv2) atmospheric reanalysis for generating a monthly precipitation forecast over Iran. The WRF model is configured with two-way nested domains of 60-20 km horizontal resolution. It is used to produce precipitation forecasts based on four different configurations and six initial conditions of CFSv.2 data (totally 24 members) for October-April precipitation over the period 2000-2019 (as hindcast). Performance of ensemble members was evaluated according to the Kling-Gupta-Efficiency (KGE) in comparison with 145 meteorological stations over Iran. Each member has a rating of 1 to 24. The weighted average method was used to calculate the average precipitation obtained from 24 members. To evaluate the performance of WRF-CFSv.2 ensemble model in monthly precipitation forecast, we used some criteria such as False Alarms Rate (FAR), Proportion Correct (PC), and Heidke Skill Score (HSS) for the verification of categorical forecast and Root Mean Square Error (RMSE), Nash-Sutcliffe Efficiency (NSE), Pearson Correlation Coefficient (PCC) and Mean Bias Error (MBE). The results showed that the average values of NSE, RMSE, MBE and PCC of ensemble model for hindcast were 0.46, 365, -5.7 and 0.67, respectively. Categorical indices indicated that the model skill in forecast of each precipitation class (below normal- less than 33rd percentile, normal- between 33rd and 66rd percentiles, above normal- greater than 66rd percentile) has an accuracy of 52%. Evaluation of the efficiency of the model for a test period (October-April precipitation over the period 2019-2020) shows that the model is able to forecast the distribution of monthly precipitation over Iran. In this case study, the results show that the forecasted monthly precipitation has a positive correlation (PCC = 0.68) with the observations. The results suggest that the WRF-CFSv.2 ensemble forecast based on 24 members can be useful for flood forecasting and water resources management, although the amount of precipitation forecast has bias in some precipitation systems.