نوع مقاله : مقاله پژوهشی
1 استادیار / پژوهشکده هواشناسی، تهران.
2 دانشجوی کارشناسی ارشد /هواشناسی، دانشگاه آزاد اسلامی واحد علوم و تحقیقات، تهران.
3 دانشیار /دانشگاه آزاد اسلامی ، واحد علوم و تحقیقات، تهران.
4 مربی/ پژوهشکده هواشناسی، تهران
عنوان مقاله [English]
Despite the fact that the quality of forecasts from numerical weather prediction (NWP) models has increased in recent years, yet exact forecast of precipitation is a difficult and challenging task. In order to obtain more accurate precipitation forecasts, efforts have been made to improve the models, formulations, and the accuracy of the initial conditions. One important alternative is to improve the model output via postprocessing.
In this paper, the WRF model was applied for a six month period from 1 November 2008 to 30 April 2009 with two nests using 45 and 15 Km grid. The model outputs were then postprocessed for 24-hour precipitation forecasts for 205 synoptic stations over Iran using two methods of the moving average (MA) and the best easy systematic estimator (BES). Data for the first three months were used for training and the rest of data were used for the test and comparison. Statistical scores including degree of mass balance (DMB), mean absolute error (MAE) and its corresponding skill score were calculated for both direct and postprocessed outputs.
Results showed that both methods improve the direct outputs of the model. The MA method decreased MAE for different stations from 5 to 50 percent. The mean of MAE decrease for all stations was about %25. In the BES method the average value of MAE for all stations is around 13 percent.