post-processing

Seasonal precipitation forecasting plays a pivotal role in water resource management and development of warning systems. This study evaluated the ensemble forecasts of three C3S models over the period 1993–2017 in Iran’s eight classified precipitation clusters for 1- to 3-month lead times. The quantile mapping (QM), the linear scaling (LS), and the gamma distribution mapping (GDM) for post-processing of precipitation forecasts, and the LS and variance scaling (VS) was used to post-process temperature forecasts. The results were then compared with the raw forecasts. It is indicated that the models performed best in western precipitation clusters, while in the northern humid cluster the models had negative skill scores. Almost all the post-processing methods were able to reduce the errors and improve the forecast accuracy in most groups. In general, after post-processing the ECMWF models had the best performance and the MF model had the worst performance. Among the precipitation post-processing methods, GDM and LS performed better, and the superiority of these methods is quite noticeable, especially in the rain-heavy groups of northern Iran (G6 and G8), which had poor raw forecasts. Regarding the post-processing of ensemble temperature forecasts, the performance of each LS and VS method is similar, they have a slight difference in increasing the accuracy of forecasts. Of course, overall, the VS method has worked a little better. The performance of post-processing methods is very effective in the cold months of the year (late autumn and winter) and slightly weaker in the hot months (summer).

Efficient forecast of precipitation and temperature with a one-month horizon can provide managers with an exceptional opportunity to plan water resources and deal with floods and droughts. The application of proper post-processing and bias correction methods can much improve the accuracy of these predictions. In this study, the S2S (Sub seasonal to Seasonal) precipitation and temperature forecasts of ECMWF were evaluated in one of the important basins of Iran. A variety of methods were used for post-processing and bias correction of these predictions, and the results were compared with different evaluation criteria. Quantile mapping (QM), Bayesian model averaging (BMA), Support vector regression (SVR), an Empirical equation for bias correction of temperature, and some hybrid methods were applied to forecasts. The BMA outperformed the other methods in improving both temperature and precipitation forecasts. Raw precipitation and temperature forecasts were only applicable in 2 or 3 months of the year, but post-processing methods were able to accurately improve precipitation in half of the months, especially rainy months. The hybrid of empirical equation-BMA in 10 months of the year was led to better results than the estimate of the next month's temperature using climatological data.

Precipitation is one of the most important meteorological phenomena and the main parameter for streamflow forecasting. Therefore, determining the amount of precipitation in the future will help to manage water resources and predict the flood. In this regard, some of the most important meteorological centers in the world provided users with Quantitative Precipitation Forecasts (QPFs) on a global scale. The availability of global ensemble forecasting models in the TIGGE database creates new opportunities for flood forecasting. In this research, the effect of post-processing on the most important global numerical ensemble forecasting models such as UKMO, ECMWF, NCEP and CMA in the TIGGE database during the years 2007 to 2014 for the Bashar river Basin investigated. Evaluations were conducted in probabilistic and nonprobabilistic approach. Initially, the four NWP models with quantile mapping methods were bias corrected. Then, by using Bayesian model averaging (BMA), the post-processing was carried out. The results of probabilistic evaluation after post-processing showed that the skill of forecasting models for the Bashar basin increased and uniform distributions were achieved in verification rank histograms. Also, the results of the probabilistic evaluation with the BSS for the combined mode of four QPF Models with BMA method at most stations were close to 0.5 and in the simple combination was close to zero, indicating that Grand ensemble has a higher skill than single models.

Precipitation forecast in monthly to seasonal time scales is one of the challenges facing the Iran meteorological organization. It is also one of the fundamental needs of water resources management in agriculture, industry and drinking water sectors. Development of numerical prediction in monthly time scale is much less than numerical short term prediction in Iran; in this regard, despite to short term weather prediction, there is no operational numerical monthly to seasonal forecast model in Iran. Lack of a reliable operational seasonal forecast system cause huge damages to water resources, agriculture and natural resources sectors in all country regions. MRI-CGCM3 is the operational dynamical seasonal forecast model which is being used in Japan Meteorological Administration (JMA). In this paper output of MRI-CGCM3 was post processed using three different techniques of multiple regressions (MR), moving average (MA) and artificial neural network (ANN) over three sites of Mashad, Sabzevar and Torbat-e-heydarieh. Post processed monthly precipitation obtained from three different methods were compared with Direct Model Output (DMO).Performance of monthly forecast has been increased by 6% up to 20% when applying post processing techniques to direct model output. Result confirmed that multiple regressions (MR) techniques have the highest performance in improvement of monthly forecast skill over selected stations among all three post processing techniques.