The Evaluation And Bias Correction Of The Summer Precipitation Over Huaihe River Basin With IAP AGCM4.1

In this paper, we try to explore the method of predicting seasonal precipitation,which consists the dynamic forecasting method and the statistical analysis method.Hindcast experiments over the Huaihe river basins during the 30 years from 1981 to 2010 has been conducted making use of the recently improved version of IAP(Institute of Atmosphere Physics) Atmospheric General Circulation Model Version 4.1 (IAP AGCM4.1) so as to assess its ability in simulating summer precipitation. For the statistical analysis method, we try to use Bayesian Merging correction method to reduce the error, which combine the observed daily precipitation data and hindcast results, and examine the applicability of these methods. The main results are as follows:We try to assess the ability of precipitation which is showed in grid. Hindcast experiments show that, this climate system model exhibits certain skill in the prediction of summer precipitation over Huaihe river basin and can predict the distribution of observation climatology. For the deterministic skills, the difference between hindcast and observation in climate is small in JJA and August, but the pattern correlation coefficients between hindcast and observation is as long as 0.9 in June, we find that the model predict precipitation more in low intensity, and access less strength in heavy rain. In the whole summer, the temporal correlation coefficient(ACC) between hindcast and observation show more skill in June. But we have to recognize that the variation in hindcast is far less than the observation's.For the evaluation of probabilistic hindcast, the ensemble spread of hindcast can represent the true variability of the observations in June and August. Otherwise, the model shows certain ability of discrimination between events and non-events. For the less rainfall, the model shows good skill in distinguishing events and non-events in south Huaihe basin. At the same time , the skill is good in the middle of Huiahe basin in June. Assessing the predictive skill of different sub-basin precipitation, JJA and July perform better.This study uses a Bayesian approach to merge climate hindcast and observation for better probabilistic and deterministic forecasting. After correction, the root mean square error of area rainfall in sub-basin has decreased significantly in June and July.The root mean square error (RMSE) of precipitation in Wangjiaba sub-basin in June reduced form 4.23mm/d to 2.97mm/d. At the same time, the ACC between observation and hindcast in Wangjiaba sub-basin in JJA increased from 0.16 to 0.26.After correction, the ACC in July became 0.31,0.27,0.16 in three sub-basins.After correcting the precipitation in the basin, results show that the deviation between hindcast and observation in June and July reduced. The RMSE decreased from above 3.6mm/d to less than 2mm/d. While for the probability assessment,reduced deviation is shown and the corrected forecast can represent the true variability (uncertainty) of the observation better.