Mahmood Ahmadi; Hasan Lashkari; Ghasem Keikhosravi; Madjid Azadi
Abstract
Abstract
The present study was conducted to simulate precipitation and temperature with the RegCM4 and LARS dynamic model in two states, with and without using the statistical post-processing technique of direct model output in the north-east of Iran (Great Khorasan) and the statistical period of 1987-2011 ...
Read More
Abstract
The present study was conducted to simulate precipitation and temperature with the RegCM4 and LARS dynamic model in two states, with and without using the statistical post-processing technique of direct model output in the north-east of Iran (Great Khorasan) and the statistical period of 1987-2011 in the annual time period. Based on the results, the annual bias average raw precipitation is equal to 53.63 millimeters and the post-processed is -11.25 in the LARS model in the study area during the 2007-2013 verification period. In summary, performing post-processing technique has been effective at 84% of the study stations in annual time scale and has reduced severely the bias error rate in most stations. Based on the results obtained from the RegCM4 model, the annual bias average raw rainfall of the RegCM4 model is calculated to be 85.3 millimeter and the post-processed to be 61.04 during the 2006-2011 verification period. Therefore, error values in most stations are very high before and after processing and the model results are not acceptable. In summary, performing post-processing technique has been effective at 75% of the research stations in annual time scale. Therefore, the absolute value of the bias error of the average annual rainfall post-processing of the LARS and RegCM4 models are equal to 13.6 and 61 respectively. The annual bias average raw temperature of the LARS model is equal to 0.096 degrees Celsius and the post-processed is -0.432. Practically, this is larger than the bias without post-processing, so post-processing operation is not effective in all stations and is only well defined in 46% of the stations. Simulation of 2 meter temperature data at the meteorological stations using the RegCM4 model as well as MA operations showed high efficiency.The annual bias average raw temperature of the RegCM4 model was -2.78 degrees centigrade which fell to -0.05 after applying post-processing technique. At all stations, the modelled annual temperature is different from observational data less than 0.1 ° C. Therefore, in the simulation of annual rainfall data, the LARS model is even more responsive than the RegCM4 model. And, in simulating the annual temperature data, the RegCM4 dynamic model shows a much better reality than the LARS statistical model.
