Optimal configuration of RegCM4.5 model for rainfall forecasting in Iran based on climatic zones (November-May), Case study: 2019-2014

Seasonal forecasting has always been one of the challenges in forecasting Iran's diverse climate. In the last one or two decades, many efforts have been made to develop and improve the climate models of the restricted area and to minimize these challenges, but the problems and challenges still remain. Convective parameterization schemes are always one of the sources of error in regional climate models that have a significant impact on model outputs. Therefore, one of the most important issues in implementing the model is choosing the appropriate convective scheme from the existing schemes. One of the methods of forecasting precipitation in our country is the use of dynamical downscaling by RegCM model. Most of the studies that have been done for this purpose in the country so far have considered single convection schemes for the whole country, the results of which have not shown a significant improvement in rainfall forecasting.

In this study, a relatively new approach was adopted, so that convection schemes were selected appropriate to the climate of the region, and then the final forecast of the entire country by regional integration of each climate zone was presented. In this paper a relatively new perspective of the climatic zones of the regions, was used for optimum configuration of the RegCM4.5 model; The study area in this study is Iran, which includes 25 to 41 degrees north latitude and 47 to 63 degrees east longitude, but the model area ranges from 30 to 70 degrees east longitude and 10 to 55 degrees north latitude. It covers important geographical features, including mountains and seas. In this study, the output of the CFSv2 global climate model originating from November 1 in each year as the boundary condition data has been used and the CRU precipitation data has been used as reanalysis data to test the output of the RegCM model. Because CRU data are averaged monthly, they are suitable for studies that examine monthly averages. CRU data have already been used by various researchers in the country to validate the output of the RegCM model. After selecting the schemes of the planetary boundary layer and surface layer, the selection of the appropriate Cumulus Parametrization Schemas(CPS) was done based on Iran's climatic classification using the Demarten index. This method is the simplest and most common method for climate classification that precipitation and temperature variables are effective in calculating climate index, and precipitation and temperature data have also been used from the CRU database. For this purpose, Iran was first divided into seven very humid, humid, semi-humid, Mediterranean, semi-arid and arid climates based on the Demarten index, and each grid points of the study area were assigned the relevant climate index. The share of each climatic class in zoning was obtained as follows; Arid 32.4%, Semi-Arid 30.1%, Mediterranean 7.6%, Semi-humid 7.6%, Humid 10.5% and Highly humid 11.8%. The study period was 5 rainy seasons 2019-2014 (November to May) that the beginning of each simulation with the initial condition data on the first of November and its end at the end of May (as the end of the rainy season in the country) in each year. The horizontal resolution considered to be 30 km regional model, the planetary boundary layer schemes and the surface layer Holslag and BATS were considered, respectively. Kuo, Grell, Emanuel, Tiedtke and Kain convection schemes were tested during this period to achieve optimal configuration.

In the first stage, mean precipitation and its RMSE from individual and integrated schemas were calculated, but due to the fact that the Emanuel and Kain schemes did not rank higher in any of the model experiments in terms of climatic classes and have more errors than others, theywere removed from the configuration selection process. The results showed that in very humid, humid, semi-humid and humid climates the Tiedtke convection scheme, in the semi-arid regions the Grell scheme and in the arid areas of the Kuo scheme had the least bias compared to other convection schemes. Therefore, the seasonal forecast of the country was presented by combining regional schemas, the average bias of which was calculated at 0.45, 0.79, 1.01 and 0.69 mm in the integrated schemes of Tiedtke, Grell and Kuo, respectively. On the other hand, in addition to calculating the ability of different schemas to predict precipitation using the RMSE index, the area under the ROC curve was also calculated in three classes less than normal (BN), normal (NN) and more than normal (AN) for different climates. For this purpose, in each precipitation layer, the number of schemas that predicted precipitation in different climates and had the largest area under the curve compared to other schemes was extracted. ROC diagrams of different schemas showed that Tiedtke and Grell schemas have the highest ability to predict less than normal, normal and more than normal rainfall classes. The results showed that the regional integrated scheme (TGK: Tiedtke, Grell and Kuo) showed an improvement of 54 to 126% compared to the individual schemas. In general, it can be said that choosing the optimal configuration based on the idea of climate-based convection scheme can increase the performance of the RegCM4.5 regional model in seasonal precipitation forecast in Iran.

Although a study with a regional climatic zones perspective was not found on Iran, but some studies have found the Tiedtke scheme suitable for our country (Alizadeh Choubari et al., 1398), which with the findings of this study in which the Tiedke scheme for four of the six climates used in this study are considered appropriate. On the other hand, Zarrin and Dadashi (1399) used the Grell scheme to study the events of the partial rainfall in Iran by RegCM4 model, which in this study was found to be suitable for semi-arid climate. In addition, it was observed that in the study period of seven months, the most RMSE error occurred in April, which is the month of transition from cold to warm season.