Data from year 2012-2015 are separated into dry (15.Oct – 14.May) and Wet season (15.May – 14.Oct). Upperair data gathering from 4 weather stations of Department of Royal Rainmaking and Agricultural Aviation (DRRAA), had been used as training data to model. Whereas, storm properties derived from DRRAA’s 4 weather radar stations and rainfall amount data from rain gauge networks of Thai meteorological department (TMD) and Hydro and Agro Informatics Institute (HAII) installed within each radar measured ranges had been used as observed data for validation (Test data). Correlation coefficient ® statistics is used for upperair indices selection. The statistical comparison of new and old model had been made using skill score of Heidke and Peirce Skill Score (HSS), Peirce Skill Score (PSS), Gandin-Murphy Skill Score (GMSS) and accuracy.
According to results of study, old models have 8 input indices in dry and wet season, upper northern part has 9 input indices in dry season and 9 input indices in wet season, central part has 27 input indices in dry season and 11 input indices in wet season, eastern part has 44 input indices in dry season and 13 input indices in wet season, and north eastern part has 10 input indices in dry season and 7 input indices in wet season. Therefore, the implication upperair indices are difference in each regions and season. Statistical skill score performed between new and old model found out that the percentage of accuracy and scores for each region of new model are higher than old models for every scores and season.