Development of daily temperature prediction model for Northeastern Thailand using artificial neural networks

Abstract

The aims of this study was to develop the daily temperature prediction models for Northeastern Thailand. Data of 23 northeastern meteorological stations from 1980-2011 were employed for models training with  10,000  data  patterns  while a 1-year observation of  2014  was  left  out  for  performance  comparison  purpose. The performance of models were calculated by the forecast accuracy on the set of previously selected days. The developed model of  daily  maximum  temperature forecast  (T_max  models)  and  daily  minimum  temperature  forecast  (T_min  models) forecast ahead 24-72 hours. As a result, MAE was in range of 0.85-1.95 for T_min and 0.85-2.68 for T_max while RMSE varied 1.16-2.32 and 1.51-3.19, respectively. R-square, moreover, showed good relationship between predicted temperature and actual temperature. Therefore, the 24-hour forecast model is better correlation than 48 hours and 72 hours, respectively.