Prediction of long-term monthly air temperature using geographical inputs

Abstract

Air temperature as a major climatic component is important in land evaluation, water resources planning and management, irrigation scheduling and agro-hydrologic planning. In this paper, the capabilities of Adaptive Neuro-Fuzzy Inference System (ANFIS) and Artificial Neural Networks (ANNs) were evaluated in predicting long-term monthly air temperature values at 30 weather stations of Iran. Monthly data of 20 weather stations were used for training and 10 stations' data were used for testing. Consequently, the periodicity component, station latitude, longitude and altitude values were introduced as input variable to predict the long-term monthly temperature values. The estimates of the ANFIS and ANN models were compared with each other with respect to root mean-squared error, mean absolute error and determination coefficient statistics. The ANN models generally performed better than the ANFIS model in the test period. For the ANN model, the maximum and minimum determination coefficient values were found to be 0.995 and 0.921 in Semnan and Bandar-e-Abbas meteorological stations, respectively. The maximum and minimum determination coefficient values were found as 0.999 and 0.876 for the ANFIS model in Shiraz and Bandar-e-Abbas stations. Copyright © 2013 Royal Meteorological Society