Predicting Temperature from Ground-based Synoptic Data in Shiraz City, Iran

Abstract

In this work, we develop a Seasonal Auto-regressive Integrated Moving Average (SARIMA) model based on a greedy algorithm for forecasting the monthly average temperature via ground-based data. The observed data collected from synoptic station of Shiraz city in Fars, Iran. Also, we employ a Triple Exponential Smoothing (TES) method to predict temperature. The SARIMA model is compared with TES and two benchmarking methods, average and persistence based on root mean square error (RMSE). Our results show that in Shiraz, the SARIMA model performs out of others in long-term monthly prediction. The RMSE for the leading 12 months ahead is 1.07°C, 1.27°C, 8.73°C and 11.99°C for SARIMA, TES, Average and Persistence method, respectively.