Estimation of Air Temperature using Data Driven Techniques Based on Best Subset Regression Model in Semi-Arid Environment

Abstract

Abstract Temperature considers one of the most important factors in the estimation of agricultural water requirements, hydrological processes and climate change studies. In order to determine the most accurate prediction model in a semi-arid environment for the daily minimum and maximum temperature (Tmax and Tmin), linear regression (LR), additive regression (AR), support vector machine (SVM), random subspace (RSS), the M5 pruned (M5P) models were compared in this study. Gharbia Governorate was selected as one of the most important governorates in the Nile Delta, Egypt, to conduct this work for the prediction of Tmax and Tmin daily. Datasets were collected from 1979 to 2014 and divided into 75% for training and 25% for testing. The best subset regression model was applied to select the model input combinations. Both minimum and maximum temperatures exhibit large magnitudes of the auto-correlation function (ACF) and partial auto-correlation function (PACF) for lag periods spanning from 1 to 8 days. In addition, as a result of the regression analysis, Tmax(t−1), Tmax(t−2), Tmax(t−3), Tmax(t−4), Tmax(t−5), Tmax(t−6), Tmax(t−8) variables showing the most appropriate statistical performance were determined as the most suitable model combination. Various statistical indicators and graphical approaches were used to select the most appropriate model. LR, M5P and SVM models outperformed the other two for minimum temperature prediction in all testing and cross-validation periods. The M5P model outperformed the LR and SVM models by effectively accommodating both high and low observed values. The M5P model outperformed the LR, AR, RSS, M5P and SVM models in predicting maximum temperatures. Compared to other models, the LR model closely emulated the performance of the M5P model in simulating maximum temperatures. The results of this study can assist decision-makers in water resource management, reservoir optimization, irrigation, and agricultural production activities.