Prediction of Climatic Parameters from Physicochemical Parameters using Artificial Neural Networks: Case Study of Ain Defla (Algeria)

Abstract

The knowledge of the climate of a region is a primordial task in that it allows predictions of climatic parameters in the future. In this study, monthly maximum and minimum air temperature (Tair,min, Tair,max), relative humidity (RH), and sunshine duration (SD) were modelled by multiple linear regression (MLR), and multilayer perceptron methods (MLP). For the four climatic parameters, the internal and external validations of MLP-ANN model showed high R2 and Q2 values in the range 0.81–0.98. The agreement between calculated and experimental values confirmed the ability of ANN-based equation to predict these parameters quickly and at lower cost.

Keywords

• neural network
• modelling
• physicochemical parameters
• climatic parameters
• physicochemical parameters