New method of prediction of climatic conditions by artificial neural networks (case of the province of Médéa)

Abstract

The predictions of climate conditions have important impacts on people's lives through agriculture, food security, water resources management, health, natural disasters and environmental degradation, etc. The main focus of this paper is to develop an accurate mathematical model using the artificial neural network (ANN) method, to predict weather conditions such as: precipitation, wind speed, humidity and average temperature, based on a database composed of: the earthsun distance, the annual evolution (time), the sunshine duration and the conversion factor, measured over a period from January 01, 2001 to December 31, 2010 for the city of Médéa (Algeria), the said data are normalized between 0 and 1, the number of neurons in the hidden layer of ANN, single layer MLP type is10 neurons. The architecture of the ANN is of MLP type with a single layer of 10 neurons with a tangential sigmoid function as activation function, and a learning algorithm of Levenberg-Marquardt type, with a mean square error RMSE = 0.1202 and a correlation coefficient R2 =0.8289. In addition, a graphical interface was developed by the Guide function programmed in Matlab in order to facilitate the use of the mathematical model established for the user