Prediction of water temperature of Sebou estuary (Morocco) using ANN and LR

Abstract

Water temperature is an important component for river water quality. Good knowledge of river thermal regime is critical for aquatic resources management and environmental impact studies. This study aims to predicting surface water temperature at a given site of Sebou estuary (Morocco) from air temperature, using artificial intelligence applied to neural networks (NN) and linear regression (LR). The models used were applied to an hourly temperature database obtained by simulating hydraulic and thermal regime of the estuary using HEC-RAS model. Temperature data (1560) was divided into training and validation series. The results showed that coefficient of determination for training was 86.26%(NN) and 79.94%(LR). For validation it was 91.23%(NN) and 80.3%(LR). Moreover the residuals generally vary between −1.5°C and +1.5°C, no trend was noted as function of estimated temperature. Hence, it is possible to predict water temperature of estuary using air temperature. Additionally, NN model gave slightly results compared to LR.