Predictive modelling of water clarity based on remote sensing data using artificial neural network (ANN): A case study in Saguling Reservoir, West Java, Indonesia

Abstract

Indonesia faced several challenges regarding water quality such as water exploitation and contamination caused by human activities. Comprehensive and sustainable water management is required to ensure its availability for the society. Ecosystem quality monitoring is needed to make sure the availability of water resource all year round by using modelling to assist. This paper presented application of Artificial Neural Network (ANN) utilizing multilayer perception model with a backpropagation algorithm to predict water clarity in Saguling Reservoir provided by PT Indonesia Power. ANN performance of predicting water clarity level were evaluated using regression analysis (R2), Mean Absolute Error (MAE) and Mean Square Error (MSE). Based on the results, prediction data during rainy season shows better performance than dry season with R2 value of 0.94, MAE value of 0.035, and MSE value of 0.0032 meanwhile dry season data of R2, MAE, and MSE are 0.83, 0.041, and 0.0045 respectively. ANN prediction model demonstrated a relatively good prediction capability of water clarity and may be used as one of references in classifying the water quality level of Saguling Reservoir quantity and quality integrated maintenance.