Assessing water quality of a lake using combination of drone images and artificial intelligence models

Abstract

Lakes serve as a source of water to meet the demand of various sectors such as urban, agricultural and recreational sectors. The purpose of this paper is to investigate the capability of using combination of multispectral drone imagery with machine learning algorithm for the assessment of water quality in an artificial lake at the Jordan University of Science and Technology (JUST) campus. Several images with different resolutions under different wavebands were captured with DJI Phantom-4 drone equipped with sensors in the blue green, red, Red Edge, and Near Infrared. At the same time water samples were also collected from ten different points in the lake to analyze physical and chemical quality parameters. The spectral reflection was used to calculate multiple water body indices, and the resulting indices were correlated to water quality parameters. The indices with coefficient of determination greater than 0.7 were used to develop various artificial intelligence models (AI) such as Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), Gradient Boosted Decision Trees (GBDT), Generalized Linear Model (GLM) and Artificial Neural Network (ANN). The results showed that among the tested models autoregressive with exogenous (NARX) ANN model has the highest prediction accuracy based on the coefficient of determination (R²) of 0.95 and relative error of 0.034. Comparison of the simulated results indicated the variability of water quality parameters with seasons and inversion accuracy was highest during the summer season. Such an approach offers a useful tool for decision-making to manage lake water quality. Future studies should include more parameters and using hyperspectral sensors for investigating quality parameters of similar water bodies.