Assessment of Groundwater Quality and Quantity Using Supervised Machine Learning

Abstract

With the urban development of Las Vegas region in the past 3 decades, the water usage has increased rapidly and, accordingly, there has been a growing interest in assessment of water quality and quantity in this region. The availability of water in Nevada is not commensurate with the water usage. Water quality assessment involves expensive and time-consuming laboratory tests and in-depth analysis of many parameters. This study implements a supervised machine learning algorithm that uses cost-free methods to predict groundwater quality characteristics, thereby helping to reduce the cost of extraction and laboratory analysis. This study employs five input parameters, namely, Temperature, Humidity, Precipitation, Soil moisture (pH), and Well depth. The study uses the “Gradient Boosting Trees” and “Gradient Descent” methods to train the data and to predict the output. The results of this study provide real-time predictions of groundwater quality and quantity parameters, including water levels, nitrate concentrations. The Mean Absolute Error (MAE) values range from 16.7 to 39.48. A limited set of input parameters is used to achieve a cost-effective solution.