Prediction of groundwater fluctuations in Chengalpattu district through machine learning

Abstract

Groundwater, found beneath the Earth’s surface in saturated zones of soil, sediment, and rock, plays a crucial role in sustaining ecosystems and supporting human activities like agriculture and industry. Monitoring and managing groundwater resources are crucial for sustainable use. As of the latest update in January 2022, Chennai, a city in southern India, has been grappling with water scarcity issues. The city has faced recurrent water shortages due to various factors, including rapid urbanization, inadequate infrastructure, depleting groundwater levels, and irregular rainfall patterns. Chengalpattu district in Tamil Nadu, India, is known for its diverse geographical features, incorporating urban and rural landscapes, and is significant for agriculture and water resource management. This study focuses on predicting variations in groundwater levels in open wells at different locations in the Chengalpattu district, assessing the effectiveness of various machine-learning models. This paper utilizes the ARIMA model provided by the stats models library. This library is widely employed for statistical modelling and hypothesis testing in Python, and encompasses a range of tools for time series analysis, including the ARIMA model. In this context. ARIMA models are employed for predicting future depth which focus on depicting autocorrelations in the data. Additionally, a package consolidating various models, including Seasonal Naïve (a straightforward forecasting method for seasonal data that serves as a reliable benchmark by relying on the observation from the same period a season ago), was incorporated in this study.

Keywords

• groundwater level
• forecasting
• fluctuations
• predictions
• machine learning.