Ahmed Makhlouf, Mustafa El-Rawy, Shinjiro Kanae, Mahmoud Sharaan, Ali Nada, Mona G. Ibrahim
{"title":"Streamlining the monitoring and assessment of irrigation groundwater quality using machine learning techniques","authors":"Ahmed Makhlouf, Mustafa El-Rawy, Shinjiro Kanae, Mahmoud Sharaan, Ali Nada, Mona G. Ibrahim","doi":"10.1007/s12665-025-12132-4","DOIUrl":null,"url":null,"abstract":"<div><p>Continuous evaluation of groundwater quality is vital for ensuring its long-term sustainability. However, traditional assessment methods for various purposes face challenges due to cost and time constraints. In this study, machine learning (ML) models, including Gaussian Process Regression (GPR), Decision Tree (DT), Support Vector Regression (SVR), and Artificial Neural Network (ANN), were employed to predict five irrigation water quality (IWQ) indices using only physical parameters (electrical conductivity (EC) and pH) and site conditions (Elevation, depth to water table, and distance to river). A dataset of 246 groundwater samples from the Eocene aquifer in Minia, Egypt, was collected and analyzed to measure groundwater quality parameters. Five combinations of the input parameters were utilized to calculate IWQ indices: sodium adsorption ratio (SAR), sodium percentage (Na %), total hardness (TH), permeability index (PI), and Kell’s ratio (KR). ML models were developed to estimate IWQ parameters based solely on physical measurements and site conditions. The results revealed that GPR, DT, SVR, and ANN strongly predicted all IWQ parameters during training. The results demonstrated that GPR accurately predicted groundwater quality, followed by DT, SVR, and ANN. The best performance of the GPR model was achieved during the fourth combination, which includes EC and distance to the river. The evaluation of GPR through the fourth combination revealed the highest accuracy with a correlation coefficient of 0.97, 0.82, 0.96, 0.87, and 0.81 in predicting SAR, %Na, TH, PI, and KR. The study emphasizes the capacity of machine learning models to efficiently employ readily available and quantifiable field data to predict IWQ characteristics. Moreover, the research findings, contributing to the second goal of the Sustainable Development Goals (SDGs), “No Hunger,” and the sixth goal, “Clean water and sanitation,” have the potential to enhance agricultural productivity and water conservation.</p></div>","PeriodicalId":542,"journal":{"name":"Environmental Earth Sciences","volume":"84 5","pages":""},"PeriodicalIF":2.8000,"publicationDate":"2025-03-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s12665-025-12132-4.pdf","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Environmental Earth Sciences","FirstCategoryId":"93","ListUrlMain":"https://link.springer.com/article/10.1007/s12665-025-12132-4","RegionNum":4,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"ENVIRONMENTAL SCIENCES","Score":null,"Total":0}
引用次数: 0
Abstract
Continuous evaluation of groundwater quality is vital for ensuring its long-term sustainability. However, traditional assessment methods for various purposes face challenges due to cost and time constraints. In this study, machine learning (ML) models, including Gaussian Process Regression (GPR), Decision Tree (DT), Support Vector Regression (SVR), and Artificial Neural Network (ANN), were employed to predict five irrigation water quality (IWQ) indices using only physical parameters (electrical conductivity (EC) and pH) and site conditions (Elevation, depth to water table, and distance to river). A dataset of 246 groundwater samples from the Eocene aquifer in Minia, Egypt, was collected and analyzed to measure groundwater quality parameters. Five combinations of the input parameters were utilized to calculate IWQ indices: sodium adsorption ratio (SAR), sodium percentage (Na %), total hardness (TH), permeability index (PI), and Kell’s ratio (KR). ML models were developed to estimate IWQ parameters based solely on physical measurements and site conditions. The results revealed that GPR, DT, SVR, and ANN strongly predicted all IWQ parameters during training. The results demonstrated that GPR accurately predicted groundwater quality, followed by DT, SVR, and ANN. The best performance of the GPR model was achieved during the fourth combination, which includes EC and distance to the river. The evaluation of GPR through the fourth combination revealed the highest accuracy with a correlation coefficient of 0.97, 0.82, 0.96, 0.87, and 0.81 in predicting SAR, %Na, TH, PI, and KR. The study emphasizes the capacity of machine learning models to efficiently employ readily available and quantifiable field data to predict IWQ characteristics. Moreover, the research findings, contributing to the second goal of the Sustainable Development Goals (SDGs), “No Hunger,” and the sixth goal, “Clean water and sanitation,” have the potential to enhance agricultural productivity and water conservation.
期刊介绍:
Environmental Earth Sciences is an international multidisciplinary journal concerned with all aspects of interaction between humans, natural resources, ecosystems, special climates or unique geographic zones, and the earth:
Water and soil contamination caused by waste management and disposal practices
Environmental problems associated with transportation by land, air, or water
Geological processes that may impact biosystems or humans
Man-made or naturally occurring geological or hydrological hazards
Environmental problems associated with the recovery of materials from the earth
Environmental problems caused by extraction of minerals, coal, and ores, as well as oil and gas, water and alternative energy sources
Environmental impacts of exploration and recultivation – Environmental impacts of hazardous materials
Management of environmental data and information in data banks and information systems
Dissemination of knowledge on techniques, methods, approaches and experiences to improve and remediate the environment
In pursuit of these topics, the geoscientific disciplines are invited to contribute their knowledge and experience. Major disciplines include: hydrogeology, hydrochemistry, geochemistry, geophysics, engineering geology, remediation science, natural resources management, environmental climatology and biota, environmental geography, soil science and geomicrobiology.