{"title":"Application of machine learning in python for temporal groundwater level prediction","authors":"Tade Mule Asrade","doi":"10.1016/j.sesci.2025.100261","DOIUrl":null,"url":null,"abstract":"<div><div>Groundwater is a critical resource for sustaining agricultural, domestic, and ecological needs in the Upper Blue Nile Basin of Ethiopia, where rapid population growth and climate variability have intensified water stress. The Temecha River Catchment, part of this basin, faces recurrent droughts and declining groundwater levels, underscoring the need for effective groundwater management strategies. One promising approach is Managed Aquifer Recharge (MAR), whose success relies heavily on accurate groundwater level monitoring and forecasting. Although MAR decisions are influenced by hydrogeological and land use factors, hydraulic head data are essential for determining recharge timing and suitability. This study evaluates the performance of five machine learning models—Gradient Boosting Regression (GBR), Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), and Linear Regression (LR)—for predicting groundwater level fluctuations in the Temecha River Catchment from 1995 to 2023. Rainfall, temperature, and evapotranspiration were used as predictive variables. Among the models, GBR showed superior performance, with a Mean Absolute Error (MAE) of 0.07 m, Nash-Sutcliffe efficiency (NSE) of 0.7934 m, Coefficient of Determination (R<sup>2</sup>) of 0.7856, and Percent Bias (PBIAS) of −2.408 %. The results demonstrate GBR's effectiveness in groundwater level forecasting and support its application in data-scarce regions to inform sustainable water resource management. However, adaptation to other regions must consider local hydrogeological and climatic conditions.</div><div>One sentence summary: This study explores the use of machine learning techniques in Python to predict temporal groundwater levels in the Temecha River Catchment.</div></div>","PeriodicalId":54172,"journal":{"name":"Solid Earth Sciences","volume":"10 3","pages":"Article 100261"},"PeriodicalIF":2.0000,"publicationDate":"2025-08-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Solid Earth Sciences","FirstCategoryId":"89","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2451912X25000340","RegionNum":4,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"GEOSCIENCES, MULTIDISCIPLINARY","Score":null,"Total":0}
引用次数: 0
Abstract
Groundwater is a critical resource for sustaining agricultural, domestic, and ecological needs in the Upper Blue Nile Basin of Ethiopia, where rapid population growth and climate variability have intensified water stress. The Temecha River Catchment, part of this basin, faces recurrent droughts and declining groundwater levels, underscoring the need for effective groundwater management strategies. One promising approach is Managed Aquifer Recharge (MAR), whose success relies heavily on accurate groundwater level monitoring and forecasting. Although MAR decisions are influenced by hydrogeological and land use factors, hydraulic head data are essential for determining recharge timing and suitability. This study evaluates the performance of five machine learning models—Gradient Boosting Regression (GBR), Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), and Linear Regression (LR)—for predicting groundwater level fluctuations in the Temecha River Catchment from 1995 to 2023. Rainfall, temperature, and evapotranspiration were used as predictive variables. Among the models, GBR showed superior performance, with a Mean Absolute Error (MAE) of 0.07 m, Nash-Sutcliffe efficiency (NSE) of 0.7934 m, Coefficient of Determination (R2) of 0.7856, and Percent Bias (PBIAS) of −2.408 %. The results demonstrate GBR's effectiveness in groundwater level forecasting and support its application in data-scarce regions to inform sustainable water resource management. However, adaptation to other regions must consider local hydrogeological and climatic conditions.
One sentence summary: This study explores the use of machine learning techniques in Python to predict temporal groundwater levels in the Temecha River Catchment.