|
|
|
|
 |
Volume 19, Issue 71 (12-2025) |
 |
|
|
|
|
Locating Areas Susceptible to Groundwater Pollution Using Maximum Entropy Model (Case Study: Badavar Basin)
|
Ali Haghizadeh *   , Zeynab Hajizadeh , Leila Ghasemi |
|
|
|
Abstract: (18 Views) |
Introduction
In arid and semi-arid regions, groundwater is a critical water source for industrial, agricultural, and drinking purposes, making its quality preservation vital for public health and sustainable development. The quality of this resource is significantly influenced by the complex interplay between human pressures such as pollution discharge and land-use changes and natural elements like topography and geology. A major environmental concern is the deterioration of groundwater quality due to excessive extraction and pollutant intrusion. Consequently, identifying vulnerable areas and conducting spatial risk zoning are essential processes in water resource management. Conventional models often have limitations in addressing these nonlinear, multivariate problems. The Maximum Entropy algorithm (MaxEnt), a powerful machine learning technique, is well-suited for this application due to its high capacity for analyzing presence-only data and modeling complex spatial patterns within a Geographic Information System (GIS) environment. The Badavar Basin, characterized by geological complexity and intensive land-use changes, is particularly vulnerable to contamination. The primary goal of this study is to employ the MaxEnt model to assess the regional pollution potential, determine the relative importance of various environmental factors, and generate a precise risk map to support management decisions. This research also highlights the importance of integrating scientific innovation with sustainable policy frameworks.
Materials and Methods
This study modeled groundwater contamination potential in the Badavar Basin using the Maximum Entropy (MaxEnt) model within a GIS framework. The input data consisted of pollution presence points, identified through sampling and monitoring of contaminated wells, and twelve environmental predictor variables. These variables were classified into three primary categories: land features (e.g., soil texture and geology), physiographic elements (e.g., slope, aspect, slope length, Topographic Wetness Index (TWI), and curvature), and distance/human factors (e.g., vegetation cover, distance from roads, distance from rivers, and land use). All variable layers were processed in GIS and converted to a raster format with a consistent resolution. The model was trained using 70% of the presence points, while the remaining 30% were reserved for testing and validation. The model's accuracy and performance were evaluated by calculating the Area Under the Curve (AUC) of the Receiver Operating Characteristic (ROC) curve for both datasets. Furthermore, a Jackknife test was performed to assess the sensitivity and relative importance of each factor, thereby identifying the primary controlling mechanisms in pollution prediction. To strengthen the methodological framework, data preprocessing ensured consistency across spatial layers and minimized potential bias. The resulting outputs included contamination risk maps, which provide valuable insights for sustainable groundwater management and regional planning initiatives.
Results and Discussion
The results from the MaxEnt model revealed the disproportionate contribution of various environmental factors to groundwater contamination likelihood. Geology was the most significant predictor, with a relative contribution of 36.7%, underscoring the importance of formation permeability and its hydrological role in controlling pollutant transport. Slope (17.7%) and distance from roads (12.6%) were the second and third most important factors, respectively, highlighting the significance of lateral pollutant transmission and anthropogenic influence across the basin. In contrast, factors such as aspect (0.2%) and elevation (0.0%) had a negligible impact.
According to the risk zoning map, areas classified with High and Very High contamination potential comprise approximately 7.6% of the basin's total area. The southern part of the Badavar Plain was identified as the zone of highest risk, attributable to its extensive alluvial deposits, gentle slopes, and the likely concentration of human activities. The model demonstrated strong performance and a good fit to the contamination patterns, as indicated by a training AUC of 0.85. However, the testing AUC score of 0.59 suggests potential overfitting and indicates a need for more comprehensive sampling and parameter tuning to improve model generalizability. These findings emphasize the necessity of integrating geological assessments with planning to mitigate contamination risks. Moreover, the spatial outputs generated by the model provide a valuable baseline for future hydrogeological investigations and targeted management strategies.
Conclusion
This study successfully demonstrated the efficacy of the Maximum Entropy (MaxEnt) algorithm as a reliable technique for spatially assessing groundwater pollution risk in the Badavar Basin. The findings reveal that geology is the predominant factor regulating contamination processes, underscoring the necessity for conservation measures focused on the main aquifers' permeable formations. Furthermore, the clear identification of the southern plain as a high-priority zone for monitoring and mitigation allows for optimal resource allocation. The model, with its acceptable accuracy on training data, proves to be a valuable tool for assessing aquifer susceptibility in regions with similar hydrological settings. Future research should aim to enhance model generalizability by resolving the performance discrepancy between the training and testing datasets, potentially through advanced parameterization and an improved spatial distribution of sampling points. Ultimately, the resulting risk map serves as a vital spatial decision-support tool, empowering water resource managers to implement targeted, proactive conservation strategies. This forward-looking approach is essential for safeguarding the long-term quality and sustainability of groundwater resources. In addition, integrating MaxEnt outputs with socio-economic and land-use data could provide a more holistic understanding of contamination drivers. Such integration would enable managers to design adaptive policies that balance environmental protection with regional development needs, ensuring resilience under future climatic and anthropogenic pressures.
|
Article number: 1 |
| Keywords: MaxEnt, environmental factors, AUC, groundwater, pollution |
|
|
|
Type of Study: Research |
Subject:
Special
Received: 2025/10/20 | Revised: 2025/12/22 | Accepted: 2025/11/17 | Published: 2025/12/22 | ePublished: 2025/12/22
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
|
Haghizadeh A, Hajizadeh Z, Ghasemi L. Locating Areas Susceptible to Groundwater Pollution Using Maximum Entropy Model (Case Study: Badavar Basin). jwmseir 2025; 19 (71) : 1 URL: http://jwmsei.ir/article-1-1214-en.html
|
|
|
|
|
|
|
