برآورد احتمال وقوع آلودگی نیترات آب زیرزمینی با استفاده از الگوریتم تقویت گرادیان شدید

Groundwater, as a vital source of fresh water, plays a fundamental role in supplying drinking, agricultural, and industrial needs in many arid and semi-arid regions worldwide. However, increased human and industrial activities have led to the exacerbation of pollution in these valuable resources. Among these, nitrate pollution, due to its high solubility and mobility in water, is recognized as one of the most serious threats to human health and aquatic ecosystems. Consumption of nitrate-contaminated water can lead to various diseases, including methemoglobinemia (blue baby syndrome) in infants and even some cancers in adults. Furthermore, the entry of nitrates into surface waters can result in eutrophication and the degradation of aquatic ecosystems. Given the importance of the issue and the necessity of protecting groundwater resources, this research was conducted with the aim of developing an integrated and comprehensive framework for predicting the probability of groundwater pollution, especially with a focus on nitrate contaminant, in the Lenjanat Plain region located in Isfahan Province, Iran. This framework, using novel modeling and spatial analysis approaches, will help identify areas susceptible to pollution and provide effective management solutions to reduce the risks associated with groundwater contamination. The results of this research can serve as a basis for future planning in the sustainable management of water resources and the protection of community health.
This study meticulously investigated nitrate concentration data in groundwater sources. To this end, crucial information was collected from 102 wells, each representing the nitrate status in the groundwater aquifers of the studied region. To analyze this extensive dataset and extract hidden patterns, the Extreme Gradient Boosting was employed. This model was chosen due to its high capability in identifying complex and non-linear relationships between variables, as well as its acceptable prediction precision. In addition to nitrate concentration data, ten key environmental and anthropogenic factors potentially influencing nitrate contamination in groundwater were identified and incorporated into the analytical model. These factors included land slope, elevation, drainage density, topographic wetness index, soil order, distance from streams, lithology, and land use. By integrating these eight factors into the Extreme Gradient Boosting model, it was possible to identify the most significant factors affecting nitrate contamination and also to spatially predict the probability of nitrate contamination in groundwater.
The results of this study clearly demonstrated the effectiveness and efficiency of the Extreme Gradient Boosting in predicting nitrate contamination in groundwater. This model, with an overall precision of 0.86, was able to distinguish the contamination status across the studied area. In addition, other performance evaluation criteria of the model also indicated its high ability to correctly identify contaminated and uncontaminated cases; so that the area under the ROC curve was equal to 0.85. Also, recall metric with a value of 0.80 indicates that the model was able to correctly identify 80% of all real infected cases. Finally, the F1-score statistic, which is a combined measure of precision and recall, with a value of 0.83, indicates a good balance between these two measures and the overall reliable performance of the model. Sensitivity analysis of the model revealed that some input variables had a significant impact on the spatial prediction process of groundwater nitrate contamination. Among the ten environmental and anthropogenic factors examined, precipitation and elevation changes were identified as the most influential and important variables in determining the spatial pattern of nitrate contamination. These findings highlight the importance of natural and geomorphological characteristics of the region in controlling the dispersion and accumulation of nitrates in groundwater and can serve as a useful guide for future studies and the development of targeted management strategies.
One of the significant achievements of this study was the generation of hazard maps, which clearly delineated areas with high nitrate contamination risk in the central part of the studied plain. These maps provide valuable tools for water resource managers and urban and rural planners to focus preventive and control measures on sensitive areas. Notably, the significant overlap of these high-risk areas with agricultural land use unequivocally confirmed the role of human activities in increasing the risk of nitrate contamination. This finding emphasizes the necessity of sustainable management of agricultural activities (such as optimal use of nitrogen fertilizers) to protect groundwater resources.