Application of novel data-mining technique based nitrate concentration susceptibility prediction approach for coastal aquifers in India. (20th April 2022)

• Record Type: Journal Article
• Authors: Pal, Subodh Chandra; Ruidas, Dipankar; Saha, Asish; Islam, Abu Reza Md. Towfiqul; Chowdhuri, Indrajit
• Is Part Of: Journal of cleaner production. Volume 346(2022)
• Publication Date: 2022-04-20
• Subjects: Water resource -- Nitrate pollution -- Coastal aquifers -- Data mining approach -- K-fold cross validation; Factory and trade waste -- Management -- Periodicals; Manufactures -- Environmental aspects -- Periodicals; Déchets industriels -- Gestion -- Périodiques; Usines -- Aspect de l'environnement -- Périodiques; 628.5
• DOI: 10.1016/j.jclepro.2022.131205
• Languages: English
• ISSNs: 0959-6526

Abstract

Abstract: In water resource management and pollution control research, prediction of nitrate concentration in groundwater gets utmost priority in the last few years. Thus, our current research work aims to identify the nitrate susceptibility areas of coastal districts of eastern India using three data mining techniques of random forest (RF), boosting and bagging approach. To make groundwater nitrate concentration susceptibility map, fifteen nitrate conditioning factors were identified using multi-collinearity analysis and identify relative importance of nitrate variability using MDA method. The resampling method of four K-Fold cross validation (CV) technique was used to preparing inventory dataset and respective modelling purpose. Seven statistics methods including receiver operating characteristics-area under curve (ROC-AUC) and Taylor diagram have been applied for evaluating the performance of all applied models. The outcomes ensure that boosting model is more efficient followed by bagging and RF. Taylor diagram also revealed Boosting (r = 0.93) is most optimal model followed by Bagging (r = 0.89) and RF (r = 0.88). From aforementioned results, our study revealed that boosting is the well performed model to delineate groundwater nitrate concentrate susceptibility map (GNCSM) in regional level which also will be helpful to worldwide researcher to find out nitrate susceptibility zone in coastal environment and it may be fruitful to the different policy makers to take accurate decision for water management in the current study area. Graphical abstract: Image 1 Highlights: K-Fold CV resampling approach was employed to partition the modeling dataset. Boosting was the best predictive model (average AUC in training 0.929 and validation 0.917). The area of very high concentration of NO3 varies from 30.84 to 44.26%.