The Predictive Capability of a Novel Ensemble Tree-Based Algorithm for Assessing Groundwater Potential

Abstract

Understanding the potential groundwater resource distribution is critical for sustainable groundwater development, conservation, and management strategies. This study analyzes and maps the groundwater potential in Busan Metropolitan City, South Korea, using random forest (RF), gradient boosting machine (GBM), and extreme gradient boosting (XGB) methods. Fourteen groundwater conditioning factors were evaluated for their contribution to groundwater potential assessment using an elastic net. Curvature, the stream power index, the distance from drainage, lineament density, and fault density were excluded from the subsequent analysis, while nine other factors were used to create groundwater potential maps (GMPs) using the RF, GBM, and XGB models. The accuracy of the resultant GPMs was tested using receiver operating characteristic curves and the seed cell area index, and the results were compared. The analysis showed that the three models used in this study satisfactorily predicted the spatial distribution of groundwater in the study area. In particular, the XGB model showed the highest prediction accuracy (0.818), followed by the GBM (0.802) and the RF models (0.794). The XGB model, which is the most recently developed technique, was found to best contribute to improving the accuracy of the GPMs. These results contribute to the establishment of a sustainable management plan for groundwater resources in the study area.

Spatial Prediction of Groundwater Withdrawal Potential Using Shallow, Hybrid, and Deep Learning Algorithms in the Toudgha Oasis, Southeast Morocco

“…Several scholars recommend integrating sampling data from areas without groundwater withdrawal in order to balance the input dataset ( [8,13,20] and references therein; therefore, the list of the water withdrawal systems (set to 1, i.e., a high GWP) inventoried was completed using the same number of locations with no water withdrawal systems (set to 0, i.e., a very low GWP), and it was randomly mapped using the function create random points in a GIS environment, resulting in a total of 884 points (Figure 1). At this step, a variation in the percentage of the data split (training/testing) has been noted among the most recent studies, in which the division of 70/30% was the most used partition [14,23,25] in addition to the partition 75/25% and 80/20%, which have been used by Namous et al [20] and Talukdar et al [9], respectively.…”

“…Park and Kim [13] XGraB Chen and Guestrin [74] eXtreme Gradient Boosting is an improved GraB algorithm with a structure offering parallel tree boosting. It employs second-order derivatives that reduce the loss function and provide more accurate trees.…”

“…Naghibi et al [21], Park and Kim [13] DLNN Kingma and Ba [75] Deep learning neural network is an advanced ANN that consists of several layers: the input layer, several hidden layers, and the output layer.…”

“…The used models were: multilayer perceptron (MLP), k-nearest neighbor (KNN), DT, SVM, voting, RF, adaptive boosting (AdB), gradient boosting (GraB), eXtreme gradient boosting (XGraB), and deep learning neural network (DLNN). These algorithms were selected due to them being the most commonly used individual models; this allowed us to compare our findings with previous results demonstrated in different environments as investigated by [13,20,23,29], etc. Furthermore, they are the most powerful boosting and deep learning models, and we used the models in order to overcome the research gap related to the comparison and validation of different types of GWP modeling algorithms.…”

“…Previous studies have shown that common indicators such as hydrological, geological, topographical, and climatic indicators were integrated according to their influence on the GWP using several mathematical and statistical approaches. Among the latter, the frequency ratio, weights-of-evidence, and logistic regression are the approaches usually used by scholars [12][13][14].…”