Groundwater Potential Mapping Using GIS-Based Hybrid Artificial Intelligence Methods.

Area of Science:

• Earth and Environmental Sciences
• Artificial Intelligence in Hydrology

Background:

• Groundwater is a critical resource for communities, agriculture, and industries.
• Accurate groundwater potential mapping (GPM) is essential for sustainable resource management.
• Basaltic terrains present unique challenges for hydrological studies.

Purpose of the Study:

• To develop and evaluate novel hybrid artificial intelligence (AI) models for GPM.
• To assess the performance of modified RealAdaBoost (MRAB), bagging (BA), and rotation forest (RF) ensembles combined with functional tree (FT) classifiers.
• To identify the most accurate AI model for GPM in the DakLak province, Vietnam.

Main Methods:

• Development of three hybrid AI models: MRAB-FT, BA-FT, and RF-FT.
• Utilized geospatial techniques with data from 130 groundwater wells.
• Employed 12 topographical and geo-environmental factors, with One-R Attribute Evaluation for feature selection.
• Performance evaluation using statistical measures, including the area under the receiver operation curve (AUC).

Main Results:

• All developed hybrid AI models improved goodness-of-fit and prediction accuracy compared to the single functional tree model.
• The MRAB-FT model achieved the highest performance with an AUC of 0.742.
• MRAB-FT outperformed RF-FT (AUC=0.736), BA-FT (AUC=0.714), and the standalone FT model (AUC=0.674).

Conclusions:

• The MRAB-FT model demonstrates significant promise as an accurate AI hybrid technique for GPM.
• Accurate GPM facilitates effective aquifer recharge and sustainable groundwater resource management.
• The study highlights the potential of advanced AI techniques in addressing hydrological challenges in complex terrains.