Assessment of Gini-, entropy- and ratio-based classification trees for groundwater potential modelling and prediction

Abstract

Artificial-intelligence and machine-learning algorithms are gaining the attention of researchers in the field of groundwater modelling. This study explored and assessed a new approach based on Gini-, entropy- and ratio-based classification trees to predict spatial patterns of groundwater potential in a mountainous region of Iran. To do this, a springs inventory was undertaken, and 362 springs were identified in the study area. A set of geo-environmental and topo-hydrological factors (slope, aspect, elevation, topographic wetness index, distance from fault, distance from river, precipitation, land use, lithology, plan curvature and roughness index) were used as predictors of groundwater. Results showed that Gini (AUC = 0.865) achieved the best results, followed by entropy (AUC = 0.847) and ratio (AUC = 0.859). Lithology was determined to be the variable with the best association with groundwater in the study area. These results indicate that all three algorithms provide robust models of groundwater potential in this mountainous region.

Highlights

• Gini, entropy and ratio were investigated for groundwater potential mapping.

• Eleven groundwater-affecting factors were considered.

• Lithology is the most important factor for groundwater potential mapping

• Gini based decision tree is the best, followed by entropy and ratio models

Keywords:

• Groundwater potential
• GIS
• data mining
• classification-tree analysis
• Iran

Data availability

Data are available upon any request

Disclosure statement

The authors declare no conflict of interest.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.