Work

Understanding multi-hazard risk dynamics between land use, climate extremes and water quality leveraging artificial intelligence

Year: 2025

Type: article

Abstract: This study develops an interpretable machine learning framework by integrating Random Forest (RF) classifiers with SHapley Additive exPlanations (SHAP) to analyze the interactions between land use, climate extremes, and water quality at a high spatial resolution in Veneto Region, Italy. The framework predicts annual impacts (2010–2022) on water quality, supporting the evaluation of ecological status in line with the Water Framework Directive. Three RF models achieved accuracies of 0.87, 0.81, and 0.85 in predicting impacts on physico-chemical elements, specific pollutants, and biological alterations, respectively. SHAP values helped interpret the model outputs by identifying key variables and their contribution to impacts on water quality. Temperature extremes, especially when combined with drought, emerged as key drivers of degradation. Natural features (e.g., vegetation, river naturality, permeable soils) buffered impacts, while intensive anthropogenic land use amplified negative effects under extreme climate conditions. This approach supports decision-making by identifying multi-hazard risks and emphasizing nature-based solutions, ecosystem restoration, and strategic land use for climate-resilient water management.

Source: Ecological Indicators

DOI: 10.1016/j.ecolind.2025.114494

Authors: Diep Ngoc Nguyen, Jacopo Furlanetto, Andrea Critto, Silvia Torresan

Institutions: Ca' Foscari University of Venice, CMCC Foundation - Euro-Mediterranean Center on Climate Change, University of Palermo

Language: English

Topic: Water Quality and Pollution Assessment

Subfield: Water Science and Technology

Field: Environmental Science

Domain: Physical Sciences

SDGs: Clean water and sanitation, Climate action, Life on land, Sustainable cities and communities

Open Access status: gold

Estimated APC paid: $3,650