Explainable Artificial Intelligence in Environmental Monitoring: A Systematic Literature Review

Le résumé fourni par la source

This repository contains the template, figures, tables and search protocol of the manuscript named "Explainable Artificial Intelligence in Environmental Monitoring: A Systematic Literature Review". This study was designed and published by the following authors: José Andrés Neira Gastón Márquez Karen Muñoz Ernesto Vivanco The documents and files in the repository are described below: Images.zip: This file contains all images that explain both the search process and the results obtained in the review. Tables.zip: This file contains the source code for the tables described in the manuscript. Template.csv: This document summarizes the metadata of all the primary studies identified in the research. Protocol.pdf: This document summarizes the review protocol used in the study for the purpose of replicability of the results. Abstract Context This Systematic Literature Review (SLR) analyzes the application of Explainable Artificial Intelligence (XAI) in environmental monitoring and pollution source detection. Following the PRISMA methodology, 26 studies published between 2020 and 2025 were selected from major scientific databases. The review identifies SHAP, LIME, Grad-CAM, and Layer-wise Relevance Propagation (LRP) as the most widely used explainability techniques across air, water, and soil quality applications. Results show that XAI improves transparency by revealing the environmental factors driving model predictions, supporting more reliable and auditable decision-making. Despite recent advances, challenges remain regarding methodological standardization and regulatory adoption. The findings highlight the potential of XAI to enhance environmental governance through transparent, interpretable, and ethically responsible AI systems. Objective Analyze the state of the art of XAI techniques applied to environmental monitoring and the detection of polluting sources, identifying the most used methods, the most relevant environmental variables, and the opportunities and challenges for the development of transparent, auditable and sustainable decision systems. Methods A Systematic Literature Review (SLR) was conducted following the PRISMA methodology. After applying inclusion/exclusion criteria and a snowball sampling strategy, 26 studies published between 2020 and 2025 on explainability applied to environmental monitoring and pollution detection were selected and analyzed. Results The review shows that Explainable Artificial Intelligence (XAI) techniques, especially SHAP, LIME, Grad-CAM, and LRP, significantly improve the transparency and interpretability of AI models applied to environmental monitoring. The studies analyzed demonstrate that XAI allows for the identification of the most influential environmental variables in predictions, facilitates the detection and location of polluting sources, strengthens confidence in the results, and supports more informed and auditable decision-making. However, challenges remain related to the lack of methodological standards and regulatory frameworks that would facilitate its adoption in environmental management. Conclusions Explainable Artificial Intelligence (XAI) allows the transformation of "black box" environmental models into transparent, interpretable, and auditable systems, improving the detection of pollutants and supporting more reliable and sustainable environmental decision-making.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.