Geospatial and Machine Learning Approaches to Forest Ecological Vulnerability Assessment: A Systematic Scoping Review and Methodological Synthesis

Abstract

Forest ecosystems are increasingly vulnerable to climatic variability and anthropogenic pressures, necessitating robust, scalable, and reproducible assessment approaches. Over the past three decades, remote sensing, geographic information systems, and machine learning techniques have been widely applied to evaluate forest ecological vulnerability; however, a consolidated synthesis of computational methods, indicator frameworks, and modelling paradigms remains limited. This study presents a systematic scoping review of global research on forest ecological vulnerability assessment using geospatial and machine learning approaches. Literature published between 1995 and 2024 was retrieved from Scopus, Web of Science, and Google Scholar following a transparent PRISMA-ScR–guided screening protocol, resulting in the inclusion of 136 studies for qualitative synthesis. The review synthesizes vulnerability indicators structured under exposure, sensitivity, and adaptive capacity, and critically examines the methodological evolution from index-based and statistical models to advanced machine learning and hybrid simulation frameworks. Results indicate a growing reliance on ensemble learning, cloud-based platforms, and high-resolution spatial data, alongside persistent methodological limitations related to model validation, spatial scale dependency, interpretability, and transferability. This study provides an integrated methodological synthesis that combines vulnerability indicator frameworks, geospatial modelling approaches, and emerging machine learning paradigms to identify key methodological trends and research gaps in forest ecological vulnerability assessment. Based on the synthesis of reviewed studies, a conceptual methodological framework is proposed to guide future forest ecological vulnerability assessments by integrating indicator classification, geospatial data processing, predictive modelling, and validation strategies. By integrating indicators, modelling approaches, and computational platforms, this review provides a guide for future ecological informatics research toward transparent, reproducible, and decision-oriented forest vulnerability assessments.