AI-BASED ENVIRONMENTAL RISK MANAGEMENT FOR CLIMATE CHANGE: CURRENT ADVANCES, CHALLENGES, AND FUTURE DIRECTIONS

Abstract

Climate change is the enhancement of the spatial complexity, magnitude and frequency of environmental hazards, thereby enhancing the need for integrated and adaptive environmental risk assessment and management. While physics-based models of climate and hazards continue to be fundamental, they may suffer from limitations in terms of computational complexity, spatial resolution, availability of heterogenous data, and timely incorporation of new and rapidly evolving environmental and socioeconomic data. The review focuses on the use of AI-aided approaches to climate risk assessment for the environment, such as machine learning, deep learning, and geospatial AI, and hybrid physics–AI models. AI can combine satellite data, GIS layers, IoT sensors, climate data, and socioeconomic data to enhance hazard prediction, specifically for floods, wildfires, droughts, extreme heat and other climate hazards. In addition to hazard forecasting, AI can support the exposure characterization, vulnerability assessment, dynamic risk estimation, risk prioritization, and early warning, preparedness, adaptation, resource allocation, and long-term risk reduction decision support. There are, however, many key challenges to overcome such as data scarcity, data quality, geographic transferability, data retrofitting, model interpretability, uncertainty quantification, model computational demands, governance, equity, and operational readiness. Hybrid physics–AI architectures, prospective validation, uncertainty-aware and explainable models, integrated multi-hazard frameworks, satellite–IoT and edge-computing systems and responsible human oversight are future areas to be prioritized for research. In summary, AI has significant potential to complement existing environmental science and help move away from reactive disaster management to dynamic, evidence and adaptive governance of climate risk.