AI for Restoring Degraded Lands: Mapping Degradation, Predicting Climate Risk and Valuing Ecosystem Services

Land degradation is a major global challenge, with up to 40% of the world's land estimated to be degraded, affecting more than 3 billion people and weakening food security, biodiversity, climate resilience and ecosystem services. In this context, the objective of this review is to examine how artificial intelligence (AI), remote sensing and geospatial analytics can support land restoration by integrating three connected domains: land degradation mapping, climate-risk prediction and ecosystem service valuation. The review focuses on AI-based approaches for identifying degradation hotspots, assessing vegetation and soil stress, predicting drought, heat, evapotranspiration and fire-related risks, and estimating ecosystem services such as carbon sequestration, soil retention, water regulation, forage productivity and biodiversity support. Methodologically, the paper adopts a thematic synthesis of peer-reviewed literature, with emphasis on machine learning, deep learning, hybrid geospatial models and multi-source data integration. The synthesis indicates that AI can improve restoration planning by strengthening spatial diagnosis, capturing nonlinear land-climate interactions, anticipating future risk and estimating likely ecosystem service gains from restoration interventions. However, the review also finds that operational adoption remains constrained by data gaps, limited field validation, scale mismatch, uncertainty and weak interpretability of complex models. The paper therefore argues for a shift from isolated AI applications towards integrated, explainable and decision-oriented frameworks that combine Earth Observation, climate data, field evidence and ecosystem service indicators. Future research should prioritise groundvalidated datasets, multi-scale modelling, uncertainty reporting, local ecological knowledge and operational decision-support systems. Such integration can help identify where restoration is most urgently needed, where it is most likely to succeed, and what ecological and livelihood benefits it can generate. This review provides state-of-the-art insights on using AI-enabled land degradation mapping, climate-risk prediction and ecosystem service valuation as decision-support tools for sustainable land management, ecosystem resilience and evidence-based restoration of degraded landscapes.