Artificial Intelligence in Climate Science: A State-of-the-Art Review (2020–2025)

Climate change is intensifying extreme weather events and posing complex challenges to human and natural systems. In parallel, rapid advances in artificial intelligence (AI) and machine learning (ML) offer new tools to tackle climate science problems. AI can analyze vast climate datasets, improve forecasts, and optimize climate solutions at scales and speeds beyond traditional methods. For example, recent analyses estimate that scaling AI applications could help reduce 5–10% of global greenhouse gas emissions by 2030. At the same time, AI can bolster climate adaptation and resilience by enhancing predictive models and decision support systems. This review surveys the state-of-the-art (2020–2025) in applying AI to climate science across five key areas: (1) extreme weather prediction and nowcasting, (2) carbon emissions tracking and estimation, (3) climate change adaptation and mitigation planning, (4) climate model emulation and downscaling, and (5) climate-related decision support systems. For each subtopic, we highlight recent developments, AI/ML methods (e.g. deep learning, graph neural networks, transformers, physics-informed models), important datasets and benchmarks, performance metrics, and technical challenges such as data sparsity, interpretability, and generalizability. Cross-cutting themes and future research directions are discussed. The goal is to provide a comprehensive, technical yet accessible overview of how AI is transforming climate science, and to identify opportunities and hurdles on the path toward robust, AI-enhanced climate solutions.