Abstract： Climate attribution aims to identify and quantify the drivers of climate change and extreme events, providing a scientific basis for understanding climate-system evolution, assessing anthropogenic influences, and supporting climate adaptation and mitigation. Traditional approaches, including regression, correlation analysis, and optimal fingerprinting, have developed relatively mature frameworks for detecting and attributing human influences on global and regional climate change. However, the climate system is characterized by strong nonlinearity, multiscale coupling, time lags, and multiple interacting drivers, making it difficult for conventional methods alone to resolve directional interactions and transmission pathways. In recent years, causal inference methods, such as Granger causality, transfer entropy, graphical models, convergent cross mapping, and counterfactual inference, have been increasingly introduced into Earth system science, extending climate attribution from signal detection toward process and mechanism interpretation. Importantly, these methods encode different meanings of causality: Granger causality reflects predictive gain, graphical models identify candidate causal structures through conditional independence, whereas counterfactual approaches focus on changes in outcomes under interventions or alternative scenarios. This review summarizes the theoretical foundations, applicability, and recent applications of major causal inference methods in trend attribution, extreme events, teleconnections, and climate-model evaluation. Current challenges include hidden variables, limited effective sample sizes, nonstationarity, and spatial heterogeneity. Future progress will require hybrid attribution frameworks that integrate physical priors, observational causal discovery, model-based interventions, and multisource, multi-method robustness assessments.