Abstract Global climate models provide essential insights into future climate conditions but operate at spatial resolutions too coarse for many regional and local applications. Statistical downscaling has emerged as a key approach to bridge this scale gap by translating large-scale climate model output into finer spatial detail using historical data. This review synthesizes the main developments in statistical modeling and stochastic simulation for climate downscaling, aiming to provide both a conceptual overview and a practical reference for advancing climate modeling towards delivering more detailed information at finer horizontal spatial scales. A broad range of methods is considered, including empirical-statistical approaches and stochastic simulation techniques, that generate ensembles capturing inherent climate variability. The review also discusses how recent advances in machine learning and ensemble-based strategies are expanding the flexibility and scope of downscaling models. Building on this foundation, the review then examines the contribution of geostatistical simulation techniques to climate downscaling, with a focus on methods that preserve spatial continuity and reproduce fine-scale patterns through conditional realizations, while underscoring the existing gap in relevant research. Throughout, attention is also given to the applicability, assumptions, and limitations of different approaches, as well as their relevance for climate impact studies and reconstruction efforts. Finally, future directions are outlined in light of emerging data-driven techniques, growing computational capacities, and the increasing need for robust fine-scale climate information.
Stylianos Hadjipetrou (Sun,) studied this question.