The study uses statistical downscaling methods to generate high-resolution climate projections for Taiwan. Due to the limitations of global climate models (GCMs) in capturing regional climate variations influenced by local topography, downscaling techniques such as Empirical-Statistical Downscaling (ESD) are necessary. Using these methods, we create a dataset to provide detailed climate projections for Taiwan under the Taiwan Climate Change Projection and Adaptation Information Platform (TCCIP). This dataset incorporates historical simulation and future projections under various greenhouse gas emission scenarios from the Coupled Model Intercomparison Project Phase 6 (Eyring et al.). The study details the methodology used in the Taiwan Empirical-Statistical Downscaling (TaiESD) project, including data collection, bias correction, and evaluation processes. Two key elements are utilized: the high-resolution Taiwan Gridded Observation dataset (TGrid) and coarse-resolution simulated datasets from GCMs. The methods of Quantile Mapping (QM) and Quantile Delta Mapping (QDM) were applied to bias-correct temperature and precipitation data. The results indicate that TaiESD effectively captures temperature trends in future climate projections, showing robust consistency with the upstream GCM data. However, precipitation projections exhibit greater variability, reflecting complex local climatic factors. The study highlights the implications for adaptation planning, emphasizing the need for high-resolution data to understand regional climate impacts. Providing a detailed statistical downscaling dataset supports decision-makers in climate adaptation strategies for various sectors, including water resources, disaster management, etc.