Assessing the potential impacts of 21st-century climate change on species distributions and ecological processes requires climate scenarios with sufficient spatial resolution to represent the varying effects of climate change across heterogeneous physical, biological, and cultural landscapes. Unfortunately, the native resolutions of global climate models (usually approximately 2 degrees x 2 degrees or coarser) are inadequate for modeling future changes in, e.g., biodiversity, species distributions, crop yields, and water resources. Also, 21st-century climate projections must be debiased prior to use, i.e., corrected for systematic offsets between modeled representations and observations of present climates. We have downscaled future temperature and precipitation projections from the World Climate Research Programme's (WCRP's) CMIP3 multi-model data set to 10-minute resolution and debiased these simulations using the change-factor approach and observational data from the Climatic Research Unit (CRU). These downscaled data sets are available online and include monthly mean temperatures and precipitation for 2041-2060 and 2081-2100, for 24 climate models and the A1B, A2, and B1 emission scenarios. This paper describes the downscaling method and compares the downscaled and native-resolution simulations. Sharp differences between the original and downscaled data sets are apparent at regional to continental scales, particularly for temperature in mountainous areas and in areas with substantial differences between observed and simulated 20th-century climatologies. Although these data sets in principle could be downscaled further, a key practical limitation is the density of observational networks, particularly for precipitation-related variables in tropical mountainous regions. These downscaled data sets can be used for a variety of climate-impact assessments, including assessments of 21st-century climate-change impacts on biodiversity and species distributions.