U.S. Downscaled Climate Data
This site contains over 120 GB of downscaled climate data for use in climate-change studies over the conterminous United States. The data is derived from gridded observed fields from the University of Washington Land Surface Hydrology Research Group and from Global Circulation Model (GCM) simulations of historical climate conditions (scenario 20c3m in IPCC4 studies), A2 future greenhouse-gas-and-sulfate-aerosols emissions scenarios, and B1 future emissions scenarios. The GCMs represented here, so far, are the National Center for Atmospheric Research's Parallel Climate Model 1 (PCM) and from the NOAA Geophysical Fluid Dynamics Lab's GFDL CM2.1 model. All data, except the gridded observations (from University of Washington) in real_obs, have been downscaled using the constructed analogs method of Hidalgo, Dettinger, and Cayan (documentary background of this method in supporting_materials/CEC-500-2007-123.pdf). Also see recent revisions.
The GCM outputs used here were obtained through the connections at https://esg.llnl.gov:8443/index.jsp, and ultimately from the original modeling groups. We gratefully acknowledge their efforts in producing and providing these simulations. Some details of the GCM model grids and other information can be found at http://meteora.ucsd.edu/cap/ipcc4.html.
The climate fields that have been downscaled for inclusion here are daily Tmax and Tmin, and daily precipitation, on a 12-km grid over the conterminous US plus parts of the Columbia River basin in Canada. Each of the binary files of downscaled (or observed) meteorological fields is a direct-access, binary format written from fortran. The daily fields each contain daily weather values at each of 54027 grid cells spread across the conterminous US, at locations given the first two columns of supporting_materials/vic.latons. More information including coding examples are available in the "supporting_materials.zip" file.
The real_obs and obs_downscaled files include leap days (February 29) in leap years; the GCM outputs do not (365 days in every year). The temperatures are in degrees Centigrade; precipitation is in mm/day. In file names, _ca_ refers to constructed analogs (downscaled), and not to California.
The folder real_obs contains the gridded historical meteorological observations of daily temperature maximum, temperature minimum, and precipitation that were used in the downscaling and downscaling verifications. The obs_downscaled folder contains the results from aggregating the gridded observations up to the coarseness of the NCAR/NCEP reanalysis fields (roughly 2.5 x 2.5 degree lat long grid) and then downscaling them back onto the original 12-km resolution grid, for use in checking the downscaling method. The folders pcm_historical and gfdl_historical contain downscaled versions of the climates simulated by the PCM and GFDL models under historical radiative (greenhouse-gas minus sulfate-aerosol) forcings (but otherwise free to evolve however they wanted). The folders pcm_a2 and pcm_b1 contain downscaled versions of the 21st Century climates simulated by the PCM model under A2 (fairly aggressive greenhouse-gas emissions scenario) and B1 (less intense greenhouse-gas emissions). See the graphic called ipcc_forcings.png in supporting_materials for a depiction of the differences between these two emissions scenarios. The folders gfdl_a2 and gfdl_b1 contain downscaled versions of the 21st Century climates simulated by the GFDL model under A2 and B1 emissions scenarios respectively.
Finally please understand that all of the GCM outputs are SCENARIOS, not predictions. In this context, a scenario is an example of how the climate could vary on day-to-day, month-to-month, year-to-year, decade-to-decade time scales, an example that is consistent with the model physics, with the particular initial conditions that the model was started from, and with the radiative forcing from greenhouse gas and sulfate aerosols emissions. A prediction would be an attempt to search through all the possible scenarios for the one that actual will coincide with the day-to-day, etc, variations of the climate in the real world. Thus, in the future simulations as in the historical simulations by each model, one cannot expect the temporal details to line up with the real chronology of individual events. Multiple simulations by the same model, only differing by their initial conditions, will yield different scenarios (or examples) of how climate might unfold. Under the conditions imposed in the simulations downscaled here, the slowest climate trends are usually directly associated with the growing greenhouse effect, but this slowest time scale is generally even longer than one or two decades. That is, the particular conditions in a given decade (say, 2050-2059) in one of these simulations may (and generally is) as much a function of the natural variability (or randomness) of the climate system as of the greenhouse responses. A dry decade in 2050s is not necessarily a symptom of the greenhouse effect; a dry decade in the 2050s followed by a wet decade in the 2060s is even less likely to be a signature of the greenhouse effect. Rather, in trying to uncover the greenhouse effects herein, pay attention to the trends not the variations around those trends.
CONTACT: Please consider emailing Michael Dettinger, if you download data for use in your studies so that he can alert you to any updates, revisions or additions.
