Wednesday, 13 January 2016
Hall D/E ( New Orleans Ernest N. Morial Convention Center)
Statistical downscaling is commonly used to derive fine resolution information from coarse climate change projections due to the high computational cost of dynamic downscaling, which also limits the feasibility of uncertainty analyses through model intercomparison. Focus will be given to two Atmospheric Ocean General Circulation Models (AOGCMs) that have been downscaled to 50 km resolution by separate groups of three Regional Climate Models (RCMs) by the North American Regional Climate Change Assessment Program (NARCCAP). The two AOGCMs show considerable disagreement in projected climate change over the central USA. The six AOGCM-RCM pairs from NARCCAP will be downscaled to 4 km resolution using the parameter-elevation regressions on independent slopes model (PRISM) dataset as observations. For daily temperature, training and validation using multiple configurations of regression models will be performed using the 32-km North American Regional Reanalysis dataset and the most accurate and robust (for climate change application) model configuration will be applied to the six AOGCM-RCM pairs (i.e., the perfect prog approach). Developed from a randomly selected eight-year training period, initial regression model configurations validate well against thirty randomly selected PRISM cells in central Missouri USA, with mean absolute errors and biases generally below 1.4oC and 0.2oC respectively over the two year validation period. Using the warmest years for the validation period to determine the robustness of the initial configurations to climate change, cold biases range from 0.05oC to 0.35oC.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner