The Integrated Precipitation and Hydrology Experiment - Hydrologic Applications for the Southeast US (IPHEx-H4SE) Part III: High-Resolution Ensemble Rainfall Products
Nogueira M, Barros AP
Environmental Physics Laboratory- CEE-Pratt School of Engineering Report EPL-2013-IPHEX-H4SE-3, doi:10.7924/G8MW2F2W
The first stage of the Integrated Precipitation and Hydrology Experiment (IPHEX) includes the 37 development of quality-controlled data sets of different hydrometeorological and landscape attributes at high spatial and temporal resolutions (respectively 1km×1km and 1 hour). These data sets will facilitate the intercomparison of hydrological models and provide support to the ground validation campaign of GPM over the Southern Appalachian region. In the present report we focus on the spatial downscaling of Stage IV precipitation data (Baldwin and Mitchell, 1996; Lin and Mitchell, 2005; see online at http://www.emc.ncep.noaa.gov/mmb/ylin/pcpanl/stage4) from 4km to 1km resolution for the period 2007-2011. First, we describe the 44 methodologies utilized to develop the various QPE products and in particular the use of modified fractal downscaling methodologies, which conserve the spatial structure of the coarse resolution while enhancing sub-grid scale variability. Three different (hourly, 1km2) precipitation datasets were produced: 1) Stage IV bilinear interpolated fields; 2) Stage IV fractal downscaled fields using (with 50 ensemble realizations for each hour); and 3) Stage IV fractal downscaled fields using a transient (with 50 ensemble realizations for each hour). The realizations 50 provided for each hour in the fractal downscaled cases should be particularly useful to ensemble hydrologic applications and analysis of uncertainty propagation. The performance of the downscaled QPE (Quantitative Precipitation Estimation) products is subsequently evaluated for selected headwater basins in the Southern Appalachians for individual events and for 5 year continuous simulations in three watersheds, which are intended to highlight that, in long-term hydrological modeling and prediction and the precipitation forcing is de facto not accurate, the uncertainty varies in time, and this is further modulated by storage, evapotranspiration and subsurface flow in the hydrological model, a highly nonlinear system. The results show improved performance of an uncalibrated hydrological model using the downscaled Stage IV product using modified fractal interpolation methods as compared to bilinear interpolation. Finally, a survey of basic skill metrics indicates that current precipitation estimates are significantly poor in the inner mountain region of the Southern Appalachians where NEXRAD (Next Generation Radar Data) data used to inform the Stage IV product is compromised, which is expected in regions of complex terrain.