Studies have shown that many severe weather events are impacted by the spatial patterns and magnitudes of land surface fluxes. In turn, the severity of flood responses to heavy precipitation are often influenced by antecedent soil moisture, groundwater and channel flow conditions. Specifically, uncertainty in initial land-surface model states can affect both convective storm triggering (and thus the amount and spatial distribution of precipitation) as well as surface runoff generation critical to determining the ultimate location and severity of potential flash flood impacts. This work examines land-surface initial state sensitivity (i.e., soil moisture status) through the evaluation of ensemble hydrometeorological forecasts for a large flash flood event that occurred in western North Carolina on 27 July 2013. Nearly a foot of rainfall falling in less than 24 hours drove flash flooding that was responsible for over 50 road closures, at least 4 destroyed homes, and over 700 damaged properties in western North Carolina. The flash flooding was also blamed for two deaths when swimmers were swept away in a fast-moving rural creek. The event occurred at the end of an already-record-setting month for rainfall in the region: most of western North Carolina received monthly precipitation amounts exceeding 500% of normal in many locations thereby exacerbating the ultimate flood response.
Initial work has already demonstrated that the coupled WRF-Hydro system is capable of producing a reasonably skillful simulation of the 27 July precipitation event and its streamflow using observed precipitation and a default configuration of model physics and initial land surface conditions. WRF simulations reveal timing and location errors representative of known NWP weaknesses in MCS representation. The probabilistic skill of quantitative precipitation forecasts and streamflow forecasts is then further explored using an ensemble comprised of (i) model physics variations, (ii) stochastic (atmospheric) dynamic and thermodynamic perturbations, and (iii) land-surface input and physical process variations. In addition to quantifying the probabilistic skill of precipitation and streamflow conditions, we will also diagnose the relative importance of the types of ensemble variation (i.e., physics vs. dynamics; atmospheric vs. land-surface) on final forecast fidelity and ensemble spread.
![](images/flag.png)
![](images/blue_ribbon.png)