2015 GSA Annual Meeting in Baltimore, Maryland, USA (1-4 November 2015)

Paper No. 211-1
Presentation Time: 9:00 AM-6:30 PM


RENGERS, Francis K.1, MCGUIRE, Luke A.1, KEAN, Jason W.2 and STALEY, Dennis M.3, (1)U.S. Geological Survey, Geological Hazards, 1711 Illinois St., Golden, CO 80401, (2)U.S. Geological Survey, Denver Federal Center, P.O. Box 25046, MS 966, Denver, CO 80225-0046, (3)U.S. Geological Survey, Denver Federal Center, P.O. Box 25046, MS 966, Denver, CO 80225, frengers@usgs.gov

The extent of the wildland-urban interface is increasing in size in the western United States, placing more humans and costly infrastructure closer to steep mountainous environments. Hydrologic modeling is crucial for the prediction of flood timing and magnitude after natural hazards in these areas. Recent advances such as LiDAR are making it possible to model hydrologic processes in unprecedented detail. This research focuses on comparing the performance of two hydrologic models with differing levels of complexity using high-resolution, lidar-derived digital elevation models.

In this study, we test these two models on two watersheds in the San Gabriel Mountains, CA. Both watersheds were burned at moderate/high severity during the 2009 Station fire, share similar soil properties, and were monitored following the wildfire. However, the watershed sizes differ by two orders of magnitude, allowing for model testing at different spatial scales. The simpler model uses the kinematic wave approximation to route water over a landscape, while the more complex model uses the full shallow water equations. In both models precipitation is spatially uniform and infiltration is simulated using the Green-Ampt infiltration equation. Both models use the same high-resolution topography for each watershed. In addition to calibrating both models to match observed overland flow, we calculate sediment transport potential using simulated overland flow depths.

Our simulation results show that the different equations we used for modeling overland flow successfully approximated the timing of observed floods. However, the model parameters required to fit observed hydrographs differ substantially. For example, the best-fit Manning’s n roughness parameters differed by a factor of 4 between the two calibrated models. This suggests that some calibrated parameters typically treated as an inherent landscape property, are actually strongly model dependent. Moreover, we show that mitigation decisions for sediment management and hazard abatement based on hydrologic modeling should consider the inherent bias of different modeling approaches. Results also provide guidance for future modeling studies that require basin-scale estimates of roughness coefficients and hydraulic conductivity following wildfire.