Main content area

Hydrologic model parameterization using dynamic Landsat-based vegetative estimates within a semiarid grassland

Kautz, Mark A., Holifield Collins, Chandra D., Phillip Guertin, D., Goodrich, David C., van Leeuwen, Willem J., Jason Williams, C.
Journal of hydrology 2019
Agricultural Research Service, Landsat, arid lands, grasslands, hydrologic models, rangelands, runoff, soil, soil erosion models, time series analysis, vegetation index, watersheds, Arizona
The use of hydrologic models to assess long-term watershed condition through repeated simulations of runoff and erosion is one common approach for rangeland health evaluation. However, obtaining vegetative data of appropriate spatiotemporal resolution for model parameterization can be difficult. The goal of this research was to assess the utility of using time-varying, Landsat-derived vegetative values to parameterize an event-based, watershed-scale hydrologic model. This study was conducted on a small, instrumented grassland watershed in the USDA Agricultural Research Service operated Walnut Gulch Experimental Watershed in southeastern, Arizona. Cloud-free Landsat scenes were acquired over the watershed for the years 1996-2014. The Soil Adjusted Total Vegetation Index (SATVI) was calculated for each image and calibrated using ground measured data to produce a time series of satellite-based foliar cover rasters. These values were used to parameterize the Rangeland Hydrology and Erosion Model (RHEM) for 26 rainfall-runoff events with corresponding observed data. Three parameterization scenarios using these data aggregated to different temporal resolutions (static, long-term mean, annual mean, and intra-annual values) were compared to a static literature-based scenario for evaluation. The linear relationship between field-measured foliar cover and SATVI showed statistically significant agreement with R2 = 0.85 and p < 0.05. Simulated runoff volume and peak flow rate using the three remotely sensed parameterization scenarios improved upon that of the literature-based scenario, with the annual mean scenario performing the best of the three temporal aggregations. The methodological framework outlined here provides a means for improved parameterization for watershed-scale modelling where vegetative data may be scarce or unobtainable for long-term analysis.