87th AMS Annual Meeting

Tuesday, 16 January 2007: 1:30 PM
Streamflow prediction in ungaged watersheds: statistical modeling with remote sensing
211 (Henry B. Gonzalez Convention Center)
Blake P. Weissling, Univ. of Texas, San Antonio, TX; and H. Xie
Antecedent soil moisture state plays a vital role in a watershed's hydrologic response to a precipitation event and is thus parameterized in most, if not all, rainfall-runoff models. Yet the soil moisture condition antecedent to an event has proven difficult to quantify both spatially and temporally. Empirical models for antecedent soil moisture have been used for years within several common runoff estimation methods, such as the popular Natural Resource Conservation Service (NRCS) 5-day antecedent moisture condition “triad” (dry, normal, and wet). While this method is ubiquitous in numerous Curve Number rainfall-runoff models, it has long been criticized for over-generalization. Soil moisture probe studies yield good data but problems lie with extrapolating from point or field scale to catchment scale. Probe studies also suffer from irregular or sparse sampling frequency. Remote sensing has offered some solutions to characterizing and quantifying soil moisture, through acquisition and analysis of microwave (radar) passive and active imagery. These data sources suffer, however, from inadequate temporal and spatial coverage and interference from vegetation canopies.

The goal of this research is to assess the potential of high temporal resolution imagery, such as that provided by the MODIS/Terra satellite, to characterize the moisture state of a watershed from remotely sensed biogeophysical variables, and to parameterize a statistical streamflow or runoff prediction model solely utilizing a precipitation record and these remotely sensed variables. A 1428 km2 rural watershed in the Guadalupe River basin of southeast Texas, a basin prone to catastrophic flooding from record-setting, intense convective precipitation events, was selected for the model training phase of this project. A multiple regression model of gaged precipitation, land surface temperature, and EVI (Enhanced Vegetation Index) accounted for 78% (R2adj = 0.78) of the variance of observed streamflow for calendar year 2004. Efforts are underway to calibrate and validate this model for other time periods within the data availability window of MODIS imagery products, and for other watersheds of varying size and similar climatic regime within the Guadalupe River and neighboring basins The success of this remote sensing approach will have implications for developing near real-time predictive models for flood risk and forecasting, and water supply management in regions of the world with ungaged watersheds and limited resources to undertake ground-based hydrologic studies.

Supplementary URL: