Empirical Evidence for the Importance of Vegetation Spatial Distribution and Temporal Dynamics in Modulating Streamflow from Macroscale Watersheds in Various Physiographic and Climatic Regimes

The complex role of vegetation in hydrology is challenging to model. Spectral vegetation index (SVI) data are now being used in spatially and temporally distributed hydrologic models to represent vegetation biophysical parameters. We sought to test whether there is empirical evidence for the inclusion of vegetation dynamics in accounting for watershed hydrology at macro scales. Observed relationships among Normalized Difference Vegetation Index (NDVI) from the Advanced Very High Resolution Radiometer (AVHRR), precipitation, and streamflow were examined and quantified for 4 watersheds from across the conterminous United States to test some fundamental assumptions regarding the role of vegetation in hydrology. Biweekly composites of NDVI spanning a 10-year period were combined with climate, soils, and topographic information for watersheds ranging in size from approximately 1,700 to 240,000 sq km. These watersheds were chosen to provide a broad crosscut of climatic, physiographic, vegetation, and land use environments at different spatial scales. Digital maps of elevation, travel time to the stream, topographic index, solar radiation index and hydrologic soil groups were created from the ancillary data and used to aggregate the NDVI over space and through time. The regression of precipitation and NDVI against streamflow explains significant amounts of variation at macro scale. In some cases, spatially distributed NDVI alone accounts for nearly as much variation as does the combination of climatic and NDVI variables. In part because the AVHRR instrument is often faulted for inconsistent spectral response through time, monthly values from one year of record or idealized curves are used as proxies for monthly leaf area index or land cover phenology in the context of hydrologic models. We averaged NDVI within each month throughout the period of record to simulate such a proxy and compared it against regression results using actual monthly NDVI values. Performance of the individual monthly values was superior - indicating that a hydrologically meaningful signal is present in the multi-year NDVI. Other spatial and temporal aggregations of the NDVI degraded the strength of observed relationships. These empirical results confirm the importance of accounting for vegetation spatial distribution and temporal dynamics in modeling the terrestrial component of the hydrologic cycle at macro scales.