| Detailed representations of vegetation processes are being added to macroscale hydrologic models. Because of model complexity, insufficient information, or both, many assumptions regarding model vegetation components have not been verified. Interactions among macroscale vegetation spatial variability and phenology have not been examined in terms of watershed hydrologic response. Spatial and temporal aggregations of remotely sensed data were used to test how well observations of vegetation spatial distribution and temporal fluctuations explain variations in macroscale watershed discharge without the use of physically based hydrologic models.; Eight years of Normalized Difference Vegetation Index (NDVI) composites were differentially stratified over space and aggregated through time for input to regressions against monthly climate and streamflow data for watersheds of different sizes (i.e., 1,700 to 250,000 km2), climates, and physiographies. The derived metrics captured vegetation fluctuations that are related to discharge, provide meaningful measurements for hydrologic process exploration, and suggest where or when vegetation fluctuations are important to macroscale streamflow.; NDVI aggregated over entire watersheds was only useful for multi-season simulations in humid, high biomass regions. At these large scales, spatial and temporal components of vegetation fluctuations are inseparable. Averaging of NDVI over such large areas convolves vegetation changes through time, creating lags among vegetation metrics and streamflow. In contrast, stratification of watershed NDVI yields additional explanation of streamflow variation, increasing variance explained by as much as 56%. While no stratification scheme outperformed all others, at this scale it is important to disaggregate watershed vegetation fluctuations through some means. Tests of sensitivity to period of analysis indicated that moisture conditions exert greater influence than possible satellite data calibration problems. Tightest linkages among vegetation and streamflow fluctuations were consistently measured during periods of low precipitation (maximum R2 = .99). Regarding temporal resolution, the monthly phenology of one year can be repeated for multi-year simulations in some cases. However, the method chosen to derive a representative phenology has important implications for model performance (R2 varied by as much as .25). Because it is difficult to determine when and where a single phenologic curve is acceptable, the use of multi-year, monthly resolution indices is recommended. |
