You are here:
Stream ecological condition modeling at the reach and the hydrologic unit (HUC) scale: A look at model performance and mapping
Citation:
WEBER, M. H. AND J. VAN SICKLE. Stream ecological condition modeling at the reach and the hydrologic unit (HUC) scale: A look at model performance and mapping. Presented at The National Map Users Conference and USGS Geographic Information Science (GIS) Workshop, Lakewood, CO, May 10 - 13, 2011.
Impact/Purpose:
The National Hydrography and updated Watershed Boundary Datasets provide a ready-made framework for hydrographic modeling.
Description:
The National Hydrography and updated Watershed Boundary Datasets provide a ready-made framework for hydrographic modeling. Determining particular stream reaches or watersheds in poor ecological condition across large regions is an essential goal for monitoring and management. To address this need, we built predictive models of stream ecological condition (i.e. water quality, biological condition, and physical habitat) across the Pacific Northwest (PNW; Oregon, Washington and Idaho) using the enhanced National Hydrography Dataset (NHDPlus) as the spatial framework. Using data from the US EPA EMAP-West survey, from the USGS NAWQA program, and from other regional monitoring efforts, we constructed a number of predictive models for several stream condition endpoints using 144 landscape metrics as predictor variables. We mapped these model results to every NHDPlus stream reach in the PNW, aggregated predictions across all reaches within each 6th-level USGS Hydrologic Unit (HUC), and mapped the HUC-averaged scores. We examine the substantial uncertainties of our predictions and suggest how they could be reduced, as well as address issues of aggregating results to coarser spatial scales (i.e. from individual stream reaches to watersheds or HUCs). We also look at differences in our models based on aggregation of stream reach results for only headwater streams in a given hydrologic unit versus results for streams flowing through multiple hydrologic units. Overall, we found landscape predictors summarized within 100m of stream reaches performed better than whole watershed predictors, and that prediction uncertainty was still substantial when examined at the 6th-field HUC level but decreased significantly when examined at the 5th and 4th field HUC level, with corresponding loss of spatial accuracy in prediction.