Enter Note Done
Go to previous page in this tab
Session
  • Presentation | H11P: Headwaters Come First: Advancements in the Science and Practice of Measurement, Protection, and Restoration of Headwater Catchments I Poster
  • Poster
  • Bookmark Icon
  • H11P-1082: Accurate predictions of perennial streams in temperate, forested, headwaters: The challenge is real
  • Schedule
    Notes
  • Board 1082‚ Hall EFG (Poster Hall)
    NOLA CC
    Set Timezone

Generic 'disconnected' Message
Author(s):
Kristin Jaeger, USGS Washington Water Science Center (First Author, Presenting Author)
Jonathan Burnett, U.S. Forest Service
Sherri Johnson, U.S. Forest Service
Steven Wondzell, U.S. Forest Service
Jason Dunham, USGS Forest and Rangeland Ecosystem Science Center
Brian Staab, US Forest Service Portland
Michael Brown, Aquatics Program - Water Resources/Soils


Understanding where small streams are located and if they flow year-round or seasonally matters for how these streams and their watersheds are managed. We developed a model trained on simple flow/no flow field observations and 13 variables that describe climate, topography, and land cover conditions to provide predictions of whether a 5-m stream sub-reach is likely to have late summer flow or may go dry. The model is applied to a high resolution, light detection and ranging-derived stream network for 426 sub-watersheds in western Oregon and represents years with average to slightly drier-than-average rainfall. We also evaluated differences in model accuracy between watersheds that did and did not have training data to provide more realistic uncertainty estimates of model predictions. Field observations for these types of models is generally limited both in number of observations and their geographic distribution. Accuracy estimates for watersheds with (83% correct) and without (67% correct) training data can help managers decide if they use model predictions or if they need to collect additional field data. In addition, this model identifies stream locations where the model is extrapolating, which can be used for identifying locations where more data collection is needed for model improvement.



Scientific Discipline
Suggested Itineraries
Neighborhood
Type
Main Session
Discussion