You are here

Arctic-Yukon-Kuskokwim Landscape Predictors of Fish Habitat

2011 Arctic-Yukon-Kuskokwim Sustainable Salmon Initiative Project: Landscape Predictors of Coho Salmon

Kelly M. Burnett2, Daniel J. Miller3, Rick Guritz4, Mark A. Meleason5, Ken Vance-Borland6, Rebecca Flitcroft2, Matthew J. Nemeth7, Justin Priest7, Nicholas A. Som8, and Christian E. Zimmerman9

2USFS, Pacific Northwest Research Station, 3200 SW Jefferson Way, Corvallis, Oregon 97331
3Earth Systems Institute, 3040 NW 57th St., Seattle, Washington 98107
4Geophysical Institute, University of Alaska Fairbanks, 903 Koyukuk Dr., Fairbanks, Alaska 99775
5Oregon Department of Forestry, Salem Headquarters, 2600 State Street, Salem, Oregon 97310,
6 Oregon State University, Forest Ecosystems and Society, Corvallis, Oregon 97331
7 LGL Alaska Research Associates, Inc., 1101 East 76th Ave, Suite B, Anchorage, Alaska 99518
8US Fish and Wildlife Service, Arcata Fish and Wildlife Office, 1655 Heindon Road, Arcata, California 95521 9 USGS Alaska Science Center, 4210 University Dr., Anchorage, Alaska 99508 14 February 2013

See online data viewer

See full report


Habitat quality and quantity are key abiotic variables driving the abundance and distribution of salmon in freshwater, but understanding about how these variables affect salmon is minimal in the Arctic-Yukon-Kuskokwim (AYK) region. This diminishes opportunities to predict effects of habitat change on salmon or to know whether freshwater factors are related to observed changes in salmon returns. Recent research indicates that statistical methods using geographically limited field surveys and spatially extensive landscape data (e.g., stream, terrain, and vegetation) can reliably estimate habitat characteristics and salmon abundance in freshwater over large areas. The objectives of this study were to: 1) develop and evaluate high-resolution terrain and hydro-geomorphic attributes, directly through remote sensing data, and indirectly through information inferred from spatial patterns in the remote sensing data; and 2) relate juvenile coho salmon abundances to hydro-geomorphic and landscape attributes to identify factors that potentially affect coho salmon distribution and productivity. The study was conducted in the Nome River of northwestern Alaska, USA. This research demonstrated the feasibility of creating highly accurate, high-resolution data over a large area of the AYK, and developed new algorithms and approaches for doing so. For the Nome River basin, we derived from ALOS PRISM satellite data a seamless ortho-rectified optical image mosaic and a 2.5-m Digital Elevation Model (DEM) mosaic (5-m vertical accuracy) that meets National Map Accuracy Standards for Alaska. We also developed processing algorithms to automate DEM production and accuracy evaluation using satellite laser altimetry data – an effective and less expensive alternative to LiDAR data. Until data from the Alaska Statewide Digital Mapping Initiative become available for the AYK region, ALOS PRISM data are a relatively low-cost source of DEMs and ortho-imagery with accuracies and resolutions sufficient to benefit fisheries management. To illustrate, we combined newly created DEMs and field data for the Nome River to delineate a highly resolved stream network along with 14 modeled attributes (e.g., stream gradient, valley-floor width, drainage area) for all channels that support salmonids. By varying the contributing-area threshold and linking to a water mask we developed from PRISM optical imagery, our algorithms produce more spatially and structurally accurate networks than other common tools for delineating streams from DEMs. Although some attributes were derived with existing approaches, we developed new methods for attributes that describe floodplain complexity, indicative of off-channel and complex edge habitats that are important for rearing juvenile salmonids. The extent to which streams freeze during winter, potentially contracting space for egg incubation and juvenile rearing, is another aspect considered. We acquired, terrain-corrected, and geocoded a time series of Synthetic Aperture Radar (TerraSAR-X) data to classify open water and ice in the Nome River. Field data were also collected during the winter to support a supervised classification. Given the ability to visually identify open water and ice on images from the TerraSAR-X data, we are optimistic about ongoing efforts to develop a statistical model for ice classification on the Nome River, as has been done with SAR data on larger rivers. A new technique we developed allowed linking field data for summer habitat, summer snorkel counts of fish, and winter ice to the 14 DEM-derived attributes and then geo-referencing to nodes in the delineated stream network. We found that juvenile coho salmon were not randomly distributed at any spatial resolution we considered, from the micro-habitat to the tributary scale. Juvenile coho using stream margins were often concentrated around the bank dens of beaver, a link that has not been well described in past studies. More juvenile coho salmon were observed than any other age-class or species of fish in the Nome River basin, consistent with the high coho salmon intrinsic potential we modeled throughout the basin. The channel length in slow-water habitat unit types varied inversely with DEM-derived channel gradient, which may help identify high-capacity stream reaches. Densities of juvenile coho salmon were much greater in slow-water than fast-water habitats, but snorkelers observed no fish in many of the slow-water habitats. This suggests that differences in fish abundance may arise from factors affecting whether or not fish occupy a unit as well as those affecting abundance in occupied units. We are exploring statistical tools necessary to model relationships between the zero-inflated juvenile coho salmon data and hydrogeomorphic and landscape characteristics as well as presence of adult salmonids as potential predators of juvenile coho salmon. Key Words: Nome River, Norton Sound, landscape