The resistance surface that formed the basis of our coastal marten connectivity model is comprised of several data layers that represent forested and non-forested land cover, waterbodies, rivers, roads, and serpentine soils.

This dataset contains the non-forested land cover data used in the resistance surface. To see actual resistance values assigned to the non-forested land cover classes in this raster when the resistance surface is compiled, see the associated spreadsheet of resistance surface data sources and resistance values.

For context, the forested land cover dataset was developed using the Old-growth Structure Index (OGSI), which is the primary estimator of habitat quality and cost-weighted distance in the connectivity model. OGSI is a parameter derived by the Gradient Nearest Neighbor (GNN) model produced by the Landscape Ecology, Modeling, Mapping & Analysis laboratory in Corvallis, OR (LEMMA 2014a). The GNN model provides fine-scale spatially explicit data on forest structure across a vast area of California, Oregon, and Washington, and is one of the very few datasets available that provides such habitat information in a consistent manner across the CA/OR state border.

Non-forested land cover types are not classified by the GNN Structure forest model; the GNN model only applies to areas with the potential to support at least 10% tree cover (LEMMA 2014a). In the GNN model these portions of the landscape are filled in with data from the U.S. Geological Survey’s (USGS) Gap Analysis Program (GAP) data on Ecological Systems(ESLF) (USGS 2011, Comer et al. 2003). We would generally expect these areas to have higher resistance to movement by coastal martens than forested habitats, but it is important not to treat them uniformly as some cover types may be passable in certain conditions (e.g. relatively small areas of native shrub-dominated habitats), whereas others are probably complete barriers (e.g. high density urban development). We identified 43 such non-forested cover types within the historical range of the coastal marten (with a few exceptions as explained in the proceeding paragraphs). Based on their description in the NatureServe web application (NatureServe 2017), we grouped these into six functional classes that we generally treated as increasing in resistance from first to last: shrub, grassland, dune, open, wet, and developed (see Appendix 1 of the report, as well as the spreadsheet of resistance surface data sources and resistance values).

To develop the non-forested land cover data, we used our modeling extent to extract the ESLF_CODE attribute from the broader GNN Structure dataset. ESLF_CODE values ranged from 1 to 9297, with the value 0 identifying pixels that did not have an ESLF value in the GNN dataset (and were thus part of the GNN forest model, as represented in the forested land cover layer).

Through visual inspection it became apparent that road and river features were registering in the data as Developed Open Space and Developed Low Intensity ESLF classifications in some parts of the landscape. This was problematic because these linear feature representations were incomplete, did not have a robust and explicit classification scheme (i.e. they were not explicitly defined as roads and rivers that could be filtered out of the data), and were not aligned to the roads and rivers data layers that were already being incorporated into our resistance surface.

To address this issue, we used the Shrink geoprocessing tool to shrink ESLF_CODE values 21 and 22 (Developed Open Space and Developed Low Intensity, respectively) by 2 cells. This had the effect of removing these linear features and replacing them with adjacent ESLF values.

This is an abbreviated and incomplete description of the dataset. Please refer to the spatial metadata for a more thorough description of the methods used to produce this dataset, and a discussion of any assumptions or caveats that should be taken into consideration.