6 minute read

Creating habitat core areas

To analyze wildlife habitat connectivity with Linkage Mapper, the habitat areas to be connected with modeled corridors must first be defined. We created a dataset of habitat cores by adapting methodology developed for ESRI’s Green Infrastructure initiative (ESRI 2017). In essence, this method uses landcover and fragmenting features (including roads, railroads, and built structures) to find contiguous patches of natural landcover above a certain size threshold. We used the Ecological Systems of Washington and 2015 Landcover of Canada to define natural land and disturbed/developed land for the U.S. and Canada, respectively. Fragmenting features were defined using the same road and railroad datasets used in the landscape resistance layer. These datasets were used as inputs to a Python model publicly available from ESRI’s Green Infrastructure website that created patches of contiguous natural land that are at least 100 acres and thicker than 200 meters at their thinnest point. We performed ten 8-neighbor smoothing passes and simplified the polygons to close small holes within patches and smooth the jagged, raster-like appearance of the edges.

The resulting cores represent intact wildlife habitat in the most general sense, and in reality, the cores may vary wildly in wildlife value, from relatively undisturbed areas in the Cascade Range spanning thousands of acres to isolated 100-acre patches surrounded by urban development near Bellingham. Linkage Mapper treats all of these cores as suitable wildlife habitat, and its tools are more interested in how wildlife are likely to navigate the landscape between the cores.

Advertisement

Analyzing wildlife habitat connectivity with Linkage Mapper

Linkage Mapper is a widely-used suite of connectivity tools that can be used to map wildlife linkages and corridors between habitat cores (McRae and Kavanagh, 2011). The Linkage Pathways tool uses input data for habitat cores and cost distance (from landscape resistance) to calculate the accumulated cost of wildlife movement from each core to all other adjacent cores. This results in a network of least cost paths, which are the optimal linkages between cores along which an animal would accrue the least possible total cost distance.

However, least cost paths are not necessarily realistic representations of how wildlife moves across the landscape. A least cost path implies that an animal has perfect knowledge of the landscape and that it will follow the optimal path from one habitat core to another. For this reason, the Linkage Pathways tool also calculates least cost corridors between each pair of adjacent cores and combines them to create a single, composite least cost corridor map for the entire study area. These least cost corridors represent the predicted movement cost to wildlife in a broader area relative to the optimal least cost path. This relative metric of connectivity is more useful for understanding the nature of the connections between habitat cores. For example, least cost corridors can show whether a given connection between two cores is channelized and restricted to the least cost path or broad and permeable with similarly low cost over a large area.

Another useful tool in Linkage Mapper is Pinchpoint Mapper (McRae 2012). This tool uses circuit theory to analyze wildlife movement within the least cost corridors created by the Linkage Pathways tool. The result identifies bottlenecks where simulated species flow is forced through a narrow area, usually due to high landscape resistance on either side. These so-called pinchpoints represent possible high-priority areas for conservation or restoration because even a small decrease in habitat value or increase in landscape resistance would have a disproportionate negative impact on connectivity.

We used the Linkage Pathways tool from Linkage Mapper (ver. 2.0) to create least cost paths and least cost corridors between our 100+ acre contiguous habitat cores using our landscape resistance layer to calculate cost distance. We then used Pinchpoint Mapper to identify circuit flow pinchpoints within least cost corridors with a cost-distance cutoff of 200,000 cost units. Both of these analyses used a 10-mile buffer around the WRIA 1, 3, and 4 study area as the processing extent to reduce edge effects.

Analyzing wildlife habitat connectivity using Omniscape

Omniscape is a newer method of mapping wildlife habitat connectivity that is a useful complement to Linkage Mapper (Landau et al. 2021). Using circuit theory, Omniscape simulates the predicted movement of wildlife across a landscape with varied levels of resistance from landcover and dispersal barriers. Essentially, circuit theory treats animals on a landscape like electrons on a circuit board with multiple open pathways, each with a different resistance level. At every point in an animal’s journey, it makes decisions on which direction to move based on the nature of the adjacent landscape around it, just as electrons flow across a circuit board following paths of least resistance, with expected current flow being inversely proportional to the resistance encountered on that particular pathway.

While traditional Circuitscape models wildlife movement using input nodes (i.e. core areas) as sources of current, Omniscape uses every cell of a landscape raster grid as both a potential current source and ground. Using a circular moving window with a user-specified radius, Omniscape analyzes each cell in turn as a target (or ground), calculates current flow to that target from all source cells within the window, and sums the resulting current maps from every cell to create a cumulative flow raster for the entire landscape (Figure 4). The input landscape resistance raster can be inverted by Omniscape and used as a current source strength raster, following the logic that low-resistance areas represent high-quality habitat and are thus more likely to be sources of wildlife. The result is a circuit flow model that essentially predicts wildlife movements of a distance less than the input radius from any location to every other location. In other words, an Omniscape analysis with an input radius of 10 kilometers would show the predicted likelihood of wildlife movement along every possible path from 0 to 10 kilometers long across the entire landscape. block size in Omniscape is a way of sub-sampling the landscape to reduce the processing load with only negligible changes to the output (Landau 2021). In this case, it divided our landscape into blocks of 25 x 25 cells (750 x 750 meters) and used only the center cell of each block as a target pixel, exponentially reducing the Omniscape runtime to days instead of weeks.

Finally, we created a normalized current flow map that divides the landscape into categories that characterize wildlife habitat connectivity in a meaningful way. Normalized cumulative current was calculated by dividing cumulative current (the predicted flow of simulated wildlife across the landscape resistance raster) by flow potential (the null model in which flow from the source layer is simulated across a null landscape where all resistance is set to 1). We then classified normalized cumulative current into distinct categories of flow types. The result is an index that describes how wildlife flow is affected by barriers and resistance on the landscape. If the normalized current flow is close to 1, then flow is similar to the null model and thus is not greatly affected by landscape barriers. If normalized current flow is less than 1, then flow is lower than expected in the null model due to barriers. If normalized current flow is higher than 1, that means flow is being redirected to that area by surrounding barriers. Although these categories are frequently used to describe Omniscape results, the values used to define each category are somewhat arbitrary, based on expert opinion, and specific to the landscape in question (McRae et al. 2016, Costa et al. 2021, Schloss et al. 2022). In this analysis, we classified normalized cumulative current into four categories:

For this analysis, we modeled connectivity in WRIAs 1, 3, and 4 using landscape resistance as the input raster for both conductance and source strength. However, for Omniscape, we created a second version of our landscape resistance raster that did not include slope as one of the resistance factors (Figure B3). This is because when we tested Omniscape with the original landscape resistance raster, the result appeared to show similar areas of highly-channelized wildlife flow in the developed/agricultural west as in the mostly natural east (Figure B4). That result, which suggested that high slopes in the Cascade Range were as much of a barrier to wildlife habitat connectivity as urban development and highways in the west, was unintuitive and somewhat disingenuous. Removing slope from the equation helped create a map that better captures the differences in wildlife habitat connectivity across the study area.

As with Linkage Mapper, we used a 10-mile buffer around the study area to reduce edge effects. For the moving window, we used a circle with a radius of 333 cells, or 9,990 meters. Due to the large radius and small cell size, we used a block size of 25 to make the analysis more manageable. Using a

This article is from: