Local Connectedness Stratified by Setting and Ecoregion with Regional Override, 2016 Eastern U.S. and Canada

A highly permeable landscape promotes resilience by facilitating range shifts and the reorganization of communities. Roads, development, dams, and other structures create resistance that interrupts or redirects movement and, therefore, lowers the permeability. Maintaining a connected landscape is the most widely cited strategy in the scientific literature for building resilience and has been suggested as an explanation for why there were few extinctions during the last period of comparable rapid climate change. This metric is an important component of resilience because it indicates whether a process is likely to be disrupted or how much access a species has to the micro-climates within its given neighborhood.

The method used to map local connectedness for the region was resistant kernel analysis, developed and run by Brad Compton using software developed by the UMASS CAPS program (Compton et al. 2007, http://www.umasscaps.org/). Connectedness refers to the connectivity of a focal cell to its ecological neighborhood when it is viewed as a source; in other words, it asks the question: “to what extent are ecological flows outward from that cell impeded or facilitated by the surrounding landscape?” Specifically,each cell of a resistance grid is coded with a resistance value base on land cover and roads, which are in turn assigned resistance weights by the user. The theoretical spread of a species or process outward from a focal cell is a function of the resistance values of the neighboring cells and their distance from a focal cell out to a maximum distance of three kilometers (the recommended distance determined by the software developer).

To calculate this metric, resistance weights were assigned to the elements of a land cover map. A variety of methods have been developed for determining resistance weights, in particular metrics of ecological similarity in community types (e.g. oak forest to oak forest assumed to be more connected than oak forest to spruce forest) have been used to good effect. However, our weighting scheme was intentionally more generalized, such that any natural cover adjacent to other natural cover was scored as highly connected. We did not differentiate between forest types, and only slightly between open wetland and upland habitats. Our assumption was that the requirements for movement and flows through natural landscapes were less specific than the requirements for breeding, and that physical landscapes are naturally composed of an interacting mosaic of different ecosystems. Our goal was to locate areas where these arrays occur in such a way as to maintain their natural relationships and the connections between all types of flows, both material processes and species movements, not to maximize permeability for a single species.

To create the resistance grid, we used landcover, roads, and railroads. For the US, the source data was the 30-meter 2011 NLCD which identifies each grid cell as belonging to one of 16 classes of land cover (Homer et al. 2015). For Canada, we used provincial landuse, forestry, & wetlands data. Although the 2011 NLCD and the Canadian Provincial datasets are the most current datasets available, we made several improvements to them that substantially improved their performance as resistance grids. These included: 1) updating the roads and railroads, 2) adding transmission line data, 3) adding dirt roads, 4) reclassifying barrens, 5) adding plantation forests 5) differentiating between hay/pasture and crop land and 5) reclassifying water polygons.

To run the local connectedness analysis on the resistance surface we decreased the grid cell resolution from 30 meters to 90 meters. This allowed us to run the analysis with a reasonable processing time (weeks) because the CAPS software program is computationally intensive. We aggregated the 30 meter cells to the 90 meter cells using the average of the 30 meter resistance weights. The final result was a grid of 90-meter cells for the entire region where each cell was scored with a local connectivity value from 0 (least connected) to 100 (most connected).