Figure 1. The number of studies published each year that used circuit theory to model connectivity for at least one terrestrial mammal.
Articulating the type of connectivity being modeled
One-hundred and ten of the 181 studies (60%) were non-specific regarding the type of connectivity they intended to model. Of the seventy-one studies that did communicate a type of connectivity, 54 (76%) indicated they were modeling genetic connectivity / dispersal facilitation, 10 (14%) specified seasonal migration, 3 (4%) specified within-home range movements, and 4 (6%) specified range shifts.
Sensitivity and uncertainty analysis
Thirty-three (18%) of the 181 studies conducted some form of SUA. Of those that conducted a form of SUA, most evaluated the sensitivity of model output to: (i) variation in resistance values used in the resistance surface layer (ii) the number of connectivity focal nodes (“points or regions between which connectivity is to be modeled”) (McRae et al. 2013), or (iii) inclusion of particular landscape variables (e.g., slope, canopy cover, distance to disturbance).
Model validation
Of the 181 studies reviewed, 80 (44%) included no validation step at all; i.e., they validated neither their input (habitat or resistance) layers nor their connectivity output (current density) layers. Fifty-five of the 181 studies (30%) validated their habitat layers that were used to generate the resistance surface layers used in the connectivity analyses. Twelve studies (7%) validated their resistance surface layers that are input into Circuitscape, and only 34 (19%) validated the results of their connectivity model (the current density output layer).
The most common type of data used in the validation step was point occurrence (60 studies; Table 1), followed by tracking (25), genetic (10), camera trap (4), and expert opinion (2). Of the 101 studies that performed some form of validation, only twenty used fully independent data, including expert opinion. Twenty-seven studies used partially independent data and the remaining 54 used non-independent data, most commonly withholding a subset of locations or paths for cross-validation. Fully independent data were used in five studies to validate their habitat layers and in three studies to validate their resistance/permeability layers. Only twelve studies out of the 181 (7%) included in this assessment used fully independent data to validate the results of their connectivity analysis, the current density output.
Table 1 . Cross-tabulation of studies according to the type of data used for validation and the degree of independence of validation data. Numbers in parentheses are for studies that validated the model output (current density layer), and remaining numbers represent studies that validated a model input (habitat or resistance layer).