"Linkage" models are usually developed based on human assumptions about wildlife response to disturbance, and assume perfect knowledge of landscapes by wildlife.  They also result in idealized "linkages" or similar emphasis areas for connectivity conservation.  They are rarely developed based on occurrence or movement information for multiple species, or validated using wildlife occurrence or movement data. In this paper we evaluated 4 connectivity models at the California extent (424,000 km2) and one at a quarter of that extent (Mojave desert). For the statistical evaluation, we used close to 200,000 observations of wildlife, as well as various climatic and landscape variables. 

We found inconsistent positive and negative correlations between linkages and wildlife occurrence and movement, with only two species (puma and black bear) of over two dozen having their movement weakly, positively correlated with the presence of a linkage. For the linkage-philes, the correlation was weaker than for elevation or latitude. We emphasize that connectivity for most wildlife species will be best conserved by protecting their movement where they choose to go, not where a model predicts. Yes, that's harder than buying the narrow green stripe on the map, but it may ultimately be what's crucial for keeping wildlife around.