Connecting models to movements: testing connectivity model predictions against empirical migration and dispersal data

Context

Connectivity has become a top conservation priority in response to landscape fragmentation. Many methods have been developed to identify areas of the landscape with high potential connectivity for wildlife movement. However, each makes different assumptions that may produce different predictions, and few comparative tests against empirical movement data are available.

Objectives

We compared predictive performance of the most-used connectivity models, cost-distance and circuit theory models. We hypothesized that cost-distance would better predict elk migration paths, while circuit theory would better predict wolverine dispersal paths, due to alignment of the methods’ assumptions with the movement ecology of each process.

Methods

We used each model to predict elk migration paths and wolverine dispersal paths in the Greater Yellowstone Ecosystem, then used telemetry data collected from actual movements to assess predictive performance. Methods for validating connectivity models against empirical data have not been standardized, thus we applied and compared four alternative methods.

Results

Our findings generally supported our hypotheses. Circuit theory models consistently predicted wolverine dispersal paths better than cost-distance, though cost-distance models predicted elk migration paths only slightly better than circuit theory. In most cases, our four validation methods supported similar conclusions, but provided complementary perspectives.

Conclusions

We reiterate suggestions that alignment of connectivity model assumptions with focal species movement ecology is an important consideration when selecting a modeling approach for conservation practice. Additional comparative tests are needed to better understand how relative model performance may vary across species, movement processes, and landscapes, and what this means for effective connectivity conservation.