Connectivity models are a key tool for predicting animal movement patterns. They often serve as the basis for conservation planning decisions, such as where to establish wildlife corridors. But all models are imperfect, so assessing how accurately model predictions reflect reality – a process called model validation – is critical for justifying model-driven decisions.
Our recent review study seeks to shed light on the practice of validating connectivity model outputs by exploring the variety of existing validation approaches and the contexts in which they have been applied.
Creech, T.G., Brennan, A., Faselt, J., Stabach, J.A. and Keeley, A.T. 2024. Validating connectivity models: a synthesis. Current Landscape Ecology Reports 9: 120-134. DOI: 10.1007/s40823-024-00102-8.
We reviewed 60 peer-reviewed articles published since 2006 that included connectivity model validation, and then created a typology of 11 distinct validation approaches used in these articles. Collectively, these approaches are diverse enough to accommodate almost any connectivity model that researchers may want to validate.
The most striking finding in our review is how rarely published connectivity models are validated. We estimated that less than 6% of the connectivity modeling papers published since 2006 have included model validation, and this rate has not increased over time.
We also identify two aspects of connectivity model validation that deserve greater attention. First, relatively few structural connectivity models (i.e., those that consider only the physical arrangement of habitat patches) have been validated in the literature compared to functional connectivity models (i.e., those that consider species-specific responses to landscape structure). Second, validation approaches have rarely been used to test the transferability of connectivity models – that is, how well a model performs when used to predict connectivity for a new geographic area, time period, species, or movement process.
While the selection of an optimal validation approach depends on many factors, we recommend five general best practices for researchers developing and applying connectivity models:
- Use validation data that match the target species and purposes of connectivity conservation measures. For instance, it would be inappropriate to use location data from typical daily movements of a habitat generalist species to validate a connectivity model designed to represent long-distance migratory movements of a habitat specialist species.
- Use validation data that are statistically independent from the data used to develop the connectivity model. Using data collected from the same individual animals or sampling sites to fit a model and then test that model can produce falsely optimistic estimates of model performance.
- Use systematic sampling strategies to minimize bias in validation data. Variation in sampling effort or detection probability that is characteristic of some commonly used validation data types (e.g., citizen science databases of live animal or roadkill locations) can lead to unreliable validation results.
- Consider the biological significance of validation results, not just their statistical significance. Statistically significant results are almost inevitable in many cases, so reporting effect size (e.g., how much better the connectivity model performs than a null model) is often more informative.
- Use multiple validation approaches to test the connectivity model. Validation approaches evaluate model performance in different ways, so using multiple approaches provides better insight into the overall performance of the model.
We hope this review will make it easier for researchers to select and incorporate validation approaches in future connectivity modeling studies.
Related Posts
Reviewing four decades of fragmentation metrics
A review of fragmentation metrics since 1990 shows how methodologies have shifted over time, and what work still needs to be done to ensure transferable and ecologically grounded research.
A practical framework for climate-informed landscape planning
Climate change refugia and landscape connectivity can be systematically incorporated into planning through a practical five-step framework.
Planning approaches for coastal connectivity under sea-level rise
A comparison of two complementary modeling approaches in Florida provides a universal pathway for resilient coastal planning and design.
