Multi-scale integrated modeling framework of animal population dynamics incorporating spatial capture-recapture and occupancy data collected from citizen science and traditional sampling
To understand and manage animal populations in rapidly changing landscapes, ecological questions have been increasing in geographic scale and temporal scope. Citizen science approaches to data collection have been effective for monitoring natural systems over large scales, and recent methodological and computational advancements have made the large quantity and often variable quality of citizen science data more tractable. Occupancy approaches, developed originally for smaller-scale, intensive sampling, have translated well to citizen-science efforts for collecting and modeling presence/absence data. Further, recent extensions to combining occupancy and spatial capture-recapture models now enable presence/absence data collected with multiple methods to be integrated into a single joint model. However, this has not been broadened to include data collected with citizen science. We demonstrate that it is possible to explicitly integrate presence/absence data collected from citizen science methods with spatial capture-recapture, occupancy, and radio-telemetry into a single integrated model to estimate population parameters. This development is particularly appropriate for management of animal populations, which requires understanding populations at multiple ecological scales and often with multiple sampling methods.
We describe a conceptual framework for a single integrated population model based on the underlying process of space usage that governs how individuals move on a landscape. The basis for this model is a spatial capture-recapture model, which we then extend to incorporate other sources of data. We argue that radio-telemetry yields the greatest amount of spatial and temporal data on space usage of individuals and that data collected with other methods, including citizen science, spatial capture-recapture, and occupancy, provide sparse or thinned versions due to imperfect detection. We develop a formal analysis using marginal likelihoods of the independent datasets, which contain shared covariates and parameters. We apply this framework to monitoring efforts for an expanding black bear population in New York, and share simulation results that demonstrate how this integrated approach is valuable for identifying patterns animal distribution, resource selection, and movement across a range of spatial scales. This multi-scale, integrated modeling approach that incorporates data from citizen science efforts and traditional intensive sampling methods efficiently uses data, increases the precision of estimates, and provides a mechanistic understanding of population dynamics.