Making survival spatial: an integrated model for territory occupancy and capture-recapture data

Jaume-Adria Badia-Boher , Michael Schaub, Mátyás Prommer, Marc Kéry

Knowledge about spatial variation in survival is central to understanding population dynamics and guiding conservation, yet assessing it is very hard. This limitation arises because capture-mark-recapture (CMR) data required for such inference must be collected over large spatial extents, which is logistically demanding and seldom possible. By contrast, territory occupancy (TO) data are typically spatially rich and widely available for territorial species, but they do not directly inform individual survival.

We developed an integrated model combining CMR and TO data. The model links site-level occupancy dynamics, governed by site persistence and colonization probabilities, to survival of the territory owner, allowing both data sources to jointly inform spatiotemporal variation in survival. To accommodate potential violations of this deterministic occupancy-survival link arising from breeding dispersal or alternative rescue dynamics, we included an estimable scaling parameter (κ). We evaluated model performance using simulation across different survival structures (constant, spatial, spatiotemporal), life histories, and CMR detection probabilities, and assessed robustness of inference when the occupancy-survival link is violated. We applied the model to long-term peregrine falcon (Falco peregrinus) data in Hungary, evaluating the effects on survival of the presence of a predator (eagle owl), and the proportion of agricultural land.

Survival estimated with the integrated model showed negligible bias and good coverage across all simulation scenarios, while substantially improving precision relative to CMR-only analyses. Precision gains were largest for spatial regression coefficients (up to 80%) and temporal variance parameters (up to 88%); they increased with model complexity and declining detection probability. When the occupancy-survival link was violated, κ absorbed the resulting discrepancy and prevented bias in survival. In the case study, integration substantially improved the precision of spatiotemporal survival estimates and revealed a negative association with eagle owl presence and a positive one with the proportion of agricultural land.

Our new integrated model improves estimation of spatial and temporal variation in survival by leveraging shared information across data sources, extending spatially explicit demographic inference to systems where CMR data alone are insufficient, and thereby allowing spatial survival inference in a broader range of populations.