Integrating recursive behavior into resource selection analyses: case study of restocked desert bighorn sheep

Recursive behavior is the repeated and systematic revisitation to the same locations or resource patches, which is ubiquitous across the animal kingdom, yet few studies have explicitly incorporated it into resource selection models. To address this gap, we investigated relationships among seasonality, recursive movements, and resource selection in restocked desert bighorn sheep ( Ovis canadensis mexicana ). We outfitted 16 individuals (9 F, 7 M) with GPS collars programmed to record locations every 3 h from 2022 to 2023. We identified (1) latent behavioral states using a Hidden Markov Model, (2) recursive behavior by quantifying revisitation frequency and residence time, and (3) selection patterns using an integrated step selection analysis. We observed a strong relationship between revisitation and residence rates. When revisitation rate (revisits/30 d) ranged from 1 to 12, residence rate (h/d) increased by approximately 1 h for every 6 revisits; above 12 revisits, residence time increased by 1 h for every 2 additional revisits. Revisitation rates were more than three times greater during the resource-stress season (3.08 [95% CI: 2.71–3.60]; 23 July–21 August) than during the establishment (0.98 [95% CI: 0.90–1.08]; 29 November–28 February) and resource-abundant (1.00 [95% CI: 0.92–1.11]; 22 August–28 November) seasons. This pattern suggests that when resources are abundant, sheep can meet nutritional and water requirements without repeatedly returning to previously used locations, potentially reducing predation risk by making movements less predictable. Selection as a function of revisitation frequently exhibited opposing trends with selection for a given landscape covariate shifting from positive to negative across the gradient of revisitation rates. At the highest revisitation rates, sheep consistently selected areas with greater NDVI across seasons with the strongest selection occurring during the resource-stress period. This pattern is consistent with a relaxed functional response in habitat specialists, whereby selection for a resource intensifies as its availability declines. Our findings advance understanding of desert bighorn sheep habitat selection by incorporating recursive behavior into resource selection models. Moreover, the methods described herein provide a foundation for future research to study patch-foraging dynamics, overgrazing, ephemeral resource availability, site fidelity, and mating opportunities.