Co-limitation by stable, dynamic and directional habitat features shapes climate vulnerability in an alpine specialist
Alpine ecosystems are among the most climate sensitive on Earth, yet logistical challenges and detection biases often impede robust assessment of alpine dependent species. We investigated habitat associations and density patterns of the Sierra Nevada subspecies of the Gray-crowned Rosy-Finch (Leucosticte tephrocotis dawsoni), an alpine obligate and regional endemic, over five breeding seasons from 2018 to 2022 using hierarchical distance sampling and mark-recapture distance sampling to explicitly account for imperfect detection and spatial heterogeneity. Density estimates tracked annual snowpack variation, ranging from 4.77 individuals/km{superscript 2} in a low snow year to 12.08 individuals/km{superscript 2} in a high snow year. Abundance was highest near persistent snow patches that provide foraging habitat and near cliffs that provide nesting substrate, and declined sharply above approximately 10% woody cover, with densities approaching zero beyond approximately 25%, indicating a steep ecological threshold. In contrast, the proportion of surveyed blocks with detections remained relatively stable across years. Together, these patterns indicate a three timescale co-limitation framework in which breeding habitat is shaped by static features (cliffs), dynamic annual drivers (snowpack), and longer-term directional change (woody encroachment). By linking population density to climate sensitive habitat features, this study provides a high-resolution abundance-based baseline for long term monitoring and offers a framework for evaluating climate vulnerability in alpine and other resource-limited systems. Open Research StatementData necessary to replicate the analyses and results presented in this manuscript will be archived in the Dryad Digital Repository upon acceptance, with no embargo on the material. The R code associated with this manuscript is not novel. All analyses use publicly available packages and functions without modification, including mrds (v2.3.0), unmarked (v1.2.5), tidyverse (v2.0.0), MuMIn (v1.47.5), and standard ggplot2 visualization tools. All code is properly cited within the manuscript and publicly available through CRAN. Complete analysis scripts will be archived in Dryad alongside the data upon acceptance to facilitate full reproducibility of results.