Climate-Driven Changes in the Habitat Suitability and Potential Co-Occurrence of Daurian Ground Squirrels and Pikas in Inner Mongolia: Implications for Brucellosis Surveillance

Where a multi-host pathogen can circulate depends on where its hosts overlap, not on any single host’s range, yet climate-change projections for zoonoses rarely model two hosts jointly or test how a warming signal is channeled toward their overlap. We addressed this gap for brucellosis, which remains endemic across the pastoral landscapes of Inner Mongolia, China, by building ENMeval-optimized MaxEnt niche models for two candidate sylvatic hosts (infection of which with Brucella in this region has not been established), the Daurian ground squirrel (Spermophilus dauricus, n = 620) and the Daurian pika (Ochotona dauurica, n = 17, with a three-method variable-screening pipeline and small-sample controls), projecting both onto current climate and eight CMIP6 scenarios (SSP1-2.6/2-4.5/3-7.0/5-8.5 × 2050s/2070s; validation AUC = 0.808 for the squirrel and 0.940 for the pika), combining them into a joint host co-occurrence index (JCI) that describes potential habitat overlap rather than disease risk, and fitting a path model that describes how climate anomaly relates to each host’s suitability and to the JCI. The two hosts diverge sharply: by the 2070s, suitable habitat contracts by 9.6–84.3% for the generalist ground squirrel but by >99.9% for the pika under the fitted model, whose projected collapse follows from a threshold in mean wettest-quarter temperature (≈18–19 °C) that is exceeded almost everywhere in the future; this loss is robust to occurrence filtering, regularization, binary thresholds and a second climate model but depends on the temperature variable and on climates beyond those sampled by the 17 records. This asymmetry carries over to the overlap: the current 66,737 km2 zone where both hosts co-occur shrinks by >99.9% under six of eight future scenarios, a collapse that is robust to how the joint index is constructed (geometric mean, minimum, or arithmetic mean all show the same pattern). The path model shows that climate anomaly is the better linear predictor of the squirrel’s own habitat trajectory (standardized β = −0.195 vs. −0.059 for the pika), whereas the pika’s habitat change has nearly double the standardized association with JCI change (β = 0.635 vs. 0.324); the latter association is largely a mathematical consequence of how the JCI is constructed, and under spatially corrected estimation, the climate effect on pika suitability is statistically indistinguishable from zero (95% CI −0.305 to 0.184). An exploratory next-generation-matrix calculation with uncalibrated parameters shows the same direction of change under most parameter combinations (the area with R0 > 1 declines under all but one) but is sensitive to its assumptions and is not a validated epidemiological prediction. These results identify the pika as the more informative indicator of change in host overlap for ecological monitoring; whether it matters more for brucellosis maintenance or spillover is unknown and requires serological, abundance and contact-rate data. The approach offers a reproducible template (joint niche modeling coupled to path analysis and exploratory transmission-threshold modeling) for exploring multi-host co-occurrence under climate change beyond this system.