Fig. 1

Conceptual diagram of how occupancy, prevalence, and load are estimates at the hibernaculum (i), swab (j), and run (k) level annually (t) using the multi-scale dynamic occupancy hurdle model. The model utilizes quantitative real-time PCR (qPCR) runs (y_ijtk) from swabs (a_ijt) used to determine the pathogen status of Myotis lucifugus at winter hibernacula (z_it). Probability of detection (p) was modeled via the probability of detecting Pseudogymnoascus destructans(Pd) on a bat (p_sbat) multiplied by the probability of Pd DNA replicating in the sample given it is present. Probability of the DNA replicating was dependent on the Cycle threshold (Ct) score and per unit (of Pd load) detection probability (p_det) which is scaled by the latent average Pd load of each bat (x^effload_ijt). We derived the load for each run (yload_ijtk) by transforming the Ct score. We then calculated the average load per swab (xload_ijt), hibernaculum (xload_it), and across all hibernacula (Mean Load) by accounting for observation error due to load variation (σ₃), load variation between individuals within a cave (σ₂), and variation in load across caves (σ₁), respectively. Effective loads (x^effload_it and x^effload_ijt) are calculated by multiplying the hibernacula or swab status by load. Dashed lines indicate false positive outcomes, which we assumed did not occur.