Epizootic tipping points: Environmental viral feedbacks predict amphibian die-offs

Virulent pathogens commonly circulate in wildlife populations without causing mass mortality, but the processes driving die-offs remain poorly understood. Prevailing frameworks emphasize individual-level susceptibility, although susceptibility factors often fail to predict population-level outcomes. We tracked ranavirus epizootics across 40 wood frog breeding ponds over three years, comparing models based on lagged viral state variables with those based on abiotic and host conditions. Models based on lagged viral-state variables received greater support for subsequent infection spread, intensification, and viral accumulation. Prevalence predicted later environmental viral concentration throughout epizootics, whereas relationships between environmental viral concentration and subsequent infection spread and intensification were substantially weakened when die-off onsets were excluded, consistent with phase-dependent feedback. Viral accumulation rates differentiated die-off from non-die-off pond-years while static pond and host features did not, suggesting that die-offs may be associated with pathogen accumulation in shared environments rather than being predictable from static pond and host features alone.