The long-term persistence and stability of host-pathogen interactions are often strongly affected by spatial distribution of the interacting populations. This is particularly true in seasonal systems where conditions for pathogen survival and host availability vary over time.  Our study occurs in the context of a trophic cascade which involves a microparasitic entomopathogenic nematode (<i>Heterohabditis marelatus</i>, hereafter EPNs) as the natural enemy of a root-feeding ghost moth caterpillar (<i>Hepialus californicus</i>) found on lupine bushes (<i>Lupinus arboreus</i>). In the absence of EPNs, root-feeding caterpillars can kill and destroy large stands of lupines, especially in dry years. EPNs, when present, suppress such ghost moth outbreaks thereby protecting the lupines. EPNs depend strongly on moisture for movement and persistence. Given the seasonal nature of the environment where moisture regimes and host availability vary dramatically, it is unclear how EPNs persist.

Our 13-year dataset shows considerable variation in EPN incidence among sites. We tested whether these differences were because of variation in survivorship. Using a known initial EPN number, we estimated the daily mortality rates for each of our sites.  A negative binomial model fit the data well (Pearson’s X2/df = 0.898). We used these mortality rates in a fully stochastic model for host-pathogen interactions in seasonal environments to calculate EPN extinction probabilities for a single season. Results from the survivorship experiment did not explain the full range of incidence patterns. Given the nature of these results, we hypothesize that dispersal is an important mechanism in explaining the discrepancy between model predictions and field observations.