A predictive model for soil seed bank outcomes in the Pyrenophora semeniperda-Bromus tectorum pathosystem

Pyrenophora semeniperda is abundant in soil seed banks of Bromus tectorum, where it kills a fraction of seeds throughout the year. The pathogen engages in a race with host seeds for endosperm resources; the pathogen success is negatively correlated with seed germination speed. We developed a deterministic model to predict pathosystem outcomes (seed death versus seed escape), using seed bank data from 80 sites collected over a 13-year period. The response variable (killed seeds in the spring seed bank) was regressed on multiple predictor variables (pathogen and host densities at seed dispersal, amount and timing of precipitation). Increased mortality was associated with high seed rain, high pathogen density, and low autumn precipitation. On xeric sites, a positive feedback loop between pathogen and host is created by a large carryover seed bank containing secondarily dormant seeds vulnerable to fungal attack and results in higher inoculum loads at seed dispersal the following year.