Demography often varies among populations, but most scientists have look at these patterns in a piecemeal way. We used a landscape level variable, relative elevation above wetlands (“dryness”) as a predictor of population variation of the perennial herb Hypericum cumulicola. Previously, we identified time-since-fire as a key driver of this species’ dynamics, but we hypothesized that dryness would explain additional variation among populations in a landscape with variable local elevation. We used data on marked plants and density plots from 1994-2014 from 14 populations, about 11K plants and 34K cases of annual vital rates, along with experiments on seed dormancy, germination and seedling survival. We used generalized linear mixed models to assess the effects of stage (first year plant or adult), plant height, time-since-fire and dryness, with random effects by population and year, on annual survival, plant growth, probability of reproduction and fecundity. Seed survival, dormancy, and germination were evaluated with general additive models. Models were then combined in an integral projection model to estimate population growth rates and other parameters. We modeled H. cumulicola occupancy in patches of various sizes and degrees of spatial isolation, comparing results to an independent dataset of patch occupancy.
Results/Conclusions
Both fire and dryness had significant and often interacting effects on germination, seed dormancy, plant survival, plant growth, probability of reproduction, and fecundity. Population growth rates peaked about 5 years post-fire and then declined sharply through about ten years, leveling out after that. Compared to wetter areas, drier sites had higher population rates shortly after fire, lower rates at intermediate times-since-fire, and similar rates when long unburned. Observed plant densities were most closely related to model output from the drier sites. These results broadly agree with prior models on the primacy of time-since-fire in driving the population dynamics of H. cumulicola, but adds detail on how the landscape affects plant demography, likely through changes in the recovery of vegetation structure. Simulations on landscapes with realistic spatial patterns of landscape patches showed effects of fire, dryness, patch size, and isolation on occupancy. Some suitable habitat patches (in terms of fire and dryness) were usually unoccupied due to small size or isolation. Modeling across a landscape gradient can help inform management. In this case, prescribed burns can occur less often in drier than wetter parts of the landscape, to support a viable metapopulation of H. cumulicola.