Predicting landscape-level distribution and abundance: Integrating demography, fire, elevation and landscape habitat configuration

Demography varies in response to multiple interactive drivers at varying scales. However, these relationships are often assessed piecemeal, particularly without consideration of drivers at the landscape level. We developed a model to predict population dynamics of an endangered, perennial herb Hypericum cumulicola across a range of landscape drivers. We estimated vital rates using 22years of annual census data from 15 populations (10,910 and 38,313 unique observations) and additional seeding experiments, considering four landscape drivers (time-since-fire, patch area, patch aggregation and patch elevation). We combined these models into integral projection models to assess population parameters. Predictions of occupancy and density were compared with an independent dataset of 33 habitat patches. We also evaluated the effect of fire-return interval on population persistence. The landscape drivers had interactive effects on vital rates and demography. Occupancy was predicted for most patches, including many that were unoccupied. Abundances were poorly predicted, primarily because projections were unreliable for patches having intermediate areas and aggregation. When these patches were removed, the model explained 42% of the variance in abundance. Projected population growth was greater and extinction risk lower under more frequent fire-return intervals, at higher elevations, and in larger and more aggregated patches. Our modelling suggests that fire can be prescribed less often in larger than smaller patches to support a viable metapopulation of H. cumulicola.Synthesis. The integration of landscape-level drivers and detailed demographic data is a valuable tool for understanding species abundances, distributions and dynamics at large scales. We evaluated the effect of interactions among fire, patch elevation, area and aggregation on the demography of a rare and endangered plant. Our findings reinforce the importance of regional dynamics. These drivers and their interactions suggest locations that can support higher vital rates, occupancy and abundance, and what fire-return intervals are optimal depending on landscape characteristics. This approach demonstrates a link between plant demography and landscape variables, although further work is needed to improve ecological predictions.