Landscape-scale patterns of shrub-species abundance in California chaparral

We examine the degree to which landscape-scale spatial patterns of shrub-species abundance in California chaparral reflect topographically mediated environmental conditions, and evaluate whether these patterns correspond to known ecophysiological plant processes. Regression tree models are developed to predict spatial patterns in the abundance of 12 chaparral shrub and tree species in three watersheds of the Santa Ynez Mountains, California. The species response models are driven by five variables: average annual soil moisture, seasonal variability in soil moisture, average annual photosynthetically active radiation, maximum air temperature over the dry season (May-October), and substrate rockiness. The energy and moisture variables are derived by integrating high resolution (10 m) digital terrain data and daily climate observations with a process-based hydro-ecological model (RHESSys). Field-sampled data on species abundance are spatially integrated with the distributed environmental variables for developing and evaluating the species response models. The species considered are differentially distributed along topographically-mediated environmental gradients in ways that are consistent with known ecophysiological processes. Spatial patterns in shrub abundance are most strongly associated with annual soil moisture and solar radiation. Substrate rockiness is also closely associated with the establishment of certain species, such as Adenostoma fasciculatum and Arctostaphylos glauca. In general, species that depend on fire for seedling recruitment (e.g., Ceanothous megacarpus) occur at high abundance in xeric environments, whereas species that do not depend on fire (e.g., Heteromeles arbutifolia) occur at higher abundance in mesic environments. Model performance varies between species and is related to life history strategies for regeneration. The scale of our analysis may be less effective at capturing the processes that underlie the establishment of species that do not depend on fire for recruitment. Analysis of predication errors in relation to environmental conditions and the abundance of potentially competing species suggest factors not explicitly considered in the species response models.