Predictive modeling of spatial patterns of soil nutrients related to fertility islands

In arid shrublands, soil resources are patchily distributed around shrub canopies, forming well-studied “islands of fertility.” While soil nutrient patterns have previously been characterized quantitatively, we develop a predictive model that explicitly considers the distance from shrubs of varying canopy sizes. In 1-ha macroplots in both the Sonoran and Mojave Deserts, we used Plant Root Simulator™ probes to measure nutrient availability along transects extending north and south from creosote bushes (Larrea tridentata). We modeled the decline of nutrients with distance from focal shrubs using hierarchical mixed models that included the effects of transect direction and shrub canopy size. Of the nutrients considered, nitrogen and potassium had the strongest response to distance from focal shrubs. In the Sonora, both depended on canopy size and had different patterns to the north versus the south. In the Mojave, potassium depended on size and direction, but nitrogen only on canopy size. We used the fitted model equations and the location and canopy size of all Larrea shrubs within the macroplots to estimate nutrient concentrations at a 20 cm resolution. This produced maps showing nutrient “hotspots” centered on Larrea. Our models predicted up to 60 % of the variation in nutrient availability the following growing season. Our models efficiently used a moderate number of sample locations to predict nutrient concentrations over a large area, given easily measured values of shrub size and location. Our method can be applied to many systems with patchily distributed resources focused around major structural landscape features.