Patterns of soil phosphorus: Concentrations and variability across an urbanizing agricultural landscape

Eutrophication, the overenrichment of aquatic ecosystems with nutrients such as phosphorus (P), is a major cause of reduced freshwater quality. The primary source of P to most lakes is nonpoint source runoff. Runoff is not generated evenly across a watershed; instead, the majority is generated in a few critical source areas. Understanding the location of critical source areas is a significant step toward managing P runoff to aquatic ecosystems.; Urban-rural gradients are one method for predicting soil P. I compare several measures of a nonlinear urban-rural gradient to other predictors of soil P in Dane County, Wisconsin, USA. Most of the factors expected to drive differences in soil P concentrations were not found to be good predictors of soil P. While there were several significant relationships, most explained only a small proportion of the variation. High variability in soil P concentrations made accurate prediction difficult.; Human activity may change variance patterns by shifting the scale at which governing processes are operating or the governing processes that are dominant at a given scale. I measured soil P concentrations and variability at three scales across four management regimes. Variance changed across scale in each management regime and across management regimes at the same scale. Variance within a field was lower in human-dominated sites than in prairies. Intensely used sites are more homogeneous and have higher P concentrations than prairies: they are more likely to be critical source areas.; In ecological studies, there are often many regressions that have similar goodness-of-fit statistics. Choosing one model as “best” may discard important information in models that fit almost as well. A technique based on bootstrapping was used to solve this problem for predictions of soil P in Dane County. Bagging, which aggregates the predictions of many well-fitting models, improved prediction accuracy and enhanced assessment of prediction uncertainty and variability.; Effective eutrophication management will require knowledge of soil P accumulation patterns. The three techniques for understanding soil P concentrations and variability that I examined demonstrated that predicting P concentrations is difficult; however, it may be possible to predict critical source areas using information about variability.