Spatial modeling of agricultural catchments with management boundaries

The objective of this research was to assess the effect of perennial vegetation strips (PVSs), installed on the contours of row cropped fields, on the spatial distribution of soil properties. These PVSs have been shown to provide ecosystem services missing in conventionally row-cropped fields including sediment control, nutrient control, as well as habitat for wildlife and pollinators. While several studies have focused on edge of field monitoring, the spatial distribution of soil properties surrounding and within the PVSs is less understood. To examine these potential differences of soil properties in relation to PVSs, soil samples were collected on hillslope profile transects that crossed PVSs and a control catchment with matching plan curvature and flow accumulation. The soil property data was analyzed in two ways. First, sample points were blocked into categorical positions relative to the PVSs and by categories of planform curvature. Second, differences in soil properties between treatments for all catchments were simulated using spatial modeling with digital terrain analysis, applying rule-based multiple linear regression. Results of this study indicate that ten years after installation, PVSs lead to a significant increase in soil pH and a significant decrease in soil test P concentrations within PVSs (α = 0.05). Spatial patterns of soil separates (sand, silt, and clay percentages) were significantly different between the test and control catchment (α = 0.05). Planform curvature had a greater effect on the distribution of soil properties than the presence of a PVS. Greater SOM variability was observed in PVS catchments; however, no significant differences in SOM concentrations were detected. Patterns of SOM detected by the data mining approach indicated that aspect in the control and planform curvature in the test catchment were more influential than spatial relation to the PVSs. Data mining using the Cubist algorithm indicated which covariates and analysis scales were most useful in predicting soil properties. Spatial modeling supported by data mining assisted in visualizing simulated differences between control and test catchments and indicating where different spatial patterns of soil properties may be expected