Landscapes commonly comprise of mosaics, patches and boundaries. Riparian boundaries are complex to delineate and characterize, with a multitude of variables available for delineation. Multiple methods exist for boundary delineation such as two-dimensional wombling, constrained classification techniques and discontinuity detection. One method that has proven to be reliable in boundary delineation with one-dimensional transect data is the moving split window (MSW) analysis. This study demonstrates the efficacy of MSW to delineate grass species turnover and environmental boundaries across two geologically dissimilar riparian zones in the Kruger National Park, South Africa. There are few studies that have delineated riparian boundaries of Kruger National Park, and none that have used the MSW analysis. MSW detects significant changes in dissimilarity indices of variables along gradients. Significant shifts in dissimilarity designate boundaries at various spatial scales dictated by window sizes. Significant boundaries emerge by altering window sizes, increasing quadrat width and removing infrequent herbaceous species. By utilizing these three methods, MSW background variance was reduced and riparian and wetland/upland boundaries were sharper and more easily defined. © 2009 Elsevier Masson SAS. All rights reserved.