An automated knickzone selection algorithm (KZ-Picker) to analyze transient landscapes: Calibration and validation

Abstract

Streams commonly respond to base-level fall by localizing erosion within steepened, convex knickzone reaches. Localized incision causes knickzone reaches to migrate upstream. Such migrating knickzones dictate the pace of landscape response to changes in tectonics or erosional efficiency and can help quantify the timing and source of base-level fall. Identification of knickzones typically requires individual selection of steepened reaches: a process that is tedious and subjective and has no efficient means to measure knickzone size. We construct an algorithm to automate this procedure by selecting the bounds of knickzone reaches in a χ-space (drainage-area normalized) framework. An automated feature calibrates algorithm parameters to a subset of knickzones handpicked by the user. The algorithm uses these parameters as consistent criteria to identify knickzones objectively, and then the algorithm measures the height, length, and slope of each knickzone reach. We test the algorithm on 1, 10, and 30 m resolution digital elevation models (DEMs) of six catchments (trunk-stream lengths: 2.1–5.4 km) on Santa Cruz Island, southern California. On the 1 m DEM, algorithm-selected knickzones confirm 93% of handpicked knickzone positions (n = 178) to a spatial accuracy of ≤100 m, 88% to an accuracy within 50 m, and 46% to an accuracy within 10 m. Using 10 and 30 m DEMs, accuracy is similar: 88–86% to ≤100 m and 82% to ≤50 m (n = 38 and 36, respectively). The algorithm enables efficient regional comparison of the size and location of knickzones with geologic structures, mapped landforms, and hillslope morphology, thereby facilitating approaches to characterize the dynamics of transient landscapes.