A Curvature-Based Method For Measuring Valley Width Applied To Glacial And Fluvial Landscapes

Valleys produced by glaciers are morphologically distinct from those created by rivers, but these differences are surprisingly difficult to automatically detect using digital topographic data. This contribution tests whether glacial and fluvial valleys can be discriminated using scaling between watershed area and valley width. We present a method for estimating valley width at each point in digital topographic data by determining the cross-valley scale at which normalized principal curvature is minimized. We assess how well this approach measures valley width in synthetic valleys using artificial valley cross sections with random topographic noise. These width estimates are validated against manual measurements in glacial and fluvial valleys. Finally, we assess the sensitivity of these quantities to the pixel dimension of the input data to determine the detection limits of the method for moderate (20 m) and coarse-resolution (>30 m) digital elevation models (DEMs). We find that valley widths are underpredicted by a factor of 2.5 to 4; nonetheless, they are well correlated with both synthetically prescribed and empirically measured widths and so can be used as reliable measures of valley width when scaled to account for this underprediction. We find clear distinctions between valley width-catchment area relationships in glacial and nonglacial valleys, indicating that this methodology might be deployed globally to characterize the distribution and extent of glacial landscapes across Earth. In addition, this technique could be used to determine anomalous downstream width changes related to processes such as valley aggradation.