Comparison of Hydrologic Algorithms for Mapping Valleys Using Digital Elevation Models

Valleys are fundamental components of many landscapes. Valley forms integrate diverse geomorphic influences, including fluvial, hillslope, and glacial processes. The geometry of a valley, such as its volume and cross-sectional shape, is commonly used to analyze long-term sediment mass balance across Earth’s surface and to interpret past geologic processes, including changes in climate and regional tectonics. Underlying lithology can also play a significant role in valley form through modification of erosion rates. To automatically map valleys, several algorithms have been developed that rely on various topographic attributes within widely available elevation data. However, significant uncertainties persist regarding the similarities and differences between these algorithms, and the feasibility of up-scaling them for large areas and large data volumes. These uncertainties currently limit the ability to quantitatively characterize and compare valley morphologies across different landscapes. To assess the prospects for regional-scale use, I compiled three algorithms for test environments. The algorithms include 1) the Valley Bottom Extraction Tool that considers slope and drainage area to identify valley bottoms; 2) the progressive black top hat algorithm that identifies valleys as local minima of topography; and 3) the elevation threshold method (“flooding” approach) that uses the elevation of the closest point in a channel to predict valley boundaries. In order to evaluate baseline performance of the algorithms, I tested outputs for three common landscape types with valleys, compared outputs to flood maps from the federal government, and examined the feasibility of applying each algorithm over large spatial scales, using the state of Virginia as an example. Emergent limitations of these algorithms were then methodically analyzed to more accurately characterize valleys from topography. I ran parameter sensitivity tests for each algorithm and closely compared outputs of mapped valleys to investigate error tendencies. I compared calibrated outputs to independently generated, manually mapped valleys to provide a slightly more objective measure of valley mapping accuracy. These analyses indicated that while each algorithm has best use cases, the progressive black top hat algorithm is the ideal choice for mapping valleys from topography due to its efficiency, accuracy, and scalability. As an example application of this algorithm, I found that within the different physiographic regions of Virginia, valley width decreases as rock strength increases. The collective investigation of these algorithms provides insight into how automatic mapping of valleys can be useful across many environmental science fields.