In current maps and geospatial datasets, representations of landforms such as mountains, hills, and ridgelines are unable to be drawn to their full extent. Due to the lack of a clearly observable boundary, the visualizations of these features are often limited to a singular point or line feature. This representation does not allow for an understanding of the true extent of landforms or the potential hierarchies that exist within the landscape. While manual attempts to delineate the extents of such features is always possible, it cannot be scaled for large areas with tens of thousands of features. In any case, there is no prescriptive way to delimit landforms, so no single set of delineated features can be considered sufficient for all people and contexts. In addition, the delineation of landforms depends on what type of landform is being searched for and the scale of delineation. Thus, this is not a deterministic process and needs to be context dependent. There needs to be flexibility and the focus should be on customizable methods, rather than canonical representations of landforms.
The author of this thesis builds upon previous work within the field of geomorphometry and semi-automated feature extraction approaches by exploring and testing the applicability of several methods for delineating landforms within a range of study areas. The goal is to assess which methods produce linear (ridges) and non-linear eminences (peaks, summits, mountains) that match common sense expectations for what these features should look like in the real world, and by extension on maps. The six methods explored within this research were proposed by Wood (1996), Jasiewicz and Stepinski (2013), Lundblad et al. (2006), Chaudry and Mackaness (2008), Sinha (2008), and Miliaresis and Argialas (1999). The methods were selected based on their popularity within the research community and/or the author’s judgment of the potential of the method for providing accurate mappings of terrain features. The first three methods represent general geomorphometric approaches and provide classification for all features across the study areas. Within this range, Wood’s and Jasiewicz and Stepinski’s methods were used to assess peak and ridge features while the Lundblad et al. method was only used to assess ridge features. The latter three methods represent specific geomorphometric approaches and only classify eminence features across the study areas. All three methods were used to assess large eminences that include multiple ridge and peak features.
All six methods were explored across three study areas within the continental US (Great Smoky Mountains (NC-TN), White Mountains (NH), and Colorado Plateau (NM)). Each method was run multiple times with a range of parameter values to fully assess the impact of parametric variation on the outputs. For the general geomorphometric method, these parameter impacts were explicitly examined. Visual analysis, based on standards designed within this research, and quantitative analysis were conducted on these results. The quantitative analysis included sensitivity and correlation analysis in RStudio and alignment analysis with GNIS features and manually delineated polygons from the USGS. For the specific geomorphometric methods, these parameters were used for exploration of the most accurate results but were not examined beyond that. The only analysis possible for these results was visual analysis with the goal of finding features that matched cognitive expectations. Excluding the visual analysis, all the creation and analysis was automated through Python scripts.
For the general geomorphometric methods, the results from this process indicated that none of them offer fully viable solutions. Wood’s morphometric features based method often resulted in a high number of reasonably small peaks with inaccurate placement. For ridge features, the method was similarly unsuccessful with narrow ridgelines which often had gaps along the line. The Jasiewicz and Stepinski’s geomorphon method yielded reasonable peak features with larger extents more often occurring with accurate placement. However, the ridge features from this method were not viable as they were incredibly inconsistent and incomplete along the ridgelines. Finally, Lundblad et al.’s method (encoded in the Benthic Terrain Modeler software) was only used to look at ridges and resulted in ridge features that had the most potential. The features were solid and continuous along the ridgelines. However, in many cases this method produced ridges that were arguably excessive in size.
For the specific geomorphometric methods, only Sinha’s method resulted in viable results. Chaudhry and Mackaness’s method was reliant on the concept of morphological variance to keep the eminences from expanding to excessive extents. However, this concept was not effective in two out of the three study areas. Sinha’s method offered 3 different approaches to constraining the eminence extents and all three present results that had potential and should be explored further. Constraining based on relative drop, slope threshold, or other prominent peaks all were successful with certain feature types and therefore could be revised to take context into consideration to produce more consistent results. Finally, Miliaresis and Argialas’s method was purely reliant on slope threshold for the expansion of features. This resulted in severe problems across all three study areas and was determined to not be a viable method for eminence classification.
Ultimately, this research has clearly illustrated the complexity of delineating topographic eminences across diverse landscapes. While some approaches appeared to have potential, none of them successfully handled all feature types, even within a single study area. This has revealed the necessity for future research to explore machine learning approaches for eminence delineation. All the methods within this research are reliant upon hardcoded values. This means that they will always prioritize one type of feature: steeply sloping eminence or gently sloping eminence. Clearly, trying to represent naturally formed landforms with hardcoded values will not be successful and therefore future research needs to take context into account in each study area. Machine learning methods may provide a reasonable solution that could be programed to take a DEM and assess each feature before applying any classification or hardcode values.