Extracting high points from statistical surfaces: A case study using digital elevation models

We introduce and test algorithm for extracting high-point locations from statistical surface data. The algorithm uses map algebra and local neighborhood analysis via three key parameters: minimum vertical gain vertical gain neighborhood, and horizontal separation neighborhood. Though the method is applicable to any x, y, z data set, we tested it on 1:250,000 digital elevation models (DEMs) for Arizona. The resulting high points were compared quantitatively with an independent data set of named summits from the USGS Geographic Names Information System (GNIS). The comparison showed that, on an aggregate basis, the extraction method can approximate the number and spatial pattern of high points when compared to the GNIS points. However, extraction by neighborhood analysis may consistently misdiagnose certain features, such as the edges of troughs (e.g., canyon rims).