DEM to Mountain Transformation of the Zagros Ranges

Key words: Terrain Feature Extraction, Region Growing Segmentation, Digital Terrain Model, GTOPO30, Continent to Continent Collision, Iran

Mountains are the elementary morphotectonic features at regional (physiographic) scale and their topographic expression commonly reflects the current style of deformation. In a previous research effort, a methodology (DEM to Mountain transformation) was developed for the extraction of mountains from the GTOPO30 digital elevation model (DEM) with 30 arc-second spacing. More specifically region-growing segmentation was applied and the ridge pixels were used as seeds while the growing criterion was based on gradient. The methodology was implemented in a study area of 82,000 km² within the Great Basin Section of the Basin and Range Province in the southwestern U.S.A. In this area, the crust is under tensional forces and is thinned by normal faulting, which results in an array of tipped mountain blocks that are separated from broad basin plains. The objective of the present research effort is to implement (and modify, if needed) the DEM to Mountain transformation to an area of compressional stress. The case study was developed for the Zagros Ranges in Iran and the study area occupied approximately 330,000 km². In the Zargos Ranges the crust shortens and thickens, producing a spectacular mountainous physiography due to the collision of the Arabian shield with Iran. In the southeast portion of the study area, broad gently sloping valleys are observed in between the mountain features. On the contrary in the northwest part more tightly spaced mountain features are present with narrow, deep sloping valleys developed between them. In order to cope with these conditions, the DEM to Mountain transformation was modified and the valley pixels were not allowed to participate in the region-growing segmentation process. More specifically, runoff simulation was implemented, and ridge pixels and valley pixels were labeled on the basis of their runoff accumulation value. Then an iterative region-growing segmentation algorithm was applied. During the first iteration, the ridge pixels formed the initial set of mountain pixels while the rest of the pixels formed the current set of non-mountain pixels. In each iteration, if a non-mountain pixel satisfied the following three conditions (a) its gradient was > 6^o, (b) the pixel was an 8-connected neighbor to the current set of mountain pixels and (c) it did not belonged to the set of valley pixels, then it was flagged as a new mountain pixel and the existing set of mountain pixels was updated. Segmentation stopped if no more pixels were added during the latest iteration. Then, small isolated islands of mountain pixels that represent either mountain remnants or artificial elevated error-peaks in GTOPO30 were removed. Additionally, small islands of non-mountain pixels standing on mountain tops and surrounded by mountain pixels were merged to the mountain terrain class. The extracted mountain objects were interpreted to be in accordance with both (a) statistical data computed for the mountain and non-mountain terrain classes and (b) the mountain features interpreted visually from a shaded relief map of the study area. The methodology allowed the extraction of mountain features in an area with different regional geomorphic (physiographic) conditions than those observed in the Basin and Range physiographic province. In the future, the mountains will be represented by a set of numerical attributes and the landscape of the study areas will be characterized on the basis of the morphometry of mountains.