Geomorphologia Slovaca et Bohemica   
Vol. 2019/1
Vol. 2018/2
Vol. 2018/1
Vol. 2017/2
Vol. 2017/1
Vol. 2016/2
Vol. 2016/1
Vol. 2015/2
Vol. 2015/1
Vol. 2014/2
Vol. 2014/1
Vol. 2013/2
Vol. 2013/1
Vol. 2012/2
Vol. 2012/1
Vol. 2011/2
Vol. 2011/1
Vol. 2010/2
Vol. 2010/1
Vol. 2009/2
Vol. 2009/1
Vol. 2008/2
Vol. 2008/1
Vol. 2007/2
Vol. 2007/1

Geomorphologia Slovaca et Bohemica


Morphometrical-morphostructural subdivision of the Western Carpathians by Object-based image analysis
.: Download PDF :.


Peter Bandura, Jozef Minár, Lucian Drăguţ:
Morphometrical-morphostructural subdivision of the Western Carpathians by Object-based image analysis
Geomorphologia Slovaca et Bohemica, 19, 2019, 1, 5 - 104.
69 figs., 16 tabs., 110 refs., 2 supplements

Geomorphological regionalisation (or subdivision) has long tradition in Slovak geomorphology. However, in view of very high demands for objective approaches in current geomorphology, the regionalisation currently in use in Slovakia suffers from subjective decisions of its authors, as it was elaborated manually. To address the demands, means of object-based image analysis (OBIA) started to be applied for automated delineation of various geomorphological features. The attempts were successful at the local level but it is still not clear to what degree this method can replace manual mapping at the regional level. Thus, the main aim of this work was to develop, using OBIA, a semi-automated method for objective delineation of the basic, third-order morphostructures of the Western Carpathians s. l., which should approximately correspond with the traditional geomorphological regions. Initial object-based segmentation of three morphostructurally meaningful geomorphometric variables (slope gradient, vertical dissection of terrain and elevation) based on the automated ESP2 tool showed as efficient, though with limitations caused by fixed segmentation scale applied globally on the highly contrasting terrain (i.e. over- or under-segmentation). To overcome them, we developed a complex object-oriented workflow consisting of the definition of the most suitable segmentation extent of our study area, objective optimisation of iterative segmentation scale based on its stabilisation (controlled by segmentation step size in the ESP tool), and the selection and removal of distinct individuals based on their mean difference in elevation to neighbouring objects. Such systematic repetition of segmentations with decreasing scale used for partitioning of systematically smaller domains proved to be important when delineation of morphostructural features variable in size and homogeneity degree is targeted. As a result, one object level represents a combination of several hierarchically-structured object levels, different in the level of detail. Compared to the traditional geomorphological regions within the Czech and Slovak part of the Western Carpathians s. s., it has quantitative compatibility of ca. 60 %. Boundaries that are not statistically compatible with the traditional regions are not considered as failure of the proposed approach, since most of them still have a morphostructural meaning either in terms of their alternative interpretation or as more- or less-detailed substitution. Some of other inconsistencies were visually attributed to uncertainties present in the traditional regions, which cannot be dealt with algorithmically, e.g. boundaries that to not follow the obvious morphological contrast or were determined on the basis of lithology. Our subdivision can be used as a more objective alternative for the traditional geomorphological regions, with strongly reduced influence of subjective decisions in the process. Further research could improve the proposed methodology, and testing of its transferability to other areas could lead to better verification of our hypotheses.