Worldwide inventory of landscapes through segmentation of global land cover dataset

Abstract

Complete global inventory of landscapes (patterns of land cover) is obtained by segmentation/classification of the CCI-LC – a high resolution global land cover dataset. The CCI-LC is first segmented into a large number of small land units. The pattern of CCI-LC categories within each unit is tightly controlled by segmentation algorithm’s merging parameter. Second, land units are classified into more manageable number (400-600) of landscape classes (LANDCs) based on composition and character of their patterns. Pattern properties are reflected in automatically generated class labels. The set of LANDCs provides a global inventory of landscapes. The final result of this work is a suite of vector files containing maps of segmentations and LANDCs at three different levels of spatial scale of pattern (9, 15, and 30 km). These maps differ from the GLC maps by having coarser spatial resolution but higher thematic resolution; they are also SQL-searchable. They have applications in macroecology to serve as proxies of vegetation structure, biotic composition, and can provide first-order information about geographical distribution of biodiversity. The method can be extended to multilayer segmentation for delineation of ecoregions.

Cite
Nowosad J., Stepinski T.F. (2017) Worldwide inventory of landscapes through segmentation of global land cover dataset. GeoComputation 2017, Leeds, UK, September 4-7, 2017
Date
Links
PDF