Towards machine ecoregionalization of Earth's landmass using pattern segmentation method

摘要: Abstract We present and evaluate a quantitative method for delineation of ecophysiographic regions throughout the entire terrestrial landmass. The uses new pattern-based segmentation technique which attempts to emulate qualitative, weight-of-evidence approach ecoregions in computer code. An region is characterized by homogeneous physiography defined cohesiveness patterns four variables: land cover, soils, landforms, climatic patterns. Homogeneous necessary but not sufficient condition be an ecoregion, thus machine first, important step toward global ecoregionalization. In this paper, we focus on first-order approximation proposed – basis cover alone. justify existence significant spatial associations between various physiographic variables. Resulting regionalization (ECOR) shown more physiographically than existing ecoregionalizations (Terrestrial Ecoregions World (TEW) Bailey's Continents (BEC)). presented has advantage being transparent objective. It can verified, easily updated, modified customized specific applications. Each ECOR contains detailed, SQL-searchable information about within it. also computer-generated label. To give sense how compares TEW and, U.S., EPA Level III ecoregions, contrast these different delineations using two sites as examples. conclude that yields somewhat similar level world, automatic means.