The influence of forest definition on landscape fragmentation assessment in Rondônia, Brazil

Estimates of tropical deforestation and forest degradation are misleading, partly because different studies make use of different forest definitions. This paper investigates the influence of this confusion on the assessment of forest extent and its spatial distribution, by means of fine-scaled land cover maps and landscape metrics. The state of Rondônia, Brazil, located in the southwestern part of the Amazon basin and known for its fishbone-like pattern of deforestation, is used as a study area. A 1:250 000 vector data product from the Brazilian Geography and Statistics Institute (IBGE), describing the land cover type in a three-step hierarchy specifying canopy density, topography, and dominant life forms, was rasterized and analyzed. Forest subcategories were aggregated into a seven level gradient, ranging from a level that is very specific and only includes dense multi-layered rain forest, to less strict levels containing open forest systems, secondary vegetation, and tree savannas. We show that there is a consistent difference between the initial class aggregation level, and all other levels, which gradually broaden the forest definition and are characterized by very distinct ecological parameters, such as a higher mean patch size, increased levels of landscape connectivity and slightly more irregularly shaped patches. We recommend a harmonization of the major forest definitions in use today, while taking care not to lose the relevant ecological information that can be extracted from its most detailed classification level.