An improved pixel counting method for arbitrary zonal statistics on GlobeLand30

Abstract

Global Land cover is essential for ecological study, climate and environmental changes, resources management, human activities as well as sustainable development. With the finer resolution products arising, e.g. GlobeLand30, quantitative analysis of the earth dynamics based on raster data sets is becoming possible. For arbitrary zonal statistics, pixel counting method is an operational practice widely used for area estimation by counting the number of classified pixels and multiplying by the area represented by each pixel, though uncertainty. For large statistical regions, the error of the projection can not be neglect. Thus an rectified PC method is developed for arbitrary zonal statistics of all raster data sets. The test result shows that the method performance is stable and easy to zonal statistics.