Simple statistical approach for computing land cover types and potential natural vegetation

Abstract

The main objective of this study was to show the potential of a simple and computationally inexpensive statistical method for the computation of land cover types (LCTs) and potential natural vegetation (PNV), which can be easily adapted to any LCT scheme used by climate models. We propose a diagnostic model (Vegetation Reconstruction by Diagnostic Equilibrium, VERDE), which is based on the cluster analysis of high-resolution datasets of observed LCT distribution and of climate variables. We discuss the reliability of this statistical approach and show that VERDE can be applied for reconstructing PNV distribution in areas such as Europe, India, and China where original vegetation has been replaced by crops and urban areas. According to VERDE, the dominant PNV consists of broadleaf deciduous trees in Central Europe, mixed savanna and grassland in Eastern Europe at mid-latitudes, and evergreen needle trees in Russia. Large areas of India are covered by savanna, and of China by grassland, mixed forest, and evergreen broadleaf trees. VERDE was applied to 5 climate model scenarios (produced by HadCM3, GFDL-CM2.0, IPSL-CM4, CSIRO-MK3, and CNRMCM3) to identify changes in potential vegetation at a global scale that would be induced by the projected climate change at the end of the 21st century. In the Northern Hemisphere, our results showed an increase in barren soils (deserts) in the areas from the tropics to the mid-latitudes, a northward shift of various types of forest, and a reduction in snowor ice-covered land and in areas occupied by shrubs and bushes (tundra) at high latitudes. Changes were smaller in the Southern Hemisphere and suggest increases in savanna in South America and shrublands in Australia.