A distribuição das formações vegetacionais do semiárido brasileiro nos últimos 21.000 anos

Abstract

In the context of South America, one of the main regions that need to be approached by paleoecological studies is the Brazilian Semiarid, an area marked by aridity in the present and located in a meeting of vegetation from different biomes. The region's past reconstructed by paleoclimatic models in the Pleistocene and Holocene periods is related to intense climatic changes in short time scales. In this work, our objective was to use the combination of bioclimatic and edaphic predictors to build models with machine learning algorithms capable of predicting the vegetation of the present and the last 21,000 years. Our results show that the gain in accuracy in the classification of vegetation formations is constant as edaphic predictors are added. The option to use subsurface horizons to generate edaphic attributes also resulted in greater local detail in the distribution of vegetation. In predictions made in past periods, the greatest change is promoted during the Heinrich Stadial 1 period, with the retraction of the areas of open vegetation formations of the Hyperxerophilous Caatinga, advance of the Arboreal Caatinga, Forested Caatinga and Deciduous Forest areas. The vegetation pattern present had its establishment observed only in the Late Holocene, which indicates that the previous humidity conditions in the Brazilian Semiarid allowed a greater occupation of forest formations during the end of the Pleistocene.