Digital soil mapping: Predicting soil classes distribution in large areas based on existing soil maps from similar small areas

ABSTRACT There is an ever-growing need for soil maps, since detailed information directly related to agricultural activities, urbanization and environmental protection. However, there a lack of large-scale maps in developing tropical countries such as Brazil. Albeit are small areas, large regions usually have undetailed maps. Considering the importance finding low-cost alternatives overcome information, main objective this work was manually create local map extrapolate it similar larger areas that information. The Anhumas River Basin, municipality Itajubá, southeast Brazil, mapped used predict soils distribution entire municipality. First, prediction model tested same basin provided sufficient results, achieving 67% global accuracy 0.62 Kappa coefficient. Second, resulting together with José Pereira Basin municipality, 54% 0.40 Low resolution parent material found confuse models; showed better results when variable removed. Minas Gerais presents general mapping units only Acrisol class its associations other classes area. predicted by identified more classes. Mapping representative extrapolating these constitute promising alternative