01555nam a2200193 a 450000100080000000500110000800800410001910000180006024500830007826001350016130000090029652009280030565000090123365300200124265300260126270000230128870000190131170000310133019235102014-12-09       2012    bl uuuu     u00u1 u    #d1 aTEN CATEN, A.  aDigital soil mappingbstrategy for data pre-processing.h[electronic resource]  aIn: THE GLOBAL WORKSHOPS ON DIGITAL SOIL MAPPING, 5., 10 - 13 Apr. 2012, Sidney. Digital Soil Assessments and Beyond: Sidneyc2012  a1 p.  aSoil maps have a great deal of ambiguity regarding the exact location of transition zones, which leads pedologists to disagree about the proper delineation of soil classes at those locations. The aim of this study was to propose a preprocessing strategy applied to digital soil mapping. Soil polygons on a training map were displaced in its inward direction by 100 and 160 m. This strategy has enabled that data covariates located near the borders of soil classes were not used for Decision Tree (DT) model adjusting. Three DT models derived from eight predictors covariates, related to the soil formation factors relief and organisms, were sampled in a complete soil map and by polygons displaced 100 and 160 m, in order to be used to predict soil classes. The DT model derived from observations distant 160 m of the boundary between polygons in the original map was less complex and shown a better predictive performance.  asoil  aDigital mapping  aStrategic information1 aDALMOLIN, R. S. D.1 aRUIZ, L. F. C.1 aMENDONÇA-SANTOS, M. de L.