02099naa a2200181 a 450000100080000000500110000800800410001910000310006024500920009126000090018352014670019265000190165965000250167865300220170370000210172570000160174677301550176213389042020-03-04       2007    bl uuuu     u00u1 u    #d1 aMENDONÇA-SANTOS, M. de L.  aSoil prediction with spatially decomposed environmental factors.h[electronic resource]  c2007  aPrediction of soil attributes and soil classes in digital soil mapping relies on finding relationships between soil and the predictor variables of soil-forming factors and processes. The predictor variables can be remotely or proximally sensed images of soil, landscape, parent material or climatic factors. Till date, most prediction methods are based on performing regression on the predictor variables directly to predict soil attributes or classes. There are problems using data layers from different sources, particularly, multicollinearity, and the fact that the relationships between soil and environmental variables can change with spatial scale. To overcome the problem of correlation between variables, principal component analysis can be performed on the predictor variables. With respect to the spatial dependency, each of these variables can be decomposed into separate spatial components and mapped separately. One of the methods of achieving this is wavelet analysis, which decomposes the variables into separate hierarchical spatial components of decreasing spatial resolution. These components could all be derived and subsequently used as separate layers in predicting soil classes or soil attributes. In this chapter, data are decomposed using the wavelet method and examples of predictions of soil classes and surface-clay content are shown, in order to evaluate the effect of using the decomposed layers in comparison with the original data.  aRemote sensing  aSensoriamento Remoto  aAtributos do solo1 aMCBRATNEY, A. B.1 aMINASNY, B.  tIn: LAGACHERIE, P.; MCBRATNEY, A. B.; VOLTZ, M. (Ed.). Digital soil mapping: an introductory perspective. Amsterdam: Elsevier, 2007. cap. 21, 269-278.