Larger geologic complexity implies larger uncertainty

Larger geologic complexity implies larger uncertainty
Alessandro Samuel-Rosa¹, Ricardo Simão Diniz Dalmolin² and Pablo Miguel³

1. Curso de Pós-Graduação em Agronomia-Ciência do Solo da Universidade Federal Rural do Rio de Janeiro, Seropédica, RJ, Brazil. (alessandrosamuel@yahoo.com.br)

2.Departamento de Solos da Universidade Federal de Santa Maria, Santa Maria, Rio Grande do Sul, Brazil. (dalmolinrsd@gmail.com)
3.Programa de Pós-Graduação em Ciência do Solo da Universidade Federal de Santa Maria, Santa Maria, Rio Grande do Sul, Brazil. (tchemiguel@yahoo.com.br)

Abstract: The prediction of soil attributes through predictive models became popular in the last decade. Land-surface parameters are the ones used at the largest extend as predictor variables. However, the relation between land-surface parameters and soil attributes is not evident in all land surfaces: it might be strong in one but weak in another. Besides, this relation usually varies across the landscape due to the other soil forming factors influence such as the parent material. Using the results of a study carried out in a geologically complex small catchment in Southern Brazil we show that soil particle-size distribution can be estimated from land-surface parameters. The prediction functions can explain more than half the overall data variance. The literature shows that such performance is superior to that of the conventional soil mapping approach. However, when we evaluate the spatial accuracy of the predictive models we see that in locations of larger geological heterogeneity the predictive models present large prediction errors. Thus, larger geological heterogeneity implies larger uncertainty. Unfortunately the geological information available is not sufficiently accurate to help improving significantly the spatial accuracy of the prediction functions.

Keywords: predictive models, linear regression models, geology, soil information user

AttachmentSize
SamuelRosaAccuracy2012.pdf138.22 KB