The effects of spatial prediction of grain size fractions on intertidal surface sediments classification

In this paper, the effects of prediction uncertainty of grain size fractions on intertidal surface sediments classification are investigated. The grain size fractions are compositional data, so the characteristics of the compositional data should be considered during spatial prediction. For efficient compositional data analysis, an additive log-ratio transformation is considered in this paper. Through a case study in Baramarae tidal flats, Korea, spatial prediction using cokriging of log-ratio transformed variables could meet constraints of compositional data and produced superior not only prediction accuracy but also classification accuracy, when compared with cokriging of untransformed fractions.