Soil Mapping in the Pantanal of Mato Grosso, Brazil, using Multitemporal Landsat TM data

Abstract

Multitemporal data sets from the Landsat Thematic Mapper (TM) were used to evaluate their applicability for exploratory soil mapping in the floodplain of the Northern Pantanal of Mato Grosso, Brazil. Fifty-four soil profiles were classified into 21 soil units according to the FAO–UNESCO system. Information layers of vegetation types and dynamics of flooding were elaborated by applying supervised hierarchical classification rules. Geomorphologic units were mapped by visual image interpretation. Multinomial logistic regression was applied to test relations between thematic layers and soil units as well as aggregated soil clusters, developing a statistical mapping model. Northern Pantanal floodplain soils show a high variability as a function of age and granulemetry of underlying sediments, as well as soil moisture and flooding regimes. GIS layers of nine vegetation formations, three geomorphologic units and three multi-temporal moisture types were elaborated. Cross-tabulations and multinomial logistic regression models indicate significant relations between FAO–UNESCO soil units and GIS layers. As soil sampling density had been low, a final predictive model was developed for the mapping of six aggregated soil clusters, obtaining a high significance level (p<0.05) for prediction. Applied methodology was found to be appropriate to develop models on soil–landscape relationships and improve information on spatial distribution of soil groupings in the Northern Pantanal.