Estimation of soil properties with geostatistical methods in floodplains

Abstract

Floodplain ecosystems reveal considerable spatial variability. As part of the interdisciplinary project ‘Generalised Robust Indication System for Ecological Changes in Riverside Wetlands’ (RIVA) we studied the spatial spread of sand-, silt-, clay- and humus content in two areas at the Central Elbe River, Germany. For the interpolation and simulation we used Conventional Soil Mapping, Ordinary Kriging, Cokriging, Collocated Cokriging and Conditional Simulation and discussed their suitability. We combined the knowledge of regional soil scientists with geostatistical methods to map soils. Some 613 sampling points were available at the main area and 184 sampling points at the second area. We propose a method which enables the consideration of additional information available at the estimation points. The following additional parameters are present: Elevation, distance from the river, groundwater floor distance and the effect of biotope type. Clear result improvements are obtained by considering elevation or groundwater floor distance applying a geostatistical method called collocated cokriging. This is particularly important for the main investigation area, since with classical kriging, satisfying results can hardly be achieved. Geostatistical methods cannot substitute Conventional Soil Mapping; they can be an excellent aid for a better understanding of soil distribution. They allow us to consider the soil as a continuum; in this issue they overcome Conventional Soil Mapping.