Variability of net soil erosion and it's association with topography in Canadian Prairie agricultural landscapes

Abstract

In this thesis, the central hypothesis is that net soil erosion varies systematically with landscape topography. To test this hypothesis, two small (84 and 180 ha) agricultural drainage basins In central Saskatchewan were studied. The concentration of 137Cs in the soil was used to estimate net erosion, integrated over the past 20 to 25 years, at nearly 300 sites in the two study basins. The topography of each study area was represented by a digital terrain model. Computer programs were used to calculate topographic variables for each element of the digital terrain models. These variables expressed the three-dimensional surface shape and relative position of each element within the landscape. Representative samples drawn from each basin showed only weak correlations between topographic variables and net erosion. A seven unit landform model of the association between net erosion and topography was proposed. The model units were main channel, tributary channels, swales, midslopes, crest slopes, level upland, and depressions. An analysis of variance showed these landform units to have significantly different net erosion ranks. While the model was applicable to both study basins, the mean net erosion for some units differed markedly between basins. Correlations between net erosion and topographic variables within the landform units were generally stronger than for the basins as a whole. The correlations suggested different processes dominate soil redistribution in different parts of the basins. An algorithm was introduced to objectively group net erosion estimates into topographically defined subsets. The classification results supported the seven unit landform model.