Effect of sampling design on characterizing surface soil fingerprinting properties

Abstract

Purpose

The characterization of soil properties is an important part of many different types of agri-environmental research including inventory, comparison, and manipulation studies. Sediment source fingerprinting is a method that is increasingly being used to link sediment sources to downstream sediment. Characterization of sediment sources is a key part of this method as it serves as the base to identify and quantify soil properties (fingerprints) that can discriminate between sources. There is currently not a standard approach to characterizing sources and the different approaches to sampling have not been well assessed.

Methods

Grid (n = 49), transect (n = 14), and likely to erode (n = 8) sampling designs were used to characterize the geochemical, colour, grain size distribution, and soil organic matter content at two sites under contrasting land uses (agricultural and forested). The impact of the three sampling designs on characterization of fingerprint properties, the relationship between particle size and organic matter content on fingerprint properties, fingerprint selection, source discrimination, and mixing apportionment results were evaluated using a range of methods including 21 virtual mixtures.

Results

The likely to erode design resulted in a unique fingerprint signature compared to the other two sampling designs. The correlation between particle size and organic matter varied between fingerprint, source, and sampling design. While the number and composition of the fingerprints selected varied between sampling designs there was a strong (100%) discrimination between sources regardless of the sampling approach. The maximum absolute difference between the virtual mixtures and the modeled proportions was 7.7%, 7.8%, and 8.9% for the grid, transect, and likely to erode sampling designs, respectively.

Conclusions

The likely to erode sampling design was not representative of the upslope areas as characterized by the grid and transect methods. Despite these differences the final apportionment results using virtual mixtures were qualitatively similar between the three sampling designs. Continued work at the watershed scale is needed to fully evaluate the importance of source sampling design on the sediment source fingerprinting approach.