Spatial differentiation of cultivated soils using compound-specific stable isotopes (CSSIs) in a temperate agricultural watershed in Manitoba, Canada

Purpose Compound-specific stable isotopes (CSSIs) of very long-chain fatty acids (VLCFAs) of plant origin were investigated in a soil and sediment tracing context in a watershed in Manitoba, Canada. Spatial and temporal variability in δ 13 C FA values and concentrations was examined at the point, transect, and field scales to determine the (1) ability to differentiate sediment sources in C3-cropped fields, (2) impact of subsampling on source tracer fingerprints, and (3) major sediment source for a downstream mixture using the Bayesian unmixing model MixSIAR. Materials and methods Analysis was performed for five agricultural fields over six sampling periods. Soil and sediment samples (320) were processed for VLCFA analyses (C20:0–C30:0, C32:0). Quantification was performed by gas chromatography–flame ionization detection (GC-FID) and 13 C determination by GC combustion–isotope ratio mass spectrometry (GCC-IRMS). Data were analyzed using weighted t tests to differentiate fields by δ 13 C FA values. The major sediment source was determined using the following steps: (1) a point-in-polygon approach to identify VLCFA tracers; (2) unmixing using MixSIAR; (3) source apportioning using VLCFA concentrations and %C. Results and discussion VLCFA δ 13 C FA values vary spatially within a cropped field due to environmental factors. Sediment source fingerprints are dependent on the variability in δ 13 C FA values and the quantitative combining of subsamples. Cropped fields that appeared homogeneous exhibited a large range in δ 13 C FA values, with variability greatest for fall and spring samples; concentrations were lowest at these times. Historical field boundaries played a role. A downstream sediment mixture (June 2013) was analyzed and found to correspond with source data from August 2012. Sediment mixture data (δ 13 C FA ) for several VLCFAs were found to fall within the source mixing polygons produced by using two cultivated fields and a riparian zone sample as sources. Conclusions Variability in δ 13 C FA values increased in fall and spring, which could affect the number of subsamples required per source. Most fields could be spatially differentiated using a weighted t test, but not necessarily using the same VLCFA chain lengths. Two spatially separated fields with similar cropping histories were difficult to differentiate, but one of the fields was more prone to VLCFA losses. Only one of several source sampling periods led to successful unmixing, suggesting multip