The use of carbon and nitrogen isotopes to study watershed erosion processes

Tracer studies are needed to better understand watershed soil erosion and calibrate watershed erosion models. For the first time, stable nitrogen and carbon isotopes (δ¹⁵N and δ¹³C) and the carbon to nitrogen atomic ratio (C/N) natural tracers are used to investigate temporal and spatial variability of erosion processes within a sub-watershed. Temporal variability was assessed by comparing δ¹⁵N, δ¹³C, and C/N of eroded-soils from a non-equilibrium erosion event immediately following freezing and thawing of surface soils with two erosion events characterized by equilibrium conditions with erosion downcutting. Spatial variability was assessed for the equilibrium events by using the δ¹⁵N and δ¹³C signatures of eroded-soils to measure the fraction of eroded-soil derived from rill/interrill erosion on upland hillslopes as compared to headcut erosion on floodplains. In order to perform this study, a number of tasks were carried out including: (1) sampling source-soils from upland hillslopes and floodplains, (2) sampling eroded-soils with an in situ trap in the stream of the sub-watershed, (3) isotopic and elemental analysis of the samples using isotope ratio mass spectrometry, (4) fractioning eroded-soil to its upland rill/interrill and floodplain headcut end-members using an unmixing model within a Bayesian Markov Chain Monte Carlo framework, and (5) evaluating tracer unmixing model results by comparison with process-based erosion prediction models for rill/interrill and headcut erosion processes. Results showed that finer soil particles eroded during the non-equilibrium event were enriched in δ¹⁵N and δ¹³C tracers and depleted in C/N tracer relative to coarser soil particles eroded during the equilibrium events. Correlation of tracer signature with soil particle size was explainable based on known biogeochemical processes. δ¹⁵N and δ¹³C were also able to distinguish between upland rill/interrill erosion and floodplain headcut erosion, which was due to different plant cover at the erosion sources. Results from the tracer unmixing model highlighted future needs for coupling rill/interrill and headcut erosion prediction models.