Field scale temporal and spatial variability of δ13C, δ15.N, TC and TN soil properties: Implications for sediment source tracing

Use of natural tracers (in particular, delta C-13 and delta(81)5N, total carbon and nitrogen) for discriminating and apportioning soil sources of eroded sediment has increased. Some critical issues, however, including the variability of such tracers over time and space, merit further investigation. This study used a 25 m grid in a 1.78 ha hydrologically-isolated permanent pasture field in SW England to assess temporal and spatial variability in the discrimination (using Kruskal-Wallis H-tests) of top (0-2 cm) and subsoil (2-10 cm) samples afforded by these tracers. Grid sampling was repeated four times (twice in each of June and July 2015) to assess the impact of temporal variability and Monte Carlo random sub-sampling of the 25 m grid was used to assess the significance of spatial variability for discriminatory efficiency. Using grouped (combining all time-periods) data, all tracers were statistically significant (p <= 0.05) for discriminating soil samples by depth. For the repeat sample collection campaigns, delta C-13 discriminated samples by soil depth in three out of four, delta N-15 in one and both TC and TN in all periods. Grouped data (all time-periods, both soil depths) yielded statistically significant results for all tracers. All tracers apart from delta N-15 yielded statistically significant tests for the comparison of topsoil tracers across the four time periods, whilst all tracers except delta C-13 were significantly different in the subsoil samples across the sampling campaigns. All tracers therefore exhibited some statistically significant temporal variation. delta C-13 and delta N-15 exhibited less temporal dependence than TC and TN. Random sub-sampling (n = 10) with 100 iterations of the tracer data grid sampling points did not alter the discriminatory efficiency of the properties. The spatial nature of the tracer data, therefore, did not compromise or confound discrimination between the soil source samples by depth or across time. Un-mixing modelling with virtual sediment mixtures based on equal proportions of the field soil source tracers from each time period and using the field soil source data from each individual period, revealed sensitivity to the temporal dependence of the tracers. Overall, the findings suggest that the temporal variation in the tracers tested requires more consideration than the spatial dimension at the scale of an individual field, which in a conventional erosion source tracing study, would represent one location of many within the target landscape or catchment.