Using artificial mixtures to test the impacts of tracer combinations and model selection on the performance of sediment source fingerprinting in a burned area

Sediment source fingerprinting can be an effective method for identifying sediment sources in wildfire-impacted areas, but the effects of tracer and model selection on robustness remain poorly understood. In this study, soil samples were collected from three potential sources (burned surface, unburned surface, and channel banks) in a wildfire-affected area, and artificial mixtures with known source proportions were created. Three types of tracers (fallout radionuclides, magnetic susceptibilities, and soil colour parameters) were tested for their sensitivity to wildfire. Ten composite fingerprints, generated through the traditional three-step procedure (TSP) as well as consensus ranking and the conservativeness index (CM) were used to assess the accuracy of two un-mixing models. These comprised one frequentist (FingerPro) and one Bayesian (MixSIAR) model. The results indicated that wildfire had substantial effects on most tracer properties, with the median concentration and variance increased by up to 104% and 374%, respectively. Among the ten composite fingerprints, the CM selection method performed best, with the average and standard deviation of the corresponding MAE being 7% and 1%, respectively. While the TSP method could achieve a near-global optimum in some cases, it was the least stable among the ten tracer sets, generating a standard deviation for the MAE of 9%. Compared to FingerPro, MixSIAR solutions calculated using composite fingerprints excluding TSP returned lower MAE values (reduced by an average of 28%). The standard deviations of MAE for MixSIAR solutions employing tracer sets, except for CM, were lower (decreased by an average of 37%), suggesting that MixSIAR delivered higher accuracy and precision for our case study. These findings offer valuable insights for future fingerprinting research in wildfire impacted areas, which can support soil conservation and catchment restoration efforts in burned regions.