Sediment source fingerprinting and the temporal variability of source contributions

Sediment source fingerprinting has been progressively developed and refined over the past 40 years or more and now represents a widely used and valuable technique, with important practical applications. However, relatively little attention has been given to the target samples and the extent to which they are able to provide meaningful information on short- or longer-term relative source contributions for a given study catchment. A key issue here is the inherent short- and longer-term temporal variability of source contributions and the extent to which such variability is taken into account by the target samples. The objective of this study was to investigate the temporal variation of source contributions from the Qiaozi West catchment, a small (1.09 km2) gully catchment located within the Loess Plateau of China. The target samples represented a suite of 214 spot suspended sediment samples collected during eight representative wet season rainfall events occurring over two years. A suite of geochemical properties was used as fingerprints and standard source apportionment calculations indicated that the gully walls contributed the most sediment (load-weighted mean 54.5%) and, together with cropland (loadweighted mean 37.3%), and gully slopes (load-weighed mean 6.6%) were the main sediment sources. The 214 individual target samples indicated that the contribution of cropland sources varied between 8.3% and 60.4%, gully walls between 22.9% and 85.8% and gully slopes between 1.1% and 30.7%, representing ranges of 52.1%, 62.9% and 29.6% respectively. In order to explore whether the temporal variability of source contributions demonstrated by the study catchment should be seen as typical, equivalent information was abstracted from 14 published studies for other catchments of varying size and located in different environments worldwide. This information demonstrated similar temporal variability of the relative contributions of the major sources, which were typically characterized by ranges of the order of 30-70%. The temporal variability associated with the estimates of relative source contributions provided by target samples has important implications for the uncertainty associated with such estimates derived using source fingerprinting techniques based on a limited number of target samples. Further attention needs to be directed to the design of sampling programmes used to collect such samples and to taking account of such uncertainty in source apportionment calculations.