摘要

This paper reports on the development of a rapid and simple spectra-reflectance-based fingerprinting approach for documenting suspended sediment sources during storm runoff events. The methodology was applied in the rural Wollefsbach catchment (4.4 km(2), NW Luxembourg). Different chemical properties of the sediment retained on glass fibre filters after filtration were predicted from VIS/NIR spectra (ASD FieldSpec-II spectrometer, 0.4-2.5 A mu m) through partial least-square regression models (PLSR). Chemical property values measured on time-integrated suspended sediment, bed sediment and potential sediment sources samples (topsoil and channel banks) collected in the catchment were used to calibrate the PLSR models. Appropriate properties were selected and used to apply the 'spectral-reflectance-based fingerprinting approach', in order to estimate the sources of the suspended sediment retained on the filters. Uncertainty associated with model predictions was assessed using an inclusive approach to mixing models based on Bayesian error estimation, incorporating tracer property variability using a Monte Carlo simulation technique. The methodology proved to be a useful, fast and easy to apply approach not only for studying the variation of suspended sediment properties during storm runoff events from spectra, but also for documenting the variation of suspended sediment sources during individual storm events. The spectral-reflectance-based fingerprinting approach possesses considerable potential, since it possesses a number of important advantages over conventional source fingerprinting techniques, in terms of cost and time and labour requirements.

  • 出版日期2010-4