Apportioning Sources of Fine Sediment in an
Agricultural Watershed using Colour Fingerprinting
Louise R.M. Barthod a, Kui Liua, David A. Lobb a,*, Philip N. Owens b,
Núria Martínez-Carreras c, Alexander J. Koiter d, Ellen L. Petticrew e,
Gregory K. McCullough f, Cenwei Liua, Leticia Gasparb
Department of Soil Science and Watershed Systems Research
Program, University of Manitoba, 13 Freedman Crescent, Winnipeg,
Manitoba R3T 2N2, Canada
b Environmental Science Program and Quesnel River Research Centre,
University of Northern British Columbia, 3333 University Way, Prince
George, British Columbia, V2N 4Z9, Canada
c Centre de Recherche Publique – Gabriel Lippmann, Department of
Environment and Agro-Biotechnologies, 41 rue du Brill, L-4422
Belvaux, Luxembourg
d Natural Resources and Environmental Studies Program, University of
Northern British Columbia, 3333 University Way, Prince George, British
Columbia, V2N 4Z9, Canada
e Geography Program and Quesnel River Research Centre, University
of Northern British Columbia, 3333 University Way, Prince George,
British Columbia, V2N 4Z9, Canada
f Centre for Earth Observation Science, Department of Environment
and Geography, 482 Wallace Building, University of Manitoba,
Winnipeg, Canada R3T 2N2
Corresponding author e-mail address: david.lobb@umanitoba.ca
(David Lobb)
Abbreviations: STCW, South Tobacco Creek Watershed; CDA,
canonical discriminant analysis; DFA, discriminant function analysis;
IQR, inter-quartile range
a
Abstract
The use of sediment color as a fingerprint property to determine
sediment sources is an emerging technique that can provide a rapid
and inexpensive means of investigating sediment sources. The present
study aims to test the ability of color parameters to distinguish
between potential sources of sediment and apportion their
contributions within the South Tobacco Creek Watershed, MB, Canada
(74 km2). The results are compared with the findings from a
conventional fingerprinting approach undertaken in the same
watershed (i.e., geochemical and radionuclide fingerprints).
Suspended sediment and potential source samples were collected at
six sites along the main stem of the creek. Canonical discriminant
analysis was used to determine the number of source groups. Sample
color of sediments and source materials was quantified from diffuse
reflectance spectrometry measurements and sixteen color coefficients
were derived from several color space models (CIE XYZ, CIE xyY, CIE
Lab, CIE Luv, CIE Lch, Landsat RGB, Redness Index). Experiments
were undertaken to test the linearly-additive nature of the color
coefficients and this was used to further refine the number of
coefficients that were suitable for source apportionment. The
discrimination power of the color coefficients was assessed using
discriminant function analysis. The selected fingerprints were included
in a Bayesian mixing model using Monte-Carlo simulation to apportion
the contribution of the different sources to the sediment collected at
each site. Consistent with the conventional fingerprinting approach, a
switch in the dominant sediment source between the headwaters and
the outlet of the watershed was observed. As the stream crosses and
down-cuts through the Manitoba Escarpment, sediment is enriched
with shale bedrock and stream bank materials from within and below
the escarpment and depleted of topsoil and/or stream bank materials
from above the escarpment.