1. Determining Sources of Suspended
Sediment in an Agricultural Watershed in
Atlantic Canada
Agriculture and Agri-Food Canada
University of Manitoba
Monica Boudreault, Sheng Li, David Lobb, Glenn Benoy and
Kui Liu
2. Sediment leads to water quality concerns
Sources of sediments vary in type and quantity
Bank erosion
Surface runoff erosion
Different land uses
Sediments transported by streams frequently
embody a mixture of sediment derived from
different source areas
Sediments
http://www.barrierreef.org/our-science/research-we-support/attributes/sediment-sourcing
3. Questions to be answered
3
Where are the sediments coming from
Is the majority of suspended sediment coming from
bank or surface erosion?
If suspended sediment is coming from surface
erosion can we distinguish between land use types
to determine the contribution from each?
Can we estimate the contributions from a specific
source at a specific time?
4. Sediment fingerprinting
It is based on the assumption that one or more
properties of suspended sediment will reflect its
origins and can be used as tracers/signatures
to track the sediment back to its source.
Sediment =
mixture of
S1 and S2
Source 1
Ts₁=10
Source 2
Ts₂=20
Tss=15:
50%=S₁
50%=S₂
Tss=18:
80%=S2
20%=S1
Unmixing
5. Traditional experiments
5
Sampling
Sediment samples: only one location --- watershed
outlet
Source samples: random locations in the watershed
for all possible sources
Assumptions
Sources are uniform within the watershed
High connectivity along the water path
6. Other experimental designs
6
What if
Not enough samples (e.g., source difficult to sample)
Watershed is heterogeneous in nature
Blockages/storages along the water path
Strategies
Taking source samples outside the watershed
Taking additional sediment samples
Different ways to analyze the data
b
b b
a
a
a
7. Simple versus complex watershed
studies
7
Simple watershed study = not spatially defined
Watershed source sampling / single sediment
sampling
Regional source sampling / single sediment
sampling
Watershed source sampling / multiple sediment
sampling
Complex watershed study = spatially defined
Upstream source definition
Reach source definition
Traditional experiment
8. Simple versus complex watershed
studies
8
Simple watershed study = not spatially defined
Watershed source sampling / single sediment
sampling
Regional source sampling / single sediment
sampling
Watershed source sampling / multiple sediment
sampling
Complex watershed study = spatially defined
Upstream source definition
Reach source definition
b
b b
a
a
a
9. Simple versus complex watershed
studies
9
Simple watershed study = not spatially defined
Watershed source sampling / single sediment
sampling
Regional source sampling / single sediment
sampling
Watershed source sampling / multiple sediment
sampling
Complex watershed study = spatially defined
Upstream source definition
Reach source definition
10. Simple versus complex watershed
studies
10
Simple watershed study = not spatially defined
Watershed source sampling / single sediment
sampling
Regional source sampling / single sediment
sampling
Watershed source sampling / multiple sediment
sampling
Complex watershed study = spatially defined
Upstream source definition
Reach source definition
11. Simple versus complex watershed
studies
11
Simple watershed study = not spatially defined
Watershed source sampling / single sediment
sampling
Regional source sampling / single sediment
sampling
Watershed source sampling / multiple sediment
sampling
Complex watershed study = spatially defined
Upstream source definition
Reach source definition
12. Study site
65% agricultural
21% forest
14% residential or
wetland area
Black Brook Watershed (BBW)
Northwest New Brunswick,
Canada
Potato production systems
13. Source Samples: Field
Soil transects extend from the top of the
stream bank, through the riparian zone
and into the field
(3-7 samples/transect)
Down-slope, mid-slope and upslope
14. Source Samples: Stream Bank
Sampled actively eroding stream
banks
3 profiles from each location
10 cm increments
15. Suspended Sediment
15
Modified version of the Phillips designed in-
situ time-integrated suspended sediment
sampler
Collected seasonally, from 2008-2014
16. Measurements/tracers
Tracers may include:
*Colour properties
*Radionuclide
Geochemical
Particle Size
Stable Isotopes
Organic constituents
Mineralogy and Magnetism
Distinct diagnostic signatures used to determine the proportion
of each source contributing to suspended sediment (assortment
of mixed signature sources)
Selection of tracers based
on the following criteria:
1. Conservative behaviour
2. Ability to distinguish between
sources
3. Linearly additive behaviour
17. Tracer 1: Spectral-Reflectance/Color
Spectral differences due to heterogeneous combination
of:
o Minerals
o Water
o Organic matter
Quantified using diffuse reflectance spectrometry
350-2500 nm wavelength spectroradiometer ASD
FieldSpecPro
Reflected light collected with a fiber optic cable
18. Tracer 2: Radionuclide Cs-137
18
Man-made radionuclide
Nuclear weapon testing
Distributed back to Earth by precipitation
Found only of surface soils
Tracer used to distinguish between
surface and subsurface soil
19. Un-mixing Model: MixSIAR
19
MixSIAR (R program)
Originally developed for ecological
studies
Used to estimate the relative
contribution of potential sediment
sources at each sediment sampling site
Bayesian isotopic mixing model
Monte Carlo methods
Statistical procedures
Canonical Discriminant Analysis
(sources)
Shapiro Wilk test (normality)
MixSIAR BIPLOT Analysis in R
(signature conservative, major sources
included)
Kruskall-Wallis H-Test (signature
different among sources)
Stepwise Discriminate Function
Analysis (DFA) (signature combination
25. Summary
25
Both bank erosion and field erosion can
contribute significantly into the sediments in
streams of the BBW
Switch in sources from headwaters to outlet
(forest for upper reaches and cultivated field
for lower reaches) can only be detected
with nested spatial framework
Strong local influences at some locations
High uncertainties with small sample sizes
26. Funding
Agriculture and Agri-Food Canada
NSERC Strategic and Discovery Grants programs
Environment Canada
Field and lab work
Serban Danielescu
Rick Allaby
Fangzhou Zheng
Sylvie Lavoie
Lionel Stevens
Meagan Betts
Tegan Smith
Zisheng Xing
Acknowledgement
Sources of sediments vary in type and quantity depending on location and erosional process responsible
Sediments transported by streams or rivers frequently embody a mixture of sediment derived from different source areas and source types within a catchment
Identification of main sources (pinpoints polluters)
Pinpointing the main causes of non-point source pollution is vital for the proper management of sediments entering water sources.
Best management practices (BMPs) can be utilized to control/reduce water quality degradation by sediments however knowledge about where the sediments are coming from is necessary.
For example, there may be many possible non-point source polluters in an area…
BBW covers approximately 1450 ha at about 7.5 km long and 3.5 km wide
It is a tributary of Little River Watershed (LRW) which accounts for 380 square-kilometres and discharges into the St. John River Valley
65% of land used for agriculture
21% forest
14% residential
Surface soil samples were taken upstream of where the suspended sediment samplers were installed. Each sub-watershed within BBW was sampled for suspended sediment and had at least one soil sample transect completed.
Soil transects extended from the top of the stream bank through the riparian zone and into the field.
Each transect was sampled to represent each section of the slope including downslope, midslope and upslope.
Surface soil sampling locations were selected based on the following criteria: 1) visible evidence of erosion or soil degradation 2) distance and potential connectivity to stream 3) access and safety Only surface soils that could potentially erode and contribute to suspended sediment were sampled.
In agricultural fields, soil samples were taken to the same depth as the tillage layer because the tilling of soil often mixes many different soil horizons changing its natural vertical profile.
The stream bank sampling procedure was based on the knowledge that over time banks will erode across the entire bank height either by instantaneous collapse or non-synchronous lower and upper bank retreat through a combination of erosion processes
Cross-sectional difference between inlet tube and pipe reduces sediment/water velocity which collects the sample
Low maintenance, installed and left for an entire season, checked every 2 weeks to ensure that stream flow and tree branches haven’t washed away samplers, Low cost as well and doesn’t require power (electrical) source like many other suspended sediment sampler types.
Sediment fingerprinting is a technique used to develop a better understanding of sediment dynamics within watersheds
Based on the assumption that one or more properties of the suspended sediment will reflect its origins and can be used as a tracer to track the sediment
Distinct diagnostic signatures of potential sources are used to determine the proportion of each source contributing to the suspended sediment which represents an assortment of mixed signature sources
Used to determine the proportion of sources in a mixture
Determining the source of erosion is incredibly important for trying to manage for it
Best Management Practices (BMPs) are used to control water pollution. They are particularly utilized and useful in agricultural landscapes that tend to have significant effects on surface water through pesticides, organic pollutants and suspended sediments.
In order to implement the right BMP’s that will be the most useful in reducing erosion, the sources of erosion must be determined, this can be done using sediment fingerprinting
Based on the assumption that one or more of the properties of the suspended sediment will reflect its origins and can be used as a tracer to track the sediment
Quantified using diffuse reflectance spectrometry
Relatively inexpensive means of characterising the physical, chemical and biological properties of soils
Spectral differences due to the behaviour of the heterogeneous combination of:
Minerals
Water
Organic matter
Best used in small catchments
Non-conservative
Reflectance spectra of sediments and source materials were quantified using a diffuse reflectance spectrometry
Martinez et al. (2010) tested the colour based fingerprinting approach on medium and large sized catchments and found that the level of source discrimination decreased as the catchment size increased. This is likely caused by heterogeneous pedology and geology, source overlap and intra-source variability (Martinez et al. 2010).
Nuclear weapon testing in the late 1950’s
Chernobyl 1986 (Northern hemisphere only)
A type of mixing/unmixing model that can be applied to solve various environmental problems
Mixing v.s. Unmixing depending on your objectives, typically for sediment fingerprinting (Un-Mixing), ecological (Mixing)
MixSIAR is often used in research pertaining to:
Diet composition, population structure and animal movements (Most common use)
Carbon sources in soils
Carbon sources for ecosystem respiration
Pollutant sourcing
Sediment fingerprinting
Originally developed for ecological studies; however, it is a basic mixing model and can be used with any fingerprint property as long as the assumptions of linear mixing are met
Used to estimate the relative contribution of the potential sediment sources at each sediment sampling site
Bayesian isotopic mixing model **because it uses the underlying probability of distribution of each statistic rather than the fixed sample means or point estimates which other mixing models use.
Because Bayesian uses variance instead of means and standard deviations the output will always differ if run multiple times
Written in open source languages R and Jags (Just Another Gibbs Sampler)
Created from the creators of MixSIR and SIAR
Sediment grouped by sampling site and year
Monte Carlo methods producing simulations of possible values of sediment source proportions using prior distributions consistent with data
Monte Carlo specifies the number of iterations to run the model at, the longer it runs the more accurate it is.
The data, data specification prior information/distributions, iterations can all be easily imputed using the easy to use GUI (graphical user interface)
Current models (MixSIAR) have covariates, hierarchical variance structures, concentration dependencies, discrimination factors and include complex variability in isotopic signatures allowing them to be more accurate then previous models.
Nuclear weapon testing in the late 1950’s
Chernobyl 1986 (Northern hemisphere only)