Geoderma 239–240 (2015) 25–33

Contents lists available at ScienceDirect

Geoderma
journal homepage: www.elsevier.com/locate/geoderma

Apportionment of suspended sediment sources in an agricultural

watershed using sediment fingerprinting
Jasmeet Lamba ⁎, K.G. Karthikeyan, A.M. Thompson
Department of Biological Systems Engineering, University of Wisconsin–Madison, Madison, WI 53706, USA

a r t i c l e i n f o a b s t r a c t

Article history: Sediment fingerprinting techniques can provide valuable information on sources of suspended sediment to facil-
Received 30 June 2014 itate effective targeting of best management practices. Research was conducted in the Pleasant Valley watershed
Received in revised form 25 September 2014 (50 km2) in South Central Wisconsin to identify sources of suspended sediment during cropping season and
Accepted 28 September 2014
snowmelt periods at a subwatershed scale. Results show that both stream banks and agriculture are important
Available online 7 October 2014
sources of suspended sediment. The contribution from agriculture and stream banks to suspended sediment at
Keywords:
the watershed outlet ranged from 45 to 97% and from 3 to 47%, respectively. During periods of high sediment
Sediment sources loading agriculture was the important source of suspended sediment at the majority of sites within this water-
Phosphorus shed, with the exception of snowmelt runoff, when stream banks were the dominant source of suspended sed-
Sediment fingerprinting iment at the watershed outlet. The average annual erosion rates in croplands and pasture land-use determined
Suspended sediment from the Revised Universal Soil Loss Equation 2 (RUSLE 2) ranged from 0 to 0.00509 t m−2 yr−1, indicating sig-
Soil erosion nificant variability among fields. Conservation practices in this watershed should be targeted to stream banks or
RUSLE 2 agriculture (prioritize fields based on RUSLE 2 average annual erosion rates) depending on the dominant source
of suspended sediment within a subwatershed. The results of this study show that both temporal and spatial var-
iability in suspended sediment sources should be considered to develop management strategies and sampling
only at the watershed outlet or capturing a few storms might not be sufficient to target locations for best man-
agement practices.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction
Eroded sediments are important non-point source pollutants caus-
ing degradation of surface water bodies. Loss of sediment-bound nutri-
ents, such as phosphorus (P), from agricultural landscapes to surface
waters results in growth of toxic algal blooms and eutrophication.
To control excessive sediment delivery to streams and implement ap-
propriate management practices effectively, sources contributing dis-
proportionate amount of suspended sediment to streams must be
identified.
Sediment fingerprinting technique has been successfully used in the
past to identify sources of suspended sediment (Davis and Fox, 2009;
Koiter et al., 2013a; Walling, 2013). This technique is based on linking
the physical or geochemical properties of the suspended sediment to
their corresponding sources within the watershed (Walling et al.,
2008) and thereby quantifying relative contribution from different
sources (Walling, 2013). Different types of sediment fingerprinting
properties that have been used to identify suspended sediment sources
⁎ Corresponding author at: USDA-ARS, Pasture Systems & Watershed Management
Research Unit, Building 3702 Curtin Road, University Park, PA 16802, USA.
E-mail address: jul59@psu.edu (J. Lamba).
include, but are not limited to, metals (e.g., Blake et al., 2012; Franz
et al., 2014), fallout radionuclides (e.g., Huisman et al., 2013; Olley
et al., 2013), stable isotopes (Mckinley et al., 2013), sediment color
(Martínez-Carreras et al., 2010), and mineral magnetic properties
(Walling et al., 1999). Sediment fingerprinting properties could be
unique for a particular watershed (Walling, 2013). Therefore, a com-
mon approach used in sediment fingerprinting studies is to select a
large number of fingerprinting properties and apply statistical proce-
dures to optimize them to best apportion suspended sediment to differ-
ent potential sources (Davis and Fox, 2009; Walling, 2013).
Sources, sinks and fluxes of sediment are highly variable in time
and space (Trimble, 1999). The contribution from different sources
to suspended sediment can change during crop growing season
(Huisman et al., 2013). For example, in spring bare soils are more
prone to erosion and susceptibility of soil movement in upland areas de-
creases as crop growing season progresses. Stream banks are more
prone to soil erosion during freeze–thaw activity (Gellis and Noe,
2013). Similarly, sediment loading at the watershed outlet changes
throughout a year. Therefore, identification of suspended sediment
sources during periods of high sediment load would help prioritize the
contributing source areas. In addition, there is a spatial dependence on
the relative contribution from various sources within a watershed
(Koiter et al., 2013b). Therefore, focusing efforts only at the watershed

http://dx.doi.org/10.1016/j.geoderma.2014.09.024
0016-7061/© 2014 Elsevier B.V. All rights reserved.
26 J. Lamba et al. / Geoderma 239–240 (2015) 25–33
outlet can result in ineffective targeting of areas for best management
practice (BMP) implementation.
Sediment fingerprinting technique using inorganic tracers can be
used to prioritize upland sediment sources based on land-use types
(e.g., croplands, pastures, woodlands), soil type, and geology. Manage-
ment practices are typically implemented at the field-scale. In a water-
shed there could be several hundred to thousand fields under a
particular land-use type. Therefore, an approach which can be used to
prioritize fields within agricultural land-use will help in effective
targeting of BMPs. A commonly used tool for field-level prioritization,
using average annual soil erosion rate, is the Revised Universal Soil
Loss Equation 2, RUSLE 2 (USDA-ARS, 2006). RUSLE 2 is readily available
and fairly easy to use in comparison to other complex models such as
Water Erosion Prediction Project (WEPP) (Flanagan and Livingston,
1995) and it is part of the P indices (Buczko and Kuchenbuch, 2007),
such as, the Wisconsin P Index (Good et al., 2012). Therefore, combined
use of RUSLE 2 with sediment fingerprinting could be effective in iden-
tifying specific fields for BMPs.
A study which considers both spatial and temporal aspects of
suspended sediment transport, provides seasonal information on
suspended sediment sources, and prioritizes fields based on average an-
nual soil erosion rate is needed to obtain a comprehensive understand-
ing of sediment transport in an agricultural watershed and to effectively
target BMPs. Past efforts have focused on identifying sources of
suspended sediment temporally (e.g., Gellis and Noe, 2013; Huisman
et al., 2013) and at a subwatershed scale (e.g., Collins et al., 2013;
Koiter et al., 2013b; Smith and Blake, 2014). However, limited work
has been done to identify suspended sediment sources as a function of
land-use during a growing season using sediment fingerprinting in con-
junction with prioritization of fields for BMPs based on average annual
soil erosion rates. Therefore, the objectives of this study were to:
(a) determine relative contributions from different potential sources
to fine suspended sediment at the subwatershed scale, (b) evaluate
how these contributions from different sources change temporally at
(a)
Site 1
Site 2
Site 4 Site 3
Fig. 1. (a) Location of the upland, stream bank and suspended sediment collection sites in the Pleasant Valley watershed. (b) Land-use distribution in the Pleasant Valley watershed.
the subwatershed scale, and (c) prioritize fields for targeting BMPs
within agricultural land-use category.
2. Methods
2.1. Study site
The 50 km2 Pleasant Valley watershed is located in the non-glaciated
area of South Central Wisconsin (Fig. 1a). The major land-uses in the
watershed are cropland (34%), Conservation Reserve Program (CRP)
grassland (unmanaged) (28%), woodland (22%), pasture (generally
managed and grazed) (8%), and pastured woodland (4%) (Fig. 1b).
The watershed has an average slope of 11% and consists primarily of
silt loam soils. The average (October, 2006–September, 2012) flow
and sediment load measured at the watershed outlet is 0.59 m3 s− 1
and 27.8 t km−2 yr−1, respectively. The Pleasant Valley branch of this
watershed is on the list of Wisconsin Impaired Waters due to degraded
habitat from non-point source pollution contributions to sediment/total
suspended solids (DNR, 2012).
2.2. Sample collection
Suspended sediment samples were collected at the outlet of three
different subwatersheds and at the watershed outlet using time-
integrated trap samplers (Fig. 1a) (Phillips et al., 2000). These time-
integrated trap samplers collect fine suspended sediment (b63 μm)
with particle size characteristics that are statistically representative of
the ambient suspended sediment (Phillips et al., 2000; Russell et al.,
2000) and have been successfully used in numerous previous studies
(e.g., Huisman et al., 2013; Smith and Blake, 2014). At each site
(i.e., monitored sub-watershed or overall watershed outlet), four
time-integrated trap samplers were installed to ensure that sufficient
sediment mass was collected for subsequent analyses. Suspended sedi-
ment samples were collected monthly (± 5 days) during the entire
(b)
 J. Lamba et al. / Geoderma 239–240 (2015) 25–33 27

study period. Exceptions included: (a) the sampling period of 1st Sep-
tember, 2011–14th October, 2011 when the samplers were retrieved
after a longer time period to ensure that sufficient sediment mass was
collected for analysis (during the previous sampling period, i.e., 28th
July, 2011–1st September, 2011, only a few sites had sufficient sediment
mass for analysis and, therefore, the samplers were deployed for a lon-
ger time during the following sample collection period), and (b) during
the sampling period of 6th February–16th March, 2011, when the entire
snowmelt period had to be sampled. The drainage area and dominant
land-use percentage of each subwatershed and range of stream bank
heights and channel bankfull widths at each suspended sediment col-
lection site are included in Table 1.
Upland soil samples (top 2.5 cm) from cropland, pasture, and wood-
land (ungrazed and grazed (pastured woodland; Fig. 1b)) were collect-
ed in a 20 m × 20 m grid with 5 m spacing and composited for analysis
at 28 different locations within the watershed (Fig. 1a). Some fields
under pasture in this watershed have been converted to croplands (per-
sonal communication Dane County Land Conservation Division, Dane
County, Wisconsin). Therefore, cropland and pasture were considered
as a single source, herein referred to as agriculture. This approach of
combining pasture and cropland into one land-use category was
employed recently by Gellis and Noe (2013). Stream bank samples
were collected from eroding banks along 150 m reaches at 24 different
locations as part of a watershed geomorphic assessment in 2009 by our
project collaborator (Faith Fitzpatrick, USGS, Wisconsin Water Science
Center). Composite sediment cores (2.5 cm deep into the eroding face
of the stream bank) within a sampling reach were collected from 3 to
6 points along the eroding bank surface (top to bottom) with a trowel.
Composite samples for each eroding bank were combined within a sam-
pling reach if there was more than one eroding bank. It should be noted
that each subwatershed was sampled independently to provide repre-
sentative source (upland and stream bank) samples for that particular
subwatershed. Apportionment of suspended sediment sources at each
site was done by considering source samples within the contributing
subwatershed upstream of the suspended sediment collection site.
2.3. Laboratory analysis
All soil and sediment samples collected were oven dried at 40 °C and
disaggregated using pestle and mortar, and dry sieved through a 63 μm
sieve. The samples were analyzed for several inorganic elements includ-
ing Li, Be, Na, Al, P, S, K, Ca, Sc, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Ga, As, Rb,
Sr, Y, Nb, Mo, Cd, Sn, Sb, Cs, Ba, La, Ce, Ti, Pb, Bi, Th, and U using ICP-MS
microwave assisted acid total digestion in Teflon bombs based on U.S.
Environmental Protection Agency (USEPA) Method 3052 (EPA, 1996)
at the Wisconsin State Laboratory of Hygiene, Madison, Wisconsin,
USA. The soil/suspended sediment samples were analyzed for total
P (TP) using concentrated sulfuric and nitric acid digestion in our
laboratory (Eaton et al., 2005). After acid digestions, a flow injection
analysis (FIA) autoanalyzer (Quik Chem Series 8000 FIA Systems,
Lachat Instruments, Loveland, CO, USA) was used to quantify ortho-
phosphate at 880 nm using the molybdate-based colorimetric method
(Murphy and Riley, 1962). The dispersed (using 50 g L− 1 sodium
hexametaphosphate) particle size distribution (assuming spherical par-
ticles) of the samples less than 63 μm was obtained using a Malvern
Mastersizer 2000 (Malvern Instruments, Worcestershire, UK). The
particle size distribution was used to determine the specific surface
area of the sediment particles.
2.4. Selection of optimum fingerprinting properties
The sediment fingerprinting properties must possess conservative
behavior during sediment transport within a watershed. Therefore, a
mass-conservative test was used to determine if the concentrations of
fingerprinting properties in the suspended sediment samples fall within
the range of the source samples' fingerprinting property concentrations
at each site (Franz et al., 2014). The fingerprinting properties passing
the mass-conservative test were subjected to a two-step statistical se-
lection procedure proposed by Collins et al. (1997). In step I, the non-
parametric Kruskal–Wallis H-test was used to test the ability of each fin-
gerprinting property to discriminate between agriculture, woodland,
and stream bank. In step II, a stepwise discriminant function analysis
(DFA) based on minimization of Wilks' lambda was applied to the fin-
gerprinting properties that passed Kruskal–Wallis H-test in step 1 to
identify optimum set of fingerprinting properties that provide optimum
discrimination between suspended sediment sources (Walling et al.,
1999).
2.5. Source apportionment of suspended sediment sources
A multivariate mixing model was used to estimate relative contribu-
tions from agriculture, woodlands, and stream banks to suspended sed-
iment (Collins et al., 2010). Relative source contributions from each
source to suspended sediment were determined by minimizing the ob-
jective function (i.e., sum of squares of the weighted relative errors):
Xm
( )2
X
n C ssi − ð PS Z Þ
s¼1 s si s
Wi ð1Þ
i¼1
C ssi
where n is the number of fingerprinting properties; m is the number of
sediment sources; Cssi is the concentration of fingerprinting property (i)
in the suspended sediment sample; Ps is the relative percentage contri-
bution from the source group s; Ssi is the mean concentration of finger-
printing property (i) in the source group s; Zs is the particle size
correction factor (i.e., ratio of the specific surface area of the suspended
sediment to the specific surface area of the source sediment) in the
source group s; and Wi is the tracer discriminatory weighting factor.
The multivariate mixing model must satisfy two linear boundary
conditions to ensure that relative source contributions from each source
are non-negative (Eq. 2) and contributions from the different sources to
suspended sediment sum to unity (Eq. 3).
0≤P s ≤1 ð2Þ
X
m
P s ¼ 1: ð3Þ
s¼1
The particle size exerts an important influence on concentrations of
elements in the sediment (Horowitz and Elrick, 1987). Therefore Zs was
used in the mixing model to account for differences in particle size be-
tween source material sediment and suspended sediment. Linear

Table 1
Land-use distribution (%), drainage area (km2) and range of stream bank heights and channel bankfull width (m) at each site.

Location Dominant land-use (%) Drainage area (km2) Channel dimensions

Cropland Grassland Pasture Pasture woodland Woodland Stream bank height (m) Bankfull width (m)

Site 1 39 23 3 1 28 10.23 0.4–0.6 2.4–2.8

Site 2 54 17 4 1 17 7.87 0.3–1.0 1.6–2.5
Site 3 25 32 12 5 20 25.10 0.6–2.8 4.0–5.2
Site 4 34 28 8 4 22 49.99 0.7–2.0 4.9–8.2
28 J. Lamba et al. / Geoderma 239–240 (2015) 25–33

particle size correction factor was used in this study, since specific sur- Table 2
face area of source and suspended sediment samples was less than Fingerprinting properties that satisfied the mass-conservative and the Kruskal–Wallis H-
test (p = 0.1) criteria at each site.
1 m2 g− 1; however, when specific surface area is greater than
1 m2 g−1 linear particle size might not be appropriate (Russell et al., Fingerprinting properties
2001; Smith and Blake, 2014). Correction factor for organic matter Mass-conservative test
was not used in this study, since recent work has shown that combined Site 1 Al As Ba Be Bi Cd Ce Co Cs Fe Ga K
use of organic matter and specific surface area correction factors may La Mo Ni Pb Rb Sc Sn Sr Th V Y
Site 2 Al Ba Bi Ca Cd Ce Co Cr Fe Ga K La
cause overcorrection of the model and produce spurious results
Mn Mo Ni P Rb S Sc Sn Sr Tl Y Zn
(Koiter et al., 2013a; Smith and Blake, 2014). The tracer discriminatory Site 3 Al Bi Cd Ce Cs Cu Fe Ga La Mn Ni P
weighting factor (Wi) was used in the mixing model to ensure that Rb S Sc Sn Sr Th Y Zn
the fingerprinting property with the greatest relative discriminatory ef- Site 4 Al Ba Bi Ca Cd Ce Co Cr Cs Cu Fe Ga
ficiency in a given composite signature exerts the greatest influence on La Mn Nb Ni P Rb S Sc Sn Sr Th Ti
Y Zn
the solutions of the mixing model (Collins et al., 2010).
Kruskal–Wallis H-test (p = 0.1)
2.6. Average annual erosion rate Site 1 Al Ba Be Ce Co Fe Ga K La Mo Ni Pb
Rb Sr Th Y
Site 2 Ba Bi Cd Mn Mo Sr
RUSLE Version 2 (RUSLE 2; USDA-ARS, 2006) was used to calculate Site 3 Al Bi Ce Cs Cu Ga La Rb S Sc Sr Sn
average annual erosion rates (2006–2012) under management opera- Th Zn Y
tions followed in fields based on information obtained from the Dane Site 4 Al Ba Ca Cd Ce Co Cs Cu Fe Ga La Ni
Rb S Sc Sn Sr Th Y Zn
County Land Conservation Division, Dane County, Wisconsin. Estima-
tion of erosion rates was done as part of the determination of the Wis-
consin Phosphorus Index (Good et al., 2012) values for individual
fields in this watershed by our project collaborator (Dr. Laura Good, De- within each subwatershed and water chemistry, could impact the ele-
partment of Soil Science, University of Wisconsin–Madison, USA). mental fingerprinting property concentrations (Walling, 2013, Davis
and Fox, 2009, Koiter et al., 2013a).
2.7. Goodness-of-fit and uncertainty analysis

The goodness-of-fit of the optimized mixing model was evaluated
by comparing the actual fingerprinting property concentration in the
suspended sediment samples with the values predicted by the opti-
mized mixing model. The average relative mean error (RME) for each
site was determined by taking the average of the relative mean error
of all fingerprinting properties within the composite signature (Collins
et al., 1998). The extent to which the average fingerprinting property
concentration for each source category used in the mixing model re-
flects the true value is an important source of uncertainty in the results
produced by the mixing model. A Monte Carlo approach was used to as-
sess this uncertainty. In this approach, one sample was randomly re-
moved for each of the three source categories and the mixing model
was run without these samples (Gellis and Noe, 2013). This procedure
involved 1000 iterations and the mean relative source contributions
were determined for each site.
3. Results and discussion
3.1. Optimum fingerprinting properties
The fingerprinting properties that successfully passed the mass-con-
servative test and Kruskal–Wallis H-test (p = 0.1) for each site are in-
cluded in Table 2. Results showed that of the 37 fingerprinting
properties selected initially, the majority of the fingerprinting proper-
ties passed the mass-conservative test and Kruskal–Wallis H-test. The
optimum set of fingerprinting properties selected by the stepwise DFA
classified ≥ 85% of the sources correctly at each site, indicating that
these selected inorganic elements provided strong discrimination
among sources (Table 3). The concentration of elements was generally
greater in stream banks than those measured in surface soils (agricul-
ture and woodlands) (Supplementary Table 1). Since stream banks are
composed of less weathered sub-surface material, concentrations of el-
ements in the bank sediments are generally greater than those in sur-
face soils (Smith and Blake, 2014). The combination and number of
optimum fingerprinting properties selected varied among different
sites. This variability in the number and combination of fingerprinting
properties among different subwatersheds has been reported else-
where (e.g., Collins et al., 2010; Collins et al., 2013; Smith and Blake;
2014). Different factors, such as the number of source samples collected
3.2. Suspended sediment sources
The sources of suspended sediment varied among different sites. At
site 1, both stream banks and agriculture were important sources of
suspended sediment (Fig. 2a), with the contributions from stream
banks and agriculture ranging from 37 to 70% and from 30 to 63%, re-
spectively. The riparian land-use in site 1 subwatershed is dominated
by woodlands (ungrazed) and CRP grasslands, which could prevent
eroded upland sediment from reaching the streams, thereby resulting
in greater relative contributions from stream banks. Cooper et al.
(1987) reported that woodlands provided an effective buffer to trap
sediment in the Cyprus Creek and Panther Swamp watersheds in
North Carolina. Similarly, CRP grasslands can trap sediment originating
from agricultural sources from reaching the streams.
At site 2, generally agriculture was the dominant contributor to
suspended sediment (Fig. 2b), with its contribution ranging from 43
to 99%. The land-use under agriculture within this subwatershed is
58%, which could cause excessive sediment delivery from agricultural
sources to streams. The greater percentage of land-use under agricul-
ture within the site 2 subwatershed in comparison with the site 1
subwatershed could have resulted in the observed trends in source con-
tributions. Contributions from woodlands to suspended sediment
ranged from 0 to 57% at this site and woodlands were not the dominant
source of suspended sediment for most of the study period. The grazed
woodlands (i.e., pastured woodlands in the land-use map; Fig. 1b) are in
close proximity to the suspended sediment collection site and riparian
land-use in the tributary (closest to suspended sediment collection
site) of the main channel is grazed woodlands (Fig. 1b). Consequently,
grazed woodlands (more susceptible for soil losses) in site 2 were
more active suspended sediment contributors than ungrazed wood-
lands present in the site 1 subwatershed.
At site 3, relative contributions from agriculture were greater than
stream banks for most of the sampling period (Fig. 2c). The contribu-
tions from agriculture and stream banks to suspended sediment ranged
from 15 to 100% and from 0 to 85%, respectively. The riparian land-use
within this subwatershed includes grazed pasture (personal communi-
cation, Dane County Land Conservation Division, Dane County, Wiscon-
sin). The grazing of pasture results in removal of vegetation exposing
bare soil that can result in more surface runoff and soil erosion, thereby
 J. Lamba et al. / Geoderma 239–240 (2015) 25–33 29

Table 3
Results of the stepwise DFA analysis at each site.

Site Fingerprinting property Wilks' lambda Percentage source samples Cumulative percentage source Tracer discriminatory
classified correctly samples classified correctly weighting

Site 1 Ni 0.0730 88.9 88.9 1.3

La 0.0183 77.8 100 1.2
Mo 0.0100 66.7 100 1.0
Sr 0.0023 77.8 100 1.2
Al 0.0003 77.8 100 1.2
Site 2 Mn 0.2004 90.0 90.0 1.1
Sr 0.0495 80.0 90.0 1.0
Bi 0.0155 80.0 90.0 1.0
Cd 0.0023 80.0 100.0 1.0
Site 3 Al 0.4124 77.8 77.8 1.3
Ce 0.3123 63.0 70.4 1.1
La 0.2510 70.4 77.8 1.2
Sc 0.1994 59.3 85.2 1.0
Site 4 La 0.3638 69.2 69.2 1.6
Sc 0.2793 63.5 76.9 1.5
Ga 0.1998 42.3 82.7 1.0
Fe 0.1518 50.0 86.5 1.2
Cd 0.1308 50.0 88.5 1.2

resulting in greater contributions from agriculture. Unrestricted cattle
access to streams (personal communication Dane County Land Conser-
vation Division) in this subwatershed could have caused greater stream
bank erosion. For example, milking cows have access to streams for
about 75–80% of the time, cows that are not milked and beef cows
have access to streams 100% of the time (personal communication Dr.
Laura Good Department of Soil Science, University of Wisconsin–Madi-
son, USA). Trimble (1994) reported that grazing of stream banks can re-
sult in greater erosion from stream banks in comparison to ungrazed
banks. Similar trends were reported by Kauffman et al. (1983) along
the Catherine Creek in North-eastern Oregon. In addition, only 37% of
the subwatershed area is under agriculture which can result in greater
relative contributions from stream banks to suspended sediment.
At site 4 (watershed outlet), agriculture was the dominant source of
suspended sediment during the entire monitoring period with only
minor contributions from stream banks (Fig. 2d). The maximum relative
contributions from stream banks at the watershed outlet occurred
during snowmelt runoff (2 February–15 March, 2012), which indicates
that stream bank erosion due to freeze–thaw processes could be
important in this watershed. Gellis and Noe (2013) also reported that

(a) 100
(b) 100
Agriculture Agriculture
Woodlands 90 Woodlands
Relative Source Contributions (%)

Stream Banks Stream Banks

Relative Source Contributions (%)

80 80

70

60 60

50

40 40

30

20 20

10

0 0
g

ep

rch
e

rch
e

ep

e
ril

ril
t
y

y
uly

uly

uly

y
t
uly

uly

uly
Oc

Oc

Oc

Oc
Ma

Au

Ma

Se

Ma
un

un

Ma

Au

Ma
un

Ma
un

un

Se

un
Ap

Ap
7S
7S

5J
5J

8J

3J

8J

3J
Ma

Ma
7J

0J

2J
7J

0J

2J

15

-14
15

-14

-16

-1
-16

-1

18

26

14
18

26

14

-13
-13

g-1
e-1

e-2

e-1
g-1
e-1

e-2

e-1

y-1

y-3

y-1
y-1

y-3

y-1

July
July

-16
-16

ril-

p-

ril-

ril-
ep
ril-

p-

ril-

ep

ril-

July
July

rch
rch

Au
Jun

Jun

Jun
Au
Jun

Jun

Jun

Se
Se

Ma

Ma

Ma
Ma

Ma

Ma

Ap

Ap

Ap
Ap

Ap

Ap

1S
1S

eb
eb

28
28

Ma
Ma

15

16

2F
15

16

2F

17
17

17

30

12
17

30

12

13

23

13
13

23

13

18

26

14
18

26

14

16
16

2010 2011 2012 2010 2011 2012

Date Date

(c) 100
(d) 100
Agriculture
Agriculture Woodlands
Relative Source Contributions (%)

Woodlands Stream Banks

Relative Source Contributions (%)

80 Stream Banks 80

60 60

40 40

20 20

0 0
June

June

June
Aug

Sep

ril
July
May

July
y

July
May
g

ep

rch
e

e
ril
y

y
t
uly

uly

uly

5 Oc

Oc
Oc

Oc

-1 Se
6 Ma
Ma

Au

Ma

Ma
un

un

Se

un

Marc
Ap

3 Ap
7S
5J

8J

3J
Ma
7J

0J

2J
15

-14
-16

-1

p-14
18

26

14

ne-15

ne-28

ne-13
ly-16

ug-17
-13

4
g-1

ay-17

ay-30

ay-12
e-1

e-2

e-1

ep- 1
y-1

y-3

y-1
July

pril-1

pril-2

pril-1
-16

arch-1
ly
ril-

p-

ril-

ril-
ep

b-16
July

rch
Au
Jun

Jun

Jun

28 Ju
Se

1 Se
Ma

Ma

Ma
Ap

Ap

Ap
1S

eb

15 Ju
28

17 Ju

30 Ju

12 Ju
Ma

16 A

17 S
18 M

26 M

14 M
13 A

23 A

13 A
15

16

2F

2 Fe
17
17

30

12
13

23

13
18

26

14

16 M
16

2010 2011 2012

2010 2011 2012
Date Date

Fig. 2. Relative contribution (%) to suspended sediment from agriculture, woodland, and stream bank: (a) site 1, (b) site 2, (c) site 3, and (d) site 4.
30 J. Lamba et al. / Geoderma 239–240 (2015) 25–33

contributions from stream banks were greater when freeze–thaw activ-

ity predominated in the agricultural and forested Linganore Creek
watershed.

3.3. Temporal variability in suspended sediment sources

The sources of suspended sediment varied temporally during the

growing season at all sites (Fig. 2a, b, c, and d). With the exception of
site 1, generally agriculture was the dominant contributor to suspended
sediment during the early part of the crop growing season (April–June)
for all the three monitoring periods (2010, 2011, and 2012). During the Site 1
early part of the crop growing season, soils are bare, which make the up-
land areas more susceptible to soil erosion. The reduction in relative
Site 2
contributions from agriculture with progression in the growing season
(as crop cover increases) resulted in a switch in the dominant contribu-
tor to stream banks during the later part of summer and early fall
(August–October). Similarly, Huisman et al. (2013) reported that chan-
Site 4 Site 3
nel sources (stream banks or stream bed sediments) were important
during later months (July–October) and upland sources were more im-
portant during early months (April–June) of a growing season in the
North Fork of the Pheasant Branch watershed in Wisconsin. The tempo-
ral variability in suspended sediment sources highlights that sampling
activities covering only a few storm events or part of a growing season
could result in incomplete assessment of the relative importance of
suspended sediment sources.
Suspended sediment loading at the watershed outlet varied
throughout the study period (Fig. 3). The loading of suspended sedi-
ment was greater during the early part of the growing season and de-
creased as the growing season progressed. During the periods of high
sediment loading at the watershed outlet, agriculture was the dominant
contributor to suspended sediment. The only exception was for snow-
melt runoff (2 February–16 March, 2012) when stream banks were
the dominant contributors. The load-weighted relative contributions, Fig. 4. RUSLE 2 average annual soil erosion rate (t m−2 yr−1) in croplands and pastures for
all the study sites.
calculated based on method provided by Walling (2005), from agricul-
ture, stream banks, and woodlands to suspended sediment at the
watershed outlet for the entire sampling period were 67%, 32%, in conjunction with the sediment loading. A dominant source of
and 1%, respectively, indicating that overall agriculture was the suspended sediment during a particular time-period might be of little
dominant source of suspended sediment. To target BMPs efficiently, in- importance if the sediment load transported by the stream during that
formation on suspended sediment sources should be considered period is very low.

200

180
Suspended Sediment Load
Suspended Sediment Load (tons)

160

140

120

100

80

60

40

20

0
e

ug

ep

ne
ril
t

y
ay

ct

y
uly

uly

uly
Oc

Ma
un

un

Se

arc

Ap
O

u
M
8M

6A

7S
5J

8J

3J
7J

0J

2J
15

-14

M
y-1
26

14
-13
y-1

g-1
e-1

e-2

e-1
l-1

y-1

y-3

y-1
-16
p-

ril-

ril-
ep
Jul

rch
pri

Au
Jun

Jun

Jun
Jul

Se
Ma

Ma

Ma
Ap

Ap
1S

eb
A

28

Ma
15

16

2F
17
17

30

12
13

23

13
18

14
26

16

2010 2011 2012

Date

Fig. 3. Temporal distribution in suspended sediment loading (tons) at site 4 during the three monitoring periods (2010, 2011, and 2012).
 J. Lamba et al. / Geoderma 239–240 (2015) 25–33 31

3.4. Fields with higher erosion rates within agricultural land-use Table 4
Relative mean error between actual and optimized model predicted suspended sediment
fingerprinting property concentration.
Average annual soil erosion rates over four years (2010–2013),
which include the suspended sediment sample collection period Year Sample collection period Relative mean error (%)
(2010, 2011, 2012) for sediment fingerprinting ranged from 0 to Site 1 Site 2 Site 3 Site 4
0.00509 t m−2 yr−1 in croplands and pastures (Fig. 4). The results indi-
2010 13th April–18th May 17 3 21 14
cate that there is a wide range in the average soil erosion rates. Soil loss 2010 18th May–17th June 16 5 18 13
tolerance values (T-value) for soils in this watershed range from 2010 17th June–15th July 11 6 11 16
0.00049 t m−2 yr−1 to 0.00124 t m−2 yr−1. Importantly, within each 2010 15th July–16th August 9 6 20 14
site, only ≤22% and b 5% of the area had average annual erosion rates 2010 16th August–17th September 14 16 5 17
2010 17th September–15th October 13 12 12 14
exceeding 0.00049 t m−2 yr−1 and 0.00124 t m−2 yr−1, respectively 2011 23rd April–26th May 14 9 17 15
(Fig. 5), indicating that BMPs should be targeted at fields with dispro- 2011 26th May–30th June 6 7 15 13
portionately high amounts of erosion per unit area. In addition to 2011 30th June–28th July 14 13 14 15
RUSLE 2 erosion rates, field-to-stream connectivity should also be 2011 28th July–1st September 21 17 14 27
2011 1st September–14th October 11 14 4 12
taken into consideration to target BMPs, since RUSLE 2 estimates aver-
2012 6th February–16th March 12 11 16 9
age annual soil erosion rates and sediment delivery information for 2012 16th March–13th April 12 11 20 12
edge-of-field and not sediment delivery to streams. It should be noted 2012 13th April–14th May 12 24 21 12
that if agriculture is the dominant source of suspended sediment within 2012 14th May–12th June 17 16 21 14
a subwatershed based on sediment fingerprinting results, that 2012 12th June–13th July 13 27 22 7

subwatershed might not necessarily contain fields with greater average

annual erosion rates compared to other subwatersheds. The relative
source contributions to suspended sediment determined from sediment
fingerprinting take into account contributions from stream banks. In-
creasing relative contributions from stream banks will result in decreas-
ing relative contributions from agriculture and vice versa. This method
of using sediment fingerprinting in conjunction with RUSLE 2 estimates
can facilitate more efficient selection of fields for BMP implementation.
3.5. Comparison between fine bed sediment and suspended sediment
sources
Fine bed sediment has been used as a surrogate for suspended sedi-
ment to identify relative contributions from different sources within a
watershed (Wilkinson et al., 2013). Our study results indicate that the
relative contributions from different sources to fine sediment deposited
on the stream bed and suspended sediment might not be always simi-
lar. For example, in a previous study in this watershed (Lamba, 2014),
the dominant source of bed sediment at the watershed outlet and at
site 3 was stream banks; however, for suspended sediment agriculture
was the dominant source at these two sites. Koiter et al. (2013b) also
70
reported that relative source contributions to bed and suspended sedi-
ments might not necessarily be the same. Suspended sediments are pri-
marily composed of finer sediment (b 63 μm) (Collins et al., 2001).
Sediment from eroded banks can be present as individual particles or
aggregates (e.g., stream bank mass failure) depending on the hydraulic
forces exerted during flow events (Wynn, 2006). Aggregates are more
likely to deposit in stream channels in comparison to individual fine
particles. Murgatroyd and Ternan (1983) found that a larger fraction
of eroded bank material gets deposited on the stream bed. Therefore,
identification of both suspended and bed sediment (if there is signifi-
cant deposition of fine sediment on the stream beds) sources is needed
to target BMPs effectively.
3.6. Goodness-of-fit and uncertainty analysis
The RME values determined to assess the goodness-of-fit of the
mixing model showed an overall good agreement between predicted
and estimated suspended sediment fingerprinting properties. For all
sites, RME values ranged from 3 to 27% and were typically less than
20% for the majority of the samples at all sites (Table 4), indicating
that the mixing model provided an acceptable prediction of the

60
Site 1
Site 2 1400
Percentage field area (%)

50 Site 3 Agriculture
1200 Stream Banks
Woodlands
Suspended Sediment
Total Phosphorus (mg kg )

40
-1

1000

30 800

600
20
400

10
200

0 0
9
25

24

Site 1 Site 2 Site 3 Site 4

04

07

09

12
00

01
00
00

00

00
.0

.0

Location
0.
0.

0.

0.
0

>0
1-
0-

7-

2-

7-
02

10
05

07
00

00

00

00

Fig. 6. Total phosphorus concentration (mg kg−1) in agriculture, stream bank, woodland,
0.

0.

0.

0.

-2 -1 and suspended sediment at site 1 (agriculture (n) = 3, stream bank = 3, woodland = 2,

Potential erosion rate (t m yr ) suspended sediment = 14), site 2 (agriculture (n) = 6, stream bank = 2, woodland = 2,
suspended sediment = 16), site 3 (agriculture (n) = 4, stream bank = 16, woodland = 2,
Fig. 5. RUSLE 2 potential erosion (t m−2 yr−1) rate and percentage of field area under the suspended sediment = 16), and at site 4 (agriculture (n) = 15, stream bank = 27, wood-
agriculture land-use within each subwatershed. land = 7, suspended sediment = 15). Each half bar represents one standard error.
32 J. Lamba et al. / Geoderma 239–240 (2015) 25–33
fingerprinting property concentrations in suspended sediment (Minella
et al., 2008). The differences in relative source contributions to
suspended sediment determined from the mixing model (using mean
fingerprinting property concentration) and the average of the Monte
Carlo simulation results were ≤9% at all sites for the entire sampling pe-
riod (Supplementary Table 3). The dominant sources of suspended sed-
iment determined from the mixing model (with mean source
fingerprinting property concentration) and from the Monte Carlo simu-
lations were similar, indicating that use of mean fingerprinting property
concentrations of the source samples in the mixing model was not a sig-
nificant source of uncertainty. It should be noted that in this study a rel-
atively simple method (from Gellis and Noe, 2013), compared to the
methods adopted by other studies (e.g., Collins et al., 2010), was used
for the uncertainty analysis. The access to private farms to collect source
samples was limited in this watershed, which constrained the number
of source samples that could be collected. The relatively small number
of source samples limited our ability to perform more rigorous uncer-
tainty analysis.
3.7. Implications for phosphorus management
Generally, TP levels in source samples from agriculture were greater
than those measured in stream bank samples (Fig. 6). At sites 1, 2, and 3
there was no significant (p N 0.1) difference in TP levels between source
samples collected from the agriculture and stream bank sites. However,
at the watershed outlet (site 4), TP levels in agriculture source samples
were significantly (p b 0.1) greater than those in stream banks. In
agriculture (sites 2, 3, 4) and stream bank (sites 1, 3, 4) source samples,
TP levels were significantly (p b 0.1) greater than those measured in
woodland source samples. In suspended sediment samples collected
over three crop growing periods (2010, 2011, 2012) TP levels at site 2
were significantly (p b 0.1) lower than those measured at site 3. At
site 3, stream banks and agriculture were the dominant sources of
suspended sediment for most of the sampling period. However at site
2, there were minor contributions from woodlands (lowest TP content
among the various source categories) in addition to agriculture. Conse-
quently, dilution effects due to contributions from woodlands could
have caused lower TP levels at site 2 than site 3. Subwatershed of site
3 had the highest tonnage of fine sediment deposited per unit area
on the stream bed (Lamba, 2014). Therefore, site 3 subwatershed
would be an excellent candidate to target fields for BMP installation.
Suspended sediment TP concentration was generally greater than
the TP concentration in source samples, indicating TP enrichment in
suspended sediment. This enrichment phenomenon is attributable to
greater specific surface area of suspended sediment samples (0.73 ±
0.01 m2 g−1) than the source samples (0.63 ± 0.02 m2 g−1). Addition-
ally, TP concentration in suspended sediment can also change depend-
ing on the soluble P concentration in flowing water (Jarvie et al.,
2005). The average TP concentration in suspended sediment at all
sites ranged from 1034 ± 55 to 1188 ± 29 mg kg−1. The TP concentra-
tions in suspended sediment during event based sampling (Lamba,
2014) at sites 1, 2, 3, and 4 ranged from 838 ± 6 to 1094 ±
141 mg kg−1 and were similar to TP concentrations in suspended sedi-
ment observed during this three year (2010, 2011, 2012) study period.
In contrast, TP concentration in suspended sediment in the North Fork
of the Pheasant Branch in Wisconsin ranged from 1357 ± 87 to
3427 ± 220 mg kg− 1 (Huisman et al., 2013). The land-use in the
North Fork watershed consisted of pasture or hay (42%) and croplands
(46%), which are the dominant sources of suspended sediment
(Huisman et al., 2013), whereas in our study watershed land under
cropland and pasture is only 42%. Therefore, less area in cropland and
pasture along with the contributions from stream bank (low TP level)
and minor contributions from woodlands (low TP level) could have re-
sulted in depleted TP concentration in suspended sediment in our study
watershed.
4. Conclusions
The results of this study show that sources of suspended sediment
can vary among different subwatersheds within a watershed. Agricul-
tural land-use was an important source of suspended sediment
early in the growing season (April–June) and relative contributions
from stream banks increased as the cropping season progressed
(i.e., August–October). Contributions from stream banks to suspended
sediment are important during snowmelt runoff sampled in the year
2012, indicating that stream banks could be an important source of
suspended sediment during winter months. During periods of high sed-
iment loading at the watershed outlet (with the exception of snowmelt
runoff), agriculture was the dominant source of suspended sediment. In
agricultural areas, BMPs (e.g., conservation tillage, strip cropping)
should be prioritized in fields with higher average annual erosion
rates. To minimize stream bank erosion, limiting cattle access (common
in this watershed) to streams and stream bank stabilization should be
recommended.
Future research should focus on quantifying stream bank erosion
rates during winter and snowmelt periods to better understand
sediment transport dynamics in this watershed. In addition, use of
watershed-scale modeling can help select BMPs based on their effec-
tiveness to reduce sediment and P loading to streams.
Acknowledgments
Funding for this project was provided by the USDA NIFA award
#2009-51130-06049, The Nature Conservancy, and the Wisconsin De-
partment of Natural Resources. We thank John Panuska, Laura Good,
Faith Fitzpatrick, and staff at Dane County Land Conservation Division,
Dane County, Wisconsin for their assistance in this study. We also
thank all the producers in this watershed for allowing soil sampling
from their lands.
Appendix A. Supplementary data
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.geoderma.2014.09.024.
References
Blake, W.H., Ficken, K.J., Taylor, P., Russell, M.A., Walling, D.E., 2012. Tracing crop-specific
sediment sources in agricultural catchments. Geomorphology 139, 322–329.
Buczko, U., Kuchenbuch, R.O., 2007. Phosphorus indices as risk-assessment tools in
the U.S.A. and Europe—a review. J. Plant Nutr. Soil Sci. 170 (4), 445–460.
Collins, A.L., Walling, D.E., Leeks, G.J.L., 1997. Source type ascription for fluvial suspended
sediment based on a quantitative composite fingerprinting technique. Catena 29 (1),
1–27.
Collins, A.L., Walling, D.E., Leeks, G.J.L., 1998. Use of composite fingerprints to determine
the provenance of the contemporary suspended sediment load transported by rivers.
Earth Surf. Process. Landf. 23 (1), 31–52.
Collins, A.L., Walling, D.E., Sichingabula, H.M., Leeks, G.J.L., 2001. Suspended sediment
source fingerprinting in a small tropical catchment and some management implica-
tions. Appl. Geogr. 21 (4), 387–412.
Collins, A.L., Walling, D.E., Webb, L., King, P., 2010. Apportioning catchment scale sedi-
ment sources using a modified composite fingerprinting technique incorporating
property weightings and prior information. Geoderma 155 (3–4), 249–261.
Collins, A.L., Zhang, Y.S., Duethmann, D., Walling, D.E., Black, K.S., 2013. Using a novel trac-
ing–tracking framework to source fine-grained sediment loss to watercourses at sub-
catchment scale. Hydrol. Process. 27 (6), 959–974.
Cooper, J.R., Gilliam, J.W., Daniels, R.B., Robarge, W.P., 1987. Riparian areas as filters for ag-
ricultural sediment. Soil Sci. Soc. Am. J. 51 (2), 416–420.
Davis, C.M., Fox, J.F., 2009. Sediment fingerprinting: review of the method and future im-
provements for allocating nonpoint source pollution. J. Environ. Eng. ASCE 135 (7),
490–504.
DNR, 2012. Purposed List of Impaired Watersavailable from Wisconsin Department of
Natural Resources, (http://dnr.wi.gov/water/impairedSearch.aspx (accessed 27
June. 2014)).
Eaton, A.D., Clesceri, L.S., Rice, E.W., Greenberg, A.E., 2005. Standard Methods for the Ex-
amination of Water and Wastewater, 21st edn. Port City Press, Baltimore, Maryland.
Flanagan, D.C., Livingston, S.J., 1995. WEPP user summary. NSERL Report No. 11. USDA-
ARS National Soil Erosion Research Laboratory, W. Lafayette, Ind.
 J. Lamba et al. / Geoderma 239–240 (2015) 25–33
Franz, C., Makeschin, F., Weiß, H., Lorz, C., 2014. Sediments in urban river basins:
identification of sediment sources within the Lago Paranoá catchment, Brasilia DF,
Brazil — using the fingerprint approach. Sci. Total Environ. 466–467, 513–523.
Gellis, A., Noe, G., 2013. Sediment source analysis in the Linganore Creek watershed,
Maryland, USA, using the sediment fingerprinting approach: 2008 to 2010. J. Soils
Sediments 13 (10), 1735–1753.
Good, L.W., Vadas, P., Panuska, J.C., Bonilla, C.A., Jokela, W.E., 2012. Testing the Wisconsin
Phosphorus Index with year-round, field-scale runoff monitoring. J. Environ. Qual. 41
(6), 1730–1740.
Horowitz, A.J., Elrick, K.A., 1987. The relation of stream sediment surface area, grain size
and composition to trace element chemistry. Appl. Geochem. 2 (4), 437–451.
Huisman, N.L.H., Karthikeyan, K.G., Lamba, J., Thompson, A.M., Peaslee, G., 2013. Quantifi-
cation of seasonal sediment and phosphorus transport dynamics in an agricultural
watershed using radiometric fingerprinting techniques. J. Soils Sediments 13 (10),
1724–1734.
Jarvie, H.P., Jürgens, M.D., Williams, R.J., Neal, C., Davies, J.J.L., Barrett, C., White, J., 2005.
Role of river bed sediments as sources and sinks of phosphorus across two major eu-
trophic UK river basins: the Hampshire Avon and Herefordshire Wye. J. Hydrol. 304
(1), 51–74.
Kauffman, J.B., Krueger, W.C., Vavra, M., 1983. Impacts of cattle on stream banks in North-
eastern Oregon. J. Range Manag. 36 (6), 683–685.
Koiter, A.J., Owens, P.N., Petticrew, E.L., Lobb, D.A., 2013a. The behavioural characteristics
of sediment properties and their implications for sediment fingerprinting as an ap-
proach for identifying sediment sources in river basins. Earth Sci. Rev. 125, 24–42.
Koiter, A., Lobb, D., Owens, P., Petticrew, E., Tiessen, K.D., Li, S., 2013b. Investigating the
role of connectivity and scale in assessing the sources of sediment in an agricultural
watershed in the Canadian prairies using sediment source fingerprinting. J. Soils Sed-
iments 13 (10), 1676–1691.
Lamba, J., 2014. Identification and assessment of critical source areas in an
agricultural watershed. Ph.D. Dissertation University of Wisconsin-Madison.
Martínez-Carreras, N., Krein, A., Gallart, F., Iffly, J.F., Pfister, L., Hoffmann, L., Owens, P.N.,
2010. Assessment of different colour parameters for discriminating potential
suspended sediment sources and provenance: a multi-scale study in Luxembourg.
Geomorphology 118 (1–2), 118–129.
McKinley, R., Radcliffe, D., Mukundan, R., 2013. A streamlined approach for sediment
source fingerprinting in a Southern Piedmont watershed, USA. J. Soils Sediments 13
(10), 1754–1769.
Minella, J.P.G., Walling, D.E., Merten, G.H., 2008. Combining sediment source tracing tech-
niques with traditional monitoring to assess the impact of improved land manage-
ment on catchment sediment yields. J. Hydrol. 348 (3–4), 546–563.
Murgatroyd, A.L., Ternan, J.L., 1983. The impact of afforestation on stream bank erosion
and channel form. Earth Surf. Process. Landf. 8 (4), 357–369.
33
Murphy, J., Riley, J.P., 1962. A modified single solution method for the determination of
phosphate in natural waters. Anal. Chim. Acta 27, 31–36.
Olley, J., Burton, J., Smolders, K., Pantus, F., Pietsch, T., 2013. The application of fallout ra-
dionuclides to determine the dominant erosion process in water supply catchments
of subtropical South-east Queensland, Australia. Hydrol. Process. 27 (6), 885–895.
Phillips, J.M., Russell, M.A., Walling, D.E., 2000. Time-integrated sampling of fluvial
suspended sediment: a simple methodology for small catchments. Hydrol. Process.
14 (14), 2589–2602.
Russell, M.A., Walling, D.E., Hodgkinson, R.A., 2000. Appraisal of a simple sampling device
for collecting time-integrated fluvial suspended sediment samples. The role of
erosion and sediment transport in nutrient and contaminant transfer. Proceedings
of a Symposium Held at Waterloo, Ontario, Canada in July 2000.
Russell, M.A., Walling, D.E., Hodgkinson, R.A., 2001. Suspended sediment sources in two
small lowland agricultural catchments in the UK. J. Hydrol. 252 (1–4), 1–24.
Smith, H.G., Blake, W.H., 2014. Sediment fingerprinting in agricultural catchments: a crit-
ical re-examination of source discrimination and data corrections. Geomorphology
204, 177–191.
Trimble, S.W., 1994. Erosional effects of cattle on stream banks in Tennessee, U.S.A. Earth
Surf. Process. Landf. 19 (5), 451–464.
Trimble, S.W., 1999. Decreased rates of alluvial sediment storage in the Coon Creek Basin,
Wisconsin, 1975–93. Science 285 (5431), 1244–1246.
USDA-ARS, 2006. RUSLE2 1.126.6.4. USDA-ARS, Washington, DC (http://fargo.nserl.
purdue.edu/rusle2_dataweb/RUSLE2_Index.htm (accessed 27 June. 2014)).
U.S. Environmental Protection Agency, EPA-Method 3052, 1996. Microwave Assisted Acid
Digestion of Siliceous and Organically Based Matrices. U.S. Government Printing
Office, Washington, DC.
Walling, D., 2013. The evolution of sediment source fingerprinting investigations in fluvial
systems. J. Soils Sediments 13 (10), 1658–1675.
Walling, D.E., 2005. Tracing suspended sediment sources in catchments and river sys-
tems. Sci. Total Environ. 344 (1–3), 159–184.
Walling, D.E., Collins, A.L., Stroud, R.W., 2008. Tracing suspended sediment and particulate
phosphorus sources in catchments. J. Hydrol. 350 (3–4), 274–289.
Walling, D.E., Owens, P.N., Leeks, G.J.L., 1999. Fingerprinting suspended sediment sources
in the catchment of the River Ouse, Yorkshire, UK. Hydrol. Process. 13 (7), 955–975.
Wilkinson, S.N., Hancock, G.J., Bartley, R., Hawdon, A.A., Keen, R.J., 2013. Using sediment
tracing to assess processes and spatial patterns of erosion in grazed rangelands,
Burdekin River basin, Australia. Agric. Ecosyst. Environ. 180, 90–102.
Wynn, T., 2006. Streambank retreat: a primer. Watershed Update, January-March 2006.
American Water Resources Association, Hydrology and Watershed Management
Technical Committee. 4 (1), 1–14.