SUSPENDED SEDIMENT FLUX DOWNSTREAM OF THE RETREATING CASTLE
CREEK GLACIER, CARIBOO MOUNTAINS, BRITISH COLUMBIA.

by

Michael Strathearn Leggat
B.Sc. (Freshwater Science), University o f British Columbia Okanagan, 2006

THESIS SUBMITTED IN PARTIAL FULFILLMENT OF
THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE
IN
NATURAL RESOURCES AND ENVIRONMENTAL STUDIES
(ENVIRONMENTAL SCIENCE)

THE UNIVERSITY OF NORTHERN BRITISH COLUMBIA
November, 2014

© Michael Leggat, 2014

UMI Number: 1526484

All rights reserved
INFORMATION TO ALL USERS
The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.

Di!ss0?t&iori Publishing
UMI 1526484
Published by ProQuest LLC 2015. Copyright in the Dissertation held by the Author.
Microform Edition © ProQuest LLC.
All rights reserved. This work is protected against
unauthorized copying under Title 17, United States Code.

ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, Ml 48106-1346

Abstract
Glaciers are a major erosive force that increase sediment load to the downstream
fluvial system. The Castle Creek Glacier, British Columbia has retreated ~1.0 km in the past
70 years. The dynamics o f suspended sediment concentration (SSC) and streamflow (Q) were
monitored independently at six sites within its proglacial zone over a 60 day period from July
to September, 2011. The time-series were divided into hydrologic days and the SSC response
to hydro-meteorological conditions was categorized using principal component analysis
(PCA) and cluster analysis (CA). Suspended sediment load (SSL) was computed and
summarized for the categories. During the 2011 study period, c. 60% of the total SSL was
derived from the glacial stream and sediment deposits proximal to the terminus o f the glacier;
during ‘storm’ events, that contribution drops to 40% as contribution from diffuse and point
sources o f sediment within the meltwater channel and proglacial zone increase. While
‘storm’ events accounted for just 3% o f the field season, SSL was 500% higher than the
seasonal average, and c. 20% of the total SSL was generated in that time.

Table of Contents
Abstract...................................................................................................................................................ii
Table of Contents..................................................................................................................................iii
List of Tables........................................................................................................................................ vi
List o f Figures.....................................................................................................................................viii
Acknowledgements...............................................................................................................................xi
1

Literature Review........................................................................................................................... 1
1.1

Introduction............................................................................................................................... 1

1.2 Research Objectives................................................................................................................. 4
1.3 Proglacial Z one.........................................................................................................................5
1.4 Hysteresis and Non-linear Relationships in Suspended Sediment Concentration and
Streamflow D ata......................................................................................................................7
1.5 Modeling Suspended Sediment Concentration.................................................................... 10
1.5.1 Modeling Suspended Sediment Yield over theParaglacial Period.................................12
1.5.2 Rationale for the Use of Turbidity Measurements to Determine Suspended
Sediment Concentration..................................................................................................... 15
1.6 Sediment Budgets...................................................................................................................16
1.6.1 Proglacial Sediment Budget Applications........................................................................17
1.7

2008 Study on Sediment Fluxes at Castle Creek G lacier.................................................. 22

2 Methods...........................................................................................................................................23
2.1

Study Area............................................................................................................................... 23

2.1.1 Field Site Observations and Site Selection...................................................................... 25
2.1.2 Catchment A rea................................................................................................................. 28
2.1.3 Longitudinal Stream Profile............................................................................................. 29
2.2 Monitoring and Sampling Strategy....................................................................................... 31
2.3 Instrumentation and Data Collection.................................................................................... 33
2.3.1 Meteorological D ata.......................................................................................................... 33
2.3.2 Water Level and Streamflow............................................................................................ 34
2.3.3 Turbidity and Suspended Sediment Concentration.........................................................36
2.4

Data Processing and Computations...................................................................................... 38

2.4.1 Time-series QA/QC........................................................................................................... 39
2.4.2 Water Level and Streamflow Rating Curves................................................................... 41
2.4.3 Turbidity and Suspended Sediment Concentration Regression..................................... 43

iv
2.5

Statistical Data Analysis........................................................................................................45

2.5.1 Meteorological Periods......................................................................................................48
2.5.2 Streamflow Driving Factor............................................................................................... 48
2.5.3 Suspended Sediment Response Shape............................................................................. 49
2.5.4 Suspended Sediment Response Magnitude..................................................................... 50
3

Results and Discussion (1) - Spatial and Temporal Patterns o f Suspended Sediment............ 51
3.1

D ataS ets.................................................................................................................................51

3.1.1 Field Season Overview......................................................................................................54
3.1.2 Streamflow Considerations............................................................................................... 57
3.1.3 Suspended Sediment Concentration Considerations...................................................... 59
3.1.4 Error and Uncertainty........................................................................................................62
3.2

Principal Component Analysis (PCA) and Cluster Analysis (CA).................................... 63

3.2.1 Meteorological Periods - Cluster Analysis..................................................................... 63
3.2.2 Streamflow Driving Factor - Principal Component Analysis........................................65
3.2.3 Suspended Sediment Response Shape - Principal Component Analysis and Cluster
Analysis..............................................................................................................................67
3.2.4 Suspended Sediment Response Magnitude —Cluster Analysis......................................69
3.3

Shape and Magnitude - Field Season Summary.................................................................. 71

3.3.1 Proximal Site 1 .................................................................................................................. 77
3.3.2 Proximal Site 2 .................................................................................................................. 78
3.3.3 Proximal Site 3 .................................................................................................................. 79
3.3.4 Middle Site..........................................................................................................................80
3.3.5 Rockback Peak Tributary..................................................................................................81
3.3.6 Distal Site............................................................................................................................83
4

5

Results and Discussion (2) - Suspended Sediment Load...........................................................84
4.1

Suspended Sediment Load.....................................................................................................84

4.2

Proglacial Suspended Sediment B udget.............................................................................. 96

Conclusions................................................................................................................................. 101
5.1

Summary............................................................................................................................... 101

5.2

Limitations............................................................................................................................ 104

5.3

Recommendations for Future Research..............................................................................107

6

Bibliography................................................................................................................................ 109

7

Appendix..................................................................................................................................... 118
7.1

Hydrometric..........................................................................................................................118

7.2

Suspended Sediment............................................................................................................ 121

V

7.3

Meteorological......................................................................................................................131

7.4 Suspended Sediment Load Summary - ‘Shape’, ‘Magnitude’ and HydroMeteorological Categories..................................................................................................132
7.5 Julian Day Calendar............................................................................................................. 133
7.6 Suspended Sediment Load Summary for the Small River Glacier.................................. 134

vi

List of Tables
Table 2.1 Catchment areas and percent glacial cover for 2011 proglacial stream sampling
sites.....................................................................................................................................29
Table 2.2 Turbidity (Tu) - suspended sediment concentration (SSC) regression equations for
the 2011 proglacial monitoring sites................................................................................44
Table 3.1 Paired turbidity (Tu) and suspended sediment concentration (SSC) sample
summary, and results of probability plot correlation coefficient (PPCC) fourth-spread
method null hypothesis test; ‘YES’ means the samples were drawn from a population
with a normal distribution.................................................................................................59
Table 3.2 Turbidity (Tu) data summary for the 2011 proglacial monitoring sites.................. 60
Table 3.3 Summary of meteorological data from upper and lower meteorological stations for
JD 195 - JD 254,2011. Four clusters o f similar meteorological data have been
assigned descriptive titles based on air temperature and precipitation.......................... 63
Table 3.4 Summary o f suspended sediment response magnitude parameters and cluster
analysis results. Values computed from daily data. Standard deviation is reported in

parentheses.........................................................................................................................71
Table 3.5 Summary of suspended sediment ‘shape’ and ‘magnitude’ analysis for 2011
proglacial monitoring locations, JD 195 - JD 254: DA Y S...........................................76
Table 4.1 Summary of average suspended sediment load (kg/day) for each sub-category.
Averages for ‘irregular’ response shape data are reported in brackets. The values
reported in the body o f the table are arithmetic means for the given category.
Weighted averages were used to account for the disproportionate number of days in
each category for the ‘shape’ and ‘magnitude’, and total summary. Table 3.5 reports
the the number of days in each category......................................................................... 85
Table 4.2 Field season summary statistics for meteorological periods determined through
principal component analysis. Streamflow is Q (m3/s), suspended sediment
concentration is SSC (mg/L), suspended sediment load is SSL (t/day), precipitation is
PT (daily total mm), and air temperature is AT (daily mean °C). Values in the table
have been computed from daily averages or totals of individual hydrologic days
(06:00 - 06:00) in the category........................................................................................ 87

Table 4.3 Percent (%) o f mean daily suspended sediment load (SSL) and streamflow (Q)
relative to the distal site (DS) during meteorological periods determined by principal
component analysis........................................................................................................... 88
Table 4.4 Meteorological summary of suspended sediment load (SSL) and streamflow (Q)
for each site. Values computed as a percentage of the seasonal total SSL (t/day) and
seasonal mean Q (m3/s). Raw values and the number o f days o f observation for each
site in each category are presented in Table 4.2............................................................. 92
Table 4.5 Suspended sediment yield (SSY) for the Castle Creek watershed during the 2011
field season. Mean daily SSY (t/km2/d) and Total SSY (t/km2) for each catchment are
reported.............................................................................................................................. 93
Table 7.1 Fourth-spread method Tu - SSC............................................................................... 122
Table 7.2 Meteorological Parameters....................................................................................... 131
Table 7.3 Total suspended sediment load summary (kg/jc days) for categories determined
through PC A and CA. Table 3.5 reports the number of days in each category..........132
Table 7.4 Julian Day (JD) Calendar for 2011 simplified to focus on typical ablation season.
The 2011 field season at Castle Creek Glacier has been shaded................................. 133
Table 7.5 Re-computed results from Orwin and Smart 2004a, pg. 1539; suspended sediment
totals have been computed as mean daily suspended sediment load (SSL) (kg/day)
and percent of total SSL observed at the downstream site. NPG - North Proglacial;
NPM - North Proglacial Middle; NPL - North Proglacial Lower; CPU - Central
Proglacial Upper; CPL - Central Proglacial Lower..................................................... 134

viii

List of Figures
Figure 1.1 Schematic exhaustion model representing sediment yield over the paraglacial
period (source: Church and Ryder 1972, pg. 3069)........................................................13
Figure 1.2 Sedimentation exhaustion curves for primary paraglacial processes: (1) rock slope
failure; (2) rockfall and talus accumulation; (3) accumulation o f large alluvial fans;
(4) rock-slope deformation; (5) modification of drift-mantled slopes; and (6)
modification o f glacier forelands (source: Ballantyne 2002a, pg. 373)........................ 14
Figure 2.1 Location of Castle Creek Glacier (adapted from Beedle et al. 2009). Monitoring
locations and sub-catchement boundaries are identified; section 2.1.1 and 2.1.2
provide explanation and methodology. Figure 2.2 includes site names and uses an air
photo of the proglacial zone as the base map..................................................................24
Figure 2.2 Proglacial zone o f the Castle Creek Glacier with the 2011 sampling locations,
sub-catchment boundaries and lower meteorological station. Turbidity and suspended
sediment data were collected at monitoring stations, while water level and streamflow
data were additional parameters collected at gauging stations. Site name

abbreviations: Proximal Sites 1, 2, and 3 (PS1, PS2, and PS3), Middle Site (MS),
Rockback Peak Tributary (RPT), Distal Site (DS).........................................................26
Figure 2.3 Longitudinal stream profile for Castle Creek proglacial meltwater channel. BF LB
- Bankfull Left Bank, BF RB - Bankfull Right Bank, CB - Channel Bottom, OF T1
though T4 - Outwash Fan Transects 1 through 4. Datum was arbitrarily set 100 m
below the highest benchmark, and zero channel distance was set as the outflow of the
proglacial lake upstream of PS2.......................................................................................30
Figure 2.4 - Schematic diagram o f the 2011 stream monitoring network at the Castle Creek
Glacier................................................................................................................................ 32
Figure 2.5 Flow chart detailing the classification procedure used to extract suspended
sediment transfer patterns (source: Orwin and Smart 2004a, pg. 1527).................... 47
Figure 3.1 Streamflow (Q) and suspended sediment concentration (SSC) time-series (5minute data interval) from six proglacial monitoring sites, JD 195 - JD 254, 2011
after QA/QC. Scale of y-axis varies according to range o f data. Figure presented over
two preceding pages. Exceedances in the SSC time-series are described in section
3.1.3

54

Figure 3.2 Hourly air temperature (AT, °C) from Lower Castle Creek Glacier meteorological
station; hourly precipitation (PT, mm) from Upper Castle Creek Glacier
meteorological station, estimated daily total precipitation after August 21, 2011.......54
Figure 3.3 Principal component loading of daily meteorological and streamflow (Q)
variables on principal component one and two explained 42% (PCI) and 30% (PC2)
o f the total variance in the principal component analysis. Distance o f the variable
from the origin indicates relative dominance of the Q generating processes; PC 1 and
PC2 were interpreted as ‘Rainfall’ and ‘Ablation’, and have been titled respectively in
the figure............................................................................................................................ 65
Figure 3.4 Principal component loading score plots for 5-minute SSC data from each
gauging station; all full hydrologic days of data were retained as variables for the
analysis. Percent of the data represented by each principal component is reported for
each site. Time, on the x-axis, is reported in arbitrary decimal days (06:00 is 0.25 of
the way through a regular day).........................................................................................68
Figure 3.5 Composite figures showing suspended sediment shape (diurnal or irregular) and
magnitude (1 = low; 2 = medium; 3 = high) classification results from principal
component analysis and cluster analysis and daily mean streamflow (Q) and
suspended sediment concentration (SSC) for each o f the proglacial monitoring sites.
PS1, PS2, and PS3 are missing days in the shape and magnitude classification due to
low water, partial days of data, erroneous data or no data (see section 3.1). Figure
continued over three preceding pages; x-axes in Julian Days....................................... 74
Figure 3.6 Daily mean air temperature (AT) and daily total precipitation presented with
results of meteorological principal component analysis (‘storm’; ‘hot and dry’; ‘warm
and damp’; ‘cold and wet’)...............................................................................................75
Figure 4.1 - Percent (%) contribution of suspended sediment load (SSL) and streamflow (Q)
relative to the total at DS over the 2011 field season and during the four defined
hydro-meteorological categories - schematic diagram.................................................. 89
Figure 7.1 Barometric pressure from Castle Creek Glacier upper and lower meteorological
stations. Rate o f change threshold used to remove erroneous raw data, time-series
averaged to give corrected time-series for use in computations.................................118
Figure 7.2 PS1 rating curve. No shifts applied......................................................................... 119

Figure 7.3 PS2 rating curve. The shift applies to data after Aug. 22,2011; the site was
moved just before the high flow event. Data considered an estimate above 10 m3/s.
120

Figure 7.4 MS Rating curve. The shift accounts for stilling well movement during the Aug.
22 event, and isapplied to data thereafter. Data considered an estimate above 10 m3/s.
120
Figure 7.5 DS rating curve. No shifts applied. The two grey rating points are outside
acceptable range o f 5%, and were not used for rating curve development. Data
considered an estimate above 10 m3/s........................................................................... 121
Figure 7.6 Turbidity and suspended sediment concentration regressions; all outliers
presented.......................................................................................................................... 126
Figure 7.7 Turbidity and suspended sediment concentration regressions. Corroded samples
removed............................................................................................................................127
Figure 7.8 - Turbidity and suspended sediment concentration regressions. Outliers removed,
95% confidence interval included. Figures presented for all six sites over three
preceding pages............................................................................................................... 130

Acknowledgements
I would like to express my sincere gratitude to the people and organizations that have
supported me through this process. My supervisor, Dr. Philip Owens, has provided me with
guidance, support, and encouragement as well as the insight, freedom and flexibility needed
to develop, steer and complete this project. Dr. Tim Stott from Liverpool John Moores
University was on site for a month o f field and laboratory work, and has continually been a
valuable and enthusiastic committee member. Dr. Brian Menounos has been a constructive
and disciplined committee member who has strengthened this project at every input.
This project would not have been possible without the generosity of AJ Downie, Scott
Jackson, Chelton van Geloven, and James Jacklin with the British Columbia Ministry of
Environment and John Rex with the British Columbia Ministry o f Forests who provided their
time to facilitate the in kind use of equipment for this project.
Financial support for this project came from an NSERC Discovery Grant and an
FRBC Research Chair Operating Grant to Dr. PN Owens and a UNBC RPA award to MS
Leggat.
Several research assistants have contributed to the field and laboratory components of
this project. In order o f appearance: Barry Forrester, Ian Eccles, Doug Roberts, Maud Barrel,
Natalie Saindon, and Sonja Ostertag assisted with field work. Barry Forrester, Ian Eccles,
Doug Roberts and Megan Harwood spent many hours in University of Northern British
Columbia Landscape Ecology Research Group soils laboratory drying and weighing
suspended sediment samples.
Dr. Stephen Dery was readily available for consultation and coordination o f field
efforts. He and his research team have had an invaluable role in the collection and analysis of
meteorological data. Dr. David Scott at the University of British Columbia Okanagan
generously made his soils lab available for part o f the laboratory analysis. Dr. Mathew
Beedle patiently guided the production o f GIS imagery; Leticia Gaspar Ferrer facilitated
corrections after the fact.
Perhaps most importantly, I have had the continued support and understanding o f my
family for the countless nights and weekends that I have been unavailable.

1

1

1.1

Literature Review

Introduction
In British Columbia (BC), glaciers cover 3% of the landmass (c. 29,000 km2) and

influence 20% of the watersheds with meltwater and sediment (Austin et al. 2008; Moore et
al. 2009; Bolch et al. 2010). Glaciers are sensitive climate change indicators that respond to
the ocean and atmospheric circulation patterns that influence winter and summer
precipitation and temperature (Dery et al. 2009; Moore et al. 2009). Glaciers store water
during cool and/or wet periods and release water during warm and/or dry periods. Glaciers
are powerful agents of erosion that abrade and fracture substrate, making it available for
various erosion processes and transport downstream (Dirszowsky 2004; Haritashya et al.
2010). The suspended sediment load o f glaciated rivers (1-90% ice covered watershed) is
higher than the global average (Gumell et al. 1996); this load can be even higher during
deglaciation as a result o f elevated meltwater production and sediment availability (Gumell
et al. 1999; Menounos et al. 2005). Currently, BC glaciers are not in equilibrium with
climatic conditions and have generally been retreating since reaching their Holocene maxima
at the end of the Little Ice Age (LIA) 150 - 300 years ago (Schiefer et al. 2007; Menounos et
al. 2009; Tennant et al. 2012). Since the LIA, regional climate has been dominated by a
warming trend and annual temperatures have increased by 0.5 to 1.5°C per century (Moore et
al. 2009). Between 1985 and 2005, glacier area in western Canada declined by 11.5% (Bolch
et al. 2010). Although current global climate models do not explicitly include glacial
evolution, it is expected that they will continue to retreat in response to the projected 1 to 4°C

2
increase in global mean surface temperature, depending on the emission scenario, over the
next 100 years (Collins et al. 2013).

In response to glacial retreat, in the near term (10’s o f years) there is likely to be an
increase in meltwater and sediment yield from proglacial rivers and a shift in the timing of
the peaks in the hydrograph and sedigraph (Kirtman et a l 2013). Moore et al. (2009) discuss
the effects of climate change scenarios on glacial hydrology. As glacial mass wanes and
watersheds become deglaciated, streamflow will approach annual net balance with
precipitation, and sediment yield will exponentially approach the ‘normal’ sub-aerial erosion
rate of non-glaciated catchments (Church and Ryder 1972). The magnitude and time scale for
the adjustment of hydrologic and geomorphic processes is dependent on climatic conditions,
the scale of the geomorphic system, glacial extent and site-specific characteristics that
determine the rate o f temporal change in streamflow and sediment availability (Tunnicliffe
and Church 2011). Many of the watersheds in BC are still responding to the Pleistocene
glaciation (Church and Slaymaker 1989).
Glacial processes can affect the timing, quantity and quality of streamflow and suspended
sediment load o f a watershed. The influence of glaciers on the hydrology, geomorphology
and sediment yield of a watershed are o f significant ecological, economical, and societal
importance. For instance:

•

aquatic ecosystems are sensitive to water temperature, and the quantity, quality, and
timing of streamflow and sediment; certain species may decline or be extirpated if
stream conditions change beyond their niche (Austin et al. 2008; Milner et al. 2009;
Moore et al. 2009);

3
•

fine sediment < 63 pm (silts and clays) are the chemically active component of the
solid sediment load; they can transport and store nutrients and contaminants (Brown
et al. 1996; Dirszowsky 2004; Hodson et al. 2004; Owens et al. 2005; Walling 2005;
Hodson et al. 2008; Haritashya et al. 2010);

•

depending on site specific and hydrologic characteristics, bedload can account for 5 65% of sediment yield from the proglacial zone and will therefore have implications
for downstream channel morphology and ecology (Bogen 1989; Hammer and Smith
1983; Church and Slaymaker 1989; Warburton 1990; Harbor and Warburton 1993;
Gumell et al. 1999; Scheifer et al. 2010);

• streamflow translates into a dollar value for hydropower operations, and the amount
o f sediment can determine the lifespan and maintenance requirements of reservoirs,
headponds, and turbines (Bogen 1989; Morehead et al. 2003);
•

there are many anthropogenic interactions that depend on the quantity and quality of
water in glacially influenced rivers including, for example, recreational boating,
swimming, and fishing; and,

• regulatory agencies that issue water licences and permits for intakes and discharges to
surface water may rely on, or have to deal with, upstream glacial influences (Moore et
al. 2009).
The importance o f glaciers, and their rapid retreat since the LIA, has caused researchers
to study the processes that influence, and are driven by, these intriguing and valuable
landscape features. The focus of this study is on the flux of fluvial suspended sediment < 2
mm through the proglacial zone.

4
1.2

Research Objectives
The overall aim of this thesis was to assess the influence of hydro-meteorological

conditions on suspended sediment flux in a proglacial creek in the Cariboo Mountains of
British Columbia during the 2011 field season. The specific research objectives were to:

1. examine the spatial and temporal response pattern o f suspended sediment
concentration in the proglacial zone; and,
2. determine the sources of streamflow and suspended sediment load under different
hydro-meteorological conditions.
In order to achieve these objectives, meteorological data were collected from two
automated weather stations in the proglacial zone, and streamflow (Q) and suspended
sediment concentration (SSC) were monitored independently at six sites along the proglacial
meltwater channels of the Castle Creek Glacier from July 14 to September 11 o f 2011. To
achieve objective 1, the shape and magnitude o f the SSC response to hydro-meteorological
conditions was categorized using principal component analysis (PCA) and cluster analysis
(CA), and sediment flux processes and source/sink areas were identified using field
observations and measurements as substantiating evidence. To achieve objective 2,
streamflow and suspended sediment load were computed and summarized for the
meteorological categories, as defined in objective 1. In addition, a suspended sediment
budget for the Castle Creek proglacial meltwater channel was defined using the results o f the
analysis in this study and the proglacial sediment parameters defined by Warburton (1990);
this helped to quantify areas of sediment sources and storage.

5
The following introductory sections review relevant work in proglacial zones and on
suspended sediment budgets; the challenges of hysteresis and non-linear relationships in SSC
and Q data; and the different modelling approaches that have been used.

1.3

Proglacial Zone
The ‘proglacial zone’ is the area surrounding a glacier that has been influenced by the

glacier, and the ‘proglacial period’ lasts until the completion o f deglaciation. ‘Paraglacial
sedimentation’ refers to non-glacial sedimentation processes directly conditioned by the
previous glaciation (Church and Ryder 1972; Ballantyne 2002a). The ‘paraglacial period’
lasts until the effects of the previous presence of ice have diminished, and erosion rates return
to that o f a non-glaciated catchment under ‘normal’ subaerial weathering processes (Church
and Ryder 1972; Ballantyne 2002a). Recently exposed sediment in the proglacial zone can be
in an unstable or metastable state; these initially unconsolidated sediments are subject to
rapid and extensive modification and erosion over the proglacial period. Sediment yield can
greatly exceed that of otherwise equivalent non-glacial landscapes throughout the paraglacial
period (Church and Ryder 1972; Church and Slaymaker 1989; Harbor and Warburton 1993;
Hallet et al. 1996; Ballantyne 2002b; Hodgkins et al. 2003).
A spectrum of geomorphic processes release and rework glacial sediment over a wide
range of timescales (Ballantyne 2002a). Immediately after exposure, the unconsolidated and
water-saturated till in the glacier forefield begins to adjust to subaerial conditions; loose
sediments consolidate as the substrate drains, and slope angles decline (Ballantyne 2002b).
More recently exposed surfaces tend to be greater sediment sources than older surfaces and
the rate of adjustment following deglaciation can be approximated by exhaustion models

6
(Ballantyne 2002a, also see section 1.5.1). Over time, the eluviation of fines, surface
armouring, reduction in surface slope and vegetation colonization act to stabilize the
proglacial zone and reduce sediment availability for fluvial entrainment and transport
(Warburton 1990; Gumell et al. 1999; Orwin and Smart 2004a; 2004b).

There are many factors that influence the rate o f erosion and sediment yield from
glaciated watersheds, and no simple linear relationship exists (Gumell et al. 1996); however,
global comparisons have found that sediment yields are higher in glaciated watersheds than
non-glaciated watersheds (Harbor and Warburton 1993; Gumell et al. 1996; Hallet et al.
1996; Richards and Moore 2003). Since the hydrology of a watershed changes as it becomes
deglaciated, sediment yield is not static through time and can vary spatially and temporally in
response to site-specific characteristics such as the underlying rock type and subglacial
deposits, the rate of glacial movement, the character of the glacial drainage system,
topography, weather and climate (Gumell et al. 1996). The size of the watershed, extent of
glaciation, abundance and distribution o f glacial sediments and their connectivity to the
fluvial system determine the duration and magnitude of the paraglacial period (Church and
Ryder 1972; Ballantyne 2002a; 2002b). Contrary to conventional global sediment yield
models that show declining sediment yield as watershed area increases (Syvitski and
Milliman 2007), the specific sediment yield for glaciated watersheds (>10 km2) in BC
increases for spatial scales up to 104 km2 as a result o f the remobilization o f sediments
deposited during the Pleistocene (Church and Slaymaker 1989; Tunnicliffe and Church
2011); however, the results for watersheds <10 km2 are more dependent on physiographic
characteristics (Schiefer et al. 2001). High sedimentation rates observed in proglacial lake
cores (i.e. derived from glacial meltwater) coincide with periods of rapid glacial retreat, but

7
also periods of rapid advance and glacial maxima due to glacial override o f previously
deposited sediment (Ballantyne 2002b; Menounos et al. 2005; Menounos et al. 2009).
Downstream sediment yields represent the combined contribution of contemporary
paraglacial erosion from active upland proglacial zones and other non-glacial erosion, as well
as the reworking o f sediment deposited more than 10,000 years ago (Church and Slaymaker
1989; Scheifer et al. 2001; Tunnicliffe and Church 2011). As such, it is difficult to partition
the relative amount o f sediment derived from the two scales o f paraglacial systems and
contemporary non-glacial processes solely from a downstream perspective (Church and
Slaymaker 1989; Ballantyne 2002a; Dirszowsky 2004).
The definition and inclusion of proglacial sediment yields in sediment budget models
can help to isolate sediment generated by contemporary glacial and paraglacial processes
from contemporary reworking of Pleistocene deposits (Harbor and Warburton 1993).
Determining accurate proglacial sediment budgets can be useful for water quality and
quantity models that include sediment yield from glacially influenced watersheds (Warburton
1990; Richards and Moore 2003; Orwin and Smart 2004a; Stott and Mount 2007).

1.4

Hysteresis and Non-linear Relationships in Suspended Sediment Concentration
and Streamflow Data
Collecting high resolution SSC and Q data requires substantial effort. This section is

included to explain the complexity of these types of data, and draws upon the experience of
past researchers to justify the need for and usefulness of independent, high resolution data
collection.

8
Empirically, SSC can be related to Q by a simple power function, but simple direct
relationships are notoriously poor because o f hysteresis loops and non-linear relationships
(Gumell and Fenn 1984; Pickup 1988; Lawson 1995; Hodson and Ferguson 1999; Swift et
al. 2002; Richards and Moore 2003; Orwin and Smart 2004b; Stott and Mount 2007). In a
given high-flow event, scatter in the SSC - Q relationship can result when the sediment wave
either lags or, more often, precedes peak Q (Naden 1988; Hodson et al. 1998). When plotted
against each other with Q as the independent variable, data where the SSC peak precedes the
peak Q shows clockwise hysteresis, while anti-clockwise hysteresis occurs when the SSC
peak follows the Q peak (Hodson et al. 1998; Richards and Moore 2003; Orwin and Smart
2004b; Eaton et al. 2010). Hysteresis data contain serial autocorrelation in the residuals,
where each data point is related to the others in the series (Naden 1988; Richards and Moore
2003), but not necessarily Q (Hodson and Ferguson 1999). Interpretation of SSC - Q plots
and hysteresis loops for a series of sites along a stream can be used to assess sediment
availability and suggest which sediment sources are contributing and when they are
contributing during a given high-flow event (Hodson et al. 1998; Hodson and Ferguson
1999; Orwin and Smart 2004b).
Clockwise hysteresis is more common, and could be considered the background
“normal” condition for geomorphically “inactive” systems where fine sediment that has
accumulated within the catchment area or the channel since the last high-flow event is
evacuated during the rising limb of the hydrograph and is then exhausted as Q peaks (Hodson
et al. 1998; Orwin and Smart 2004b; Eaton et al. 2010). The falling limb of the hydrograph
tends to have a much lower SSC, and two equivalent Q values separated by a short period of

9
time can have instantaneous suspended sediment loads that differ by more than an order of
magnitude (Pickup 1988; Orwin and Smart 2004b).
Spatial separation o f the sample site from the source o f sediment, and episodic mass
movement on the falling limb of the hydrograph - commonly a bank failure from
undercutting during the event - can result in a “late” or seemingly random pulse o f sediment
after the initial peak has passed (Pickup 1988; Morehead et al. 2003; Orwin and Smart
2004b). This situation can create anti-clockwise hysteresis loops in the SSC - Q plot.
Geomorphically “active” point sources of sediment that are connected to the stream
network, such as a bank failure, can be discharged in declining pulses of sediment through
subsequent high flow events over a season or several years (Morehead et al. 2003; Eaton et
al. 2010). Additionally, they may become active sediment sources under specific hydrologic

or weather conditions such as high Q, spring melt, or heavy rainfall but otherwise be inactive.
Diffuse areas of exposed sediment, as found in the proglacial zone, act as transient sediment
sources (Warburton 1990; Hodson et al. 1998; Orwin and Smart 2004b). The amount of
sediment eroded from a particular area is highest immediately following exposure and
declines over time as mass movement processes reduce surface slope and sediment
availability. These processes can operate at diurnal, weekly, sub-seasonal, and seasonal time
scales or over longer periods of time (Warburton 1990; Hallet et al. 1996; Hodson and
Ferguson 1999; Richards and Moore 2003; Orwin and Smart 2004b).
The processes described above result in a complex relationship between SSC and Q for
both “inactive” and “active” geomorphic systems. “Inactive” systems will tend to show more
prominent clockwise hysteresis where available fine sediment is entrained during the rising

10
limb of the hydrograph and then exhausted as the hydrograph peaks. Geomorphically
“active” systems tend to have a more complex SSC - Q relationship (Richards and Moore
2003). In proglacial zones, the abundance of unconsolidated fine sediment can cause SSC to
track Q more closely (Hodson et al. 1998; Hodson and Ferguson 1999). However, the
abundance of active sediment sources in these systems will tend to create irregular response
patterns, which can include sustained or late pulses o f sediment on the hydrograph and anti­
clockwise hysteresis. Suspended sediment concentration can change dramatically in a short
period o f time, and though Q can be a trigger, the suspended sediment response can be
unpredictable and thus the two variables are best recorded independently.

1.5

Modeling Suspended Sediment Concentration
The hysteresis and non-linearity that arise from seasonality, source exhaustion,

disturbance regimes, and spatial separation of the sample site from the sediment source
described above make lumped parameter models that relate SSC to Q inaccurate in supplylimited systems. Such lumped parameter suspended sediment rating curves assume that SSC
depends on Q, or that SSC is transport-limited. However, SSC in most fluvial systems
depends on the rate of erosion within the catchment and the rate o f supply to the channel, not
Q (Pickup 1988; Ritter et al. 2002). As an exception, in certain locations unconsolidated
sediments may be more transport-limited, and in these situations empirical suspended
sediment rating curves may be more applicable (Hodson et al. 1998; Swift et al. 2002; Stott
and Mount 2007).
Since the SSC peak tends to occur on the rising limb of the hydrograph, various time
lag, hydrograph separation, cross-correlation, log transformation and multiple regression

11
techniques have been used to predict SSC from Q to correct the suspended sediment rating
curve (Hodson and Ferguson 1999). Gumell and Fenn (1984) tested several of these methods
to improve the SSC - Q rating relation for a proglacial system, but none were able to
significantly improve upon the estimates made by the ordinary rating curve. They found that
a Box-Jenkins transfer function based on a longer time series o f SSC and Q data was more
accurate and suitable for real-time forecasting o f SSC based on Q (Gumell and Fenn 1984).
However, a significant amount of SSC data is required to establish and maintain such
models, a continuous record of Q is also required, and the forecasted SSC is relative to the
previous SSC.
Building on the past models, Syvitski and Milliman (2007) introduced the BQART
model as a global predictor of sediment flux to the oceans. When the model was applied to a
database o f 488 rivers that drain into the oceans it was able to account for 96% o f the long­
term sediment yield (Syvitski and Milliman 2007). Glacial sediment is a relatively minor
component of this large-scale model, accounting for only 1% o f the signal, but they noted
that this component would be much more important during and just after glaciation (Syvitski
and Milliman 2007). Geographic factors (such as topographic relief, watershed size, geology,
latitude, etc.) accounted for 65% of the variation between rivers, while climatic and
anthropogenic factors accounted for 14% and 16%, respectively. Syvitski and Milliman
(2007) state that more advanced models are required to account for the magnitude and
timescale of paraglacial and deglacial processes, and that the model includes a general glacial
erosion factor as a function of glacial coverage in the watershed. However, specific research
has found that glacial cover alone is not a good predictor o f suspended sediment yield
(Harbor and Warburton 1993; Gumell et al. 1996) and that other factors need to be included,

12
such as the underlying rock type and sub-glacial deposits, rate of glacial movement, character
o f glacial drainage system, and the topography and physiography o f the basin (Gumell et al.
1996; Scheifer et al. 2001).

1.5.1

Modeling Suspended Sediment Yield over the Paraglacial Period
Church and Ryder’s (1972) description of the combined effect of deglacial processes

over the paraglacial period shows that in-stream sediment yield initially increases following
the start o f deglaciation and then decreases exponentially to the ‘normal’ subaerial erosion
rate of a non-glacial landscape (Figure 1.1). The inflection point on the falling limb o f the
model occurs as deglaciation is completed. This model represents the exhaustion of available
glacigenic sediment over time through mass movement processes and reduction of surface
slope by surface wash, frost sorting, eluviation of fines, and stabilization by vegetation
colonization (Ballantyne 2002a). The axes of the conceptual model proposed by Church and
Ryder (1972) are scale independent because the response time and magnitude reflect the
spatial scale and site-specific characteristics of the catchment. In general, larger systems will
have a nested effect o f several individual response curves from glacially influenced tributary
watersheds and take longer to reach the subaerial norm (Church and Slaymaker 1989; Harbor
and Warburton 1993). While primary paraglacial sedimentation can last decades to centuries
in small, alpine watersheds (Ballantyne 2002a; Orwin and Smart 2004b), some of the largest
rivers in BC continue to have suspended sediment loads greater than the ‘geologic norm’ as
secondary paraglacial processes continue to rework Quaternary sediment deposits (Church
and Slaymaker 1989; Scheifer etal. 2001; Ballantyne 2002b).

13

Commencement
of degtadaiion

End of
period

Geological ’norm' - f- -f
time

Figure 1.1 Schem atic exhaustion model representing sediment yield over the paraglacial period (source:
Church and Ryder 1972, pg. 3069).

Embedded within Church and Ryder’s (1972) paraglacial sediment yield model is a
spectrum of geomorphic processes which includes both primary and secondary paraglacial
sedimentation. Primary paraglacial geomorphic systems are directly glacially conditioned
(e.g. rock-slope failure due to deglacial stress release, modification o f moraines by slope
processes, entrainment of glacigenic deposits by rivers), whereas secondary paraglacial
systems encompass the reworking o f in situ glacigenic sediment and paraglacial sediment
stores (e.g. talus, debris cones, outwash fans, valley fills) (Ballantyne 2002a). The work of
several researchers on primary paraglacial geomorphology is combined into one diagram to
compare the rate of paraglacial sediment exhaustion for different processes (Figure 1.2;
Ballantyne 2002a). Note that the diagram must be treated with caution because available data
are sparse, collected from different locations and are, therefore, not necessarily universally

14
applicable (Ballantyne 2002a). The y-axis in Figure 1.2 is presented as a proportion of the
total, and the contribution o f each process to sediment yield will not be equal. All primary
paraglacial sedimentation processes follow an exponential decay model following
deglaciation (Ballantyne 2002a), which is encompassed by Church and Ryder’s (1972) model
that extends over the paraglacial period.

%

100

0

500

1000

1500

2000

2500

3000

Time since deglaciation in years

Figure 1.2 Sedim entation exhaustion curves for prim ary paraglacial processes: (1) rock slope failure; (2)
rockfail and talus accumulation; (3) accumulation o f large alluvial fans; (4) roek-slope
deform ation; (5) modification o f drift-m antled slopes; and (6) modification o f glacier forelands
(source: Ballantyne 2002a, pg. 373).

The processes in Figure 1.2 are primary paraglacial processes that are not
replenished; the model assumes steady-state conditions with no episodic events or other
changes to the system that would rejuvenate or renew the availability of sediment.
Additionally, secondary non-glacial reworking of the sediment is not considered in this
diagram. Fluvial processes are the dominant non-glacial force that reworks paraglacial
sediment deposited on land (Ballantyne 2002a). In the proglacial zone o f a retreating glacier,
the sediment exposed each year starts at time zero and begins its progression through the

15
exhaustion model, resulting in dynamic proglacial sediment availability over space and time.
The most rapid paraglacial process following deglaciation is the modification o f glacier
forelands. Mass movement, eluviation of fines, and redistribution and sorting o f sediment are
often complete within a few decades, and as vegetation colonization occurs normal subaerial
weathering rates are approached within c. 200 years (Church and Ryder 1972; Ballantyne
2002a). However, secondary paraglacial sedimentation can continue for much longer and is
related to watershed size, climatic conditions, and the extent of glaciation and glacial deposits
(Church and Slaymaker 1989; Scheifer et al. 2001).

1.5.2

Rationale fo r the Use o f Turbidity Measurements to Determine Suspended
Sediment Concentration
The spatial and temporal complexity of suspended sediment entrainment, transport

and storage precludes predictive models o f SSC (Naden 1988). Because o f this complexity, it
is necessary to have a high sample frequency in monitoring programs that aim to quantify
suspended sediment load (SSL) (Gippel 1989; Warburton 1990; Gumell et al. 1996; Navratil
et al. 2011). Additionally, studies that aim to quantify sediment sources, sinks and processes
in a catchment or to establish a sediment budget will likely need multiple monitoring sites
(Richards and Moore 2003; Orwin and Smart 2004a; 2004b). Due to the high effort and cost
of collecting and analysing physical water samples to measure SSC, turbidity (Tu) is often
used as a surrogate measure and sample frequency can be near continuous (Orwin and Smart
2004a; Orwin and Smart 2005; Navratil et al. 2011).
Turbidity is a measure of light penetration and is affected by particle size, shape and
composition, as well as bubbles (turbulence), water colour, and algae (Gippel 1989; Orwin
and Smart 2005; Ginting and Mamo 2006). Turbidity-SSC regressions are especially

applicable in low biologic productivity systems, such as proglacial zones (Swift et al. 2002;
Orwin and Smart 2004a; Stott and Mount 2007). However, there can be a significant amount
of uncertainty in the Tu-SSC relation which culminates from, and is propagated through,
various aspects o f the data collection and analytical procedure (Navratil et al. 2011).

Hysteresis in Tu-SSC data can occur, and has been attributed to changing flow
conditions at the monitoring location and different properties o f different sediment source
materials as they are transported past the monitoring location (Orwin and Smart 2005;
Ginting and Mamo 2006; Navratil et al. 2011). Unless the SSC sample frequency is high (30
minutes or less), this hysteresis pattern appears as scatter in the Tu-SSC relationship,
resulting in a wider spread o f the residuals, and a lower R2 value. However, source materials
may be relatively homogenous in the small proglacial catchment area that is the focus o f this
study, which should limit the amount of Tu-SSC hysteresis from differing source materials.

1.6

Sediment Budgets
While the transport of bedload sediment is hydraulically controlled (i.e. stream

competency determines the credibility and transport distance o f sediment), the transport of
suspended sediment is typically supply controlled (i.e. sediment delivery mechanisms can be
more important than stream competency in determining the amount o f material transported in
suspension). Suspended sediment can be transported a great distance downstream (Ritter et
al. 2002); thus, exported from the proglacial zone once entrained.
Fine sediment (i.e. rock flour in glacial meltwater) is the chemically active portion of
the solid load transported by a river and is an important vector for the transfer and fate of
nutrients and contaminants through both terrestrial and aquatic ecosystems (Brown et al.

17
1996; Owens et al. 2005). For fluvial geomorphologists, developing a suspended sediment
budget that quantifies the nature, importance and interaction between sediment production,
mobilization, transport, storage and yield is a precursor to developing effective sediment
management and control strategies aimed at reducing diffuse or point source pollution by fine
sediment (Slaymaker 2003; Hodson et al. 2004; Owens 2005; Walling 2005; Walling and
Collins 2008). The spatial and temporal variability o f suspended sediment processes within a
catchment can make quantification with direct measurement techniques such as field
observations, photogrammetry, erosion pins, profilometers, sediment traps and erosion plots
exceedingly complex (Walling and Collins 2008). Thus, there is considerable motivation to
find more effective and efficient ways to assemble the necessary data to construct reliable
sediment budgets (Harbor and Warburton 1993). As there is no well-defined single procedure
that is universally applicable, researchers have focused on integrated approaches that use a
combination o f complementary techniques to discriminate and quantify sediment sources,
fluxes and storage within sediment budget frameworks (Warburton 1990; Harbor and
Warburton 1993; Owens 2005; Walling and Collins 2008).
1.6.1

Proglacial Sediment Budget Applications
Proglacial sediment budgets tend to be simpler than equivalent budgets for temperate

catchments because of their smaller size, relative lack o f vegetation and abundance of
unconsolidated material, which make the sources, sinks, and transfers o f suspended sediment
easier to define. However, a common problem with the definition of proglacial erosion and
sedimentation processes is the short time scale o f many of the studies due to the high effort
involved with collecting data from these often remote locations with extreme climatic
conditions (Warburton 1990; Gumell et al. 1996; Hodson et al. 1998; Hodgkins et al. 2003;

18
Orwin and Smart 2004a). The definition o f sediment budgets over short time scales is subject
to partial definition when the recognition and quantification of budget processes are being
conducted simultaneously, as such, pilot studies have high value (Warburton 1990).
Additionally, Warburton (1990) has noted that data which represent only part of an ablation
season may be misleading, as, for example, sediment evacuation during spring freshet may
be followed by sediment storage later in that same season. These difficulties are confounded
by the fact that each individual ablation season can be markedly different (Hodgkins et al.
2003; Jobard and Dzikowski 2006; Stott and Mount 2007; Cockbum and Lamoureux 2008;
Haritashya et al. 2010), and that sediment sources evolve through the ablation season
(Hodson et al. 1998; Swift et al. 2002; 2005; Haritashya et al. 2010). Walling (1978)
suggests that 10 years of monitoring are required before the sediment transport system o f a
catchment can be adequately characterized; however, Warburton (1990) comments that
because o f the rapid rate of change and condensed frequency of geomorphic events in the
proglacial zone, multi-year studies are subject to many of the same limitations.
The proglacial zone can be a significant source and sink of sediment (Warburton
1990; Harbor and Warburton 1993; Hodson et al. 1998; Hodgkins et al. 2003; Richards and
Moore 2003; Orwin and Smart 2004a), and will depend on site-specific characteristics
(Harbor and Warburton 1993; Gumell et al. 1996). The function of the proglacial zone as a
source and sink of sediment operates at a range of timescales as glacierised catchments
respond to weather patterns, and seasonal and climatic trends. Abnormally warm ablation
seasons have been shown to increase sediment yield from the proglacial zone (Stott and
Mount 2007; Cockbum and Lamoureux 2008), whereas, cooler seasons have been found to
increase sediment storage within the proglacial zone (Hodgkins et al. 2003; Richards and

19
Moore 2003). The distribution and intensity o f monitoring sites is important in determining
and interpreting suspended sediment fluxes because as distance from the glacier increases, so
does the potential for sediment storage and remobilization, and therefore modification of
transfer patterns (Harbor and Warburton 1993; Orwin and Smart 2004a). Warburton (1990)
defined the basic sediment balance equation for the coarse and fine components o f a
proglacial sediment budget as:

Y = SL + TR + M + GL + AVS

(1)

where: Y is sediment yield; SL is direct hillslope inputs; TR is tributary channel
inputs; M is the input from moraine deposition; GL is the glacial stream input; and AVS is
the change in valley sandur.
Input to the proglacial zone from the glacier can encompasses several sources of
meltwater and sediment, including ice melt, snow melt, and subglacial meltwater that entrain
supraglacial and subglacial sediment. Glacial stream input is an important component o f the
proglacial suspended sediment budget (Hammer and Smith 1983; Warburton 1990; Swift et
al. 2002; Haritashya et al. 2010; Orwin and Smart 2004a). Variability among the results of
proglacial SSC and Q studies may depend on the characteristics o f the glacier, the proglacial
zone and underlying geology (Harbor and Warburton 1993; Gumell et al. 1996), as well as
antecedent conditions and weather and climate patterns during data collection (Richards and
Moore 2003; Cobum and Lamoureux 2008; Moore et al. 2009). It is important to note that
subglacial, supraglacial and englacial drainage networks evolve, and the timing and amount

of fluvial sediment load can gradually drift or change suddenly within a single melt season
(Swift et al. 2002), or over multiple seasons (Swift et al. 2005; Haritashya et al. 2010). When
studying proglacial suspended sediment response patterns, the evolution o f these drainage
networks is significant, and may explain shifts in the pattern as well as seemingly random
spikes or decreases in suspended sediment load as the meltwater flow path becomes more or
less efficient, and gradually or suddenly exhausts, accesses or abandons sediment sources
(Hodson et al. 1998). The evolution of proglacial SSC and Q patterns through the ablation
season has led researchers to divide data into categories of similar conditions to infer and
summarize the processes, controls, and driving factors (Hodson et al. 1998; Richards and
Moore 2003; Orwin and Smart 2004a).
Richards and Moore (2003) monitored SSC and Q during two ablation seasons at
Place Creek Glacier, in the Canadian Coast Mountains and an adjacent catchment that was
almost unglaciated. The aspect of the glaciated catchment was northwest with an area o f 13
km2 and 26% glaciated. Using the relatively unglaciated catchment for comparison, they
divided the ablation season into four sub-seasons that reflect the Q generation processes: 1)
nival; 2) nival-glacial; 3) glacial; and 4) autumn recession. They found that fine sediment was
temporarily stored within the fluvial network between the proximal and distal site at low
flow, and then re-entrained at higher flow and the response of the catchment to rainfall
changed over the ablation season in relation to snow cover and antecedent conditions.
In Orwin and Smart (2004a), the spatial and temporal patterns of suspended sediment
flux in proglacial channels of the Small River Glacier, BC, were assessed and ascribed to
weather phenomenon. Their analysis, based on that of Hannah et al. (2000), combined
principal component analysis (PCA) and cluster analysis (CA) (see section 2.5) as an

21
objective way to characterize the ‘shape’ and ‘magnitude’ o f the suspended sediment timeseries and divide the data into four hydro-meteorological categories: 1) hot and dry; 2) cold
and wet; 3) snowmelt; and 4) storm (Orwin and Smart 2004a). Field observations were used
to identify sediment sources, paraglacial sedimentation processes and temporary storage
within the study site. The study design and analysis procedure for the proglacial zone of the
Small River Glacier by Orwin and Smart (2004a) was emulated for this 2011 study on the
Castle Creek glacier proglacial zone. As such, some additional detail about their study site
and methods are included for reference; similar information for this study on the Castle
Creek Glacier can be found in sections 2.1 and 2.2.
The Small River Glacier (SRG) is located in the Canadian Rocky Mountains
(53°11’N, 119°30’W), c. 47 km east south-east of the community of McBride. The SRG is a
small cirque glacier with a south-east aspect. The geology is dominated by limestone with
interstratal dolomite and shale units of the Mural and Mahto Formations (Orwin and Smart
2004a). Their study area in the upper basin was 6.86 km2 and c. 50% glaciated with an
elevation range o f 1750 to 2600 m a.s.l. (Orwin and Smart 2004a). The proglacial zone of the
SRG was c. 2.0 km2 with an elevation range o f 1750 - 2200 m a.s.l. Average stream
gradients were c. 15% on two parallel streams and c. 21% on a third meltwater stream
(Orwin and Smart 2004a). Meteorological data were collected from a central proglacial
monitoring station, and SSC and Q data were collected from a network of nine monitoring
sites along three proglacial meltwater channels from Julian Day (JD) 288 - JD 238 (July 7 August 26, 2000). Approximately 10 salt dilutions per site were used to establish stagedischarge rating curves, and SSC was measured in 50 - 100 water samples from each site to
field-calibrate the turbidity meters.

22
1.7

2008 Study on Sediment Fluxes at Castle Creek Glacier
In order to gain a better understanding o f suspended sediment fluxes downstream of a

retreating glacier, a 34-day study was conducted in the Castle Creek proglacial zone in July
and August of 2008 (Stott et al. 2009). The study found that SSL and Q were controlled by
rainfall and snow/ice melt, sometimes independently and sometimes in concert. The results
were used to estimate a 43 ± 2 t km'2 increase o f suspended sediment yield between a
proximal and distal site separated by c. 600 m of stream distance, and found that the distal
site had more scatter in the SSC - Q relationship than the proximal site. The study
highlighted the potential importance o f the proglacial zone in modifying SSL and limitations
o f the SSC - Q relation. Importantly, the study identified that further investigation with a
more detailed sampling strategy was required in order to constrain the role o f hydro­
meteorological conditions on suspended sediment fluxes in the proglacial zone.
In this 2011 study, key limitations identified in the 2008 study were addressed by
expanding the network of proglacial monitoring sites, increasing the study length, and
independently monitoring Q and SSC to allow for a more comprehensive and quantitative
analysis o f proglacial suspended sediment flux patterns.

23

2

M eth o d s

2.1

Study Area
The Castle Creek Glacier (CCG) is an alpine valley glacier located c. 35 km south-

southwest o f the community of McBride, in the Cariboo Mountains o f British Columbia,
Canada (53°2’N, 120°24’ W) (Figure 2.1). The snout o f the glacier receded 700 m between
1959 and 2007, and c. 1.5 km since its Little Ice Age (LIA) maximum (Beedle et al. 2009).
In 2005, the CCG had an area of c. 9.8 km2 (Bolch et al. 2010). Based on 2011 imagery, the
CCG had an area of c. 9 km2; its length and elevation range were c. 6 km and 1870 to 2850 m
a.s.l., respectively (Figure 2.1; also see section 2.1.2). The glacier flows northeast, while the
aspect o f the accumulation zone is north. The underlying geology is a vertical outcrop o f the
Windermere Supergroup, which represents a deep-ocean basin turbidite system that formed
700 million years ago (Amott per s. comm. 2011). It was pushed up from the ocean floor
about 100 million years ago during the formation o f the Rocky Mountains. This bedrock
outcrops extensively in the east north-east area of the proglacial zone, and on the west side of
the proglacial zone in an area above the terminal lobe of the CCG. Additionally, there is an
outcrop along the distal end o f the lower till apron, which connects with the east north-east
outcrop where the creek flows through a small gorge that marks the end o f the immediate
proglacial zone. Meltwater from the CCG flows southeast for c. 7 km as it leaves the
proglacial zone, and then turns abruptly and flows generally northeast, draining into the
upper Fraser River basin near McBride after c. 34 km.

24

OUetor \

^^■neeuww
s^5
,

'^0l4

US**..
WASHINQTON

i^aapp*

Gaugng Station
0

Monrtcxtng Staton

^

Meteorological Stator
Sub^alcftmte* Boundary
C atte Croak Gtacwr
O terG taatfs
100 m Contour
Stream*

20*3C m

120*27^/

Figure 2.1 Location of Castle Creek Glacier (adapted front Beedle et al 2009). M onitoring locations and
sub-catchem ent boundaries are identified; section 2.1..1 and 2.1.2 provide explanation and
methodology. Figure 2.2 includes site names and uses an air photo o f the proglacial zone as the
base map.

25

2.1.1

Field Site Observations and Site Selection
The scope of this thesis was on the meltwater channels and sediment sources in the

proglacial area from the snout o f the glacier c. 1870 m a.s.l. to the small gorge c. 1800 m
a.s.l. The watershed area above the gorge was c. 16 km2 and was c. 60% glaciated in 2011
(see section 2.1.2), stream distance was c. 1.2 km with an average slope of c. 3% (see section
2.1.3).
The area immediately downslope from the snout of the glacier was characterized by
low relief till sheets, outwash fans, abandoned meltwater channels, and bedrock outcrops.
The till deposits on the west side of the meltwater channel have been substantially eroded and
modified by several abandoned meltwater channels incised to varying depths (up to 10 m)

which end at abandoned outwash fans. The east side of the meltwater channel was
characterized by two relatively intact till sheets separated by an outwash fan complex (Figure
2 .2 ).

26

120*26’30”W

I

Gauging Station
53*4*0“N

0

120*26'0"W

Monitoring Station

o I

A

Meteorological Station

<•> ■

—

Sub-catchment Boundary
HP I
I)S

MS

53*3'30“N

PS I

120*26‘30"W

120»2«,0"W

Figure 2,2 Proglacial /o n e of the Castle Creek Glacier with the 2011 sam pling locations, sub-catchm ent
boundaries and lower meteorological station. Turbidity and suspended sedim ent data were
collected at monitoring stations, w hile water level and streamflow data were additional param eters
collected at gauging stations. Site name abbreviations: Proximal Sites 1 ,2 , and 3 (PS I. PS2, and
PS3), M iddle Site (M S), Rockbaek Peak Tributary (RPT), Distal Site (DS).

27

The main meltwater stream emanated from a subglacial channel portal on the
northwest side o f the glacial terminus and flowed steeply to a proglacial lake on the west side
of the terminus. The outflow from this proglacial lake was mostly bedrock controlled, as was
the confluence with another bedrock controlled meltwater stream flowing from a proglacial
lake centered at the glacial terminus. From this point, stream slope increased and the channel
was bound by bedrock on the left bank and the over-steepened, slumping bank of the upper
till sheet on the right bank. In this reach, fine sediment had been washed away leaving behind
large cobble and boulders that armour the channel bottom and right bank. The steep single
thread channel split at the top of an outwash fan. Much of the left side of the fan was
abandoned and elevated from the current channels. The majority o f flow was along the right
side o f the fan where a vertical bank (c. 2 m) had been cut along the base o f the upper till
sheet. There was a gradual decrease in slope and corresponding decrease in stream power and
particle size; near the bottom o f the fan the stream became braided with wide and shallow
channels. A small tributary from the east side o f the glacier entered at multiple points from a
low gradient outwash plain as dispersed flow converged and curved west around the distal
end o f the outwash fan (Figure 2.2).
The cumulative meltwater from the CCG flowed generally north with the low
gradient till sheet on the right bank, and a series of abandoned meltwater channels and
outwash fans on the left for c. 400 m. A prominent moraine directs the stream sharply to the
east around the distal end of the till sheet. The confluence with a tributary that drained the
western side of the watershed and a small cirque glacier/rock glacier on the north aspect o f
Rockback Peak occurred as the new heading was achieved. From this point, the channel was

28
relatively straight with a gravel-cobble bed that continued generally east in a single thread to
the small gorge, which marked the end o f the alluvial proglacial zone, and the study area.
Six monitoring sites were established to assess the spatial and temporal pattern o f
suspended sediment flux in the CCG proglacial zone (Figure 2.2). Three sites proximal to the
glacier monitored suspended sediment input to the proglacial channel from the glacier and
the area immediately proximal to the terminus: Proximal Site 1 (PS1) sampled a small ice
marginal stream draining the east side of the glacier; Proximal Site 2 (PS2) sampled the main
meltwater stream downstream of a proglacial lake on the north-west site o f the glacial
terminus; and Proximal Site 3 (PS3) sampled a stream flowing north-west from a small
proglacial lake roughly centered at the glacial terminus. The Middle Site (MS) sampled the
main meltwater channel, downstream of the outwash fan. The Rockback Peak tributary
(RPT), which entered the Castle Creek meltwater stream downstream of MS, was sampled c.
100 m from the mouth. For the Distal Site (DS), stream gauging was done upstream o f the
small gorge; however, to ensure complete mixing o f water from the RPT with that o f Castle
Creek, the Tu and SSC sample point was located in the bedrock-controlled reach downstream
o f the small gorge. The sites PS2, MS, and DS were in approximately the same location as
the three monitoring sites that were used in the 2008 study.
2.1,2

Catchment Area
The six monitoring locations were chosen to isolate sediment source and storage areas

within the proglacial channel network in order to ascribe the total sediment yield to the
proglacial suspended sediment variables (Warburton 1990; Equation 1). Watershed
boundaries, catchment area, and glacial cover for the monitoring sites were delineated and
computed using a digital elevation model and air photos (Figures 2.1 and 2.2, and Table 2.1).

29
Table 2.1 Catchm ent areas and percent glacial cover for 2011 proglacial stream sam pling sites.
Site
Castle Creek Glacier
PSl
PS2
PS3
MS
RPT
DS

Total Area
(km2)
8.96
1.24E
9.366
1.73E
12.69
2.66
15.68

Glaciated
(km2)
0.14E
7.19E
1.64E
8.96
0.5E
9.46E

Un-glaciated
(km2)
0
1.1
2.17
0.09
3.73
2.16E
6.22E

% Glaciated
100
11
77
95
71
20
60

E - E stim ated a re a

The glacial catchment areas for the three proximal sites were estimated based on the
flow lines and topography of the glacier in air photos, and field observations such as stream
volume; which cause the total and glaciated area to be flagged as estimates. The cirque
glacier/rock glacier in the RPT catchment was estimated based on air photos and field
observations, which cause the glaciated and un-glaciated areas to be flagged as estimates for
RPT. The glaciated and un-glaciated areas at DS were flagged as estimates because the
estimates for RPT were included in the values; however, the effect on the percent glacial
cover at DS would be negligible.
2.1.3

Longitudinal Stream Profile
A longitudinal profile of the main meltwater channel from PS2 to DS was collected to

help identify sediment storage and source areas based on slope (Figure 2.3). The benchmarks
used for vertical control o f the water level loggers (see section 2.3.2) were tied-in during the
longitudinal stream profile to establish relative elevation above the arbitrary datum set 100 m
below the highest benchmark.
From PS2 to the confluence with PS3, the slope was a minimum o f 0.5% and a
maximum of 10%. Below the confluence, the slope increased to 14%, but declined to 7% by
the top of the outwash fan. Four transects were surveyed down the outwash fan (OF)

30
upstream o f MS to capture both the inactive area (elevated from the active channel) and the
active area.

105 ~i

♦ BF LB

100 ?

E
3

□ BF RB
ACB

95 90 -

PS3

ra

V
n
>

XOFT1
XOFT2

♦

♦

O 85 <u
> 80
o
.Q
75 c
o

•

OOFT3
4 OF T4

%
±

70 -

PS1 MS

01 65 UJ 60 -

RPT DS

55
-100

100

300

500

700

900

1100

1300

C han n el D istan ce (m)

Figure 2.3 Longitudinal stream profile for Castle Creek proglacial meltwater channel. BF LB - Bankfull
Left Bank, BF' RB - Bankfull Right Bank, CB - Channel Bottom, OF T1 though T4 - O utwash
F'an Transects I through 4. Datum was arbitrarily set 100 m below the highest benchm ark, and
zero channel distance was set as the outflow o f tlie proglacial lake upstream o f PS2.

The inactive area, OF T4, that began at an abandoned apex was characterized by
larger clasts, the stream slope was 6.0% at the top, but declined to 2.5% by the bottom. The
existing channel had scoured below and around the apex of OF T4 and had a slope of 7%
which gradually declined to where the channel braids. The slope of the braided channels
through OF T l, OF T2, and OF T3 was 2 ± 0.3%. Flow converged at the base of the outwash
fan just upstream of MS. Stream slope remained gentle (1.5 ± 1.0%) for c. 250 m from the
base of the fan, past MS, and then increased (2.5 ± 0.5%) just upstream o f where RPT
entered the main meltwater channel. Stream slope was relatively gentle and consistent (1.5 ±

31
0.3%) to DS, which was the distal site of this study and the end o f the recently exposed (< 60
years) proglacial channel.

2.2

Monitoring and Sampling Strategy
Direct measures of Q and SSC are time consuming and therefore surrogates are

needed to achieve a sample frequency that will capture the level o f detail necessary to assess
the short term variability and rapid fluctuations o f these independent variables (Lawson
1995). Water level (WL) and Tu were recorded at a 5-minute interval as surrogates for Q and
SSC, respectively. Discrete measures o f Q and SSC were collected to develop site-specific
rating relations which were then used to compute 5-minute data sets for Q and SSC. The
product of these time-series is suspended sediment load (SSL):

Q (m3/s) x SSC (mg/L) x 0.3 (kg/5 min) = SSL (kg/5 min)

(2)

Summary computations used SSL in kilograms per 5-minute time step. Suspended
sediment load summaries were divided by catchment area to give suspended sediment yield
(SSY) in units of Mass per Time per Area.
During the 2008 pilot study, the confluence o f PS2 and PS3 was subglacial (Figure
2.2), and the need for an additional monitoring site was not identified prior to arriving at the
field site in 2011. In light of this development, the six monitoring sites were prioritized based
on flow volume, and the available equipment was distributed accordingly (Figure 2.4).

32

▼ Water Level
■

I I
DS

RPT

Pump Sampler
Shared Pump
Sampler

MS

CCG Meltwater
Channels
Tributary

Meltwater
Channel

PS2

PS3

PS1

Proglacial Lake

Figure 2.4 - Schematic diagram o f the 2011 stream m onitoring network at the Castle Creek Glacier.

It was possible to equip all six sites with Tu probes for the duration o f the field
season. However, only five automated pump samplers and four water level loggers were
available. The three sites along the main meltwater stream (PS2, MS, and DS) and the
tributary (RPT) were equipped with pump samplers. The remaining automated pump sampler
was split between PS1 and PS3. After some initial samples were collected at PS1, the
sampler was moved to PS3 for the remainder o f the field season. Water level and stream
gauging was conducted at PS1, PS2, MS, and DS; the Q record for RPT and PS3 was
deduced (see section 2.4.2).

33
2.3

Instrumentation and Data Collection

2.3.1 Meteorological Data
The meteorological conditions o f a given region drive the hydrology of that region,
which includes the accumulation and ablation of glacial ice. Over the ablation season,
proglacial Q can be dominated by snowmelt, ice melt, or precipitation to varying degrees in
response to meteorological conditions. Dr. S.J. Dery and his research group have been
collecting precipitation, air temperature, total solar radiation, wind speed and barometric
pressure data (among other variables) at two sites in the study area (Figure 2.1): 1) in the
proglacial zone at c. 1815m a.s.l.; and 2) on a ridge adjacent to the terminal lobe at c. 2105
m a.s.l. These meteorological data were imperative to the analysis and interpretation of the
fluvial and geomorphological data collected by this study.
The tipping bucket rain gauge at the upper CCG meteorological station was damaged
during the onset of a storm event on Julian Day (JD) 234, and data for this event and the
remainder of the field season were not collected from this site. Unfortunately, the lower
meteorological station was not equipped with a precipitation gauge in 2011. Daily
precipitation records from three nearby meteorological stations (Cariboo Lodge near
Valemount, Environment Canada ID No. 117393; Crescent Spur, Environment Canada ID
No. 1092120; and McBride (upper) snow pillow, BC Ministry of Environment ID No.
1A02P) were collected, weighted by proximity to Castle Creek, and used to estimate the
precipitation record following JD 234. Fortunately, this period was dominated by high
pressure systems (JD 247 - JD 254), and it was only necessary to estimate precipitation for
three days: the storm event on JD 234 (31 mm) that damaged the upper CCG rain gauge, and
moderate precipitation on JD 241 (8 mm) and JD 244 (7 mm).

34
Barometric pressure was collected at both meteorological stations and was needed to
isolate water level from the absolute pressure record collected by the submersible pressure
transducers {section 2.3.2). The lower station was closer in elevation and proximity to stream
monitoring sites, and was therefore more representative. Unfortunately, data from this station
were prone to erroneous spikes and using the data directly would have propagated these
errors to the water level records. Corrections to the barometric pressure record (see section
2.4.1 and Appendix 7.1) were made prior to the computation of water level (Equation 3,
section 2.3.2).

2.3.2

Water Level and Streamflow
Water level (WL) data loggers were fixed vertically in stilling wells using stainless

steel bolts and nylon coated aircraft cable. A metric gauge plate was fixed to the stilling well
and the assembly was fastened with hosed clamps to rebar driven into the streambed at the
sample location. Specific sites for stilling wells were chosen with a suitable downstream
control to provide a stable relationship between WL and Q. During site visits, WL was
recorded from the gauge plate on the outside of the stilling wells at the beginning and end of
Q measurement. Measurement cross sections were selected for having a single channel and
relatively straight velocity vectors at the range of WL suitable for wading, but were not
necessarily adjacent to the stilling well.
Wireless Hobo U20 pressure transducers (Onset Computer Corporation, Bourne, MA,
USA) were used to record WL at the four gauging sites: PS1, PS2, MS, and DS. The data
loggers at PS2, MS and DS had a 9 m range with a 0.002 m resolution (± 0.005 m), the
logger at PS1 had a 4 m range with a 0.001 m resolution (± 0.003 m) (Onset 2013). The
Hobo U20 data loggers record absolute pressure and need to have local (within 2 km)

35
barometric pressure removed to isolate water pressure, which is proportional to water level
(Equation 3):

Abs. Pres. (kPa) - Baro. Pres. (kPa) = Water Pres. (kPa) « Water Level (m)

(3)

This initial computation was performed within the Hoboware Pro software (Onset
Computer Corporation, Bourne, MA, USA); the WL time-series were exported as comma
separated files for processing and Q computations (section 2.4).
Streamflow (Q) measurements were collected using a Swoffer 2100 impeller type
current meter on a top-set wading rod (Swoffer Instruments, Inc., Seattle, WA, USA).
Following the mid-section method described in the Resource Inventory Standards Committee
Hydrometric Manual (RISC 2009), Q measurements take 30 to 40 minutes and ± 5 - 15%
error can be expected, varying with flow conditions (Navratil et al. 2011). During the field
season, 10 Q measurements were made at DS, eight at MS, eight at PS2, and four at PS1;
targeting the wadeable range o f streamflow for each site. The near continuous WL records
were converted to Q in units of volume per time using rating curves that were established for
each site using discrete rating points (see section 2.4.2, and Appendix 7.1).
A key component of WL monitoring is establishing and maintaining vertical control
so that the WL record can be corrected in the event that the vertical reference point changes
during the monitoring period (i.e. the pressure transducer moves) (RISC 2009). The stilling
wells were surveyed-in to local benchmarks (painted points on boulders) at the beginning of

36
the season. Ideally, WL would have been measured from these benchmarks during each field
visit to ensure the reference point had not changed; however, given time constraints and the
amount o f field equipment that was being transported between sites, such a task was
unrealistic. The stilling wells were re-surveyed at the end of the field season, and when there
was an event that obviously affected the vertical reference point.
2.3.3

Turbidity and Suspended Sediment Concentration
Past researchers have found that suspended sediment in proglacial streams tends to be

well mixed, as long as consideration is given to site-specific hydraulics (such as backwater or
recirculating eddies) and to avoid the influence o f upstream tributaries (Gumell et al. 1992;
Richards and Moore 2003). A USGS DH-48 was used to ensure that SSC and Tu sample
locations were representative within the stream cross-section. Once the sites were selected,
rebar was driven into the stream bed and 30 mm pipe was fastened over the rebar. For sites
with a stilling well, the second rebar was driven into the stream bed in a suitable, but slightly
different location.
At each monitoring station, the Tu probe and intake hose for SSC samples were
mounted together on a floating apparatus attached to a piece o f larger pipe (40 mm) that used
the smaller (30 mm) pipe over the rebar as a slide guide (Appendix 7.2). Hoses and wires
were suspended under tension from the top of the rebar to pump samplers and Tu data
loggers located a “safe” distance up the channel bank. Each site was powered by a 12 V
deep-cycle battery charged by a 10 W solar panel. As water level changes, the proportional
depth of the sample point changes, which can lead to uncertainty in the procedure (Navratil et
al. 2011). The floating apparatus kept the intake hose and Tu probe at a set depth from the
water surface; ideally, above the streambed and the effect of coarser sediment transported in

37
saltation. As water level increases, the proportional depth from the surface o f the sample
point within the water column decreases, which was acceptable since suspended sediment
(typically < 2 mm) and wash load (< 0.063 mm) tend to be well mixed within the water
column.
Each o f the six sites was equipped with a Hobo U 12-008 data logger (Onset
Computer Corporation, Bourne, MA, USA) programmed to record DC voltage output from
an Analite 195 Turbidity (Tu) probe at a 5-minute interval. The Analite 195 turbidity probes
(McVan Instruments, Scoresby, Australia) use 90° optics and employ infrared light in
accordance with ISO7027 to measure Nephelometric Turbidity Units (NTU). The range of
the Analite 195 is 0 - 400 NTU, which equates to 0 - 4 V output, and linearity is 1% in the 0
- 1 V range (McVan 2003). The record from the Analite 195 Tu probes was limited by the
range of the Hobo U12 data loggers, which is 0 - 2.5 V with an accuracy of ± 2 mV or ±
2.5% o f absolute reading (Onset 2013). Since SSC was the target variable, converting the Tu
record from V to NTU was not necessary.
Automatic water samplers (ISCO 6700 Teledyne Technologies, Inc., Lincoln, NB,
USA) were deployed to collect physical water samples to measure SSC. The sample interval
and strategy varied from discrete 800 mL samples every 2, 3,4, and 6 hours, to 800 mL daily
composites of a 100 mL intake every 3 hours. The sample frequency was dependent on the
capacity of the field team, and scaled back to daily composite samples once enough (100 150) discrete samples were collected to establish a Tu-SSC relationship for each site. Site
specific relationships were developed for each of the Tu probes from measured SSC in water
samples. These site- and Tu probe-specific relations were then used to calculate a near
continuous record of SSC from the near continuous record o f Tu (section 2.4.3). The daily

38
composite samples collected from JD 218 - JD 254 were used as a check and backup data set
for the calculated SSC record.
Water samples were measured with a graduated cylinder (to determine volume) and
vacuum-filtered in the field through pre-dried and pre-weighed Whatman ashless 8 pm filter
papers, labelled, and stored. Upon returning to the UNBC Landscape Ecology Laboratory,
the samples were unpacked, dried for 24 hours at 105 °C, re-weighed, and SSC was
calculated as a mass per volume (Equation 4):

(Dry sediment and filter (mg) - Dry filter (mg)) / Sample volume (L) = SSC (mg/L)

(4)

The error associated with this gravimetric method has been estimated at c. 4%, but
was likely higher in this study since samples were not dried in the field (Gumell et al. 1992;
Orwin and Smart 2004a). Past researchers have found that there is little (c. 4%) to no
statistical difference between 8 pm filter papers and 4 pm filter papers because the effective
pore size is rapidly reduced as sediment clogs the 8 pm filters (Gumell et al. 1992; Hodgkins
et al. 2003; Orwin and Smart 2004a). Furthermore, during the 2008 study at Castle Creek it
was determined that the use of 8 pm filters underestimated the total flux o f sediment in the
range 63 to 0.45 pm by about 7% (Stott et al. 2009).

2.4

Data Processing and Computations
After reviewing the time-series data, it was identified that JD 195 - JD 254 (July 14 -

September 11, 2011) had the most consistent data coverage, and all data sets were trimmed to

39
this 60 day period prior to computations and analysis (Julian Day calendar included in
Appendix 7.5).
Time-series quality assurance and quality control (QA/QC), rating curve
development, computations and summaries were done using the Aquarius Whiteboard TimeSeries Software (Aquatic Informatics, Inc., Vancouver, BC, Canada). Data summaries were
exported from Aquarius as comma separated files for statistical analyses (see section 2.5) in
Microsoft Office Excel and IBM SPSS Statistics 20.0 (IBM Corporation, Armonk, NY,
USA).
2.4.1

Time-series QA/QC
The raw time-series data sets were reviewed and erroneous data were corrected or

deleted and filled using Aquarius Whiteboard. In some cases the erroneous data were left in

as the best possible information. The following paragraphs describe the quality assurance and
quality control (QA/QC) review and revision of raw time-series data.
The 5-minute barometric data from the lower meteorological station contained several
erroneous spikes (Appendix 7.1), which were removed using an upper rate of change
threshold of 1 mbar per 1 hour (Dery per s. comm. 2011). Missing data were linearly
interpolated and then averaged with the record from the upper meteorological station. The
validity o f the corrected file was confirmed with Dr. SJ Dery and used to compute raw WL
records from the absolute pressure recorded by the four Hobo pressure transducers.
In the 5-minute interval WL data, 28 cases of partial ice-damming in the vicinity of
PS2 were recorded. During these events, ice calving into the proglacial pool became
grounded in the outflow channel and was slowly moved downstream by the force of the

40
water as bed deformation and gradual melting allowed. Stott and Grove (2001) report similar
observations in data from the Skeldal River, Greenland. In the WL data, these events
appeared as a sudden increase that was not apparent at other sites, followed by a period of
stability and then a sudden drop or gradual decline back to the extrapolation o f the timeseries before the sudden increase. These events typically lasted several hours, but less than a
complete diumal cycle and predominantly (68%) occurred on the falling limb o f the diurnal
hydrograph. Some ice calving events were apparent at PS2 and MS as a surge o f water from
ice dropping into the proglacial pool passed though the proglacial channel, but these were not
apparent at DS. In two instances, PS2 remained stable after a sudden increase; this was
interpreted as ice impact and offset corrections were applied to the data following the events.
On JD 233, following a severe ice impact and wave that knocked over the ISCO water
sampler, site PS2 was moved 50 m downstream to a location that was less susceptible to ice
impact, ice-damming and waves from calving ice. The record from the original site was
matched with the record from the new site using benchmarks to maintain vertical control.
In the Tu data, sudden changes in the records following field visits when the lens was
cleaned were minimal, indicating that drift due to bio-fouling was negligible. There were
occasional spikes or troughs in the data that could have been erroneous. However, since SSC
can change drastically in a short period o f time, corrections were only applied for obvious
errors or if justified by field notes, and all other data were left in the time-series. The range o f
the Tu meter was exceeded at PS1, PS3, RPT and DS. These data were left in the time-series
for the analysis as the best available information. The duration, character, effect and resolve
o f Tu exceedances will be elaborated upon in the computational, analysis and discussion
sections that follow (see sections 3.1.3 and 4.1).

41
Once the preliminary QA/QC was completed, a 7-point moving average was used to
smooth the WL and Tu time-series data, keeping the sample interval at 5-minutes. Discrete
measurements of Q and SSC were then paired with WL and Tu values from the smoothed
time-series data. These paired values were used to develop WL - Q rating curves (section
2.4.2) and Tu - SSC regressions (section 2.4.2). In various stages o f the analysis, 5-minute
data, hourly data, and daily data with respective standard statistics (min, max, mean, total,
standard deviation, standardized z-score) were used or used as inputs for further statistical
analysis (section 2.5).
2.4.2 Water Level and Stream/low Rating Curves
At each gauging site, the discrete Q measurements were paired with the mean WL
during the measurement to give a rating point for the development of a rating curve, which
was then used to compute a Q time-series from the 5-minute interval WL time-series. Rating
curves were developed in accordance with the Water Survey of Canada Hydrometric Manual
(WSC 2012). Two rating points were not included in the development of the rating curve for
DS because they fell outside of the acceptable range (± 5%) o f the data from the curve, and
lacked sufficient justification to ‘shift’ the rating curve (WSC 2012). All other rating points
were acceptable. The rating curves used to compute the Q time-series for each gauging site
are presented in Appendix 7.1. The dynamic nature of the proglacial stream means that the
channel profile and control for each site will be prone to frequent ‘shifts’ or changes in the
relation between Q and WL; as such, the rating curves developed for the 2011 season are not
valid for subsequent seasons.
Since high-flow measurements are difficult to obtain because o f safety concerns and
because high-flow events are typically o f a short duration, it is common to extrapolate the

42
rating curve based on the available data (WSC 2012). According to WSC (2012), rating
curve extensions are considered “valid” up to twice the maximum gauged streamflow, and an
“estimate” for flow that is greater than that value. Sites along the main meltwater channel o f
Castle Creek (PS2, MS, and DS) were unsafe to wade when streamflow was greater than c. 5
m3/s. In accordance with WSC (2012), rating curves were developed based on the available
data and then extended linearly in log-log space. Based on WSC (2012), the Castle Creek
rating curve extensions were considered “valid” up to c. 10 m3/s and an “estimate” for flow >
10 m3/s.
Field observations o f channel geometry, stream slope, high water marks, and water
velocity estimates from high-flow events were used as a check during rating curve
development. Channel geometry included breakpoints within the channel and bankfull as
determined by cross-sectional surveys at lower flow. Stream slope was calculated from the
longitudinal profile (see section 2.1.3). Flagging tape on rocks was used to identify the high
water marks, and velocity was estimated by visually tracking and timing floating debris
during high flow.
Two o f the streams (RPT and PS3) were not gauged because additional equipment
was not available. The Q time-series for RPT was calculated as:

RPT Q = DS Q - MS Q

(5)

43
Similarly, the Q time-series for PS3 was calculated as:

PS3 Q = MS Q - (PS2 Q + PS1 Q)

(6)

This method lumps all inflow or outflow between the gauging locations onto one
parameter; the stream being deduced. It strictly assumes that there were no other tributary or
ephemeral channel inputs, and does not account for hyporehic zone or groundwater
interactions. Based on field observations, this assumption was mostly valid with the
exception of storm events that resulted in contribution from the ephemeral channel network,
direct contributions and/or overland flow. Since the stream bed was predominantly
deformable sediment, some interaction with the hyporehic zone can be expected, which
would have varied with flow conditions through the season. However, hyporehic and
groundwater interactions were not quantified.
2.4.3 Turbidity and Suspended Sediment Concentration Regression
For this study, the target parameter was SSC, and Tu probes were field calibrated to
site-specific conditions using discrete SSC samples to develop regression equations.
Automated water samplers were programmed to collect discrete time based {see section
2.3.3) water samples for SSC analysis. The SSC samples were paired with corresponding Tu
readings for each site, and the fourth-spread (or quartile) method (Jacobs and Dinman 2013)
was used to quantitatively exclude outliers from the data set before developing site-specific
Tu-SSC relations. The method assumes that the data were normally distributed and drawn
from a representative population. For each site, the ratio of SSC/Tu was ranked, and the

44
difference between the 75th percentile rank and the 25th percentile rank was the “fourthspread”. The median (50th percentile rank) plus and minus 1.5 of the “fourth-spread” was
used to define the upper and lower limits, respectively, for the data set. The ranked data that
fell outside o f these limits were considered to be outliers, and excluded from further analysis.
A probability plot correlation coefficient (PPCC) was computed for each site and the critical
value (CV) at the 5% significance level was obtained from a PPCC CV table (Filliben and
Devaney 2013) for the given sample size. When the PPCC is greater than the CV, the null
hypothesis that the data came from a population with a normal distribution cannot be
rejected (Filliben 1975; Filliben and Devaney 2013). Summary tables of the fourth-spread
method, normal probability plots and Tu-SSC regression plots can be found in Appendix 7.2.
Once the outliers were removed, the remaining data were used to develop regression
equations (Table 2.2) to compute SSC from the 5-minute interval Tu record.
Table 2.2 Turbidity (Tu) - suspended sedim ent concentration (SSC) regression equations for the 2011
proglacial monitoring sites.

PS1
PS2
PS3
MS
RPT
DS

Equation

95% C.l.

R value

Sample N

SSC = 304.6 • Tu - 54.2

57.6
2.8
9.1
3.6
20.5
5.4

0.43
0.79
0.85
0.67
0.76
0.77

18
156
81
176
175
169

SSC = 252.5 * Tu + 8.6
SSC* 184.0 *Tu - 28.0
SSC = 213.8 * Tu + 8.1
SSC *468.9* To-106.8
SSC = 413.0 * T u - 23.3

The difference in the equations was attributed to site-specific conditions, sediment
source characteristics and the individual characteristics o f the turbidity probes (Navratil et al.
2011; section 1.5.2 and section 2.3.3). The 95% confidence intervals for the regressions show
that there is greater uncertainty with PS1, PS3 and RPT. As is evident from the equations,
estimated SSC could be negative in low turbidity conditions (Table 2.2). Partial days of

45
negative value data occurred at PS1 and PS3 during low flow. These data were removed from
the analysis, and the resulting data gaps were not filled. Missing data due to low flow was
considered acceptable as sediment is predominantly transported by high flow (Pickup 1988).

2.5

Statistical Data Analysis
During the ablation season, glacially influenced hydrologic data have a diurnal

pattern in response to daily temperature cycles. Warm midday temperatures cause increased
snow and ice melt, which slows as temperatures cool overnight. When analysing daily data,
dividing days at midnight tends to include part o f the falling limb of the diurnal hydrograph
from the previous day with the next day, which can influence the analysis of daily data
(Orwin and Smart 2004). To mitigate this problem, the approximate time o f minimum daily
flow was used to divide the data into hydrologic days. This time was specific to this study
and would change depending on proximity o f the study site to the glacier and characteristics
o f the watershed. Based on the observation of peaks in the hydrograph, the transit time from
the proximal site to the distal site in the Castle Creek proglacial study catchment was c. 20
minutes. The daily minimum flow occurred between 06:00 and 09:00, and 06:00 was chosen
to divide the time-series data sets into hydrologic days for the analysis.
The statistical analysis o f this proglacial hydrologic and suspended sediment data
follows the analysis of a similar data set by Orwin and Smart (2004a), which is based on an
objective proglacial hydrograph classification technique developed by Hannah et al. (2000).
The analysis uses two multivariate statistical techniques - principal component analysis
(PCA) and cluster analysis (CA) - to reduce large time-series data sets into categories o f
similar data while maintaining as much of the underlying structure o f the data as possible.

46
Hannah et al. (2000) use the analysis to categorize hydrologic days based on the ‘shape’ and
‘magnitude’ of the hydrograph; a method reapplied successfully by Swift et al. (2005).
Hannah et al. (2000) state that the analysis is applicable to any time-series data with an
underlying cyclic structure, and Orwin and Smart (2004a) expanded the analysis to included
proglacial suspended sediment data, which tends to have a diurnal structure similar to
proglacial streamflow data. Through the analysis they were able to infer controls on the
pattern o f proglacial suspended sediment flux at the Small River Glacier (SRG) using four
separate classification procedures (Figure 2.5).

47

DISCHARGE DRIVING
FORCES

SUSPENDED
SEDIMENT
RESPONSE

SUSPENDED
SEDIMENT
MAGNITUDE

METEOROLOGICAL
PERIOOS

Calculate daily
magnitude
indices

Calculate daily
magnitude indices
Oman>
T ,# ,.
Rainfall^*
SolarRadtou)

SHAPE
5 min suspended
sediment data

(S>nwn»Sp»t il.ShM, .
Smng*. ^Srrto, Smi.

PCA to reduce
data to underlying
structure

PCA to reduce
data to underlying
structure

Plot of PC Scores
to identify nature
of the underlying
shape structure

Slow))
Standardization
of intfces using
z-scores
...._ .......... + . ........

Standardization
of indices using
z-scores

+

CA of indices

CA of indices

CA on PCA
loadings to
identify days of
‘like-shspe'
response

Identification of
‘Kke-magnltude'
days

Identification o(
Ikemeteorological
days

Classification of
"shape" classes

Classification of
"magnitude*
classes

Classification of
days

1I Z

Combined shape and magnitude plot
with discharge, sediment load and
meteorological data

Assessment of driving forces behind
suspended sediment shape and
magnitude response

Figure 2.5 Flow chart detailing the classification procedure used to extract suspended sedim ent transfer
patterns (source: Orwin and Smart 2004a, pg. 1527)

The meteorological, Q, and SSC data were categorized using a combination of PCA,
CA, and interpretation of data matrixes where “cases” refer to rows o f data categories down
the y-axis and “variables” refer to columns o f data categories across the x-axis (Orwin and
Smart 2004a). Mathematically, the data matrix for this PCA must have more rows (cases)
than columns (variables) (Hannah et al. 2000). The next sections describe the statistical
analyses procedures that were performed on the Castle Creek proglacial data set for this

48
project; the analysis protocol of Orwin and Smart (2004a; Figure 2.5) was followed in order
to generate comparable results. The PCA and CA analyses were run in IBM SPSS version
20 . 0 .

2.5.1

Meteorological Periods
A CA was run on meteorological data to group the daily data into categories of

similar conditions. The CA o f meteorological data included cases o f daily values for the
variables: mean, maximum, and minimum air temperature; total precipitation; mean relative
humidity; total solar radiation; and mean wind speed. The data were standardized (z-scored)
and the CA was run using Ward’s Method (Tabachnick and Fidell 1989). An agglomeration
dendrogram was plotted and used to determine the number of meaningful clusters within the
data. The raw data within each cluster were reviewed, and descriptive titles (i.e. ‘hot and
dry’, ‘warm and damp’, ‘cold and wet’ or ‘storm’) were assigned, which were broadly similar
to those assigned by Orwin and Smart (2004).
2.5.2

Streamflow Driving Factor
To determine the main driving forces of streamflow (i.e. from glacial meltwater or

precipitation), the input matrix for PCA had daily average Q for each site, total precipitation,
and solar radiation, average wind speed, and air temperature minimum, maximum and mean
as variables, and hydrologic days as cases. The PCA was run using a VARIMAX orthogonal
rotation with standard retention criteria. Low communality variables were removed from the
analysis and the PCA was re-run on the remaining variables. The Kaiser-Mayer-Olkin
measure o f sampling adequacy (Tabachnick and Fidell 1989) was used to assess the
correlation matrix and suitability o f the data set for PCA. Parallel analysis was used to
identify the statistically significant eigenvalue for the data (O’Connor 2000). Components

49
with significant eigenvalues were retained and used to assess the driving factors of
streamflow and the proportion of variance in the data explained by each component. A bi­
plot of the two dominant components was generated to assess the driving factors of
streamflow and descriptive titles (i.e. ‘ablation’ or ‘rainfall’) were assigned after assessing
the data explained by the component.

2.5.3

Suspended Sediment Response Shape
To assess the underlying suspended sediment response shape, an independent PCA

was run on a data matrix with hydrologic days as variables and a 5-minute time step as cases
for SSC data at each site. The PCA was run using a VARIMAX orthogonal rotation with
standard retention criteria. Parallel analysis was used to identify the statistically significant
eigenvalue for the data (O’Connor 2000). For each site, a scree plot was generated to confirm
the break point in the principal components, and that the components with eigenvalues > 1
were retained. Principal component loading scores were plotted against time to reveal the
underlying shape of the 5-minute SSC data for each site.
Days with similar suspended sediment response shape were identified by running a
hierarchical CA on the principal component loading scores using Ward’s Method.
Observations were standardized (z-scored) to remove major variations in SSC magnitude.
Low communality variables were removed and an agglomeration dendrogram was plotted to
visually identify the number of clusters. The shape structure o f the raw data in the clusters
was examined and appropriate titles (i.e. ‘diurnal’ or ‘irregular’) were assigned.

50
2.5.4

Suspended Sediment Response Magnitude
The magnitude classification o f daily suspended sediment response shape was

determined by running a CA on a data matrix with daily SSC mean, minimum, maximum,
range, standard deviation (in mg/L) and daily total SSL (in kg/day) as variables, with
hydrologic days as cases for each site. Data were standardized (z-scored) prior to running the
CA using Ward’s Method, and an agglomeration dendrogram was plotted to visually identify
the number o f clusters. The magnitude structure of the raw data in the clusters was examined
and appropriate titles (i.e. ‘low’, ‘medium’ or ‘high’) were assigned.

51

3

Results and Discussion (1) - Spatial and Temporal Patterns of
Suspended Sediment
In this chapter, objective 1 is addressed by examining the spatial and temporal

patterns of suspended sediment flux in response to hydro-meteorological conditions in the
proglacial zone. Principal component analysis (PCA) and cluster analysis (CA) are used to
categorize and summarize the 5-minute time-series SSC and Q data while maintaining as
much o f the underlying structure and response pattern as possible.

3.1

Data Sets
The 5-minute time-series data for streamflow (Q) and suspended sediment

concentration (SSC) are presented in Figure 3.1, and air temperature (AT) and precipitation
(PT) are presented in Figure 3.2; these figures will be a useful reference for the field season
summary in section 3.3. The Aquarius software that was used to produce the figures does not
use Julian Days; refer to Appendix 7.5 for a JD calendar.

52

PS1 - Q and SSC 5min Data
• Q (mA3 /s)

SSC (mg/L)

800

1,2

r W

.......

3
%

17/07/2011

24/07/2011

31/07/2011

07/08/2011

14/08/2011

21/08/2011

28/08/2011

UTC-08:00

PS2 - Q and SSC 5 min Data
. Q (m*3/s)

SSC (mg/L)

14

800

12

10
400
8

3

6

*
,

200

4

2

0

0
17/07/2011

24/07/2011

31/07/2011

07/08/2011

14/04/2011

21/08/2011

28/06/2011

044)9/2011

UTC-08:00

PS3 - Q and SSC 5 min Data
• Q (mA3 /s)

SSC (mg/L)

3

3

17/07/2011

24/07/2011

314)7/2011

07 /0 8 0 0 1 1

14/08/2011

l/TC -0800

21/08/2011

284)8/2011

044)8/2011

53

MS - Q and SSC 5 min Data
- Q (mA3 /s)

SSC (mg/L)

20

400
15

10
200

SSC (mg/L)

(V€v«*)0

300

5

100

0
17/07/2011

24/07/2011

31/07/2011

07/04/2011

14/04/2011

21/04/2011

24/04/2011

04/04/2011

11/04/2011

UTC-OfrOO

RPT - Q and SSC 5 min Data
I

•Q

SSC (mg/L)

10

1200

0

1000

{ 800

600
4

SSC (mg/L)

(* /t.“0 O

o

400

2

200

0
11/04/2011

31/07/2011

UTC-08:00

DS - Q and SSC 5min Data
• Q (m A3 /s )

S S C (m g /L )

1000

20

800

400

b 200

0

0
17/07/2011

24/07/2011

31/07/2011

07/04/2011

14/08/2011

UTC-08:00

21/04/2011

24/04/2011

04/09/2011

11/09/2011

SSC (mg/L)

(VC^u)O

800

54
Figure 3.1 Streamflow <Q) and suspended sedim ent concentration (SSC) time-series (5-m inute data
interval) from six proglacial m onitoring sites, JD 195 - JD 254, 2011 after Q A /Q C Scale o f y-axis
varies according to range o f data. Figure presented over two preceding pages. Exceedances in the
SSC tim e-series are described in section 3.1.3.

AT Lower CCG (*C)

13/ 07/2011

23 / 07/2011

02 / 08/2011

12/ 08/2011

22/ 08/2011

01 / 09/2011

11 / 09/2011

Figure 3.2 Hourly air tem perature (AT, °C) from Lower Castle Creek Glacier m eteorological station;
hourly precipitation (PT, mm) from Upper Castle Creek Glacier m eteorological station, estimated
daily total precipitation after August 21, 2011.

3.1.1

Field Season Overview
The 2011 field season at CCG captured a mixture of conditions that generated a

complex hydrologic and geomorphic response pattern. The following general overview uses
category names that are described and summarized in detail subsequently (sections 3.2 and
3.5). The first third of the field season, JD 195 - JD 214, switched between ‘cold and wet’
and ‘hot and dry’ conditions every few days, transitioning through ‘warm and damp’
conditions in the process and included a ‘storm’ event on JD 211 which delivered c. 30 mm
of precipitation over 24 hours. Discharge and SSC responded to this event at all sites.
Following JD 214, there was a period of ‘hot and dry’ conditions where Q and SSC
data followed a diurnal pattern at most o f the sites. During this time, SSC data at PS3

55
exceeded the range of the Tu meter. These data were likely erroneous as a result o f influence
from the stream bed since the effect of a high SSC input was not discemable at downstream
monitoring locations and the SSC at the other sites was consistent with low-flow conditions.
‘Cold and wet’ conditions returned on JD 226 for four days, generating a small ‘irregular’
increase in Q and SSC data at most sites which was followed by low Q and SSC.
The SSC exceedances from JD 215 - JD 225 at RPT may have been the result of low
water allowing the sample point to be influenced by material being transported in saltation
near the stream bed. Unfortunately, the field team was not on site during this period. On JD
227 the field team returned, and found the RPT sample point close to the stream bed,
sustained Tu range exceedances recorded by the logger, and a high amount of sediment in the
water samples; the sample location was moved to a deeper location on JD 228. However, on
JD 227, the water level was lower than it had been in the time the field team was absent and
the Tu was within the range o f the sensor, thus, the data from JD 215 - JD225 was accepted
and computed as exceedances, rather than removed as erroneous.
Warm weather arrived on JD 231 which was followed by a ‘storm event’ on JD 234.
The warm weather that preceded the event caused meltwater Q to be high; c. 6 m3/s at DS
over the 24 hour period before the storm began. The tipping bucket rain gauge was damaged
during the onset of the storm; however, an estimated 31 mm of precipitation (see section
2.3.1) was delivered by four intense squalls over a six hour period, starting at 16:00 on JD
234. This storm event caused the highest Q and SSC during the field season at all sites. Peak
flows at DS were estimated at c. 24.7 m3/s at 21:45 PST and SSC exceeded the range o f the
Tu meter for nearly three hours at the peak o f the event. Five grab samples were collected
over a 2.25 hour period (19:30 - 21:45) as the event peaked. Two were collected from

56
ephemeral channels that drained directly into Castle Creek, one upstream o f RPT, and one
downstream of RPT; they were measured at c. 3600 mg/L and c. 2000 mg/L respectively.
The grab sample from RPT was measured at c. 4200 mg/L; and the main flow o f Castle
Creek upstream and downstream of RPT was measured at c. 1300 mg/L and c. 2700 mg/L,
respectively. These peak values are comparable to those observed by other researchers
(Gumell et al. 1996; Hodson et al. 1998), but much less than the 12000 mg/L reported by
Orwin and Smart (2004a).
‘Hot and dry’ conditions followed the ‘storm event’ on JD 234, but Q and SSC data
were ‘irregular’ at most sites for a few days as storm flows subsided. Cold air temperatures
from JD 242 - JD 245 caused Q and SSC to be, generally, low. On JD 241, the estimated 8
mm of precipitation was apparent at all of the monitoring sites in the water level and
suspended sediment records. The estimated 7 mm of precipitation on JD 244 increased Q at
PS1 and RPT, but a corresponding increase in SSC was not recorded for the event at these
two sites. This event was evident at DS, but became more muted at sites further upstream
along the main channel and for PS3. The dominance of ice cover in the catchment o f the
main Castle Creek channel and PS3 will make those sites less responsive to low intensity rain
events than RPT and PS1, which have a higher proportion of bedrock and less ice cover in
their catchment area.
Following JD 246, ‘hot and dry’ conditions began, and persisted until the end o f the
field season, generating low magnitude, but consistent ‘diurnal’ Q and SSC patterns. The
minimum flow for the field season at DS o f 1.37 m3/s occurred from 09:25 - 10:10 PST on
JD 247. Following this minimum, warm weather generated meltwater and increased Q and
SSC.

57
As the field season progressed the diurnal peaks in the hydrograph at DS became
more pronounced. The time of the daily minimum did not appear to change markedly, but the
time from minimum to peak became progressively shorter; the daily peak in the hydrograph
occurred at c. 18:00 at the start of the field season, c. 17:00 in the middle, and c. 16:00 by the
end of the field season. This trend was likely a result o f increasing dominance o f ice melt
over snowmelt in meltwater as the annual snow pack thins, reducing albedo and meltwater
travel time, as well as the seasonal evolution o f meltwater channels and flow paths within the
glacier (Swift et al. 2005; Jobard and Dzikowski 2006; Haritashya et al. 2010). The sediment
peak tended to occur on the falling limb o f the hydrograph, especially during the warm sunny
weather at the end o f the field season.
3.1.2

Streamflow Considerations
The percent o f data at DS that exceeded the maximum gauged flow (5 m3/s) was

13%. While the uppermost 0.6% of the data exceeded the limit of “valid” extension (i.e. 10
m3/s), and were considered an “estimate”, the 12.4% of the data that were between 5 and 10
m3/s can be considered “valid” based on the rating curve extension (WSC 2012). Estimated
data occurred during the peak o f both high flow events at DS. PS2 and MS data were similar;
however, there were fewer estimated data points for these sites.
‘Offset shifts’ were applied to the rating curves for PS2 and MS part way through the
field season based on changes to the WL reference point and Q measurements made after the
event (see Appendix 7.1). On JD 233, an ice calving event knocked over the water sampler at
PS2 and bent the stilling well over. The site was moved on the following day in the hours
before the onset of the main high flow event for the 2011 season on JD 234. The data
between the ice calving event and moving the site were corrected in the time-series rather

58
than applying a temporary shift to the rating curve. The shift that was applied to PS2 was
determined by rating points collected after moving the site, and was confirmed with survey
notes from the established bench marks. The stilling well at MS was bent over (c. 50°) during
the high-flow event on JD 234. An offset shift was applied to data from the peak o f the event
and thereafter based on rating points collected after the event; again, survey notes were used
to confirm the magnitude o f the shift. There were no shifts applied to the rating curves for DS
or PS1, and survey notes indicated a negligible change in WL reference point through the
field season. Control at DS was the bedrock gorge; control at PS1 was maintained by large
boulders. Section control at MS and PS2 may have been affected by scour and aggradation
along the channel during the field season, especially during high-flow events. Unfortunately,
both of these sites had a change in the WL reference point near the high-flow event and the
scope and amount of data collected following the event permits only speculation on the
influence o f these processes.
The discharge record for two o f the six sites was deduced from the gauged sites
(Equations 5 and 6, section 2.4.2). Ideally, sites RPT and PS3 would have had their own
water level loggers, but since additional equipment was not available, this deductive method
was the next best option. Based on field observations during the study period, other channel
inputs were negligible except during storm events, when precipitation rate exceeds
infiltration rate and overland flow converges in ephemeral channels. During these isolated
events, the amount of flow entering the system from these channels was variable and difficult
to estimate, but likely less than 5% o f the flow in the monitored stream network.

59
3.1.3

Suspended Sediment Concentration Considerations
The fourth-spread method was used as a quantitative basis to exclude outliers from

the analysis before developing Tu-SSC regression equations (section 2.4.3). The ISCO pump
samplers have 24 one liter bottles, thus, a full set o f discrete water samples was 24; however,
sample sets were not always foil as a result o f field operational procedures. Four SSC sample
sets were subject to corrosion during sample storage, and all plotted as outliers in the T u SSC data for PS2, PS3, and MS (Table 3.1).
Table 3.1 Paired turbidity (Tu) and suspended sedim ent concentration (SSC) sample sum m ary, and
results o f probability plot correlation coefficient (PPCC) fourth-spread method null hypothesis
test; ‘YES' means the sam ples w ere drawn from a population with a normal distribution.
PS1
Tu-SSC Samples
Outliers (Corroded)

n/a

PS2

PS3

MS

RPT

OS

205

110

- 202

179

183

47

22

23

n/a

n/a

2

6

3

4

14

Outliers

;6

Actual Sample (N)

18

156

82

176

175

169

Fail to reject H,

YES

YES

YES

YES

NO

YES

■

There were relatively few outliers in the remaining data (Table 3.1); between 2% and
8% of the samples for all sites with the exception of PS1. PS1 has a low number of paired
Tu-SSC samples because of equipment limitation and the necessity to prioritize available
equipment (section 2.2).
Only four outliers were removed from the RPT Tu-SSC data set, but since the PPCC
was less than the CV at the 5% significance level for the sample number, the null hypothesis
that the data were drawn from a population with a normal distribution was rejected (Filliben
and Devaney 2013). Therefore, the quantitative method o f excluding outliers from this data
set was not statistically sound. A plot o f the raw Tu-SSC data can be found in Appendix 7.2;

60
it is obvious that the RPT data are not normally distributed, which may be a result o f multiple
sediment sources in that catchment or errors during sample collection. Given the uncertainty
in the RPT data, the fourth spread method of excluding outliers from the Tu - SSC data was
not modified for this site. The PPCC for the other five sites was greater than the CV at the
5% significance level for their sample number, and so the null hypothesis cannot be rejected.
A detailed table on the fourth-spread method and normal probability plots can be found in
Appendix 7.2. To maintain consistency in the analysis, linear regressions were used to
compute 5-minute SSC time-series from Tu for all sites (Table 2.2, section 2.4.3).
Three sites had a substantial amount o f data that exceeded the range of the turbidity
monitoring instrumentation (Figure 3.1 and Table 3.2). The difference in the maximum SSC
value as determined by the regression equation for each site reflect both site specific
conditions and the response of the individual Tu probe. Maximum SSC values were included
in the Tu-SSC analysis as a better option than omitting Tu monitoring range exceedances
(Table 3.2).
Table 3.2 Turbidity (Tu) data summary for the 20.11 proglacial monitoring sites.
Data Record
Total days with data record
NO. of days JD195 »JD254
Partial days
Partial day exceedance *
Missing days
Full day exceedances *
I Useable days within JD195-JD254
%of record useable
Number of 5min data points
Number of 5mbt exceedances
%of data within Tu range
MsxSSCfmg/l}*
Superscript key: * la te s ta rt o f d a ta collection;
reg ressio n e q u atio n .

Site
PS1
60
58
10
48
2a
4
58a
97
16134
6332
61
707.1

PS2
63
60
3b
_

.

lb
■

56b
89
16534
0
100
640.0

PS3
62
60
2b
13
3b
2C
53 bc
85
15567c
1707
89
427.4

low w a ter; ‘ tw o days of e rro n e o u s d a ta excluded;

MS
62
60

RPT
63
60

DS
64
60

-

-

-, ~ :
60
97
17280
0
100
542.6

22

-

1

-

-

2
60
95
17280
2864
83
1065.1

>
60
94
17280
33
100
1009.1

m ax SSC value as c o m p u te d by

61
The amount of Tu data exceedances at DS was less than three hours in total (0.2%).
This occurred during a storm peak on JD 234. A grab sample at DS on JD 234 at 19:30 had a
SSC of 2736 mg/L, which was c. 1 hour before the peak on the hydrograph. PS1 and PS3
were much smaller streams than the main stem o f Castle Creek, and thus the effect of Tu
exceedances at these sites on the overall sediment budget was somewhat limited. The
tributary RPT entered Castle Creek between MS and DS. This stream had a high sediment
load and strong influence on the suspended sediment data at DS; therefore, the contribution
of sediment from RPT was needed to determine the sediment budget in the reach between
MS and DS. For consistency in the PCA and CA analyses across the sites {section 2.5 and
3.2) exceedances (i.e. maximum computed SSC (Table 3.2)) were left in the SSC time-series
and included in hydrologic daily averages. However, for the computation o f suspended
sediment load and the suspended sediment budget {Chapter 4), exceedances in the RPT timeseries were estimated.
Discrete and composite SSC samples from RPT were used along with the SSC record
from MS and DS to estimate exceedances in the RPT SSC data between JD 215 and JD 236.
The time-series was adjusted to fit with the best information available using multipoint
corrections in Aquarius. The difference in suspended sediment load between MS and DS was
used for reference, but the assumption that all additional SSL in the reach MS-DS was from
RPT was false under certain conditions. The Tu exceedances from JD 215 - JD 225 at RPT
were not well supported by independent samples and could be erroneously high as a result of
stream aggradation. It was not possible to quantify or correct this potential error, and more
confidence can be placed in the RPT estimates after JD 228 because of the new sample
location and additional grab samples.

62
3.1.4

Error and Uncertainty
Navratil et al. (2011) assessed nine different uncertainty components in the Tu

approach to SSC monitoring using Monte Carlo simulations and found uncertainty associated
with automatic pump samplers, stream discharge measurements, and Tu fluctuations at short
time-scales to be the greatest limitations. They also identified technical limitations o f Tu
meters, Tu and water level sample frequency, representativeness of the SSC sample point
within the cross section over the range o f flow conditions, varying contribution from
different sediment source areas, technical field problems, field sampling and laboratory
procedures, and the calibration of the Tu-SSC relationship as other sources of uncertainty
(Navratil et al. 2011). These uncertainty components may be correlated with one another and
propagate through analytical computations.
The error in Q data is expected to be 5 - 15% depending on flow (Navratil et al.
2011). The cumulative error in Tu-SSC data is expected to be similar (Gumell et al. 1992;
Orwin and Smart 2004a; Stott et al. 2009), but would vary with SSC and Q conditions
(Richards and Moore 2003; Navratil et al. 2011). The 95% confidence intervals are reported
for the Tu-SSC regressions in Table 2.2 (section 2.4.3), and it is clear that the sites along the
main channel produced better data than the smaller and tributary streams. Thus, the sum total
o f the error for the computed SSL was estimated to be between 15 and 30% of the reported
values for the three sites (PS2, MS, and DS) along the main CCG meltwater channel, but
would likely be higher for the three sites (PS1, PS3, and RPT) where monitoring was
compromised due to limited equipment, equipment limitations, or equipment failure. The
precise values reported in this study include this error and uncertainty, which is similar to the
26% error in the proglacial sediment budget by Warburton (1990).

63
3.2

Principal Component Analysis (PCA) and Cluster Analysis (CA)

3.2.1 Meteorological Periods - Cluster Analysis
The CA of meteorological data allowed the field season to be divided into four
categories that, upon reviewing the raw data within the category, were described based on
precipitation and air temperature conditions. Those categories and the percent o f the field
season that they represented were: ‘Cold and Wet’ (17/60 days, 28%), ‘Warm and Damp’
(15/60 days, 25%), ‘Hot and Dry’ (26/60 days, 43%), and ‘Storm’ (2/60 days, 3%). These
categories were used for comparison o f streamflow and suspended sediment response under
different meteorological conditions (section 3.3). Table 3.3 summarizes the field data that
were used in the meteorological analysis; a similar table with additional parameters is
included in Appendix 7.3. Mean daily air temperature and precipitation have been presented
as a time-series in Figure 3.6 following the summary results o f the PCA and CA, Figure 3.5
{section 3.3)
Table 3.3 Summary of m eteorological data from upper and lower m eteorological stations for JD 195 JD 254, 2011. Four clusters o f similar m eteorological data have been assigned descriptive titles
based on air tem perature and precipitation.
Meteorological Station:
Parameter
Meteorolaglcat
Cluster

Lower

Lower

Lower

U pper

Lower

Lower

U pper

AT min
(*C)

AT max
CC)

AT mean
CC)

PT mean
(mm/day)

Ret.
Hum. (K)

W5pd.
mean
(m/s)
4.1

Cold and Wet
(17 days)

Average

2.0

7.3

4.8

8.0

84.4

T.Sol.
Rad.
(W/m*)
14553

Std.Dev.

1.4

1.3

1.2

5.9

4.3

5133.2

1.1

Warm and
Damp
(15 days)
Hot and Dry
(26 days)

Average
Std.Dev.

4.3
1.3

10.7
1.S

7.9
1.2

2.5
2.5

72.5
7.9

18195
3657.9

3.6
1.0

Average

7.4

14.7

10.9

1.2

58.6

22041

3.6

Std.Dev.

1.2

1.2

1.0

2.0

7.2

3717.6

1.0

Storm
(2 days)

Average
Std.Dev.

5.1
0.6

12.7
1.9

8.9
0.5

31.1
0.1

79.0
5.1

14243
6241,7

5.1
0.8

64
Nine of the ‘hot and dry’ days occurred in early September when the approaching
autumnal equinox limited the amount of daily insolation and the potential for ablation.
Additionally, by this point in the field season, the annual snowpack had mostly retreated
from the proglacial zone and ablation zone of the glacier, leaving primarily ice melt to
augment streamflow. Had this ‘hot and dry’ weather occurred earlier in the field season when
the days were longer and annual snowpack was still present, the Q and SSC response could
have been much different. Without these nine days in the data set, the field season was nearly
balanced between the three main categories o f meteorological conditions.
Orwin and Smart (2004a) used similar titles to describe the meteorological conditions
during the 2000 field season at the Small River Glacier (SRG). However, since their
monitoring period started 7 days earlier and the proglacial zone o f the SRG has a greater
elevation range and is steeper than the CCG, they found 16 days fell into a ‘snowmelt’
category. Comparing all meteorological categories between the two data sets, relative
humidity and solar radiation were higher, and wind speed was much higher at the CCG in
2011 than at the SRG in 2000. Air temperatures were similar during ‘cold and wet’ periods,
slightly higher at the CCG during ‘hot and dry’ periods and slightly lower at the CCG during
‘storm events’. Interestingly, meteorological parameters during ‘warm and damp’ conditions
at the CCG in 2011 were similar to ‘snowmelt’ conditions at the SRG 2000. ‘Warm and
damp’ days at the CCG were distributed through the 2011 field season, and, based on the title
of the category, it is assumed that ‘snowmelt’ days occurred early in the field season at the
SRG; however, Orwin and Smart (2004a) do not specify when the ‘snowmelt’ days occurred
in their study,

65
3.2.2

Stream/low Driving Factor - Principal Component Analysis
The PCA of Q and meteorological conditions reduced the data to its underlying

components. The two dominant eigenvalues > 1 were used to generate a bi-plot, and
descriptive titles were assigned (Figure 3.3). The Kaiser-Meyer-Olkin (KMO) measure of
sampling adequacy index for the correlation matrix was 0.532 which indicated that the PCA
was a suitable analysis. As a rule o f thumb, if the KMO is > 0.5, PCA is a suitable analysis
(Tabachnick and Fidell 1989).

! AT_M«*
A T _ M e « n T A T Min

S .R ad_Sum

O

PS2
° 0 DS
_MS O

c

o

.. W S R D J tfe w ..

O

n
<

0.0

Rainfall

Figure 3.3 Principal com ponent loading o f daily meteorological and streamflow (Q) variables on
principal com ponent one and two explained 42% (P C I) and 30% (PC2) o f the total variance in the
principal com ponent analysis. Distance o f the variable from the origin indicates relative
dom inance o f the Q generating processes; PCI and PC2 were interpreted as •Rainfall’ and
‘Ablation*, and have been titled respectively in the figure.

The two components that were retained from the analysis explained 72% of the total
variance in Q data (Figure 3.3). The first component was interpreted as ‘rainfall’ or stormy
conditions and explained 42% of the variance in Q. The second component was interpreted as

66
‘ablation’ and explained 30% of the variance in Q. Distance from the origin (0.0, 0.0) was
interpreted as dominance of the driving factor on Q pattern for the site. Orwin and Smart
(2004a) found that the two component solution explained 77% of the total variability in the
data from the 2000 field season at the SRG; 55% was attributed to ‘ablation’, and 22% was
attributed to ‘rainfall’. In the CCG analysis the days that were represented by ‘rainfall’ were
not necessarily days with substantial precipitation, they may have just not scored as
‘ablation’ driven days because they were overcast, cool and/or windy; thus, stormy
conditions may be an equally applicable title for the component. The greater influence o f the
‘rainfall’ component on Q in the CCG analysis may be a result o f the later field season (JD
194 - JD 254 at CCG vs. JD 188 - JD 238 at SRG), and thus a lower influence of annual
snowmelt ablation in the 2011 Q data at CCG than in the 2000 Q data at SRG (Orwin and
Smart 2004a).
In general, all of the sites plot strongly positive on the ‘rainfall’ axis, but show less
variation from the origin on the ‘ablation’ axis. Sites PS2, MS, and DS were along the main
stem of the Castle Creek meltwater channel, and all plot close together, and were strongly
influenced by ‘rainfall’ and moderately influenced by ‘ablation’ for the 2011 data set. As the
distance from the glacier increased, the influence o f ‘ablation’ on Q patterns decreased and
the influence of ‘rainfall’ increased; which was consistent with the results of Orwin and
Smart (2004a). Although the PS3 catchment had the greatest percent glacial cover, it was less
influenced by ‘ablation’ and ‘rainfall’ than the sites along the main channel, which suggests a
more stable source of flow from deeper within the glacier than the active ablation zone. Two
sites, PS1 and RPT, plot negatively on the ‘ablation’ axis. For PS1, this was interpreted as a
stronger influence of ‘rainfall’ on Q than ‘ablation’ due to the small proportion o f glaciated

67
catchment area. For RPT, the negative ablation response could be due to independent timing
of ablation generated peaks, a muted response from the cirque glacier/rock glacier, or a small
proportion of glacial cover.
The precipitation variable plotted positively on the ‘rainfall’ axis and negatively on
the ‘ablation’ axis while solar radiation plotted positively on the ‘ablation’ axis and
negatively on the ‘rainfall’ axis; which was interpreted as ablation was generated by sunny
days, and cloudy days generated precipitation. The air temperature variables indicate a strong
positive relation on the ‘ablation’ axis and near neutral on the ‘rainfall’ axis; which was
interpreted as warm weather generated ablation, and rainy weather was not necessarily cool.
The wind speed variable plots neutral on the ‘ablation’ axis, which could be a result o f net
balance in the data, rather than no effect, and positive on the ‘rainfall’ axis, indicating that
wind speed increased during rainy or stormy weather. The trend of katabatic winds would
have been more strongly observed at the lower meteorological station, this trend was muted
by averaging the wind speed data from the upper and lower meteorological stations. Wind
speed from the lower meteorological station alone would have likely plotted more positively
on the ‘ablation’ axis.
3.2.3

Suspended Sediment Response Shape - Principal Component Analysis and Cluster
Analysis
Three components were retained in the PCA that was run on the 5-minute SSC data

for each site. Principal loading scores were generated and plotted against time to reveal the
underlying shape of the components (Figure 3.4). Time on the x-axis is reported in decimal
days counting up from zero, and data are reported for the hydrologic day (06:00-06:00). For
instance, the first sample is at 06:00, which is 6/24, or 0.25 of a day.

68

PSl

MS

0.6

0.8

1.0

1.2

0.8

w y
P C I (49% )

PC2 (1 8 % )

P C I (32% )

PC3 (9% )

PS2

PC2 (2 0 % )

- PC 3 (8% )

PC2 ( 2 2 % ) ---------

- PC3 (8% )

RPT

0.4

P C I (4 1 % )

PC 2 (1 3 % ) — — — - P C 3 (7% )

P C I (2 6 % )

PS3

0.8

0.8

P C I (32% )

PC2 (2 6 % ) -

— -

- PC 3 (11% )

P C I (34% )

P C 2 (28% )

— — --P C 3 (8 % )

Figure 3.4 Principal com ponent loading score plots for 5-m inute SSC data from each gauging station: all
full hydrologic days of data were retained as variables for the analysis. Percent o f the data
represented by each principal com ponent is reported for each site. Time, on the x-axis, is reported
in arbitrary decimal days (06:00 is 0.25 o f the way through a regular day).

For this analysis, it was necessary to exclude partial days of data, but data
exceedances were left in using the maximum value (refer to section 3.1.3, Table 3.2). The
percent o f the data that were represented by each principal component (PC) is reported for
each site in Figure 3.4. Since the analyses were run independently for each site, the shape of
the PC was not necessarily comparable across sites and the days that were represented by
PCI at one site may not be represented by PCI at another site. PCI and PC2 represented an
average of 36% and 21 % of the data, PC3 represented an average o f 8.5% o f the data, and an

69
average of 35% of the data was not represented by the any o f the three principal components.
All three principal components appeared to have a relatively well defined pattern for PS1,
PS3, MS, RPT and DS. The PC3 pattern appeared to be more stochastic for PS2, and appears
to be double peaked at PS3, MS, RPT and DS. The results presented by Orwin and Smart
(2004a) were similar: PCI, PC2 and PC3 represented an average o f 37%, 20%, and 10% of
their suspended sediment data, respectively, and PC3 also showed a more irregular response
pattern.
A CA was run on the principal component loading scores, and the two cluster solution
categorized days as either ‘diurnal’ or ‘irregular’ SSC response shape. Comparisons of the
CA results with those from the regression score loading plots (Figure 3.4) confirmed that
PCI and PC2 roughly represented the ‘diurnal’ data as a percentage; c. 60% at DS. Most of
the data represented by PC3 at PS1, PS3 and MS may be included with the ‘diurnal’ data.
However, based on percentage, the data represented by PC3 at PS2, RPT and DS appeared to
be categorized as ‘irregular’ data following the cluster analysis. The data that were not
represented by any o f the three principal components in the first part of the analysis were
categorized as ‘irregular’. Similarly, Orwin and Smart (2004a) found that, on average, 75%
of their data were categorized as ‘diurnal’ following the CA.
3.2.4

Suspended Sediment Response Magnitude - Cluster Analysis
The CA of the SSC magnitude parameters (see section 2.5.4) separated the daily data

into ‘high’, ‘medium’ and Tow’ categories, and was a useful tool for looking at how the
magnitude o f SSC changed over the field season at a particular site. Sites MS and DS were
dominated by Tow’ magnitude response days; PS2 was dominated by ‘medium’ and Tow’
response days; PS3 was split across the three magnitude categories; RPT had more ‘high’ and

70
‘low’ magnitude days than ‘medium’ days; and, PS1 had ‘high’ and ‘medium’ magnitude
response days. Orwin and Smart (2004a) report that on average 80% o f their data fell into the
‘low’ magnitude category and 20% fell into the ‘high’ magnitude category.
Since each CA was independent from the other sites, the scale o f the magnitude
analysis varied, which limited the ability to compare the results of this analysis across sites.
The mean daily SSC for a ‘high’ magnitude day at PS2, MS, and DS was 195 mg/L, 336
mg/L, and 449 mg/L, respectively (Table 3.4), while ‘medium’ magnitude days were 112
mg/L, 161 mg/L, and 238 mg/L, respectively. A given magnitude classification will have a
different value for each site because o f the differences in the SSC time-series. For example,
the number of ‘medium’ and ‘high’ magnitude response days at PS2 was greater than at MS,
which was a result of lower peak sediment loads at PS2 which allowed the scale o f the
analysis to be focused on a smaller range than at MS. Essentially, a small number of very
high SSC data could stretch the scale so that the majority of the data fall into a lower
magnitude category; in which case, the detail of the time-series data could become lost or
obscured. Differences in scale between the sites were also reported for this analysis on the
data from the SRG by Orwin and Smart (2004a), which they attributed to sediment
availability in the contributing catchment area. The analysis could easily be misinterpreted;
however, the scale o f the SSC data and the results of the magnitude analysis became
somewhat o f a moot point when SSL was calculated (Chapter 4).

71
Table 3.4 Summary o f suspended sediment response magnitude param eters and cluster aualysis results.
Values com puted from daily data. Standard deviation is reported in parentheses.
Site

PS1

PS2

SSC

Avg.

Avg.

Mean

Magnitude
Class

SSC.*
(mg/L)

SSCM
(mg/L)

High

43 3 (1 6 8 )

7 0 7 (0 )

64 4 (6 3 )

MS

DS

Avg.

s s c ..

ssc u .

s s c * * ,.

s td . s s c * .

SSu*

(mg/L)

(mg/L)

(mg/L)

(ratio)

(kg/Day)

284(168)

8 0 (5 5 )

1.0 (2 .0

5 6 19(3698)

36

Days
(N)

683(52)

292(99)

66 7 (5 4 )

233(36}

2313(577}

21

Lew

0**

122

38

146

41

6.0

640

1

H ig h

119(44)

356 (84)

196 (58)

236 (50)

5 8(14)

2.2 (0.8)

108893(68359)

5

M e d iu m

80(1 7 )

176 (32)

112(19)

96(35)

20 (8)

1.3 (0.6)

29072(9564)

26

65(10)

111(18)

81(13)

4 5(14)

9 (3 )

0 .7 (0 3 )

17323(5572)

25

174(92)

393(70)

306(90}

219(81)

6 3 (3 3 )

2.0 (1.6)

13387(6898}

21

7211(2628)

IS

2817(2305}

17

■rt-S.
ra p i

Tot.
Days
(Nt)

16(40}

Medhim

102(32)

319(74)

199 (49)

216(60)

5 4 (2 0 )

2.3 (0.8)

lew

4 3 (2 0 )

192(78)

92(40}

149(71)

3 3 (1 6 )

4 -9 ( 3 3 )

305

84

1.6

301577
6 8 1 2 0 (3S730)

8

H ig h

193

499

336

M e d iu m

88 (30)

311 (72)

161 (39)

233 (53)

4 7(12)

2.7 (0.7)

74(17)

147 (35)

103 (24)

73(26)

16(6)

1.0 (0.4)

28630(13223}

51

58

56

53

60

H ig h

7 3 2 (1 0 0 )

1056(26)

9 37(102)

304(173}

8 4 (5 5 )

0.S (0.4)

S623S (24639)

31

M e d h im

2 2 7 (6 7 )

922(145}

4 67(155)

6 9 5(128)

156(42)

3.4 (1.4}

19008(9717)

12

tew

2 2 2 (0 5 )

5 3 5 (1 2 0 )

m m

294(75)

6 6 (2 1 )

1.5 (0.7)

13497(7291)

17

H ig h

211(108)

968 (45)

449(152}

757 (71)

189(50)

5.0 (3.0)

367279 (249474)

M e d iu m

162 (41)

396(116)

238 (53)

233(120)

51 (25)

1.6 (1.2)

100484(49369)

17

97(2 2 )

210 (55)

143 (32)

113 (45)

26(12)

1.2 (0.5)

41631(15893}

40

Low

3.3

Avg.

Medium

lo w

RPT

Avg.

27(196}

Low

PS3

Avg.

60

60

Shape and Magnitude - Field Season Summary

This section compiles and draws upon several parts of the data collection and analysis
presented earlier, and uses field observations to describe the results o f the statistical analyses
(section 3.2) for the 2011 field season. The composite figures in Figure 3.5 summarize daily
data and the results of the PCA and CA that were used to categorize the ‘shape’ and
‘magnitude’ of the suspended sediment response for each o f the sites. Daily mean air
temperature and total precipitation are presented with the meteorological CA results for
comparison purposes (Figure 3.6). Sites PS1, PS2 and PS3 were missing days in the SSC

72
shape and magnitude classification due to low water, partial days o f data, erroneous data or
no data (see Table 3.2, section 3.1.3).

Qt u .2 ,
8 1&
02
2 1
■ Diurnal Shape

Q Irregular Shape

0.4

Daily M ean Q (m3/s)

800

Daily M ean SSC (mg/L)

0.35

700

0.3

600

0.25

500

'SS
E.

£ 0.2

uiSl
sA
400

Jg 0.15

300

C

a>

0.1

n
0)
2
200 >

0.05

100

io
O

0
CT)

8 S 8 8 2 s 0r-t0

m

00

IN

m

in

PSl

3
*w
o
u .3 2
iR ££
£
V)
m
2 l

■o ■

B

■3 ■ ■

m

m

ta

— m —BE3-

- S 3 3 3 — 0 E 3 -Q —
■ Diurnal Shape

B Irregular Shape

9

350

*ifl 8

300

7

I 6

Daily M ean Q (m 3/s)

Daily M ean SSC (mg/L)

250

200

S4
s

>

150

3

v /

E

l/l
«c01

s

‘5
Q

1
0
8
PS2

100

2

'Si

Tl
O
O O C
ro

ID

00

rH

<N

r-*

IN

O
cn
1ro0 cnm
m
cn

in

oo

SSC
Magnitude

Daily Mean Q (m3/s )
o

n»

^

01

oo

o

nj

I-*

NJ

Daily M ean Q (m3/s )

U)

194

197:
200:
203 :
206 :
209 :

212:
215:
2 1 8

:

221:
224 :
227
230
233

236:
239:
242 :
2 4 5

;

248 :

251:
254 :

J M
UJ U
6
O ol/l o in
h* K
o
in o
in g
Daily Mean SSC (mg/L)

Daily M ean SSC (mg/L)

SSC
Magnitude

74
-m -

3
*0o1
u
.2 2
m
m£
re
5 i

-Q—I

-O E 9

pixwa

H

0 0

■ Diurnal Shape

0 Irregular Shape

1.8

Daily M ean Q (m 3/s)

Daily M ean SSC ( m g / l ) r

1.6

1000

1.4

E. 1.2
0 1

v N

e

800
600

re
01 0.8
2 0.6

400

re 0.4

200

O 0.2
0
CTl

RPT

1200

r-

CM

8

m

vo Oi
O O
O

00

■ D iu rn a l S h a p e

O aily M e a n Q ( m 3 /s )

CN

p-.

PM

o

PO

m
l£
> <
Ti
pr>
PPI
m

<N

in

U
ift
A
C
re

a)
s
>

re

O

m

£3 Ir r e g u la r S h a p e

Daily M ean SSC (mg/L)

Figure 3.5 Composite figures showing suspended sediment shape (diurnal or irregular) and m agnitude <1
= low; 2 = medium: 3 = high) classification results from principal com ponent analysis and cluster
analysis and daily mean streamflovt (Q) and suspended sedim ent concentration (SSC) for each o f
the proglacial m onitoring sites. PS1, PS2, and PS3 are m issing days ill the shape and m agnitude
classification due to low water, partial days of data, erroneous data or no data (see section 3.1).
Figure continued over three preceding pages: x-axes in Julian Days.

75
Q S torm

i Hot a n d Dry

B W arm an d Damp

18

r 35
Daiiy M ean AT (*C)

^

16

H

14

<

12

C

10

ro

Daily Precip. (mm)

30

20

d
u

15 I
io £

£
6
'(5 4

n

5

2

0

E

25 £

« 8
a

B Cold and W et

to
© O (N
«H

°

0
rt
00

Figure 3.6 Daily mean air tem perature (AT) and daily total precipitation presented with results of
meteorological principal com ponent analysis ('storm ’; 'hot aud dry’; ‘warm and dam p’: ‘cold and
w et’).

The three sites along the main Castle Creek channel (PS2, MS and DS) show a
similar ‘shape’ and ‘magnitude’ pattern dominated by ‘irregular’ SSC data on ‘cold and wet’
days and ‘diurnal’ SSC data on ‘hot and dry’ days (Figure 3.5, Table 3.5). The SSC response
at these three sites tracks air temperature and precipitation, but, particularly for MS, the
response was often not strong enough to cross the threshold to a higher magnitude category.
While both ‘storm’ days were classified as ‘diurnal’ for PS2, MS and DS, the three or four
days following the storm were ‘irregular’ as storm flows subsided. Successive downstream
sites along the main channel show the influence o f the tributary streams along the way.
Similar to the results o f the main meltwater stream at CCG in 2011, Orwin and Smart
(2004a) state that the response pattern from three sites along the North Proglacial Stream at
the SRG in 2000 was dominated by ‘low’ magnitude ‘diurnal’ data, and that ‘high’
magnitude ‘irregular’ response data were closely associated with precipitation events. They
also found that the SSC magnitude response at a high elevation site was only elevated by

76
temperature peaks (Orwin and Smart 2004a). Compared to the SRG study, all sites at the
CCG were relatively similar in elevation; as such, this high elevation response pattern was
not observed. The SSC response at RPT was similar to the response at the Central Proglacial
Lower site in the SRG study (Orwin and Smart 2004a), where cool temperatures generated
‘diurnal’ response data and high temperatures generated ‘irregular’ response data.

Table 3.5 Sum mary o f suspended sedim ent ‘shape’ and ‘m agnitude’ analysis for 2011 proglacial
m onitoring locations, .11) 195 - .11) 254: DAYS
Site

PS1

PS2

Cluster

Cold and Wet

Warm and Damp

Hot and Dry

Storm

Days(N)

Days (Nt)

Classification
Diurnal (Irregular)
High
Medium
Low

(17 days)
15(1)
1 0 (1 )
.. ■ 4 .....

(IS days)
1 2 (2 )
8 (2 )

(2 days)

^

60
45(13)
23(13)
21: « ..:

60
58
36
2 i. :

1

(26 days)
18(8)
5(8)
13
■. ..

D iurnal (Irregular)

1 (1 5 )

7 (7 )

High

(1)
1 (6 )
(8)

(1)
4 (1 )
3 (5 )

13(4)
7(1)

9(4)
4(1)

13(8)

Medium
Low

5(3)

5(3)

5(1}
(2 )

D iurnal (Irregular)

2 (1 5 )

10 (5)

23 (3)

High
M edium

-

-

-

(1)
1 0 (4 )

3

Low

(3)
2 (1 2 )

Diurnal (Irregular)
High
Medium
Low

14(3)
4(1)
3(1)
7(1)

11(4)
5(3)
-

1(6)

(1)

6(1}

(2 )

D iurnal (Irregular)

3 (1 4 )

8 (7 )

23 (3)

2

36 (24)

60

High

(1)
1 (3 )
2 (1 0 )

-

6 (2 )

2

2 (1 )
1 0 (7 )
24 (16)

3
17
40

Medium
Low

PS3

Diurnal (Irregular)
High

MS

RPT

DS

Medium
Low

4

3 (2 )
5 (5 )

(2 )
(2 )
...

. . ,j_. . ......

■

■■■■%
■■ ■'

l

2 1 (3 )
1

2

31 (25)

2

12 (2 )

-2
2

3 (2 )
17 (9)
1 1 (1 4 )

56
5

8 (1 )

26
25

37(16)

53

15(6)

21

11(4)
1 1 (6 )

15
17

2

37 (23)

60

1
1

1
4 (4 )

1
8

20 (3)

-

32 (19)

51

4(22}
3(14)

1 (1 )
1

30(30)
13 (18)
4(8)
13(4)

60
31

2(4)
6(1)

17 (1)

--

12

17

The results of the ‘shape’ and ‘magnitude’ analysis of suspended sediment response
data can be explained by the characteristics of the contributing watershed area. The following
sections examine the observed response patterns at each of the six sites with brief reviews of
the catchment characteristics and field observations for context.

77
3.3.1

Proximal Site 1
Proximal Site 1 (PS1) had a small flow volume compared to that of the main stem of

the Castle Creek. The influence of its typically high SSC was substantially diluted a short
distance from the confluence with the main channel on the low gradient outwash fan
upstream of MS (Figure 2.2). It drained an estimated c. 1 km2 area along the east-northeast
side of the terminal lobe, which included lateral moraine deposits and the ice-marginal
interface; however, the catchment of PS1 was only 11% glaciated (Table 2.1). Much o f the
deglaciated catchment area was dominated by bedrock and metastable till deposits, but the
area proximal to the glacier was dominated by unconsolidated glacial till.
At PS1, 78% of the suspended sediment data during the field season had a ‘diurnal’
response shape (Figure 3.5, Table 3.5). The sediment load of the small meltwater stream was
enriched as it flowed along the unconsolidated till-dominated ice-marginal interface, and its
high sediment concentration fluctuated with ablation and precipitation (Figures 3.1, 3.5, and
3.6). The pattern was more apparent as the duration of values that exceed the range o f the Tu
meter became progressively shorter later in the field season. Both ‘storm’ events were
classified as ‘irregular’ as a result of sustained ‘high’ magnitude SSC. There was a noticeable
decrease in the SSC diurnal range following each storm event, and data fell completely
within the range of the Tu meter for most of the time after the second storm event on JD 234.
The declining magnitude of suspended sediment response over the ablation season was a
result of reduced activity, stabilization or exhaustion of sediment sources along the meltwater
channel. Additionally, seasonal snowmelt and ablation potential decline as the autumnal
equinox approaches. Any disturbance or redirection of the active channel would likely

78
reactivate or rejuvenate sediment sources, and glacial advance or retreat could substantially
reset sediment availability for the next ablation season.
The four ‘high’ magnitude ‘irregular’ response days that occurred from JD 204 to
JD207 were a result of full day exceedances driven by ablation, but may have shown a
diumal response shape if data were within the range of the Tu meter. Interestingly, the data
show that on JD 244, a ~10 mm rain event and cool air temperatures caused Q to increase
while SSC decreased, although the pattern remained ‘diumal’. The JD 244 - JD 245 data
were considered valid because all other sites, with the exception o f PS3, show a
corresponding Q response and low SSC response to this event (Figure 3.1). Upstream of the
monitoring station on PS1, there was a tributary that drained a slope dominated by bedrock
and metastable till deposits. Low air temperature would have reduced ablation and meltwater
production, while rainwater draining from the tributary catchment would have been relatively
clean, thus diluting the sediment load of meltwater flowing from the area proximal to the
glacier. Other studies have reported a similar proglacial SSC response to rain events when
temperatures are low (Sawada and Johnson 2000; Orwin and Smart 2004a)
3.3.2

Proximal Site 2
Proximal Site 2 (PS2) monitored the main flow of the Castle Creek proglacial stream,

c. 200 m downstream from where it emanated from a meltwater channel portal on the
northwest side of the terminal lobe o f the glacier (Figure 2.2). The estimated catchment area
was c. 9 km2 and 77% glaciated (Table 2.1). Because o f the distance between the meltwater
channel portal and the monitoring location, there was the opportunity for recently exposed,
unconsolidated sediment deposits proximal to the glacier and adjacent to the meltwater
stream to enrich the sediment load of meltwater emanating from the glacier. Orwin and Smart

79
(2004a) observed a similar limitation with proximal monitoring sites located a short distance
from the glacier. Ideally, the two sediment sources would be isolated by locating monitoring
sites immediately adjacent to the glacier; however, site conditions evolve through the
ablation season and the complex geomorphology o f the ice-marginal environment limits site
selection. Small Q and SSC spikes were observed on the falling limb of the hydrograph,
mostly during warm weather, when ice calving into the proglacial pool c. 75m upstream o f
PS2 caused waves that generated a pulse o f water and entrained unconsolidated sediment
from the deposits adjacent to the pool; Stott and Grove (2001) observed a similar process on
the Skeldal River in Greenland.
Suspended sediment response ‘shape’ and ‘magnitude’ at PS2 tended to mirror the
trend of ablation driven streamflow, but unconsolidated sediment sources proximal to the
glacier were activated by precipitation. About 55% o f the SSC data had a ‘diumal’ shape and
the majority were categorized as ‘medium’ magnitude (Figure 3.5, Table 3.5). During the
‘storm event’ on JD 234, available sediment became exhausted during the peak flow and the
SSC was much lower on the falling limb than on the rising limb of the event hydrograph
(Figure 3.1).
3.3.3

Proximal Site 3
Proximal Site 3 (PS3) drained a proglacial lake that was perched on top o f a till sheet

in front of the terminal lobe of the glacier (Figure 2.2). Most o f the relatively low gradient
catchment was exposed within the previous c. 8 years (Beedle et al. 2009); however, the
stability of the deposit appears to have increased rapidly. The estimated catchment area was
c. 2 km2 and 95% glaciated (Table 2.1). The source of meltwater to this proglacial lake was
primarily beneath the terminus of the glacier, but some meltwater was received from streams

80
that flowed along unconsolidated sediment at the ice—till interface. Additionally, channels
observed on the relatively low gradient area surrounding the lake suggest that runoff from
snow melt and storm events can also transport sediment into the lake. The stream that drains
this proglacial lake was bedrock controlled and entered Castle Creek c. 130 m downstream
from the PS2 monitoring location. The suspended sediment response at this site appeared to
influence the SSC data at MS, especially during the first part of the field season.
Site PS3 showed a dynamic suspended sediment ‘shape’ and ‘magnitude’ response
through the field season, responding to ablation, seasonal snowmelt and rainfall (Figures 3.1,
3.5, and 3.6). The extremely high SSC recorded from JD 218 - JD 221, may have been
influenced by sediment derived from the stream bed and was not supported by downstream
monitoring of SSC at MS; it was, therefore, considered erroneous. While the two full days of
Tu exceedances (JD 219 and JD 220) were removed before the analysis, the two partial days
(JD 218 and JD 221) were left in, but there was low confidence in the accuracy o f these
results. Suspended sediment concentration covaried with air temperature and rainfall until the
‘storm’ event on JD 234. Data with high SSC values were sustained for some time as Q
dropped following the event; possibly a function o f the amount o f time needed for sediment
laden water to be flushed out of (or settle in) the lake. The shape o f SSC data at this site were
70% ‘diumal’, while magnitude responded to air temperature and precipitation peaks. Similar
to PS1, as the ablation season progressed, the sediment supply appeared to become
increasingly exhausted or transport limited.
3.3.4

Middle Site
The Middle Site (MS) was located at the bottom of an outwash fan complex,

downstream from the confluences of the three proximal sites. The catchment area was c. 13

81
>>

km and 71% glaciated (Table 2.1). In addition to the catchment areas of the proximal sites,
there was a bedrock dominated area to the west, the till sheet that extended downslope from
the terminus of the glacier, and two outwash fans that merged c. 100 m upstream of the site.
Active mass movement along the right bank of the main channel, between the PS3
confluence and the PS1 confluence, was triggered by snowmelt, precipitation and high flow
events.
The suspended sediment response ‘shape’ at MS was 62% diumal, and closely
mirrors that o f PS2 which accounts for c. 85% o f the flow at MS (see section 4.1, Table 4.3).
Slightly higher peaks in the SSC data at MS tend to cause the overall magnitude
classification to be lower than PS2, 85% of the data at MS fell into the ‘low’ magnitude
category (Table 3.5). Higher SSC peaks during ‘storm events’ may be the result of sediment
sources in the channel, along the channel banks or in the un-glaciated catchment area being
activated. Site PS3 was capable o f influencing the SSC at MS, which was primarily evident
prior to the ‘storm’ event on JD 211. The input from PS1 has a high daily mean SSC and
persistent diumal pattern, but, because o f its low Q, the influence o f its response pattern
cannot be discerned in the data at MS.
3.3.5

Rockback Peak Tributary
The Rockback Peak Tributary (RPT) drained the northeast aspect of Rockback Peak.

The estimated catchment area of RPT was c. 3 km2 and less than 20% glaciated (Table
2.1).The cirque glacier/rock glacier in the top o f the catchment had active sediment sources
depositing debris via avalanches and rockslides from the mountain above onto its surface.
Meltwater was divided over the east ridge of the cirque; a portion flowed east into the Castle
Creek catchment, and a portion flowed west into the adjacent drainage. Aggradation o f the

82
outwash fan on the ridge had recently caused RPT to take a more northern route off the ridge.
Sediment from this site was characteristically black in contrast to the gray-brown sediment
and bluish rock-flour from CCG; this contrast was observed and photographed on site, as
well as recorded in suspended sediment and sediment source samples. As the cirque
glacier/rock glacier retreats and the alluvial path along the ridge aggrades, this sediment
source may divert to the west, away from the Castle Creek catchment. However, for the 2011
field season, RPT was a major source o f fine sediment for Castle Creek.
The sediment load of RPT was primarily from the cirque glacier/rock glacier and
actively eroding deposits on the adjacent ridge; the response pattern was not necessarily
synchronous with the Castle Creek system. An ‘irregular’ suspended sediment response
shape was dominant during ‘hot and dry’ periods, while other periods tended to have a
‘diumal’ response shape. The SSC and Q data responded positively to warm temperatures
and precipitation events throughout the season. Following the two ‘storm’ events, SSC
remained elevated for a period of time, which was probably a result of sediment slumping
into the channel during the event and slowly being evacuated; Orwin and Smart (2004a)
report similar processes and response patterns for one o f their sites at the SRG. Warm
temperatures caused SSC to exceed the range of the Tu meter for the majority of data from
JD 215 - JD219 and JD 223 - JD 225. The range o f the meter was also exceeded during the
warm weather and ‘storm’ event that occurred during JD 233 - JD 236. Following the
‘storm’ event on JD 234, SSC levels remained ‘high’ for eight days, and then dropped
suddenly to a ‘low’ magnitude response during ‘cold and wet’ weather on JD 241.
Interestingly, on JD 244 streamflow increased in response to a precipitation event but SSC
decreased, which was the same response observed at PS1. Similar to PS 1, a tributary stream

83
upstream of the RPT monitoring station drained the bedrock and metastable till mantled slope
below the sediment laden ridge where RPT begins; runoff from this tributary may have
diluted the SSC in RPT during the event.

3.3.6

Distal Site
The Distal Site (DS) was essentially the combined response o f RPT and MS;

additional sediment sources were limited to ephemeral channels that were only activated
during intense precipitation. These channels and sediment sources would also be active
during spring runoff, which was mostly completed before the start o f the 2011 field season.
At this monitoring site, the catchment area was c. 16 km2 and 60% glaciated (Table 2.1).
Additional areas that were not included in the RPT or MS catchments include a low relief till
sheet along the right bank of the channel, abandoned outwash channels that carve through
thick till deposits along the left bank upstream of RPT, and a moraine and till mantled
bedrock outcrop downstream of RPT. ‘Shape’ classification at DS was very similar to MS,
60% diumal, while ‘magnitude’ classification at DS was slightly higher than MS, which was
most likely a result of the RPT tributary (Table 3.5). The ‘magnitude’ response at DS
increased with ablation during the early season, but the response diminished later in the field
season which was attributed to declining contribution of annual snowmelt.

84

4

Results and Discussion (2) - Suspended Sediment Load
The analysis of meteorological conditions, streamflow driving factors, and the ‘shape’

and ‘magnitude’ o f the suspended sediment response has divided the data from the six 2011
monitoring stations into groups of similar data (sections 3.2 and 3.3). In this chapter,
objective 2 is addressed by computing SSL from the 5-minute time-series SSC and Q data
following Equation 2 (section 2.2). The SSL time-series are then summarized into hydrologic
daily averages and organized into the sub-categories as determined through PCA and CA for
each site {section 3.2). Building on this analysis, a suspended sediment budget is defined for
the 2011 field season for the Castle Creek proglacial zone using the key variables defined by
Warburton (1990).

4.1

Suspended Sediment Load

The arithmetic mean o f the average daily SSL (kg/day) was computed for each hydro­
meteorological, ‘shape’ and ‘magnitude’ sub-category (Table 4.1). When summarizing the
categories, weighted averages were used to account for the disproportionate number o f days
in each category; Table 3.5 reports the number of days in each category. A similar table with
SSL totals (kg/x days) can be found in Appendix 7.4.

85
Table 4.1 Summary o f average suspended sedim ent load (kg/day) for each sub-category. Averages for
‘irregular’ response shape data are reported in brackets. The values reported in the body o f the
table are arithmetic means for the given category. W eighted averages were used to account for the
disproportionate number o f days in each category for the ’shape' and ‘m agnitude’, and total
summary. Table 3.5 reports the the number o f days in each category.
Site

PS1

PS2

PS3

MS

RPT*

DS

Cluster

Cold and Wet

Warm and

Hot and Dry

Storm

Weighted Average

Classification

(17 Days)

Damp (15 Days)

(26 Days)

(2 Days)

summary

Olumal (Irregular)

*927(5829)

3567(4736)

2866(4188)

(17200)

3740(6400}

High

5430(5829)

4320(4736)

4044(4188)

(17200)

5177(6400)

S619

MerSum

2240

2061

2413

-

2313

2313

tow

640

-

-

-

640

640

D iu rn a l (I rre g u la r)

2 0 7 7 4 (2 3 4 5 1 )

2 6 8 7 4 (3 0 7 0 2 )

2 7 5 9 5 (3 0 8 2 2 )

142879

3 4 6 5 0 (2 6 3 6 6 )

30952

H igh

(7 4 6 0 6 )

(1 0 9 5 2 7 )

74572

142879

1 2 0 1 1 0 (9 2 0 6 7 )

108893

M e d iu m

2 0 7 7 4 (2 6 3 4 5 )

3 7 9 5 1 (2 3 2 6 3 )

2 8 0 1 3 (3 2 9 1 1 )

-

2 9 9 2 6 (2 7 4 6 2 )

29072

Low

(1 4 8 8 6 )

1 2 1 0 5 (1 6 4 2 5 )

2 1 0 9 6 (2 6 6 4 3 )

--

1 8 6 4 4 (1 6 2 7 5 )

17323

.

Diurnal (Irregular)

7625(8329)

8568(5267)

4642 (10877)

High

10932(15130)

11378(10915)

11824(12552)

Medium

.^8460:;:

5818(9982}

low

2829(6062)

6320(9060)
(547)

D iu rn a l (Irre g u la r)

3 3 1 0 9 (2 9 6 8 3 )

2 4 0 8 9 (5 8 3 9 6 )

High

-

-

4336

79600 ...

7994(8838}

8249

:'»600:

13659(12709)

13387

6286(9752)

7211

T '.

1978(43571

2817

3 2 7 9 1 (4 1 5 1 4 )

191799

3 9 0 5 1 (3 7 4 6 8 )

38444

--

301577

301577

301577

1268(6863)

■ ■■■

M e d iu m

(5 3 9 0 5 )

(1 3 8 1 3 4 )

54363

82021

6 1 2 7 8 (7 4 9 6 2 )

68120

Low

3 3 1 0 9 (2 3 6 2 7 )

2 4 0 8 9 (3 8 4 6 2 )

2 9 5 5 5 (4 1 5 1 4 )

--

2 8 0 6 9 (2 9 5 7 4 )

28630

Diumal (Irregular)

30471(13807)

26962(59615)

34744(30566)

343410(42929)

40185(3317(1

36681

44*17(78371)

41200(38843)

; 343410

72795(4*275)

56235

15374 (15588)

(42929)

18371(19327)
14288(10926)

19008
13*97
74577

High

64308(18039)

Medium

19370(18153)

tow

15892(5228)

12416(3349)

(17563)

D iu rn a l (Irre g u la r)

5 0 2 3 3 (6 1 0 7 0 )

5 3 4 4 8 (8 5 1 7 1 )

5 8 5 2 7 (7 0 1 2 2 )

444324

7 8 1 4 0 (6 9 2 3 1 )

H igh

(2 1 3 1 8 7 )

--

-•

444324

4 4 4 3 2 4 (2 1 3 1 8 7 )

367279

M e d iu m

5 6 0 3 4 (9 5 4 7 6 )

8 2 6 8 0 (2 0 3 1 8 5 )

9 3 8 4 1 (7 4 1 5 6 )

--

8 6 7 1 2 (1 2 0 1 5 9 )

100484

Low

4 7 3 3 2 (3 5 5 3 7 )

3 5 9 0 8 (3 7 9 6 6 )

4 6 0 6 3 (6 2 0 S 4 )

-

4 4 0 5 3 (3 7 9 5 3 )

41613

‘ e x ce e d a n ce s in th e RPT d a ta set h ave b e en e stim a te d

The information in Table 4.1 illustrates where, under what conditions and how much
sediment was generated, transported, stored and evacuated from the watershed during the
2011 field season. To simplify the information, it could be presented as percentage of total;
however, the distribution o f days across categories varied by site and was specific to the 2011
field season, which would make comparisons between locations or over different field
seasons difficult, Hannah et al. (2000) also report this limitation with the analysis.

86
As described in sections 3.2.3 and 3.2.4, ‘shape’ and ‘magnitude’ parameters were
essentially driven by hydro-meteorological conditions at the time o f monitoring. Thus, the
most applicable division o f the field season for the suspended sediment budget was into the
hydro-meteorological periods. From there, similar computations of totals and averages can be
made, but reported in a simplified format that will also be more useful for modelling
applications, assessing subsequent years of data, or comparing results with other sites. Table
4.2 presents the SSL data divided into meteorological categories. The values in the table were
computed from daily means or totals as appropriate; max and min values were not
instantaneous values, they were max and min o f the daily means in the category. Table 4.2 is
a good reference for interpreting the other tables presented later in this discussion, and a wide
range of summary computations are possible.

87
Table 4.2 Field season summary statistics for meteorological periods determined through principal
com ponent analysis. Streainflow is Q (m '/s), suspended sedim ent concentration is SSC (mg/L),
suspended sedim ent load is SSL (t/day), precipitation is PT (daily total mm), and air tem perature
is AT (daily mean °C). V alues in the table have been com puted from daily averages or totals o f
individual hvdrologic days (06:00 - 06:00) in the category.
AT

M in

0.25

PT

PS1

M3

PS2

ssc

SSL

a

SSC

SSL

0.1*4

IM

5,1*

1.96

16.1

15.4

90.2

i n

1190

950

6.44

360

213

104

30.0

0.574

525

V S

3.60

165

59.1

17

17

17

17

17

u

17

17

17

W

n /a

n /a

511

«*»

e/9

4M

n /a

n /a

1005

251

1.21

69.5

49.1

'f t J * '

1.61

75.9

10.6

SSC

SSI

■: Q

SSC

J

SSC

a

M m

SJu2

M M

1.7?

65.9

]

11.2

74.6

QMT

427

t u

532

1*7

21*

7Jt

3.04

17

17

Q

SSC

sst

a

SSC

SSL

0.75

M 8

IM "

0040

1.51

60.1

7.92

10J

4.67

177

OS

APT

M S

1

„

Mm

C.7

19.1

MM

1 70*

1
2

Mh o

2.9*

7.95

M O

005

4.95

2.55

94.3

23.3

0 J6S

8

~

17

17

U

1*

15

17

16

16

3

^

»

n /a

135

" /•

70.7

n/«

n /a

372

e/»

1.5*

6.0*

M O

212

3.12

0.172

29.9

16.7

0.171

u*

: M*

1.03

34*

20.9

M M

m

M in

2.5*

0.0

M O

M l'

O JA

1.16

60.9

9.6

fttn r

447

• ' *29

1.29

650

7.21

' 49*2

19*

|

Mw

9.47

5.35

U K

M l

? .»

5.3*

194

11.0

0.629

IM

IM

7.10

220

139

4 .7 M

1994

11*

7.52

409

271

^

M ean

5.54

2.45

eon

MS

i n

279

99.2

29.3

IM S

219

:* •»

3.23

109

35 S

0.4*2

?*■

: 0 .7

3.69

1*0

69.3

15

15

'' M

1*

1*

15

14

14

»

II

15

IS

15

IS

15

15

15

15

15

n /a

36.9

■ ttf9

: 'M i

n /a

n /a

403

« /*

942 '

n /a

n /a

533

nA

V*

W

n /a

n /a

1024

1.17

2.61

' 40*7: ■

102

U S

1.34

37.6

26.9

0.190

10*

&S7

1.45

42.2

342

<4X07

527

8 4

1.54

93.7

64.9

M in

1.97

0.

o jm

UK

U P *'

1.36

74.2

15.6

04*0

s u

O JM

1.55

76.4

10.5

4*35

201

IM

1.79

95.6

13.5

m m

15.4

*.54

00*7

907*

M0

4.92

169

74.6

0.5*9

4»

117J0

5.29

1*7

j

90.0

4701

M U

: SLt

5.46

294

14*

. Mmm

10.3

1.19

M m

422

u ?

2*4

107

21.0

0 J6S

17*

7.02

42 7

US

i

33.6

44»

AS

SU

3.69

177

59.9

»

M

26

24

24

2*

21

26

26

26

' M .I

n /a

n /a

672

* !•

81

n /a

n /a

1556

0.547

23.6

115

0.147

3.50

133

46.2

. *<*
S ttl.

0*v.

'

;

■'

:

m
•

■ »

'

§
3B

Sum
S id .

Oev.

.

a

26

26

26

2*

147

n /a

n /a

*79

n/a

110

M2

0.704

245

14.1

040*

: 2S4: :

117

0.73

46.7

26.6

0.7M

m

27J

4.51

1**

92.0

M M

« i

424

5.14

323

169

240

0J9*

m '

it*

9.7

336

302

IM

23*7
C

90

11.0

663

719

143

. 4 * 0 ? : : : 356

2M

7.10

262

192

tis

1475

1»

6.05

493

444

2

2

1

2 '

*

2

2

2*

25

u

n /i

SO.*

* /*

1.12

2.03

4022

IM

M in

7.52

31.0
U

' 'M U ;

■ MO

; t u

M an

7.93

31.2

. 0.150

' **»

*M

5.51

304

M ean

7.72

570 .

171

6.00

219

1

2

n /a

oA

*M

" /•

n /a

2*6

0.17

0 JM

12.1

M l

3.S5

121

137

0.19*

0

Q4MB

R t

4*49

1.16

60.1

7.92

' M m y [ ■■■ 0 4 4 :

■

21

x| ~ ^.....
Sum
S td .

Oev.

e

31.1

;: t

D *v»

S tm t
S td .

Oev.
MM

^

15.4

0.29

0.25

S td .

2

2

1

:

RS

i

:.9 M

n /a

n /a

394

! * ^B ■/ /

U 7

467

10S

155

5«1

0 jS 4 -

1.2*

65.0

7.2

10.6

•M

11.0

662

719

16.7

3.91

195

74.6

' 40

60

60

60

2X01

n /a

n /a

4475

442

1.45

90.9

96.6

60

60

90

m

* /•

417

n /a

n /a

2307

0 /9

n/B

111

6.73

1.32

43.9

41.2

0440

: 9A

105

31.0

M 9*

207

50

50

5*

•;

60

56

56

N

.

n /a

n /a

1733

1.11

39.9

33.5

E - E stim ate; e - Exceedance

75.9

60

2.95

4M *

1.61

5*

4SS

*.40

495

H

901

■: m

ut

: 745

MM

M m

n sftf

450*

4.42

6.54

3*9

39.4

5.94

3.59

240

115

tti

■ ■ ‘M l.

4.11

432

427

«/*

XU

4 »

M m

E

M M ;-:

U t

240

265

9*9

302

304

n/«

n /a

33*

*.S1

10/: : ' »

n /a

9.7

20.9

•/

« .

23S7
i

707.

M ean

Oev.

31.2

■

;

0400

15.4

Sum

62.2

M B ..

■;

'

I

M ax

|

6 JM

'

::

■

'

.

88
Table 4.3 and Figure 4.1 report SSL and Q for the monitoring network as percent
contribution to the total at DS for each hydro-meteorological category. Mean values were
used for the computations rather than totals to reduce the influence o f missing days at the
proximal sites; raw values are reported in Table 4.2.

Fable 4.3 Percent (%) o f mean daily suspended sedim ent load (SSL) and stream ilow (Q) relative to the
distal site (DS) during meteorological periods determined by principal com ponent analysis.
Meteorological

Cold and Wet

Warm and Damp

Hot and Dry

Storm

Seasonal

period

28% (17 days)

25% (IS days)

44% (26 days)

3% (2 days)

100% (60 days)

% of mean

%of mean

% of mean

% of mean

% of mean

% of mean

% of mean

% of mean

% of mean

% of mean

SSL at DS

QatDS

SSL at DS

QatDS

SSL at DS

QatDS

SSL at DS

QatDS

SSL at DS

QatDS

m

84

3.4

&4

:: 1.9

.5 ,5 :....

1.8

3.9

■■■■ 3.6

5.8

2.4

PS2

39.4

70.8

42.2

75.6

46.7

77.0

32.2

74.5

41.6

74.8

PS3

13 a

104

11.1

9.9

' 11.7

9.9

6.7

11.0

114

10.0

IPS1 + PS2 + P53

61.0

84.3

S8.7

87.4

63.9

88.7

42.8

89.1

58.5

87.2

MS

50.7

84.4

52.0

87.5

56.4

88.6

43.2

88.2

SIS

87.1

Site

RPT

4 6 .5

15.9

52.3

13.1

52.1

11.7

43.4

14.3

49.2

13.4

lMS*fVT

S7S

100.4

m s

100.6

108.5

100.3

86.6

102.5

100.7

100.5

DS

100

100

100

100

100

100

100

100

100

100

0

(S91S8)

(3.601

(68252)

(3.69)

(S986S)

(3.69)

(444324)

(8.05)

(74S77)

(3.81)

* M ean SSL (kg/day) an d Q (m 3/s ) have b e en included fo r DS for back calculation p u rp o ses; also re fe r to T able 4.2.

Streamflow was assumed to be conservative throughout the monitoring period in
order to compute Q for RPT and PS3 {section 2.4.2, equations 5 and 6), as such, there is net
balance between upstream and downstream sites (Table 4.3 and Figure 4.1). Suspended
sediment load inputs totalling more (or less) than 100% o f the SSL at a downstream
monitoring site indicate sediment storage (or erosion) within that reach. Based on the
longitudinal stream profile (Figure 2.3, section 2.1.3) and site observations, in channel
storage had the greatest potential to occur on the low gradient outwash fan complex
immediately upstream of MS, and there was not likely a significant amount o f suspended
sediment stored within the reach MS-DS during any meteorological conditions.

89

Field Season Summary
100

49.2

Suspended
Serfiment Load

■0.7

RPT
1 3 .4

100

Streamflow

87.1

MS

CCG M eltwater
C hannels
Tributary
M eltwater
C h annel

41.61 74.8

11.1

24

10.0

PS3

PS2

P ro g la cia lL a k e

PS1

Storage (-)
Erosion (+)

Cold and W et

W a rm a n d D am p
100

OS

RPT

15.9

100

52.3

OS

RPT

J3J

100

100
87.5

MS

MS

13.21 1 0 1

PS3

PS2

8.4

34

PS1

42.21 75 5

99

11.1

PS2

1.9

5.4

PS1

PS3

Hot and Dry
100

RPT

RPT

----- DS

14.3

11.7

100

100

56.4

MS

MS

46.71 77.0

11.7 9.9

PS2

PS3

53

1.8

PS1

3 1 2 I 74.5

PS2

6.7

11.0

PS3

3.9

PS1

F igure 4.1 - Percent (% ) contribution o f suspended sedim ent load (SSL) and streamflow (Q) relative to
the total at DS over the 2011 field season and during the four defined hydro-m eteorological
categories - schematic diagram.

90

Differences in source contribution to the total SSL at DS during the defined hydro­
meteorological categories are apparent in Table 4.3 and Figure 4.1. While the mean daily
SSL and Q were similar during ‘cold and wet’ and ‘hot and dry’ conditions at the catchment
outlet (Table 4.3), the percent contribution to the total from the monitoring locations varies
(Figure 4.1). Glacial melt decreased during ‘cold and wet’ conditions, precipitation activated
some in channel and proglacial sediment sources, but low stream competency allowed
storage on the outwash fan complex. There is more contribution from PS2 during ‘hot and
dry’ conditions in response to ablation Figure 4.1. Comparing between ‘warm and damp’ and
‘hot and dry’, Q values were very similar, but the mean SSL is greater for ‘warm and damp’
conditions and less SSL is derived from the proximal sites.
The seasonal average SSL at MS was 7% less than the input from the three proximal
sites, which indicated net channel storage between the monitoring locations. The amount of
storage on the outwash fan varied slightly over the three main hydro-meteorological periods;
however, during the observed ‘storm’ events, the output from the three proximal stations was
slightly less than the SSL at MS. The increase was attributed to sediment contribution from
channel bed and bank erosion between the proximal sites and MS associated with high water
levels and stream competency. As a seasonal average, the combined SSL of MS and RPT
was nearly equal to the SSL at DS, which meant that on average the MS-DS reach was a
transport reach. ‘Hot and dry’ and ‘warm and damp’ meteorological conditions show that
there was, respectively, 8.5% and 4.3% storage within the MS-DS reach, and ‘cold and wet’
meteorological conditions show 2.8% increase in the MS-DS reach (Figure 4.1). This flux is
well within the range of uncertainty o f the results (section 3.1.4). During ‘storm’ events, the

91
SSL for the MS-DS reach increased to 13.4%, which was probable given the contribution
from channel bed and channel bank erosion, and the contribution from ephemeral channels
that drained diffuse proglacial sediment sources. The reach MS-DS had the most ephemeral
channels; during the JD 234 ‘storm’ event grab samples from these streams were measured at
c. 4000 mg/L.
As a seasonal average, RPT supplied 49% of the SSL, but only 13% o f the Q at DS,
while the respective values for the three proximal sites were 59% and 87% o f the SSL and Q
Figure (4.1). These values not only highlight the dominance of the glacier on SSL and Q, but
also the strong influence of a relatively small sediment laden tributary in the overall
suspended sediment budget. Aside from RPT, PS2, the main meltwater channel emanating
from the glacier, contributed the majority of the SSL throughout the field season, varying
between 32% during ‘storm’ events and 47% during ‘hot and dry’ periods.
As a result of Tu range exceedances, there was greater uncertainty within the RPT
analysis. There were 12 days in the RPT SSC time-series that had exceedances filled based
on discrete and composite daily SSC samples {section 3.1.3). These 12 days account for 6%,
33%, 19%, and 50% o f the RPT data for ‘cold and wet’, ‘warm and damp’, ‘hot and dry’, and
‘storm’ event periods, respectively. The estimates are reasonable, given the available
information, but compromise the accuracy in the MS-DS reach. The deductive method used
to compute Q for RPT strictly assumed that there were no other inflows in the MS-DS reach,
which was false during intense precipitation events when ephemeral channels became active
and contributed to streamflow; similarly, spring snowmelt could also concentrate in
ephemeral channels and deliver sediment from diffuse proglacial sources to glacial meltwater

92
channels (Richards and Moore 2003; Orwin and Smart 2004a). This assumption in the Q
computation affects SSL at RPT and creates additional uncertainty in the reach MS-DS.
In Table 4.4, values are computed based on the seasonal total SSL and seasonal mean
Q to show the contribution during the specified meteorological category for each o f the sites.

Table 4.4 M eteorological sum m ary of suspended sedim ent load (SSL) and streamflow (Q) for each site.
Values computed as a percentage o f the seasonal total SSL (t/day) and seasonal mean Q (m ’/s).
Raw values and the number o f days o f observation for each site in each category are presented in
Table 4.2.
Meteorological

Cold and Wet

Warm and Damp

Hot and Dry

Storm

Seasonal

period

28% (17 days)

25% (15 days)

44% (26 days)

3% (2 days)

100% (60 days)

% of

%of

% of

%of

%of

% of

%of

%of

Total SSL

Mean Q

seasonal

seasonal

seasona

seasonal

seasona

seasonal

seasona

seasonal

(xlO3 kg)

(m3/s)

SSL

Q

SSL

Q

SSL

Q

SSL

Q

PS1

32

38

21

19

34

34

14

2S1

0.09

PS2

21

25

23

25

39

43

16

7

1733

2.85

K3

29

27

24

24

: 35

41

13

8

437

0.38

JPS1 +PS2 +PS3

24

26

23

24

37

43

16

7

2421

3.36

Site

;

:

n

:

MS

22

26

23

24

38

. 43

17

7

2307

3.32

RPT

21

32

24

24

37

37

18

8

2201

0.51

OS

22

24

3S

■ 42

20

7

447S

3.81

DS-RPT

24

24

33

43

22

7

2274

3.30

23
26

22

.

% seasonal Q values weighted based on number of days of observation

The two ‘storm’ days accounted for 3% o f the duration of the field season, c. 7% of
the Q (with the exception of PS1), and 13 - 20% of the SSL (Table 4.4). Interestingly, 20%
of the SSL at DS occurred during the ‘storm’ events, which was c. 4% higher than PS2 and
MS, and 2.5% higher than RPT. The higher sediment load at DS was likely due to the
contribution o f sediment from diffuse proglacial sources via ephemeral channels in the reach
MS-DS. Sites PS1 and PS3 had a lower percent of their total SSL transported during ‘storm’
events than PS2, which was likely because of the dominance of bedrock in the PS1
catchment and glacial ice in the PS3 catchment. During the field season, PS1 and PS3 had a

93
substantial and sustained influence from meltwater flowing over fresh moraine deposits,
whereas PS2 emanated from a subglacial meltwater channel portal with a well-established
channel bed. Site MS shows influence from PS1 and PS3 in each of the meteorological
periods. However, the response pattern of MS was very similar to PS2 because 75% and 86%
o f the seasonal SSL and Q, respectively, was derived from PS2. The SSL at PS2 showed a
greater increase with ‘storm’ events than the other proximal sites because unconsolidated
extra-channel sediment sources proximal or adjacent to the glacier were activated by
precipitation and increased streamflow (Table 4.4).
Daily suspended sediment yield (SSY) in Table 4.5 was computed by dividing the
mean SSL for a given period (Table 4.2) by the watershed area (Table 2.1, section 2.1.2). For
total yield, the mean daily SSY values were multiplied by the respective number o f days of
observation in the category (note that some sites were missing days, as reported in Table 4.2).
Table 4.5 Suspended sedim ent yield (SSY) for the Castle Creek watershed during the 2011 field season.
M ean daily SSY (t/kin'/d) and Total SSY (t/km 2) for each catchm ent are reported.
Cold and Wet

Site

Warm and Damp

Hot and Dry

Storm

Seasonal

(17 days)

(IS days)

(26 days)

(2 days)

(60 days)

Mean daily (Total)

Mean daily (Total)

Mean daily (Total)

Mean daily (Total)

Mean daily (Total)

1*1(1.24)

4.01 (64.3)

3.00 (42.0)

2.64 (68.6)

13.9(27.7)

3.49(202)

PS2 (9.36)

2.48 (39.7)

3.08(43.1)

2.99 (71.8)

15.3 (30.6)

3.30 (185)

PS3(1.73}

4.49(76.3)

4.37 (56.8)

4.05 (85.0)

17.1(34.2)

4.76(253)

MS (12.69)

2.37 (40.3)

2.80 (42.0)

2.66 (69.3)

15.1 (30.3)

3.03 (182)

RPT (2.66)

10.3(175.9}

13.4 (201)

11.7(305)

72.6(145}

13.7(827)

DS (15.68)

3.77 (64.1)

4.35 (65.3)

3.82 (99.2)

28.3 (56.6)

4.75 (285)

(kmJ)

The total SSY from RPT during the 2011 field season was more than double that of
any o f the other sub-catchments. However, since RPT was a relatively small sub-catchment,
the seasonal total SSY at DS was only 36% greater than MS. Interestingly, the SSY of the

94
proximal sites tended to be greater than the SSY o f MS. However, SSY between MS and DS
increased with catchment area, which, as expected (Church and Slaymaker 1989; Schiefer et
al. 2001; Gumell et al. 1996 Tunnicliffe and Church 2011), disagrees with conventional
sediment yield models (Syvitski and Milliman 2007). During ‘cold and wet’ conditions the
SSY at PS1 was greater than PS2, while it was less than PS2 for the other three
meteorological conditions. Suspended sediment yield at MS was typically less than any o f
the proximal sites, indicating sediment storage within the proglacial channel network
upstream of MS. The effect of ‘storm’ events on SSY was striking: compared to the seasonal
mean, the minimum increase was 260% for PS3 and the maximum increase was 496% for
DS. The increase for RPT, MS, PS2 and PS1 was 430%, 400%, 360%, and 300%
respectively. The downstream trend of increasing relative SSY during ‘storm’ events was
likely because o f ephemeral stream inputs from diffuse proglacial sediment sources during
the events. Warburton (1990) also found that a large proportion o f the SSY can be generated
in a short period of high stream competency, and Orwin and Smart (2004a) also found that
sediment was evacuated during storm events on both of their monitored streams. This
triggered response from the proglacial zone should be expected to continue, in declining
magnitude, until the end o f the paraglacial period (Church and Ryder 1972; Church and
Slaymaker 1989; Gumell et al. 1996; Ballantyne 2002a; 2002b).
In summary, ‘hot and dry’ conditions dominated 44% of the 2011 field season and
generated the greatest portion of SSL and Q (Table 4.4), and thus, the greatest total SSY
(Table 4.5). However, the mean daily SSY for ‘hot and dry’ conditions was less than the
mean daily SSY for both the ‘seasonal average’ and ‘warm and damp’ conditions at all sites.
‘Hot and dry’ had similar streamflow to ‘warm and damp’ days but generated lower mean

95
daily SSL at the catchment outlet (Table 4.3). With the exception of RPT, SSL and Q were
primarily derived from the glacier and, with the exception o f ‘storm’ events, sediment storage
occurred on the outwash fan complex upstream from MS (Table 4.3). ‘Storm’ events
activated diffuse sediment sources within the proglacial zone and rapidly increased SSY at
the catchment outlet to c. 500% o f the seasonal mean. The range of SSY observed under
different conditions at the CCG sites (Table 4.5) fits within the range o f observations for
alpine glacier systems with comparable catchment areas reported by Gumell et al. (1996);
and was greater than that found in a small high arctic glacier basin in Svalbard by Hodson et
al. (1998), and greater than that found in Greenland reported and summarized by Stott et al.
(2014). Data presented by Orwin and Smart (2004a) from the SRG have been divided into
similar categories which permits a closer comparison.
The SSL results presented by Orwin and Smart have been recomputed in a format that
is comparable with this present study (Appendix 7.6). Unit area SSY has not been computed
for the SRG results as the catchment area for the individual sites was not reported. At the
SRG, similar to this study, the least amount of sediment was generated by the glacier and
entrained from the proglacial zone during ‘cold and wet’ conditions, and sediment was stored
in low gradient reaches (Orwin and Smart 2004). ‘Storm’ events increased SSL by more than
400% of the seasonal mean in their north stream, but just c. 100% in the central stream.
Although the results show that sediment was stored on the north stream during ‘hot and dry’
conditions, there was a c. 600% increase in SSL on the central stream. Orwin and Smart
(2004a) attribute the increase to elevated meltwater flow that mobilized sediment stored
within the channel and triggered episodic bank collapses. Their combined evidence showed
that short-term storage and release of sediment in the proglacial channels at the SRG

96
controlled the suspended sediment response pattern. Richards and Moore (2003) also found
that fine sediment was temporarily stored between proximal and distal sites at low flow, and
then re-entrained at higher flow.
Overall, the differences between the results from the CCG and SRG can be partially
explained by the meteorological conditions at the time of monitoring (i.e. the SRG data set
was more driven by ‘ablation’ (55%) than the CCG data set (30%), and data collection
started earlier in the ablation season). However, considering the differences between the two
streams that were monitored at the SRG and the results from the CCG, the influence of
catchment characteristics is highlighted. Particularly the slope and elevation range o f the
proglacial zone, the type and size of glacier, the stability of proglacial sediment deposits and
their connectivity to the fluvial system; which are factors that have been identified previously
(Harbor and Warburton 1993; Gumell et al. 1996).

4.2

Proglacial Suspended Sediment Budget
A sediment budget is a useful management tool for assessing where, when and how

much sediment is being entrained, transported, stored and evacuated from a watershed to the
downstream aquatic environment (Slaymaker 2003; Owens 2005; Walling and Collins 2008).
However, sediment budgets can be extremely difficult to establish; even for the c. 16 km2
proglacial catchment in this study. Due to limited field capacity and time constraints, detailed
examination o f sediment sources was not possible. Field data and observations were used to
infer sediment flux processes between the sites. With the information collected, a basic
proglacial suspended sediment budget was developed to ascribe SSL to various sources based
on the parameters defined by Warburton (1990) for the coarse and fine components o f a

97
proglacial sediment budget (Equation I, section 1.6.1). It was assumed that the error was
normally distributed within the data and that the relative accuracy was valid; however,
limitations and caveats apply regarding precision {section 3.1).
The suspended sediment contribution from the three proximal sites was a combination
of direct input from the glacial meltwater stream (GL), and input from moraine deposits at
the terminus (M). Based on field observations, PS1 was predominantly M, while PS2 was
mostly (80%) GL and PS3 may be split equally. Based on stream slope and field
observations, change in valley sandur (AVS) occurred on the outwash fan complex upstream
of MS. With the exception of ‘storm’ events, there was sediment stored upstream of MS,
which can be defined as the difference between the SSL input from the proximal sites and the
SSL measured at MS. Direct hillslope inputs (SL) were observed along the right bank o f the
meltwater channel, upstream from the outwash fan and immediately downstream from the
confluence of PS2 and PS3. The SL contribution was typically small, but episodic increases
can be expected when triggered by high streamflow, precipitation, overland flow or spring
snowmelt and freeze-thaw cycles. Tributary channel inputs (TR) to the CCG meltwater
stream predominantly came from RPT, and were substantial throughout the field season. The
total yield (Y) from the proglacial catchment was measured at DS. Following this premise,
the seasonal suspended sediment load budget for the 2011 Castle Creek Glacier proglacial
zone was defined as:

Y (100%) = GL (39%) + M (20%) + AVS (-7%) + SL (0.5%) + TR (49%)

(7)

98
Values were used directly or subdivided, as stated in the preceding paragraph, from
the seasonal percentage of mean daily SSL (Table 4.3) and, following the same premise, a
sediment budget could be drawn from Table 4.3 for any o f the meteorological categories.
Equation 7 has a slight (1.5%) surplus of sediment input, thus, suggesting additional storage
of sediment in the proglacial zone; however, given the error and uncertainty of the data
collected, this level of precision is false accuracy and it may be reasonable to accept a net
balance.
In the overall suspended sediment budget for the Castle Creek proglacial zone, TR
from RPT was the dominant source of sediment. Aside from RPT, virtually all of the
sediment enters the proglacial stream through GL and M in the active meltwater channels at
the snout o f the glacier, accounting for 59% of the SSL input. For the seasonal budget, AVS
was computed as -7%, showing sediment storage on the fan. The balance o f contribution
shifts moderately under different meteorological and streamflow conditions (Tables 4.3, 4.4,
and 4.5); however, ‘storm’ events were remarkably different and sediment contribution from
the proximal sites (GL and M) became less important as SL and AVS increased. When TR
from RPT is removed from the budget, the three proximal sites contributed a minimum of
76% during ‘storm’ events, a maximum of 133% during ‘hot and dry’ and a seasonal average
of 115% of the SSL at DS. Thus, M and GL contributed 100% of the SSL measured at DS
during the 2011 field season and excess SSL went into storage as AVS, with the exception of
‘storm’ events when 24% of the SSL is derived from SL and AVS. The CCG pilot study in
2008 found that there was an enrichment of 35% between MS and DS. This enrichment was
most likely TR from RPT (which was not quantified in 2008), not diffuse sediment sources
in the proglacial zone or the proglacial channel. In 2011, the enrichment in the same reach

99
was c. 49% as a seasonal average, which was all attributed to TR from RPT. As stated
previously, the exception is ‘storm’ events when ephemeral channels are actively
contributing SSL in the MS-DS reach.
Intensive field measurements were conducted by Warburton (1990) to define the
proglacial fluvial sediment budget for JD 134 - JD 211 o f the 1987 ablation season at the Bas
Glacier d’Arolla, Switzerland. The sediment yield was measured at proximal and distal ends
of a 300 m proglacial reach. At the distal site, the catchment area was c. 8 km2 and 70%
glaciated. Using various sampling approaches, Y, SL, TR, M and AVS were measured or
estimated. Proglacial sediment sources contributed 23% of the sediment received at the
catchment outlet, and 95% of that contribution was generated from bank and channel erosion
o f valley sandur during a short period of meltwater flooding from JD 197 - JD 199. While
SL and TR accounted for a small percentage of the total SSY, the GL and M contribution
accounted for c. 77%. The AVS was of overwhelming importance in modifying the sediment
load from GL and four basic fluvial process subsets were identified: 1) channel marginal; 2)
channel; 3) hillslope; and 4) slopewash. Since the GL component was estimated by
quantifying the other variables and subtracting their total from the overall sediment yield, Y,
the budget was not truly “closed”, and the cumulative error in the measurement o f the other
terms o f the equation made the estimate precise to only ± 26% (Warburton 1990).
Work by Hammer and Smith (1983) on the Hilda Glacier in Alberta found that the
proglacial area between a monitoring site at the snout of the glacier and a site 1 km
downstream supplied c. 50% of the total SSL. Orwin and Smart (2004a) found that SL and
AVS in the proglacial zone were the source of 80% o f the suspended sediment flux for the
central stream, and 30% for the north stream during the 2000 ablation season at the SRG

100
(Appendix 7.5). They cite sediment availability within the proglacial channels, SSC and Q of
glacial inputs, and contribution from extra-channel sediment sources as key differences
between the streams they monitored.
The SRG is a small (c. 7 km2) cirque glacier with a relatively steep and small
proglacial zone (c. 14% and 2 km2) compared to that of the CCG (c. 3% and 6 km2) which is
an alpine valley glacier (c. 16 km2). Also, the deglacierized study area at the SRG had a
greater elevation range (c. 450 m) compared to the CCG (c. 70 m). The differences in the
characteristics of the glacier and study site may partially explain the contrasting results.
However, inter-annual variability o f hydro-meteorological conditions and antecedent
conditions such as seasonal snowpack can strongly influence proglacial Q and SSC which
would affect the results o f the analyses and thus comparisons between different sites and
years of data (Gumell et al. 1996; Richards and Moore 2003; Swift et al. 2005; Cockbum
and Lamoureux 2008; Haritashya et al. 2010).

101

5

Conclusions

5.1

Summary
Divergence between the Q and SSC time-series data sets is common for all sites

(Figures 3.1 and 3.5), and it is clear that there is no simple relationship between Q and SSC;
thus it is clear that the two variables need to be monitored independently. If Q and SSC were
not monitored independently, there would be significant shortcomings in the analysis and the
level o f detail in the results and discussion section would not be possible. Additionally, the
computation of SSL from SSC and Q is much more accurate when the variables are
independent.
Objective 1 was to examine the influence of hydro-meteorological conditions on the
spatial and temporal pattern o f suspended sediment flux in the proglacial zone. The time
series were divided into hydrologic days and principal component analysis and cluster
analysis were used to categorize days o f like conditions and similar suspended sediment
response ‘shape’ and ‘magnitude’. Each field season and glacier will be different, and this
2011 data set is a snap shot in time of the conditions in the proglacial zone of the Castle
Creek Glacier. From this field season, the following conclusions can be made for the sites
along the main Castle Creek meltwater channel:

•

‘warm and damp’ conditions showed a mixed SSC response pattern that was
influenced by antecedent conditions;

•

‘hot and dry’ conditions tended to generate a Tow’ magnitude, ‘diurnal’ SSC
response;

102
•

‘cold and wet’ conditions tended to generate ‘medium’ or ‘low’ magnitude,
‘irregular’ SSC response;

•

‘storm’ events generated a ‘high’ magnitude, ‘irregular’ response and increased the
daily mean SSL by more than 500% of the seasonal mean.
The two small proximal streams, PS1 and PS3, had more ‘diurnal’ data in the ‘cold

and wet’ and ‘warm and damp’ conditions, but tended to have more ‘irregular’ data during
‘hot and dry’ days than the sites along the main meltwater channel. While the response
‘magnitude’ at RPT during ‘cold and wet’ and ‘warm and damp’ conditions tended to be
lower than PS1 and PS3, the response ‘shape’ was similar. During ‘hot and dry’ conditions,
there were more ‘irregular’ data at RPT. The ‘magnitude’ analysis was performed
independently for each station and was influenced by the range o f the data within the input
matrix. It was difficult to make comparisons across sites from this analysis, but the SSC data
for each site were successfully categorized into ‘high’, ‘medium’, or ‘low’ magnitude
response with respect to the rest of the SSC data collected at the site.
Objective 2 was to determine the sources of Q and SSL under different hydro­
meteorological conditions. Aside from the tributary input, RPT, the data collected during the
2011 field season show that glacial stream inputs and the area proximal to the snout of the
glacier, exposed within the last few years, were the dominant source o f suspended sediment
for the majority of the field season. More than 20% o f the total SSL evacuated during the
field season occurred during two ‘storm’ event days that represented 3% o f the field season.
During ‘storm’ events, diffuse and point sediment sources throughout the proglacial zone and
meltwater channels were activated by intense precipitation or high streamflow. Based on
these results, we can expect SSL to be sustained with episodic pulses o f high SSL from the

103
proglacial zone until the completion of deglaciation. Once the catchment has become
deglaciated, we can expect episodic pulses of elevated SSL associated with storm events and
snowmelt that activate exposed sediment sources in the previously glaciated area. These
episodic pulses of high SSL are likely to continue in declining magnitude until the
paraglacial period has ended and erosion rates are no different than that of an un-glaciated
mountainous catchment (Ballantyne 2002a; Richards and Moore 2003).
This project has built upon the work o f past researchers (e.g. Church and Ryder 1972;
Hammer and Smith 1983; Warburton 1990; Gumell et al. 1992; Harbor and Warburton 1993;
Hodson et al. 1998; Hannah et al. 2000; Swift et al. 2002; Hodgkins et al. 2003; Richards
and Moore 2003; Orwin and Smart 2004a, 2004b; Swift et al. 2005; Jobard and Dzikowski
2006; Stott et al. 2007; Stott et al. 2009; Moore et al. 2009) adding to the growing body o f
knowledge surrounding deglacial and proglacial processes and paraglacial sedimentation
from contemporary glacial environments (Ballantyne 2002b). Most glaciers have experienced
rapid recession and/or volume loss since the end of the LIA, and that trend will likely
continue at an accelerating rate, even under the most conservative climate projections
(Collins et al. 2013; Kirtman et al. 2013). Glaciers are a major erosive force that increase
sediment load to the downstream fluvial system, especially during periods of rapid advance
or retreat (Menounos et al. 2009). The rate o f glacial erosion will likely decrease as the
overburden o f ice wanes (Church and Ryder 1972; Ballantyne 2002a); however, exposed
sediment and channels in proglacial zones have been found to be a significant or dominant
source of fine sediment during the ablation season in glaciated catchments (Gumell et al.
1999; Orwin and Smart 2004a; 2004b). Fine sediment affects water quality and is the
chemically active component of sediment transported by the fluvial system (Hodson et al.

104
2004; Owens et al. 2005). Water quantity and quality are central to life on earth and the
information collected during this project may provide useful input parameters for paraglacial
sedimentation in water quality modeling for glacially influenced tributaries o f socially and
ecologically important watersheds, such as the Fraser River basin.

5.2

Limitations
Equipment limitations have compromised the quality of the data that were collected

for this study. The two primary issues were the limited number of WL loggers, and the
limited range o f the Tu data loggers. The four WL loggers available for the study were
strategically distributed across six monitoring locations and the WL record for two sites, RPT
and PS3, was deduced as the difference between the records collected at the other sites. This
method strictly assumes that there is no other source o f water entering the channel between
the sites and neglects hyporehic interactions. The range of the Tu probes was limited by the
data loggers that they were connected to; respectively, their range was 0 - 4V and 0 - 2.5V.
Thus, the maximum recorded value was 2.5V, which equates to a site specific SSC (Table
2.2). The range was exceeded at four of the sites (Table 3.2) and missing data was estimated
based on composite samples.
Another limitation was encountered with the stream gauging procedure; the main
meltwater channel was unsafe to wade when streamflow was greater than 5 m3/s. This
limitation was partially overcome through rating curve extensions and slope-area estimation
of peak flow during a high water event. Salt dilution gauging may have been a suitable way
to measure flow up to 20 m3/s (Hudson and Fraser 2005). Obtaining higher Q measurements
would have increased the confidence in, or eliminated the need for rating curve extensions.

105
However, the amount o f data that was affected by this limitation was relatively small. The
distal site was the worst case, where 13% of the data were greater than the maximum gauged
flow, but considered ‘valid’ based on the rating curve extension, and 0.6% o f the data were
considered an ‘estimate’ since they were greater than two times the maximum gauged flow
(WSC 2012).
Aside from the equipment limitations, the study design would have been improved by
moving the monitoring site on RPT to the main meltwater channel just upstream o f the
confluence with RPT. Doing this would have reduced the amount of uncertainty in the reach
MS-DS. The difference between DS and the hypothetical site just upstream of the confluence
with RPT would have been used to deduce the contribution from RPT. This method would
have included any other contributions between the RPT confluence with Castle Creek and DS
with RPT, but would have avoided much of the Tu range exceedances. Site maintenance and
data collection for RPT was made difficult by high sediment load, low stream volume, and
stream bed aggradation. The computations and data analysis for RPT have been the most
challenging and required the most exceptions to analytical and computational procedures of
all of the sites.
The hydrologic data for RPT were deduced as the difference between MS and DS,
which lumps all inputs within the MS-DS reach onto one parameter, RPT. This method is
flawed when ephemeral streams are flowing into the MS-DS reach. During storm events, the
computations showed that these ephemeral streams contributed c. 25% o f the SSL at DS.
However, there was no streamflow attributed to these streams in the computations. Field
observations were used to estimate the flow volume at 5% of the flow at DS. Overestimating
the flow from RPT would have caused the SSL to be overestimated in the computations, and

106
thus the contribution from ephemeral streams to be underestimated. The SSC data collected
at RPT and Q derived from the difference between MS and DS has helped refine suspended
sediment sources within the MS-DS reach, but there is room for improvement.
The collected data for DS and MS show that there was a SSL enrichment of c. 49%
and a c. 13% increase in Q in the reach, which varied with flow conditions and as the season
progressed. Most of this SSL came from RPT. However, during ‘storm’ events and, likely,
early season snowmelt, ephemeral channels that drain diffuse unstable sediment sources in
the proglacial zone were also a significant source of sediment to the MS-DS reach.
Error and uncertainty in the data and the analysis were mostly addressed with caveats,
using values reported by manufacturers or the work of past researchers as guidance. The 95%
confidence intervals were computed and reported for the SSC data; however, the uncertainty
terms for SSL values were not reported. These terms will be computed and included for
future publications.
The field season did not start early enough to capture spring snowmelt due to
equipment logistics and availability of the field team. At the beginning o f data collection for
the 2011 field season (JD 195), much o f the annual snow had already melted from the
proglacial zone. It would have been valuable to capture the entire ablation season in the data
set.
The diurnal hydrograph was divided at 06:00; however, 08:00 may have better
represented the average time of minimum flow. The effect of this difference on the results
may be minimal, but it might have made parts of the ‘shape’ and ‘magnitude’ analysis more
clear.

107
5.3

Recommendations for Future Research
The proglacial zone o f the CCG has a relatively low gradient in comparison to that of

the SRG, and it is possible that these two projects could serve as relative endpoints in
proglacial sediment modelling for the upper Fraser River basin. As stated previously, and by
other researchers, the conditions vary widely from year to year, and additional years o f data
may be necessary to test the repeatability o f the results. For this type of data, a consistent
statistical analysis technique will improve the comparability of results. The PCA and CA
technique adapted by Orwin and Smart (2004a) from Hannah et al. (2000) was an effective
method for classifying and categorizing the data; however, parts of the SSC shape and
magnitude analysis were subjective, gave unclear results or were not comparable across sites.
The most useful and widely applicable part of the analysis may be dividing and summarizing
the SSL data based on hydro-meteorological conditions. When the data were summarized in
this way it was easier to compare like-conditions from different locations, or to draw
information for climate and sediment flux models, such as BQART (Syvitski and Milliman
2007), to help project sediment yield from similar glacially influenced catchments.
The focus of this study was on the sources and flux o f suspended sediment within the
CCG proglacial zone. However, simplifying the study design to collect SSC and Q data from
one site at the outlet of several different proglacial zones over one or a few field seasons
could provide a valuable and interesting data set for comparison purposes and regional
modelling applications. Having SSL data from the same field season at the outlet o f several
different proglacial zones within a specific region would reduce the effect o f inter-annual
variability of hydro-meteorological conditions when making comparisons and allow the
effects of catchment characteristics on suspended sediment production to be isolated. The

108
PCA and CA analysis that was followed for objective two of this study would be an
applicable and useful comparison tool. Such a project may permit the development o f more
spatially precise and temporally responsive proglacial suspended sediment production
parameters for modeling applications.

109

6

Bibliography

Amott B. 2011. Personal Communication on site at Castle Creek Glacier, British Columbia, August 2011.

Austin MA, Buffet DA, Nicholson DJ, Scudder GGE, Stevens V (eds.). 2008. Taking Nature’s Pulse: The
Status o f Biodiversity in British Columbia. Biodiversity BC, Victoria, BC. 268 pp. Available at:
www.biodi versitvbc. ore.

Ballantyne CK. 2002a. A general model o f paraglacial landscape response. TheHolocene 12 : 371 - 3 7 6 .

Ballantyne CK. 2002b. Paraglacial geomorphology. Quaternary Science Reviews 2 1 : 1935 - 2017.

Beedle MJ, Menounos B, Luckman BH, Wheate R. 2009. Annual push moraines as climate proxy. Geophysical

Research Letters 36, L20501.

Bogen J. 1989. Glacial sediment production and development o f hydro-electric power in glacierised areas.

Annals of Glaciology 13: 6 - 11.

Bolch T, Menounos B, Wheate R. 2010. Landsat-based inventory o f glaciers in western Canada, 1985 - 2005.

Remote Sensing of the Environment 114 : 1 2 7 - 137.

Brown GH, Tranter M, Sharp MJ. 1996. Experimental investigations o f the weathering o f suspended sediment
by alpine glacial meltwater. Hydrological Processes 10 : 579 - 597.

Church M, Ryder JM. 1972. Paraglacial sedimentation: A consideration o f fluvial processes conditioned by
glaciation. Geological Society o f America Bulletin 83: 3059 —3072.

Church M, Slaymaker 0 . 1989. Disequilibrium ofH olocene sediment yield in glaciated British Columbia.

Nature 337: 452-454.

Cockburn JMH, Lamoureux SF. 2008. Hydroclimatic controls over seasonal sediment yield in two adjacent
high arctic watersheds. Hydrological Processes 2 2 : 2013 —2027.

110

Collins M, Knutti R, Arblaster J, Dufresne J-L, Fichefet T, Friedlingstein P, Gao X, Gutowski WJ, Johns T,
Krinner G, Shongwe M, Tebaldi C, Weaver AJ, Wehner M. 2013. Long-term climate change:
projections, commitments and irreversibility. In: Climate Change 2013: The Physical Science Basis.

Contribution o f Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on
Climate Change. Stocker TF, Qin D, Plattner G-K, Tignor M, Allen SK, Boschung J, Nauels A, Xia Y,
Bex V, Midgley PM (eds.). Cambridge University Press, Cambridge, United Kingdom and N ew York,
N Y , USA.

Dery SJ. 2011. Personal Communication at University o f Northern British Columbia.

Dery SJ, Stahl K, Moore RD, Whitfield PH, Menounos B, Burford JE. 2009. Detection o f runoff timing changes
in pluvial, nival and glacial rivers o f western Canada. Water Resources Research 45: W 04426, DOI:
10.1029/2008W R006975.

Dirszowsky RW. 2004. Bed sediment sources and mixing in the glacierised upper Fraser River watershed, eastcentral British Columbia. Earth Surface Processes andLandforms 29: 533 - 552.

Eaton BC, Moore RD, Giles TR. 2010. Forest fire, bank strength and channel instability: the ‘unusual’ response
ofFishtrap Creek, British Columbia. Earth Surface Processes and Landforms 35: 1167 - 1183.

Filliben JJ. 1975. The probability plot correlation coefficient test for normality. Technometrics 17: 1 1 1 - 1 1 7 .

Filliben JJ, Devaney J. 2013. Critical values o f the normal PPCC distribution. NIST/SEMATECH e-Handbook

of Statistical Methods, http://www.itl.nist.gov/div898/handbook/eda/section3/eda3676.htm. Accessed
01-03-2014.

Ginting D, Mamo M. 2006. Measuring runoff-suspended solids using an improved turbidometer method.

Journal of Environmental Quality 35: 815 - 823.
Gippel CJ. 1989. The use o f turbidimeters in suspended sediment research. Hydrobiologia 176/177: 465 - 4 8 0 .

Gumell AM, Fenn CR. 1984. Box-Jenkins transfer function models applied to suspended sediment
concentration-discharge relationships in a proglacial stream. Arctic and Alpine Research 16: 93 - 106.

Ill
Gumell AM, Clark MJ, Hill CT, Greenhalgh J. 1992. Reliability and representativeness o f a suspended
sediment concentration monitoring program for a remote alpine proglacial river. International

Association of Hydrological Sciences Publication 210 : 191 - 200.

Gumell AM, Hannah D, Lawler D. 1996. Suspended sediment yield from glacier basins. Erosion and Sediment

Yield: Global and Regional Perspectives (Proceedings o f the Exeter Symposium, July 1996). JAHS
236 , 1996.

Gumell AM, Edwards PJ, Petts GE, Ward JV. 1999. A conceptual model for an alpine proglacial river channel
evolution under changing climatic conditions. Catena 38: 223 - 242.

Hammer KM, Smith ND. 1983. Sediment production and transport in a proglacial stream: Hilda Glacier,
Alberta, Canada. Boreas 12 , 91 —106.

Hannah DM, Smith BPG, Gumell AM, McGregor GR. 2000. An approach to hydrograph classification.

Hydrological Processes 14: 317 —338.

Hallet B, Hunter L, Bogen J. 1996. Rates o f erosion and sediment evacuation by glaciers: a review o f field data
and their implications. Global and Planetary Change 12:213 - 235.

Harbor J, Warburton J. 1993. Relative rates o f glacial and nonglacial erosion in alpine environments. Arctic and

Alpine Research 2 5 : 1 - 7 .

Haritashya UK, Kumar A, Singh P. 2010. Particle size characteristics o f suspended sediment transported in
meltwater from the Gangotri Glacier, central Himalaya - An indicator o f subglacial sediment
evacuation. Geomorphology 122 , 1 4 0 - 152.

Hodgkins R, Cooper R, Wadham J, Tranter M. 2003. Suspended sediment fluxes in a high-arctic glacierised
catchment: implications for fluvial sediment storage. Sedimentary Geology 162: 105 - 117.

Hodson A, Gumell A, Tranter M, Bogen J, Hagen JO, Clark M. 1998. Suspended sediment yield and transfer
processes in a small High-Arctic glacier basin, Svalbard. Hydrological Processes 12 : 73 - 86.

112
Hodson AJ, Ferguson RI. 1999. Fluvial suspended sediment transport from cold and warm-based glaciers in
Svalbard. Earth Surface Processes and Landforms 24: 957 - 974.

Hodson A, Mumford P, Lister D. 2004. Suspended sediment and phosphorus in proglacial rivers: bioavailability
and potential impacts upon the P status o f ice-marginal receiving waters. Hydrological Processes 18:
2409 - 2422.

Hodson A, Anesio AM, Tranter M, Fountain A, Osborn M, Priscu J, Layboum-Parry J, Sattler B. 2008. Glacial
ecosystems. Ecological Monographs 78(1): 41 - 6 7 .

Hudson R, Fraser J. 2005. Introduction to salt dilution gauging for streamflow measurements part IV: the mass
balance (or dry injection) method. Streamline Watershed Management Bulletin 9 / 1 : 6 - 12.

Jacobs JL, Dinman JD. 2013. Systematic analysis o f bicistronic reporter assay data.
http://dinmanlab.umd.edU/statistics/tutorial.html#Q. Accessed 02-04-2013.

Jobard S, Dzikowski M. 2006. Evolution o f glacier flow and drainage during the ablation season. Journal of

Hydrology 330: 6 6 3 - 6 7 1 .

Kirtman B, Power SB, Adedoyin JA, Boer GJ, Bojariu R, Camilloni I, Doblas-Reyes FJ, Fiore AM, Kimoto M,
Meehl GA, Prather M, Sarr A, Schar C, Sutton R, van Oldenborgh GJ, Vecchi G, Wang HJ. 2013.
Near-term climate change: projections and predictability. In: Climate Change 2013: The Physical

Science Basis. Contribution of Working Group I to the Fifth Assessment Report o f the
Intergovernmental Panel on Climate Change. Stocker TF, Qin D, Plattner G-K, Tignor M, Allen SK,
Boschung J, Nauels A, Xia Y, Bex V, Midgley PM (eds.). Cambridge University Press, Cambridge,
United Kingdom and N ew York, NY, USA.

Lawson DE. 1995. Sedimentary and hydrologic processes within modem terrestrial valley glaciers. In Modem

Glacial Environments Processes, Dynamics and Sediments. Menzies J (ed.). Glacial Environments:
Volume 1, Butterworth-Heinemann; Ch. 11: 337 - 362.

McVan 2003. Analite 180 and 190 series manual. Copyright 1995-2003 - McVan Instruments PTY. LTD.

113
Menounos B, Clague JJ, Gilbert R, Slaymaker O. 2005. Environmental reconstruction from varve network in
the southern Coast Mountains, British Columbia, Canada. The Holocene 15: 1163 - 1 1 7 1 .

Menounos B, Osborn G, Clague JJ, Luckman BH. 2009. Latest Pleistocene and Holocene glacier fluctuations in
western Canada. Quaternary Science Reviews 28: 2049 - 2074.

Milner AM, Brown LE, Hannah DM. 2009. Hydroecological response o f river systems to shrinking glaciers.

Hydrological Processes 23: 62 - 77.

Moore RD, Fleming SW, Menounos B, Wheate R, Fountain A, Stahl K, Holm K, Jakob M. 2009. Glacier
change in western North America: influences on hydrology, geomorphic hazards and water quality.

Hydrological Processes 23 :4 2 - 61.

Morehead MD, Syvitski JP, Hutton EWH, Peckham SB. 2003. Modeling the temporal variability in the flux o f
sediment from ungauged river basins. Global Planetary Change 39: 95-110.

Naden PS. 1988. Modeling Geomorphic Systems, Chapter 8: Models of suspended sediment in natural streams,
edited by M. G. Anderson, John W iley & Sons Ltd.

Navratil O, Esteves M, Legout C, Gratiot N, Nemery J, Willmore S, Grangeon T. 2011. Global uncertainty
analysis o f suspended sediment monitoring using turbidimeter in a small mountainous river catchment.

Journal o f Hydrology 398: 246 —259.

O ’Connor BP. 2000. SPSS and SAS programs for determining the number o f components using parallel
analysis and Velicer’s MAP test. Behavior Research Methods, Instrumentation, and Computers 32(3):
396-402.

Onset 2013. Hoboware User’s Guide, www.onsetcomp.com.

Orwin JF, Smart CC. 2004a. Short-term spatial and temporal patterns o f suspended sediment transfer in
proglacial channels, Small River Glacier, Canada. Hydrological Processes 18: 1521 - 1542.

Orwin JF, Smart CC. 2004b. The evidence for paraglacial sedimentation and its temporal scale in the
deglacierizing basin o f Small River Glacier, Canada. Geomorphology 58: 175 - 202.

114
Orwin JF, Smart CC. 2005. An inexpensive turbidimeter for monitoring suspended sediment. Geomorphology
6 8 :3 -1 5 .

Owens PN. 2005. Conceptual models and budgets for sediment management at the river basin scale. Journal of

Soils and Sediments 5: 201 - 212.

Owens PN, Batalla R, Collins AJ, Gomez B, Hicks DM, Horowitz AJ, Kondolf GM, Marden M, Page MJ,
Peacock DH, Petticrew EL, Salomons W, Trustrum NA. 2005. Fine-grained sediment in river systems:
environmental significance and management issues. River Research and Applications 21: 693 - 717.

Pickup G. 1988. Modeling Geomorphic Systems, Chapter 7: Hydrology and sediment models. Anderson MG
(ed.). John W iley & Sons Ltd.

Resource Inventory Standards Committee (RISC). 2009. Manual o f British Columbia Hydrometric Standards.

Prepared by the Ministry of Environment Science and Information Branch. V 1.0.
http://www.ilmb.gov.be.ca/risc. Accessed 07-06-11.

Richards G, Moore RD. 2003. Suspended sediment dynamics in a steep, glacier-fed mountain stream, Place
Creek, Canada. Hydrological Processes 17: 1733 - 1753.

Ritter D, Kochel C, Miller J. 2002. Process Geomorphology. Fourth Edition. McGraw-Hill Higher Education.

Schiefer E, Slaymaker O, Klinkenberg B. 2001. Physiographically controlled allometry o f specific sediment
yield in the Canadian Cordillera: a lake sediment-based approach. Geografiska Annaler 83(A): 55 - 65.

Schiefer E, Menounos B, Wheate R. 2007. Recent volume loss o f British Columbian glaciers, Canada.

Geophysical Research Letters 34: L16503. doi: 10.1029/2007GL030780, 2007.

Scheifer E, Hassan MA, Menounos B, Pelpola CP, Slaymaker O. 2010. Interdecadal patterns o f total sediment
yield from a montane catchment, southern Coast Mountains, British Columbia, Canada.

Geomorphology 118: 2 0 7 - 2 1 2 .

Shea J. 2011. Personal Communication at University o f Northern British Columbia, Prince George.

115
Slaymaker O. 2003. The sediment budget as conceptual framework and management tool. Hydrobiologia 494:
7 1 -8 2 .

Stott TA, Grove J. 2001. Short-term discharge and suspended sediment fluctuations in the proglacial Skeldal
River, NE Greenland. Hydrological Processes 15:407 - 423.

Stott TA, Mount NJ. 2007. Alpine proglacial suspended sediment dynamics in warm and cool ablation seasons:
Implications for global warming. Journal o f Hydrology 332: 259-270.

Stott TA, Owens PN, Forrester BJ, Lee J. 2009. Suspended Sediment Fluxes in the Castle Creek Glacier
Proglacial Zone, Cariboo Mountains, British Columbia. Innovations in Practice 3:49-70.

Stott T, Nuttall A-M, Biggs E. 2014. Observed run-off and suspended sediment dynamics from a minor
glacierised basin in south-west Greenland. Geografisk Tidsskrift-Danish Journal o f Geography , doi:
10.1080/00167223.2013.862911.

Swift DA, Nienow PW, Spedding N, Hoey TB. 2002. Geomorphic implications o f subglacial drainage
configuration: rates o f basal sediment evacuation controlled by seasonal drainage system evolution.

Sedimentary Geology 149: 5 - 1 9 .

Swift DA, Nienow PW, Hoey TB, Mair DWF. 2005. Seasonal evolution o f runoff from Haut Glacier d’Arolla,
Switzerland and implications for glacial geomorphic process. Journal of Hydrology 309: 133 - 148.

Swada M, Johnson PG. 2000. Hydrometeorology, suspended sediment and conductivity in a large glacierised
basin, Slims River, Yukon Territory, Canada (1993 - 94). Arctic 53: 101-117

Syvitski JPM, Milliman JD. 2007. Geology, geography, and Humans battle for dominance over the delivery o f
fluvial sediment to the coastal ocean. The Journal o f Geology 115: 1 - 19.

Tabachnick BG, Fidell LS. 1989. Using multivariate statistics, second edition. Harper Collins Publishers, Inc.

Tennant C, Menounos B, Wheate R, Clague JJ. 2012. Area change o f glaciers in the Canadian Rocky
Mountains, 1919 to 2006. The Cryosphere 6(6): 1541-1552.

116
Tunnicliffe JF, Church M. 2011. Scale variation o f post-glacial sediment yield in Chilliwack Valley, British
Columbia. Earth Surfaces Processes and Landforms 36: 229 - 243.

Walling DE. 1978. Reliability considerations in the evaluation and analysis o f river loads. Zeitschrift fur

Geomorphologie, Supplementband 29: 2 9 - 4 2 .

Walling DE. 2005. Tracing suspended sediment sources in catchments and river systems. Science of the Total

Environment 344: 159 - 184.

Walling DE, Collins AL. 2008. The catchment sediment budget as a management tool. Environmental Science

and Policy IT. 1 3 6 -1 4 3 .
Warburton J. 1990. An alpine proglacial fluvial sediment budget. Geografiska Annaler 72(A): 261 - 272.

Water Survey o f Canada (WSC). 2012. Hydrometric manual: data computations. Weather and Environmental
Monitoring Directorate, issued under the authority o f the Assistant Deputy Minister, Meteorological
Services o f Canada. qSOP-NA037 (Beta Version) 2012-12-17. Water Survey o f Canada.
http://www.wsc.ec. gc. ca

117

118

7

Appendix

7.1

Hydrometric
Barometric Data:

Atmos. Pros. CCG 2011
Atm. PfM. Uppor CCQ (raw)

Atm. P m . Upper CCG (corr.)

Atm. Pms. Lower {raw}

800

830

795 -

825

790

820

mbar

E

e

785

780

810

778

80S

Jul 12

15

18

21

24

27

30

2011

Aug 05 08

11

14

17

20

23

02
2011

26

29

Sep

04

07

10

13

01
2011

Figure 7.1 Barometric pressure from Castle Creek Glacier upper and lower m eteorological stations.
Rate o f change threshold used to remove erroneous raw data, dnie-series averaged to give
corrected tiine-series for use in computations.

Rating Curves:
The relationship between Q and WL can change over time, and thus, it is necessary to
apply ‘shifts’ to the base rating curve to compensate for changes in water level reference
point, section control, or channel control. A change in the WL

reference

pointwill affect the

entire relationship by the same valueand can be compensated for with a single offset shift.
Section control often defines the lower end o f the rating curve and is affected by scour and
aggradation processes that have a greater influence on the lower end of the relationship.

Changes to section control can be compensated for with “knee-bend” shifts; however, the
point at which the shift blends back onto the base rating curve may be difficult to define with
limited data. Channel control defines the upper end o f the rating curve and refers to the
general configuration of the stream such as flood plain elevations, breakpoints, channel slope
and width which tend to be more stable over time.

(1.4

0.38
0.36
0.34

£
•

0.32
0.3

£

5 0.28
0.26
0.24
0.22
0.2
0.0

0.2

0.4

Discharge (mA3/s)

Figure 7.2 PS1 rating curve. No shifts applied.

0.6

120

1.2

f

0.8

«o>

| 0.6
0.4

0.2
Discharge (mA3/s)

Figure 7.3 PS2 rating curve. The shift applies to data after Aug. 22, 2011; the site was moved just before
the high flow event. Data considered an estim ate above 10 nr/s.

1.2

0.8

?
3> 0.6

a
0)

0.4
0.2

Discharge (mA3/s)

Figure 7.4 M S Rating curve. The shift accounts for stilling well m ovement during the Aug. 22 event, and
is applied to data t hereafter. Dat a considered an estim ate above 10 nr/s.

121

0.8

E
V 0.6
CD

3
<n

0.4

0.2

Discharge (mA3/s)

Figure 7.5 DS rating curve. No shifts applied. The two grey rating points are outside acceptable range o f
5% , and were not used for rating curve developm ent. Data considered an estim ate above 10 mVs.

7.2

Suspended Sediment
Floating apparatus used to collect Tu and SSC samples:

122
Tu - SSC computations
Table 7.1 Fourth-spread method Tu - SSC

Site

PS1*

PS2

PS3*

MS

RPT

DS

q(100)

707.41

2506.653

585.65

3205.77

997.77

970.72

q(75)

399.40

356.01

263.59

279.09

467.36

418.37

Median, x~

289.27

292.22

172.27

232.70

371.96

371.39

q(25)

270.67

254.29

138.96

198.35

249.41

311.89

q(o)
fourth
spread
upper
outlier
boundary
lower
outlier
boundary
Number of
samples (n)
Outliers?
Actual
Sample Size
(N)
mean, xbar

177.31

128.22

100.03

96.50

111.08

76.46

128.73

101.72

124.63

80.74

217.95

106.48

482.36

444.81

359.21

353.81

698.88

531.11

96.18

139.64

-14.68

111.58

45.04

211.67

24

205

110

202

179

183

6
18

49
156

28
82

26
176

4
175

14
169

VAR(xbar)
STDEV(xbar)

279.73

274.82

161.15

230.09

361.57

361.32

2815.30
53.06

1392.81

1155.56

2249.15

4864.57

37.32

33.99

47.43

17315.3
131.59

12.51

2.99

3.75

3.57

9.95

69.75
5.37

0.9939

0.9960

0.9883

0.9943

0.9904

0.9940

0.9452

0.9913

0.9842

0.9921

0.9921

0.9919

Standard
Error,
se(xbar)
PPCC
Critical
Value (0.0S)
Normal?
Min.
Sample Size
(N~)

YES

YES
56

YES

YES
29

69

NO
66

YES
204

58

If Probability Plot Correlation Coefficient > Critical Value then we fail to reject that
the data is drawn from a population with a normal distribution.
*additional outlier (s) removed qualitatively prior to “F-S Method”, PS1 =1, PS3=8

123

Normal Probability Plots:

500

PS1

450
400
350

o

300

1/i

12

150
400

-

2

-

1

0

1

Expected z-score

450
PS2

400
050

00O

Expected z-score

2

124

300

PS3

250

000

3

2

1

0
Expected z-score

MS

050

100

Expected z-score

1

2

3

OJ

TSS/Tu Ratio

UJ

TSS/Tu Ratio

3D

fsj

a. o

LO

Expected z-score

n

fp
O
P

o

u>

P

126

PS1
24
n/a
6
18
Yes
305x-54
0.43
57.6

SSC Samples
C orroded samples
Outliers
Actual Sample (N)
Normality
Equation, y=
R value
95% C.I. (Ay)

PS2
205
47
2
156
Yes
252x+9
0.79
2.8

PS3
110
22
6
81
Yes
183x-28
0.85
9.1

RPT
179
n/a
4
175
No
469x-107
0.76
20.5

MS
202
23
3
176
Yes
213x+8
0.67
3.6

DS
183
n/a
14
169
Yes
413x-23
0.77
5.4

PS1

PS2

y - SSC (mg/L)aoB
acc

x-Tu(mV)

*"

«c

sec

o
+ Series

13

2

&

2.3

PS3

2

23
MS

so
4DC -

Note:
y-axis
scale
varies

SC
0

2

IS

13
1%

RPT

4se
200
S

2

ss

I

2

IS

8

03

Figure 7.6 Turbidity and suspeuded sedim ent concentration regressions; ail outliers presented

127

PS1

PS2

x - Tu (mVj

~T~

+ Series

23

13

53

1

13

2

13

MS

* Outliers

S O -

Note:
y-axis
scale
varies

i
M OB.

RPT

5

9 4

2

53

*

13

2

DS

♦

53

2

2

13

9

IS

2

Figure 7.7 Turbidity and suspended sedim ent concentration regressions. Corroded sam ples removed.

25

128

Proximal Site 1
1000
900
g

800

c
o

700

ra
c
V
u

RJ = 0.4308

600
♦

i

500

u

400

c

y = 304.59X - 54.185

PS1 Raw

U pper 95%

V

C.I.

E

Lower 95%

200

C.I.

100

■Linear (PS1
Raw)

0

1

0.5

1.5

2

2.5

3

Turbidity (V)

Proximal Site 2
350

|

300
E,

1§

250

c

200

y = 252.53x + 8.6323
R2 = 0.7901

PS2 Raw

100

- - U p p er 95%

VST**

♦♦

■o

- - Low er 95%

0.2

0.4

0.6
Turbidity (V)

0.8

1.2

Linear (PS2
Raw)

129

Proximal Site 3
_ 500

' ♦♦ "% f ..............
-

^ 450

I 400

I 350

♦

♦

*

y = 182.95X - 28.028
R2 = 0.8533

I 300

♦

8 250

♦
♦

| 200

#♦

PS3 Raw

♦ ♦
U pper 95%

C.l.

I 100

c«
a

♦

\ **

♦ ♦

1 150

♦

......... S & s *.......

■ . ........ ..

...........—

.....

Lower 95%
C.l.

50

0
0.5

1.5

• Linear (PS3
Raw)

2.5

Turbidity (V)

M iddle Site
250

y = 2 1 3 . 8 x + 8.0523
R2 = 0.6697

S 200

150
♦

MS Raw

100
U pper 95%
C.l.

T3

50

- - Lower 95%
C.I.
Linear (MS
R aw )

0
0

0.2

0.6

0.4
Turbidity (V)

0.8

1

130

Rockback Peak Tributary
1600
y = 4 6 8 .8 7 x - 106.81

1400

RJ = 0.7623

§ 1200
£ 1000
800

♦

RPT raw da ta

♦ ♦ ♦

% 600

U p p er 95% C.l.
♦♦

400

Lower 95% C.l.

- Linear (RPT raw
d a ta )

0

1

0.5

1.5

2

2.5

3

Turbidity (V)

Distal Site
600

ao

E 500

y = 4 1 3 .0 4 x - 23.256
R2 = 0.7723

400
OS Raw

300

u p p e r 95% C.l.

| 200

- -

TJ
i 100

0.000

Linear (DS
Raw)

*<*

0.200

Lower 95% C.l.

0.400

0.600

0.800

1.000

1.200

Turbidity (V)

Figure 7.8 - Turbidity and suspended sedim ent concentration regressions. Outliers removed, 95%
confidence interval included. Figures presented for all six sites over three preceding pages.

131
7.3

Meteorological
Table 7.2 M eteorological Parameters
j Met. Stn.

Upper

Upper

Upper

Lower

Lowe

Lowe

Uppe

r

r

r

Lower

Lower

Upper

W.Sp

j Paramete

AT

AT

AT

AT

AT

AT

PT

Rel.

T.Soi.

1r

min

max

mean

min

max

mean

mean

Hum.

Rad.

d.
mean

Cold and

Min

-1.9

1.6

-0.2

-0.3

4.9

2.5

0.8

76.3

6265

2.0

Max

3.9

14.2

6.7

4.4

9.2

6.7

19.1

90.1

22772

6.0

j Average

1.0

5.6

3.0

2.0

7.3

4.8

8.0

84.4

14553

4.1

j Std.Dev.

1.6

2.7

1.5

1.4

1.3

1.2

5.9

4.3

5133

1.1

j

Min

0.2

5.4

2.6

2.4

8.8

5.6

0.0

61.5

9326

1.7

Max

6.4

126

95

61

13.5

9.7

6.9

88.6

25597

5.5

1 Average

3.0

8.2

55

4.3

10.7

7.9

2.5

72.5

18195

3.6

j Std.Dev.

1.9

1.9

1.8

1.3

1.5

1.2

2.5

7.9

3658

1.0

j

Min

4.2

10.8

7.7

4.9

12.2

8.9

0.0

40.0

13792

2.4

j

Max

13.3

18.2

15.4

9.9

18.6

13.2

8.6

73.0

28782

7.5

\ Average

7.5

13.1

10.3

7.4

14.7

10.9

1.2

58.6

22041

3.6

| Std.Dev.

1.9

1.9

1.8

1.2

1.2

1.0

2.0

7.2

3718

1.0

5.7

11.5

7.9

4.5

14.6

9.4

31.2

73.9

20485

4.3

4.7

9.3

7.5

57

10.9

8.4

31.0E

84.0

8001

5.9

j Average

5.2

104

7.7

5.1

12.7

8.9

31.1

79.0

14243

5.1

i Std.Dev.

0.S

1.1

0.2

06

1.9

0.5

0.1

5.1

6242

0.8

j

Wet
(17days)

Warm and
Damp
(15 days)

Hot and
Dry
(26 days)

Storm

j

JO 211

(2 days)

j

JO 234

132
7.4

Suspended Sediment Load Summary - ‘Shape’, ‘Magnitude’ and HydroMeteorological Categories

Table 7 3 Total suspended sedim ent load summary (kg/.v days) for categories determined through PCA
and CA. Table 3.5 reports the number o f days in eaeh category.
Station

Cluster

Cold and Wet

Classification
PS1

PS2

PS3

RPT

DS

Hot and Dry

Storm

Category Sum

Sum,

Damp

Diurnal {Irregular!

73906(5829)

42658(9472)

51S94 (33503)

(34400)

168158(83204)

251362

(34400)

High

64305 (5829)

34414(9472)

20219(33503)

118938(83204)

202142

Medium

8961

8244

31375

48580

48580

tow

640

-

~

640

640

Diurnal (Irregular)

20774(351764)

188116(215052)

579493 (92464)

285759

1074142(659280)

1733422
544464

High

(74606)

(109527)

74572

285759

360331(184133)

Medium

20774(158067)

151802 (23263)

336154(65821)

-

508730(247151)

755881

Low

(119091)

36314(82262)

168767(26643)

--

205081(227996)

433077

Diurnal (Irregular)

99130(33315)

77113(21069)

60350(87018)

59199

295792(141402}

437194

; ■

76527(15130)

45512(10915)

23649(50209)::

■■59199

204887(76254)

281141

Medium

8460

31601(9060)

29092(29946}

-

69153 (39006)

tow

14143(181851

(1094)

7609(6863}

21752(26142}

108159
47894

Diurnal (Irregular)

66218(445241)

240890(291981)

754191 (124541)

383598

1444897(861763)

2306660

'

MS

Warm and

High

-

-

-

301577

301577

301577

Medium

(161716)

(138134)

163088

82021

245109(299850)

544959

Low

66218 (283525)

240890(153847)

591103 (124541)

--

898211(561913)

1460124

Diurnal (Irregular)

426592 (41420)

296581 (238462)

138975(672463)

343410(42929)

1205558(995273)

2200831

. High

257233(18039)

222085 (235113!

123601(543805)

;: 343410

946329(796957)

Medium

58110(18153)

15374(93531)

(42929)

73484(154613)

1743286
228097

. ■ ■■toast' ';

111249 (5228)

74495(3349)

(35127)

v:

185744(43704)

229448

Diurnal (Irregular)

150697(854988)

427579(596198)

1346120(210365)

888649

2813045(1661651)

4474596

High

(213187)

-

--

888649

888649 (213187)

1101836

Medium

56034 (286429)

248039(406369)

563048 (148311)

--

867121(841109)

1708230

Low

94663(355372)

179540(189829)

783072 (62054)

--

1057275(607255)

1664530

v

133
7.5

Julian Day C alendar

Table 7.4 Julian Day (JD) Calendar for 2011 simplified (o focus on typical ablation season. The 2011
field season at Castle Creek Glacier has been shaded.

May
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

JD
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151

June
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

Julian Day Calendar
JD July JD Aug.
152
1 182
1
153
2 183
2
154
3 184
3
4 185
155
4
156
5 186
5
157
6 187
6
7 188
158
7
159
8 189
8
160
9 190
9
161
10 191
10
162
11 192
11
12 193
163
12
164
13 194
13
165
14 195
14
15 196
166
15
167
16 197
16
168
17 198
17
169
18 199
18
170
19 200
19
171
20 201
20
172
21 202
21
173
22 203
22
174
23 204
23
24
175
24 205
176
25 206
25
177
26 207
26
178
27 208
27
179
28 209
28
180
29 210
29
181
30 211
30
31 212
31

JD
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243

Sept.
: 1.
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

JD
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273

134
7.6

Suspended Sediment Load Summary for the Small River Glacier

Orwin and Smart reported total suspended sediment load (kg) for common days o f
monitoring, and computed percentage increase based on the uppermost site that was
monitored. Their summary differed from the summary presented for the CCG in that daily
mean SSLs and weighted averages were used for summaries to avoid bias in the results due
to missing days of data and use as much of the data as possible. Table 7.5 presents the results
o f Orwin and Smart (2004a) in a format that is comparable with results presented in section
4.1.
Table 7.5 Re-computed results from Orwin and Smart 2004a, pg. 1539; suspended sediment totals have
been com puted as mean daily suspended sedim ent load (SSL) (kg/day) and percent o f total SSL
observed at the downstream site. NPG - North Proglacial; NPM - North Proglacial Middle; NPL
- North Proglacial Lower; C P I - Central Proglacial Upper; CPL - Central Proglacial Lower.
Meteorological

Cold and Wet

Snowmelt

Hot and Dry

Storm

Seasonal

period

25% (7 days)

25% (7 days)

43% (12 days)

7% (2 days)

100%(28 days)

Site

% of Total

mean SSL

SSL

%of Tota

mean SSL

SSL

% of Total

mean SSL

SSL

%of Total

mean SSL

SSL

% of Totai

mean SSL

SSL

NPG

#1

764

43

1100

87

1250

55

NPM

134

1129

103

2600

119

1704

98

12400

108

2548

NPL

ioo

843

100

2529

100

1430

100

1270)

100

2363

CPU

42

232

38

486

18

367

47

1735

26

461

CPL

100

551

100

1286

100

2500

100

3680

100

1794

7000

64

1502