Alluvial Floodplain Classification and Organization in Low-Relief Glacially Conditioned River Catchments

Roger Thomas James Phillips

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Department of Geography University of Toronto

Roger Thomas James Phillips

Doctor of Philosophy

Department of Geography University of Toronto

2014 Abstract

The imprint of late Pleistocene glaciation on river systems is an essential theme in Canadian geomorphology. Existing ideas about glacial legacy effects tend to focus on mountainous environments, which are different from the low‐relief physiography of the Laurentian Great

Lakes region. This study investigates river landforms in southern Ontario to develop an improved conceptual fluvial landscape model, reflecting glacial legacy and post‐glacial fluvial processes. The analysis is based on an original dataset, including basic channel observations from over 500 field sites, alluvial floodplain properties from 109 sites, and published physiographic mapping from digital sources.

Glacial signatures are evident in river profiles extracted from a digital elevation model (DEM) for 22 river catchments in southern Ontario. Stream power and slope–area analysis stratify river slopes by glacial landform types and demonstrate significant differences between rivers incised into glacial moraines versus plains. A stream length–gradient index provides a relative measure of how river profiles are oversteepened or understeepened by glacial landforms relative to a theoretical graded profile.

ii Four first‐order alluvial floodplain classifications are presented using k‐means clustering analysis. Predictive variables are explored using PCA and discriminant analysis, producing two principal components: (1) stream power‐resistance and (2) floodplain sedimentology (or floodplain sand equivalent, FSE). The river classifications from this study are consistent with previous literature, but special consideration is required for the inherited sources of cobble and sand materials.

General glacial–fluvial landform relationships can broadly be clustered into: 1) topographic and sedimentological glacial legacy effects; 2) landforms resulting from isostatic and lake baselevel change, and 3) superimposed patterns of Holocene fluvial sediment supply. The adapted fluvial landscape model and conceptual framework presented for rivers in low‐relief glacially conditioned landscapes have the potential to enhance interdisciplinary and applied science in the areas of biogeochemistry, geoarcheology, conservation ecology, and environmental water resources management.

iii Acknowledgments

It doesn’t hurt to start with your PhD supervisory committee. Joe Desloges, Sarah Finkelstein,

Bill Gough, Brian Branfireun, and Nick Eyles, thank you greatly for all your academic and intellectual advice over the years. Joe, you have been an extraordinary mentor in science and teaching. I marvel at your work ethic and dedication to students. Sarah, your enthusiasm and thoughtfulness in teaching are inspiring; and thanks to you I will never forget counting diatoms under the microscope for hours in PGB. Bill, you may not have started as a geographer, but I very much admire your passion for geography now. Brian, our pre‐comps discussions of representative elementary areas were compelling and formative. Nick, our tour of southern

Ontario geological sites in 2008 was invaluable to my learning.

I am grateful to André Roy for his positive and constructive external review of this dissertation and for our insightful discussions during the thesis defense. Thanks also to Mike Church for reading parts of this thesis, and specifically for his review and comments on Chapter 3 resulting in considerable enhancements to the final thesis.

Thanks to the many field assistants over the period of 2010–2013, namely Stephanie Mah,

Beata Opalinska, and Joyce Arabian; as well as James Thayer, Jennifer Henshaw and Tina Hui who also donated some of their own field data. Thanks also to the many public agency staff from OMNR and Conservation Authorities who responded to our information requests, including Brynn Upsdell, Kari Jean, and Ross Wilson (ABCA); Peter Dragunas and Tony Difazio

(CCCA); Shannon Wood and David Pybus (SVCA); Glenn Switzer (NVCA); Muriel Andreae and

Rick Batterson (SCRCA); Joe Gordon and Brian Widner (KCCA); and Kent Todd (OMNR).

iv This research was supported by the Natural Sciences and Engineering Research Council of

Canada (NSERC‐CGS graduate scholarship) and research funding from the University of Toronto to J.R. Desloges.

Again to my patient PhD advisor Joe Desloges, my sincerest gratitude for your guidance and support over the last 6 years. As the saying goes, or it goes without saying, I could not have done this without you. The thesis might be a bit long, or I guess in the words of Mike Church, a bit overdone. So thank you for accepting a few literary indulgences.

Oh well, my mother Doris never minded burnt toast. But she also said I could be rather verbose. I suppose she would be proud of me no matter what. Thanks mom.

To my late father Reginald Phillips (1927–2013) I owe my perseverance for learning.

To my wife Kate, thank you for your loving commitment, for your editorial reviews, and for remembering that Snuffbox has an “n” in the name.

Dedication For Kate, Amelie, and Lachlan. Yes Amelie, I am done my work now. Can you find the three little book mice?

v Table of Contents

Abstract ...... ii Acknowledgments ...... iv Table of Contents ...... vi List of Tables ...... viii List of Figures ...... ix List of Appendices ...... xi

Chapter 1 Introduction ...... 1 Research Statement ...... 1 1.1 Glacial Legacy Effects on Fluvial Systems ...... 4 1.2 Fluvial Process and Landform Interactions ...... 9 1.3 The Laurentian Great Lakes Region ...... 17 1.3.1 Modern human impacts ...... 21 1.4 Summary of Study Approach ...... 21 1.4.1 Statement of authorship and publication status ...... 23

Chapter 2 Glacially conditioned specific stream powers in low‐relief river catchments of the southern Laurentian Great Lakes ...... 24 2.1 Introduction ...... 25 2.2 Theoretical Background ...... 26 2.2.1 The graded river concept ...... 26 2.2.2 Specific stream power approach ...... 30 2.3 Regional Setting ...... 33 2.4 Materials and Methods ...... 36 2.5 Specific Stream Power Inputs ...... 41 2.5.1 Discharge regime models ...... 41 2.5.2 Bankfull width regime models ...... 44 2.5.3 DEM longitudinal profile extraction and slope generalization ...... 47 2.6 Results and Discussion ...... 51 2.6.1 Specific stream power mapping ...... 51 2.6.2 Profile analysis and SL/K index...... 54 2.6.3 Slope–area analysis ...... 59 2.6.4 Glacial conditioning of stream power ...... 64 2.7 Conclusions ...... 66

Chapter 3 Alluvial floodplain classification by multivariate clustering and discriminant analysis for low‐relief glacially conditioned river catchments ...... 70 3.1 Introduction ...... 71 3.1.1 Interdisciplinary floodplains ...... 71 3.1.2 Genetic floodplains ...... 73 3.1.3 Hyperdimensional floodplains ...... 75 3.2 Study Area ...... 77 3.3 Data Collection and Methods ...... 80 vi 3.3.1 Floodplain dataset ...... 80 3.3.2 Multivariate analysis ...... 83 3.4 Results ...... 85 3.4.1 Field floodplain classifications ...... 85 3.4.2 K‐means clustering analysis ...... 88 3.4.3 Principal component analysis ...... 89 3.4.4 Discriminant analysis ...... 95 3.5 Discussion ...... 100 3.5.1 Floodplain parsimony...... 100 3.5.2 Floodplain alphabet ...... 104 3.6 Conclusions ...... 108

Chapter 4 Glacial legacy effects on the spatial organization of alluvial floodplain types in the Laurentian Great Lakes region ...... 110 4.1 Introduction ...... 111 4.2 Study Area ...... 112 4.2.1 Post‐glacial baselevel changes ...... 113 4.3 Data and Methods ...... 118 4.3.1 Field and GIS data ...... 119 4.3.2 Isostatic paleo‐DEM and river incision ...... 122 4.3.3 Radiocarbon and OSL ...... 123 4.4 Results and Discussion ...... 125 4.4.1 Isostatic and baselevel variations ...... 125 4.4.2 Reach classification and river profiles ...... 130 4.4.3 Post‐glacial fluvial adjustments ...... 142 4.5 Conclusions ...... 146

Chapter 5 Conclusions ...... 148 Introduction: “The Beginning of the End” ...... 148 5.1 Thesis Questions ...... 149 5.1.1 Thesis question #1: Glacial signatures and river profiles ...... 149 5.1.2 Thesis question #2: Alluvial floodplain classifications ...... 151 5.1.3 Thesis question #3: Glacial legacy and landform organization ...... 154 5.2 Adapted Fluvial Landscape Model ...... 156 5.3 Research Significance ...... 159 5.3.1 Interdisciplinary significance ...... 160 5.3.2 Contribution to applied geoscience ...... 163

References ...... 164

vii List of Tables

Table 1.1: Comparison of fluvial depositional units and floodplain building process ...... 13 Table 1.2: Summary of floodplain classifications by Nanson and Croke (1992) ...... 14 Table 1.3: Summary of authorship and publication status ...... 23

Table 2.1: Area–discharge regime models for southern Ontario ...... 43 Table 2.2: Area–width regime models (bankfull width) for southern Ontario ...... 46 Table 2.3: Profile analysis results for Hack and Flint equations ...... 55

Table 3.1: Statistical transformation and normality tests of 20 reach variables ...... 80 Table 3.2: KMC analysis results based on 9 selected floodplain variables ...... 89 Table 3.3: PCA eigenvalues of the correlation matrix for 12 variables ...... 93 Table 3.4: PCA squared cosines of the 12 variables for the first four principal components ...... 93 Table 3.5: KMC‐DA confusion matrix for jackknife cross‐validation of floodplain types ...... 97 Table 3.6: Select examples of floodplain sites from 12‐variable DA jackknife cross‐validation .. 98 Table 3.7: KMC‐DA percent correct classification matches from jackknife cross‐validation ...... 99 Table 3.8: Three‐variable classification domains for primary floodplain archetypes ...... 104 Table 3.9: Comparison of floodplain classifications from this study with previous classifications ...... 105

Table 4.1: Number of field sites per catchment...... 120 Table 4.2: Summary of floodplain classifications for southern Ontario rivers...... 121 Table 4.3: Radiocarbon samples from southern Ontario alluvial river floodplains for this study ...... 124 Table 4.4: Summary of general landform relationships for glacially conditioned catchments . 143

Table 5.1: Summary of floodplain classifications for southern Ontario rivers ...... 153 Table 5.2: Glacial–fluvial landform relationships for southern Ontario river catchments ...... 155

viii List of Figures

Figure 1.1: Laurentian Great Lakes watershed boundary and study area of southern Ontario .... 3 Figure 1.2: Fluvial landscape models for fluvial landforms and sedimentary cascades ...... 7 Figure 1.3: Aerial images of river channel patterns from Western Canada and southern Ontario ...... 12 Figure 1.4: Generalized water level curves for the three eastern Great Lakes ...... 18 Figure 1.5: Summary of paleo‐hydrography with emphasis on eastern Great Lakes ...... 19

Figure 2.1: Watershed for the Laurentian Great Lakes of North America ...... 34 Figure 2.2: Map of study area showing major river drainages selected in southern Ontario ..... 36 Figure 2.3: Flow chart of GIS and spreadsheet analysis for DEM slope generalization and stream power mapping ...... 38 Figure 2.4: Area–discharge regime model statistical regression of empirical data ...... 43 Figure 2.5: Area–width regime model statistical regression of empirical data ...... 45 Figure 2.6: Example DEM profile from the Ausable River ...... 48 Figure 2.7: Stream power mapping of select major river drainages in southern Ontario ...... 52 Figure 2.8: Histogram of specific stream power results in southern Ontario ...... 54 Figure 2.9: Longitudinal profiles for select rivers in southern Ontario ...... 57 Figure 2.10: Slope–area plots of 146 river reaches in southern Ontario ...... 60 Figure 2.11: Slope–area plots with reach bed material and reach planform classifications ...... 63 Figure 2.12: Generalized patterns of bedload transport based on modeled stream power ...... 65

Figure 3.1: Study area of southern Ontario ...... 77 Figure 3.2: Schematic cross‐sections and mapping of alluvial floodplain types ...... 86 Figure 3.3: Photographs of channels representing four interpreted floodplain types ...... 87 Figure 3.4: PCA correlation circles with variable projections ...... 90 Figure 3.5: PCA ordinations for floodplain types based on first three principal components ..... 95 Figure 3.6: First two factors (F1 vs. F2) and correlation circle for DA test for 12‐variables ...... 97 Figure 3.7: Group‐split cross‐validation DA analysis for 3‐variable model ...... 100 Figure 3.8: Floodplain classifications for low‐relief glacially conditioned river catchments ..... 102 Figure 3.9: Four floodplain classifications with specific stream power versus FSE ...... 103

Figure 4.1: Study area of southern Ontario ...... 113 Figure 4.2: Post‐glacial water‐level curve for Lake Huron basin ...... 117 Figure 4.3: Isoline maps and histograms of isostatic tilting for each study catchment ...... 126 Figure 4.4: Post‐glacial baselevel highstands in Lake Huron basin ...... 128 Figure 4.5: Reach classification for the Ausable River and select tributaries ...... 131 Figure 4.6: Reach classification for the Saugeen River and select tributaries ...... 135 Figure 4.7: Reach classification for the Nottawasaga River and select tributaries ...... 139 Figure 4.8: Radiocarbon results from alluvial floodplain deposits for this study ...... 145

Figure 5.1: Glacial legacy signatures of moraines and plains for rivers in southern Ontario .... 150 Figure 5.2: Adapted fluvial landscape model for low‐relief glacially conditioned catchments . 157

x List of Appendices

Appendix A: Raw Floodplain Dataset and Example Photographs……………………………………………186

Appendix B: Optically Stimulated Luminescence Dating Potential of Quartz for Holocene Alluvial Deposits in the Southern Laurentian Great Lakes Glaciated Region……………………………………….194

Appendix C: Interdisciplinary Conference Presentations 2009–2012……………………………………..205

Appendix C.1: Snuffbox and the Three Bars: Investigating Geomorphological Approaches to Assess the Distribution of Freshwater Mussel Species at Risk in the Lower Ausable River (Latornell 2012)…………………………………………………………………………………………………205

Appendix C.2: The Geomorphology‐Ecology Balance of Designing Headwater Channels (Natural Channel Systems 2010)…………………………………………………………………………………211

Appendix C.3: “Well my heart’s in the Highlands”: Historical Patterns of Specific Stream Power in Urbanized Highland Creek (CAG‐ONT 2012)………………………………………………..216

Appendix C.4: Tightening the River Meander‐Belt: Application of a Topographic Erodible Corridor Concept Using DEM Raster Analysis – A Case Study of Highland Creek, Ontario (AGU 2009)…………………………………………………………………………………………………………………221

Chapter 1 Introduction

1 Research Statement

Rivers are idealized. They are natural phenomena, but are idealized in that our understanding

must balance simplicity with complexity — generalization with nuance. How well do prevailing

scientific theories represent the environmental complexity and geomorphological diversity of

rivers? The glacial legacy of Quaternary ice ages can be a considerable source of discontinuity

imposed on fluvial systems relative to theories of equilibrium and self‐organization. While the

inscription of glaciation is prominent in much of the Canadian landscape, the low‐relief

topography of the southern Laurentian Great Lakes region is a relatively untested environment

with respect to glacial legacy effects on the spatial organization of fluvial processes and river landforms. As a result, fundamental concepts in fluvial geomorphology must often be highly contextualized.

In such environments, applied geoscience in fluvial geomorphology and river engineering often cannot rely on classic theories which idealize river systems. For example, understanding catchment‐scale patterns of sediment supply and transport is essential in stream restoration, channel design, and erosion hazard assessment projects. Similarly, understanding the spatial organization of fluvial processes is important in interdisciplinary ecological research on aquatic

and riparian habitat, now with growing interest for the sake of sensitive and endangered species. In a low‐relief glacially conditioned landscape not only are steep bedrock‐hillslope and mass wasting processes absent in the headwaters, but idealized trends in channel slope and sediment grain‐size are frequently interrupted by inherited glacial topography and sediments, disrupting systematic patterns of fluvial sediment transport.

The broad purpose of this thesis is to investigate the nature of fluvial systems and alluvial floodplains in southern Ontario conditioned to varying degrees by glacial landforms, glacial sediments, and post‐glacial landscape history. Fluvial‐floodplain systems of the Laurentian

Great Lakes region, and southern Ontario in particular (Figure 1.1), have developed through a

complex history of environmental changes which are expressed at a variety of spatial and

temporal scales. In paleoenvironmental reconstructions, and potential insights gained with

respect to future environmental change responses, documentation of this complexity is the

necessary point of scientific explanation and prediction. However, in a traditional view of

scientific parsimony, theoretical expectations of geomorphic process‐landform relationships are

also anticipated to yield some explanation of the spatial distribution of contemporary (or late

Holocene) river landforms and genetic floodplains (Nanson and Croke, 1992).

With respect to the low‐relief glacially conditioned river catchments of the southern Laurentian

Great Lakes region, the following thesis questions have been proposed and refined:

1. To what degree are glacial signatures imposed on the longitudinal profiles and stream power distributions of river systems in this low‐relief landscape?

2. Are natural groupings of channel morphology, alluvial landforms, and fluvial process domains in this environment consistent with previous river and floodplain classifications from other environments, and in view of scientific parsimony which environmental variables are most effective at explaining and predicting distinct classes?

3. How are the spatial arrangements of river landforms and process domains spatially organized in the context of glacial landforms and post‐glacial landscape histories, and how have the fluvial systems responded to the glacial legacy over the Holocene?

Figure 1.1: Late Pleistocene glaciation of North America, with the limit of the Laurentide Ice Sheet at 18 ka BP as per data from Dyke et al. (2003). Inset map of modern Laurentian Great Lakes watershed boundary and study area of southern Ontario (SO), with Niagara escarpment (NE), Oak Ridges moraine (ORM), and Algonquin Arch, Precambrian basement bedrock (AA).

The rationale for this research arises from the need to adapt existing ideas about glacial legacy effects to the low‐relief physiography and river systems of the Laurentian Great Lakes region — an environment which has received little systematic attention with respect to fluvial landscape

evolution. This research is also important to inform applied geoscience in fluvial

geomorphology, which would benefit from developing a consistent framework to explain

spatial patterns of river and stream morphology in this context.

The theoretical foundation of this thesis is outlined in the following sections of Chapter 1,

starting with a review of some established concepts in our understanding of glacial legacy

effects on fluvial systems. Fundamental theories of fluvial process and landform interactions

are then discussed with respect to river morphology, channel patterns, floodplain

classifications, and fluvial process domains. A brief review of the post‐glacial history of the

Great Lakes region is then provided to set the stage for an investigation of rivers and alluvial floodplains in southern Ontario. The overall study approach is then summarized at the end of

Chapter 1 to layout the general structure of the thesis. 3

1.1 Glacial Legacy Effects on Fluvial Systems

A recurring discourse in fluvial geomorphology is the effect of past glaciations on river systems.

A notable contribution is that of Church and Ryder (1972) who propose the term paraglacial to describe how fluvial processes are conditioned by glaciation. The paraglacial concept emphasizes the increased sediment supply available to fluvial sediment transport processes

following glaciation and consequently the time period required to effectively exhaust glacial sources of debris. More generally, Slaymaker (2009) discusses this in terms of fluvial responses in “disturbed” landscapes and in terms of how glaciated landscapes respond to non‐glacial

conditions. What emerges from this perspective is that most post‐glacial fluvial adjustments

remain incomplete since the last Pleistocene glaciation (Eyles and Kocsis, 1989; Brardinoni and

Hassan, 2006; Collins and Montgomery, 2011; McCleary et al., 2011; Snyder et al., 2013).

A practical question then is what constitutes an effective framework for measuring glacial

legacy effects? A promising answer seems to hinge on equilibrium theory in geomorphology

(Thorn and Welford, 1994), and specifically relies on the idealized concept of graded river

profiles with systematic increases in drainage area and discharge matched by systematic

decreases in channel slope and bed material grain size. Longitudinal variations in river

morphology or channel pattern can then be expected in a downstream succession of fluvial

process‐landform relationships. In terms of equilibrium responses, river profiles and morphologies may in theory adjust to the prevailing physiographic conditions of topography, geology, and climate. While it is recognized that equilibrium states in geomorphic systems are scale‐dependent in time and space (Schumm and Lichty, 1965), the assumption is that an equilibrium river profile, often envisioned with a smooth concave‐up geometric form, provides

a conceptual benchmark from which to gauge external factors which complicate fluvial landscape evolution.

Particularly, since the work of Hack (1957), the equilibrium graded river profile has been demonstrated as a useful measure to investigate landscape diversity, specifically with respect to tectonic, lithological, sedimentological, and base level controls on long‐term river profile

adjustments. The focus being on deviations from a theoretically graded equilibrium‐profile as

evidence of underlying non‐fluvial influences in the landscape. This conceptual approach has

also been applied to glacial legacy effects on fluvial systems (Fonstad, 2003; Brardinoni and

Hassan, 2006; Collins and Montgomery, 2011; McCleary et al., 2011). It is expected that systematic downstream trends in fluvial systems may be interrupted by the lingering effects of

glacial landforms, including both erosional and depositional features. Brardinoni and Hassan

(2006) refer to these features as glacial signatures within stream profiles. A similar concept which has been eluded to by others is the idea of fluvial discontinuities (such as in sediment flux), specifically for glacially conditioned environments (Snyder et al., 2013) and more generally (Burchsted et al., 2014). A complication of this approach is that topographic signatures and sedimentological discontinuities in fluvial systems may be the result of any number of complex environmental factors, of which glacial legacy effects may be just one. As

such, the spatial scale and strength of other potential controlling factors (e.g., tectonic uplift,

structural faulting, mass wasting, geologic variability and base level controls) must be carefully

considered for different environmental contexts.

A number of practical measurement and analysis methods have been previously tested with

respect to glacial legacy signatures on river profiles. Based on deviations from expected

equilibrium stream profiles, some notable methods include stream power maps (Fonstad,

2003); slope‐drainage area plots (Brardinoni and Hassan, 2006; Collins and Montgomery, 2011); gradient‐index profiles (or the SL/K index; McCleary et al., 2011); and grain‐size prediction

models (Snyder et al., 2013). Effectively, each of these methods detect slope‐profile anomalies

relative to a theoretical equilibrium slope profile. Each of these methods will be discussed

further in Chapter 2.

Glacial signatures in river profiles and discontinuities in fluvial processes then provide a basis to

assess the downstream spatial organization of channel morphology. Reach‐scale process

domains (Montgomery, 1999) provide a convenient spatial unit by which to delineate fluvial

landforms within the context of relic glacial landform effects. For mountainous catchments in

British Columbia (Canada), Brardinoni and Hassan (2006) present data showing how the

topographic signatures of glacial macro‐forms (e.g., cirques, hanging‐valleys) complicate the

sequence of reach‐scale channel morphology (e.g., cascade, step‐pool, riffle‐pool) due to

imposed variations in fluvial processes and patterns of hillslope mass transfer (e.g., colluvial‐

alluvial coupling and decoupling). Investigating the transition between mountainous

headwaters and lowland alluvial rivers in Washington state (USA), Collins and Montgomery

(2011) document the “valley‐scale” imprint of late Pleistocene glacial features on the

organization of alluvial river landforms (e.g., channel patterns, sinuosity, channel width). These

two examples from British Columbia and Washington demonstrate how morphological

sequences of fluvial landforms can be organized by larger‐scale glacial landscape features.

Thus, the incomplete Holocene response of fluvial systems to the topographic and

sedimentological legacy of late Pleistocene glaciation reflects a long‐term disequilibrium in

modern catchment‐scale erosion and sediment transport patterns (Church and Slaymaker,

1989). The downstream succession of fluvial landforms can also be discussed in terms of

sediment cascades (Burt and Allison, 2010) defined as the transfer of sediments through the fluvial landscape, including their source, transport, temporary storage (e.g., in alluvial landforms), and ultimate sink. In terms of paraglacial landscape responses, much of the discussion of sediment cascades tends to focus on a mountains–to–marine succession (Figure

1.2A) of sedimentary land systems (Ballantyne, 2002). However over the continental scale of

Pleistocene glaciation in North America (Figure 1.1), less emphasis has been placed on the vast

low‐relief areas of the continental craton where sediment cascades can be considered largely truncated with respect to the ultimate end‐member sources and sinks of sediment. Specifically within the Laurentian Great Lakes region, the extent and complex architecture of glacial drift together with the large inland lake system leads to a truncated moraines–to–lake‐mud succession of sedimentary cascades and fluvial landforms (Figure 1.2B).

Figure 1.2: Fluvial landscape models for fluvial landforms and sedimentary cascades. (A) Classic mountains–to–marine fluvial landform succession (based on a similar diagram of the fluvial system from Charlton, 2009). (B) Truncated moraines–to–lake‐mud fluvial landform succession, this study (dashed curves represent theoretical equilibrium graded profile surface for mountainous catchments).

As a general terminology, glacial conditioning of fluvial systems represents the lasting

impression of late Pleistocene glaciation, including the paraglacial legacy of sediment

availability and the inherited glacial signatures on river profiles relative to a theoretical self‐

organized equilibrium state for fluvial landscapes. Ferguson (1981) describes this condition as a

persistent passive disequilibrium. More specifically, the balance between equilibrium‐

disequilibrium timescales for fluvial response is largely dependent on spatial scale (Schumm

and Lichty, 1965). This idea has been acknowledged in the paraglacial literature where it is expected that small catchments may adjust rapidly to post‐glacial conditions, while at the scale

of large river basins and regional landscapes the paraglacial timescale may potentially span

entire interglacial periods (Church and Ryder, 1972; Slaymaker, 2009; Ballantyne, 2002).

However, it also follows that reach‐scale fluvial processes can exhibit quasi‐equilibrium states

within the overall persistent disequilibrium of much slower river profile adjustments. Quasi‐

equilibrium states of rivers were considered on a theoretical basis by Langbein and Leopold

(1964) where they rationalize the competing tendencies of rivers to adjust rapidly within reaches and to respond slowly in river profiles over geologic time. This idea is further demonstrated for post‐glacial fluvial landscapes by Brardinoni and Hassan (2006) and Collins

and Montgomery (2011) where reach‐scale channel morphologies are organized by non‐fluvial

topographic signatures and glacial features at the catchment scale. In other words, fluvial

process‐landform equilibriums (or quasi‐equilibriums) can in theory be nested within a long‐ term disequilibrium state of a river profile. Reach‐scale equilibriums provide a basis to consider

fluvial process and landform interactions in terms of previous literature on river morphology, channel patterns, floodplain classifications, and fluvial process domains.

1.2 Fluvial Process and Landform Interactions

The morphology of a natural river channel is the product of interactions between the variable

flow of water within the channel and the potentially erodible boundary materials, with

consideration of upstream supplies of sediment. Fluvial processes then make‐up the

instantaneous physical mechanics of fluid forces in terms of sediment entrainment, transport,

and deposition; as well as the cumulative patterns of sediment movement (and selective transport) which are responsible for molding and building the alluvial landforms of the river channel. As such, fluvial landforms may be considered the product of a suite of fluvial processes. However, what is less clear is the degree to which these process‐landform

interactions can be considered in equilibrium (i.e., in terms of stationary environmental

controls) balanced with the relative imprint of historical contingency (i.e., importance of

preceding landforms and system trajectories; see Section 1.1 discussion of glacial legacy and equilibrium). These concepts and questions are of course fundamental in geomorphology.

River Morphology and Channel Patterns

A common starting point for a discussion of river morphology is the meandering‐braiding

paradigm of Leopold and Wolman (1954) who discriminated river planforms based on only

discharge (Q) and channel slope (S). Although the classic Q‐S planform discrimination‐analysis

provides a compelling theoretical explanation of river channel patterns based on a threshold

stream power (where stream power, Ω, is proportional to QS), it is understood that the

erodibility of the channel boundary, and namely sediment grain sizes (or calibre), is another

essential factor required to constrain channel pattern predictions (e.g., Ferguson, 1987). It is also recognized that multiple‐channel planforms (or multi‐thread) are not all strictly subject to

braiding processes of dynamic bar formation (i.e., below the bankfull elevation). Specifically, 9

the term anabranching as per Nanson and Knighton (1996) evolved from descriptions of

anastomosing rivers as multiple‐channel planforms where channels are separated by semi‐ stable islands which are elevated above the active channel and are flooded less frequently than braid‐bars (e.g., island surfaces are close to or above the bankfull elevation).

The broadly used term anabranching has also been appropriate to describe other transitional types of multi‐thread channel patterns. For example, wandering gravel‐bed rivers of Western

Canada (Church, 1983; Desloges and Church, 1989) exhibit transitional properties of braiding, meandering, and anabranching as paraglacial sediment supplies are gradually being exhausted

since the last glaciation (e.g., Church and Slaymaker, 1989; see discussion Section 1.1).

Although formerly encompassing many types of anabranching, anastomosing rivers are now

generally considered most typical of extremely low‐gradient rivers in aggrading environments such as estuary deltas, mountainous trenches with cross‐valley alluvial fans (Tabata and Hickin,

2003); and areas of isostatic rebound (Smith, 1983) for example.

The over‐simplification of the original Leopold and Wolman (1954) meandering‐braiding threshold has also been addressed with further consideration of boundary erodibility. A notable example is that the braiding threshold in Q‐S for sand beds is considerably lower

compared to gravel‐bed rivers (e.g., Simpson and Smith, 2001). Further, bank strength has

been shown to be an important factor in determining channel stability and associated

adjustments in channel width (w), with consequential changes in specific stream power (ω ~

Ωw‐1) and thus sediment transport potential (Ferguson 1987; Eaton et al., 2004). While only some studies directly consider bank strength, recent treatments of the meandering‐braiding threshold are more sophisticated, with notable examples being from the work of van den Berg

(1995) and Kleinhans and van den Berg (2011) based on specific stream power (ω) and bed

material grain size (D). However, Kleinhans and van den Berg (2011) are careful not to

overstate the discrimination of planform types suggesting that their equations are indicators of transitions (‘transitionators’) between distinguishable bar patterns (i.e., scroll‐bars versus

chute‐bars), rather than hard thresholds between meandering and braiding planforms.

Thus from the planform view, environmental factors which govern fluvial processes, such as

stream power (Q‐S) and sediment calibre (D), have been shown to explain variations in channel

pattern. While such process‐landform associations may have some relevance in the context of southern Ontario rivers (Figure 1.3), the low‐relief landscape of the Laurentian Great Lakes

region does considerably constrain patterns of stream power and sediment supply. Effectively,

multiple‐channel planforms are rare and single‐thread rivers dominate the landscape. As such,

investigation of river morphologies in such environments requires more in‐depth consideration

of the alluvial landforms and their sedimentological characteristics.

Alluvial Floodplain Classification

River planform types (or channel patterns) have provided a basis to interpret alluvial floodplain

landforms, with the seminal article by Nanson and Croke (1992) summarizing a framework for

genetic floodplain classifications. However, interpretation of modern and ancient river

deposits also has a long history in geomorphology and geology in terms of fluvial facies models

(e.g., Miall, 1985; Hickin, 1993). Church (2006) highlights the interconnection between channel

morphology and bed material sediment transport by classifying rivers based on Shields number

and relative roughness. These parameters summarize stream competence and the relative

roles of bedload and suspended sediment load which contribute to building alluvial floodplains.

Figure 1.3: Google Earth™ aerial images of river channel patterns from Western Canada (left) and southern Ontario (right) based on planform continuum of Church (1992). A) Braiding Sunwapta River, Alberta (Goff and Ashmore, 1994). B) Braiding Lower Maitland River, southern Ontario (43°45'17" N., 81°42'20" W.). C) Wandering gravel‐bed Bella Coola River, British Columbia (Church, 1983; Desloges and Church, 1989). D) Anabranching Grand River, southern Ontario (Croil, 2002). E) Meandering Milk River, Alberta (Simpson and Smith, 2001). F) Meandering Thames River, southern Ontario (Stewart and Desloges, 2013).

Miall (1985) presents a detailed framework for interpreting fluvial deposits in terms of building

three‐dimensional stratigraphical models using a suite of what he terms architectural‐elements.

Nanson and Croke (1992) on the other hand offer a simplified list of six alluvial floodplain

formation processes, including three main and three secondary processes. Table 1.1 summarizes the eight architectural‐elements of Miall (1985) and the six floodplain formation processes of Nanson and Croke (1992). While there is overlap in the first four definitions of fluvial deposits, the remaining depositional architectural‐elements of Miall (1985) hint at the diversity of depositional units relative to the more generalized floodplain formation processes

of Nanson and Croke (1992).

The genetic floodplain classifications of Nanson and Croke (1992) are basically constrained by

two factors: 1) specific stream power (ω), which summarizes the available forces of sediment transport; and 2) cohesive versus non‐cohesive boundary materials, which summarizes in very

general terms the erodibility of the alluvial boundary in terms of sediment calibre and bank strength. A condensed version of the Nanson and Croke (1992) classification scheme is provided in Table 1.2.

Table 1.1: Comparison of fluvial depositional units and floodplain building process from key sources in the literature.

Miall (1985) Nanson and Croke (1992) * Architectural‐Elements Floodplain Formation Processes 1. Lateral accretion deposits = 1. Lateral point‐bar accretion 2. Overbank fines = 2. Overbank vertical‐accretion 3. Gravel bars and bedforms ~ 3. Braid‐channel accretion main 4. Channels ~ 4. Abandoned‐channel accretion secondary 5. Sandy bedforms 5. Oblique accretion 6. Laminated sand sheets 6. Counterpoint accretion 7. Foreset macro‐forms 8. Sediment gravity flows

* Most recently revised in Miall (2010).

Table 1.2: Summary of floodplain classifications by Nanson and Croke (1992).

Classification Stream Power Description Examples (ω, Wm‐2) Class A > 300 Disequilibrium floodplains which Single‐thread, often confined, High Energy erode in response to extreme events, dominantly vertical accretion types, Non‐Cohesive typically located in steep headwater include catastrophic cut and fill areas where channel migration is floodplains prevented by valley confinement

Class B 10–300 Equilibrium floodplains formed by Braiding, wandering, and laterally Medium Energy regular flow events in relatively migrating meandering rivers ‐2 Non‐Cohesive unconfined valleys Note: meandering, ω =10–60 Wm

Class C < 10 Floodplains formed by regular flow Single‐thread to anastomosing Low Energy events along laterally stable single‐ Note: anastomosing is a low‐energy Cohesive thread or anastomosing low‐gradient form of anabranching (Nanson and channels Knighton, 1996)

Given that the low‐relief landscape of the Laurentian Great Lakes region is anticipated to constrain patterns of stream power and sediment supply, it is expected that high energy Class A floodplains of Nanson and Croke (1992) will not typically be found in southern Ontario (Note: a few exceptions may be found for localized stream reaches along the Niagara escarpment).

Further, the discharge variability, stream power, and sediment supply characteristics of braiding

and wandering rivers (two sub‐orders of Nanson and Croke (1992) Class B floodplains) do not

typically occur within the study area, limiting the potential development of these planforms.

Other than a few known examples of local anabranching (e.g., Croil, 2002), this generally leaves single‐channel (or single‐thread) rivers as the dominant floodplain types in southern Ontario, including laterally active meandering rivers (Nanson and Croke (1992) Class B, suborder B3) and more stable low‐gradient single‐thread channels with cohesive boundaries (Nanson and Croke

(1992) Class C, suborder C1).

In terms of floodplain formation processes (or fluvial architectural‐elements), typical Nanson

and Croke Class B3 meandering river floodplains are primarily built of some combination of lateral point‐bar accretion and overbank vertical‐accretion (Table 1.1). For Nanson and Croke

Class C1 floodplains, overbank vertical accretion is the dominant process. Both Class B3 and C1

floodplains of Nanson and Croke (1992) may also contain minor contributions from oblique

accretion (a sort of steep‐angle cohesive lateral accretion) and abandoned‐channel accretion.

However, in the context of glacial legacy effects in southern Ontario, it is unclear to what

degree these floodplain formation processes hold‐up and it is unknown if the inherited

sediments and topography of the landscape have produced any distinct departures from the

Nanson and Croke (1992) classification framework.

Fluvial Process Domains

A reach of river has long been considered an appropriate spatial scale to organize river

landforms as coherent landscape features derived from distinct assemblages of fluvial

processes (Leopold and Wolman, 1957). However, particularly in the context of environmental heterogeneity the concept of river reaches is juxtaposed with the perspective that fluvial systems are multi‐dimensional physical and biological continuums varying gradually in space and time (Montgomery and Buffington, 1997). While these two perspectives may be at odds, each retains its own value in interpreting fluvial landscapes.

As effectively codified in the work of Montgomery (1999), stream reaches have been rationalized not only as a convenient spatial scale to describe river landforms, but also as

conspicuous spatial units characterized by different suites of dominant geomorphic processes

(i.e., fluvial process domains). As such reaches are intended to represent relatively

homogeneous units within the reach‐patchwork of a fluvial drainage network. In abstract statistical terms, it is expected that the morphological variability between reaches will be greater than the variability within reaches. Reaches as defined by Montgomery (1999) are

typically at least 10–20 channel widths in length and may generally be delineated by variations

in topography, geology, vegetation, and hydrology (e.g., tributary confluences).

Montgomery (1999) further suggests that because reaches tend to have similar geology and topography (or slopes), that reaches provide a basis to define what he calls lithotopo units.

Slope–drainage area plots for reaches can then provide a first‐order approach to spatially

discriminate between distinct fluvial process domains, which is essentially a means of stratifying reach‐scale morphology based on stream power (cf. Flores et al., 2006). Montgomery (1999) even advocates that process domains provide an effective framework to interpret fluvial responses to environmental complexity and landscape disturbance (e.g., glacial legacy effects,

Collins and Montgomery, 2011).

The reach framework allows for inspection of the spatial arrangement and structure of

individual reach morphologies within a drainage network mosaic of process domains. It follows

that exploring the spatial organization of fluvial process‐landform groups across a landscape such as southern Ontario, and in the context of complex glacial legacy effects, should provide

insights into patterns of post‐glacial fluvial adjustment. Indeed, the late Pleistocene and

Holocene history of fluvial systems in the Laurentian Great Lakes region is complicated by a

variety of environmental changes following glacial retreat.

1.3 The Laurentian Great Lakes Region

The Laurentian Great Lakes are a significant landscape inscription of northern hemisphere glaciations. The story of the Great Lakes has been an exercise in paleoenvironmental reconstruction by scientists for more than a century. Since the late Cenozoic initiation of northern hemisphere glaciations (Raymo, 1994), ice has subsequently scoured large areas of the North American continent. In particular, the Great Lake basins have been carved into bedrock strata of the Paleozoic Michigan sedimentary basin, where weaker strata have been preferentially scoured and resistant bedrock dominates existing islands and peninsulas (Larsen and Schaetzl, 2001). Reworking the landscape of previous Pleistocene glaciations, the melting

Laurentide Ice Sheet of the last glacial period has left behind widespread evidence of the pro‐

glacial and post‐glacial lake systems which evolved in its wake (Larsen and Schaetzl, 2001).

The evidence of paleo‐Great Lakes evolution is complicated by dynamic interactions between changing meltwater inputs, changing topography, and changing climate (i.e., precipitation,

temperature, wind) (Teller, 1995). Areas previously scoured by glacial ice lobes provided some

physical constraints for basins of the proto‐Great Lakes, however initial impoundment during

deglaciation was also accentuated by glacial isostatic depression of the earth’s crust (Peltier,

1994). The processes of deglaciation beginning approximately 15 ka BP were not characterized

by a continuous retreat of ice, but rather by fluctuations in the ice margin superimposed on a

millennium scale melting trend (Teller, 1995; Larsen and Schaetzl, 2001). The evolution of the

Great Lakes can be generalized into four broad lake phases, similar to those discussed by Lewis

et al. (2008). These four phases include: glacial lakes; extreme lakes; Nipissing lakes; and

modern lakes (Figure 1.4). The paleo‐lakes with glacial retreat are also illustrated in Figure 1.5.

Paleo‐Lake Names 1. Lake Maumee 2. Lake Ypsilanti 3. Lake Whittlesey 4. Lake Iroquois 5. Early Lake Algonquin – Kirkfield 6. Early Lake Erie 7. Early Lake Ontario 8. Lake Algonquin – Main 9. Lake Stanley/Hough 10. Mattawa highstands 11. Nipissing lakes 12. Modern lakes

Figure 1.4: Generalized water level curves for the three eastern Great Lakes (Huron, Erie, Ontario) relative to modern lake elevations (Adapted from Anderson and Lewis, 1985; Barnett, 1985; Eschman and Karrow, 1985; Pengelly et al., 1997; Lewis et al., 2007), including generalized lake phases (Adapted from Lewis et al., 2008) and paleo‐lake names.

During the glacial lake phase (~14–10 ka BP), pro‐glacial lakes were formed within the basins by

ice‐damning along the margins of the ice sheet, with initial drainage south to the Mississippi

basin (Teller, 1995). These drainage connections likely represented important entry routes to

the early Great Lakes for flora and fauna from the south (Lewis et al., 2008).

During the transition from glacial to post‐glacial conditions (~12–7 ka BP), the lake levels in each basin were adjusting to changing topography (i.e., glacial isostatic rebound) and changing climates, while still receiving indirect pulses of meltwater from glacial lakes to the north (Lewis et al., 2008). This extreme lake phase was most significant in the Lake Huron basin which was characterized by dramatic fluctuations in water levels (Lewis et al., 2008). In comparison, the

Lake Erie and Ontario basins were dominated by low water levels during this phase, with slowly

rising lake levels associated with adjustment to changing paleoclimate and differential isostatic

rebound of their outlet sills (Lewis et al., 2008).

Figure 1.5: Summary of paleo‐hydrography with emphasis on eastern Great Lakes (Adapted from data by Dyke et al., 2003). Chronology in uncalibrated radiocarbon years before present.

The rising Nipissing lake phase (~7–4 ka BP) followed complete deglaciation of the Laurentide

Ice Sheet with continued adjustments to paleoclimate and differential isostatic rebound (Lewis

et al., 2008). Effectively, inferred increases in precipitation and continued uplift in northeastern

areas of the Great Lakes region resulted in rising lake levels and drainage pattern adjustments

(Lewis et al., 2008). In particular, the lower lakes (i.e., Erie and Ontario) began to receive water

from the upper lakes (i.e., Huron, Michigan, and Superior) by drainage through the Port Huron

outlet starting at about 6 ka BP (Lewis et al., 2008). Water levels then slowly lowered to

modern lake levels (~3–0 ka BP) with incision at the Port Huron outlet and a slight reduction in

climate precipitation in the late Holocene (Lewis et al., 2008).

Late Pleistocene deglaciation in the Great Lakes region exposed glaciogenic sediments to renewed fluvial processes of erosion, sedimentation, and alluvial floodplain development over

the Holocene period. Post‐glacial drainage patterns were inherited from physiographic features of glacial erosion and deposition, as well as from antecedent bedrock controls including

structural features (e.g., Algonquin Arch, Figure 1.1) and neotectonic joints (Eyles, 2002). As discussed above in Section 1.1, glacial legacy effects on fluvial systems generally include

paraglacial influences on sediment availability as well as inherited topographic signatures of

glacial landforms in river profiles. Consequently, reach‐scale fluvial processes and landform

relationships may be spatially organized by larger scale glacial features in the landscape.

However, other paleoenvironmental factors have also influenced post‐glacial fluvial

adjustments over the Holocene, including lake level history, paleoclimate variations, and

differential isostatic rebound. Thus for catchments of the Laurentian Great Lakes region, the

spatial arrangement of river landforms is complicated by diverse paleoenvironmental histories,

with many confounding effects potentially expressed at several spatial and temporal scales.

Still, documenting this spatial complexity relative to theoretical concepts of equilibrium and 20

self‐organization in fluvial geomorphology is expected to provide insights into a more generalized framework for understanding fluvial process‐landform associations in low‐relief glacially conditioned landscapes.

1.3.1 Modern human impacts

Consideration of modern human impacts is acknowledged as a necessary component in the study of river landforms in southern Ontario. While glacial features in the landscape are

expected to impose physiographic controls on southern Ontario rivers, the impacts of European settlement starting in the 1700s and 1800s are associated with widespread deforestation, extensive agricultural development, and local river engineering projects. Post‐settlement

alluvium eroded from cleared land surfaces, grade changes from mill and flood‐control dams,

and channel erosion from runoff in urbanized areas have resulted in modern river adjustments.

Even so, it is expected that late Holocene alluvial records can largely be distinguished from modern impacts in river channel and floodplain landforms. Ultimately, modern human impacts will impart additional variability in a regional dataset of river landforms. While these impacts

may confound glacial influences, the persistence of glacial features can still be tested based on

interpretation of pre‐settlement river landforms and alluvial records.

1.4 Summary of Study Approach

The discussion and literature review presented in Chapter 1 provide the theoretical foundation

to investigate fluvial process‐landform relationships in the context of complex glacial legacy

effects for the low‐relief landscape of southern Ontario. The purpose of this thesis is to

investigate the nature of fluvial systems and alluvial floodplains conditioned to varying degrees

by the inherited glacial landforms and sediments; as well as by other post‐glacial environmental

changes. To engage the proposed thesis questions, this dissertation is structured in a series of

three papers (i.e., thesis‐by‐papers), ultimately building to synthesize the study conclusions in

the final chapter. As such, each paper (Chapters 2, 3, and 4) will begin with its own abstract and introduction.

 Chapter 2 investigates the glacial signatures contained within river profiles extracted from a digital elevation model of southern Ontario. Anomalies in channel slope and stream power are assessed relative to a theoretical graded profile and are compared to glacial landforms using stream power mapping, slope‐area analysis, and a stream length‐gradient index (i.e., an SL/K index).

 Chapter 3 tests alluvial floodplain classifications to characterize the morphological and sedimentological variations of single‐channel planforms in southern Ontario. The classifications are intended to represent distinct morphological groups with differing fluvial process domains in the landscape.

 Chapter 4 examines the spatial arrangement of morphological groups to interpret post‐ glacial fluvial adjustments in the context of glacial landforms and post‐glacial landscape histories for select catchments and drainage networks within the study area.

The concluding discussion in Chapter 5 presents a theoretical framework for understanding the

spatial organization of fluvial process‐landform relationships in low‐relief glacially conditioned

environments such as southern Ontario, as adapted from the theoretical foundations derived in

other environments and presented in previous literature. It is intended that this conceptual framework will provide an improved approach to explain the geomorphological diversity of rivers and alluvial floodplain landforms in complex glacially inherited landscapes, particularly those with modest relief. The significance of the research will also be discussed, particularly in

terms of its relevance to applied geoscience in fluvial geomorphology and to interdisciplinary research.

1.4.1 Statement of authorship and publication status

This thesis is submitted in conformity with the requirements for the degree of Doctor of Philosophy by the author Roger T.J. Phillips. The introductory and concluding chapters 1 and 5 (respectively) have been written by the author for the purposes of this dissertation. The core chapters 2, 3, and 4 make up the substantive research articles of this thesis. As such, this research has been completed in collaboration with, and with support from, the author’s supervising professor Joseph R. Desloges. The authorship contributions and publication status of each research paper (as of June 2014) are outline in Table 1.3 below.

Table 1.3: Summary of authorship and publication status. Research Paper #1 (Chapter 2)

Glacially conditioned specific stream powers in low relief river catchments of the southern Laurentian Great Lakes Authorship Contribution: R.T.J. Phillips (90%): primary author of thesis; project development; field and lab work; data collection and analysis. J.R. Desloges (10%): supervision of project development (theory and methodology); funding and editorial support. Journal: Geomorphology Publication Status: Published February 1, 2014 (Volume 206, pg. 271–287). Available online: October 19, 2013.

Research Paper #2 (Chapter 3) Alluvial floodplain classification by multivariate clustering and discriminant analysis for low‐relief glacially conditioned river catchments Authorship Contribution: R.T.J. Phillips (90%): primary author of thesis; project development; field and lab work; data collection and analysis. J.R. Desloges (10%): supervision of project development (theory and methodology); funding and editorial support. Journal: Earth Surface Processes and Landforms Publication Status: Submitted March 7, 2014 (Comments received May 28, 2014). Revisions pending.

Research Paper #3 (Chapter 4) Glacial legacy effects on the spatial organization of alluvial floodplain types in the Laurentian Great Lakes region Authorship Contribution: R.T.J. Phillips (90%): primary author of thesis; project development; field and lab work; data collection and analysis. J.R. Desloges (10%): supervision of project development (theory and methodology); funding and editorial support. Journal: TBD Publication Status: In preparation.

Chapter 2 Glacially conditioned specific stream powers in low-relief river catchments of the southern Laurentian Great Lakes

2 Chapter 2

Abstract: Fluvial systems of the southern Laurentian Great Lakes region are carved into a complex glacial landscape shaped by continental ice and meltwater of the late Pleistocene. These glacially conditioned river catchments are typically small with drainage areas < 104 km2. A 10‐m digital elevation model (DEM) is used to map the spatial distribution of stream gradient for 22 major river catchments of peninsular southern Ontario, which drain to base levels in the lower Great Lakes (Huron, St. Clair, Erie, and Ontario). Raw data from the DEM show stream gradients that exhibit multiscale variance from real and from artifact sources. Based on a vertical slice and multiple‐pass moving‐window averaging approach, slope data are generalised to the river reach scale (1–2 km) as a representative spatial scale for fluvial processes operating over Holocene timescales. Models of specific stream power are then compared with glacial landform and surface geology mapping. Inherited glacial signatures in river slope appear as deviations in a stream length‐gradient index (SL/K index), where river reaches are frequently oversteepened or understeepened. Based on a slope–area analysis, and complementary to theories of channel pattern discrimination, constant stream power curves (with power‐law exponent of ‐0.4) provide a first‐order approach to stratify river reaches in terms of glacial conditioning and expected planform morphologies. However, multiple‐channel planform types are rare and localized in southern Ontario, indicating that oversteepened reaches with high stream powers may often be moderated by (1) sediment calibre, with cobble‐beds from inherited glacial sediments; and/or (2) relative bank strength, with limited channel widening particularly in gravel and sand‐bed channels. Further discrimination of glacially conditioned fluvial process domains will ultimately require consideration of alluvial floodplain characteristics in addition to general observations of river morphology and channel pattern.

2.1 Introduction

Fluvial systems of the southern Laurentian Great Lakes region are geologically young, carved

into a landscape of complex glacial drift and subtle topography. A product of Quaternary

glaciations, the Great Lakes were formed by glacial erosion of Precambrian bedrock of the

Canadian Shield and overlying Paleozoic sedimentary strata of the Michigan basin. At the

southern margins, late Pleistocene deglaciation deposited an extensive landscape of glacial

landforms and sediments, including moraines and glaciolacustrine sequences, in the wake of

the melting Laurentide Ice Sheet (Larson and Schaetzl, 2001).

Glacial conditioning of stream longitudinal profiles has been documented by recent studies for

mountainous landscapes in terms of the topographic reorganization of fluvial processes, as well

as the coupling (or decoupling) of fluvial with hillslope and mass wasting processes (Fonstad,

2003; Brardinoni and Hassan, 2006; McCleary et al., 2011). The study of glacially conditioned

fluvial catchments within the geologic context of past continental ice sheets in nonmountainous

regions has received less attention, and any existing research has yet to be synthesized to a

landscape scale. This is particularly true of the southern Laurentian Great Lakes region where

the topography is subtle and the sedimentary architecture of the glacial palimpsest is vast and

complex.

This paper investigates glacial signatures within river profiles of southern Ontario to extend

previous research on glacial conditioning from mountainous landscapes to the low‐relief

catchments of the southern Laurentian Great Lakes. From river longitudinal profiles, the spatial

properties of stream power are mapped within the context of glacial landforms and sediments.

Previous approaches to evaluate environmental controls on river slope, and consequently

channel morphology, are tested with respect to the concept of stream power, in terms of

channel patterns (e.g., meandering vs. braiding; Leopold and Wolman, 1957) and fluvial process

domains (Montgomery, 1999). To evaluate glacial signatures, some theory of fluvial profile

evolution must be acknowledged, traditionally defined as the graded river profile.

2.2 Theoretical Background

2.2.1 The graded river concept

The idealized concept of graded rivers is a compelling paradigm with a long history in the

discipline of fluvial geomorphology (see Chorley, 2000, for historical retrospective). While a

graded stream is by tradition envisioned as a geometrical manifestation, a smooth longitudinal

profile concave to the sky, its scientific establishment as a product of physical channel

processes is largely founded on Mackin’s (1948) definition that it represents the slope of transportation — the hydraulic gradient of the river is adjusted to transport the sediment

supplied to it. Thus changes to the quantity and calibre of sediment supply should cause

adjustments in slope produced by channel aggradation or degradation to compensate for the change, giving the concept of grade a strong connection to theories of geomorphic equilibrium and disequilibrium (e.g., Schumm and Lichty, 1965; Thorn and Welford, 1994).

In general terms, the graded river state trends downstream with systematic increases in discharge and decreases in channel slope (and sediment size), so it has also been considered to

be closely associated with spatial distributions of stream power (i.e., the product of discharge

and slope) (Knighton, 1999). Considerable research relies on the concept of grade or at least

on the theoretical expectation of a smooth concave‐up profile, most often with the assumption

that it can be mathematically represented as an exponential curve (Hack, 1957, 1973; Seeber

and Gornitz, 1983; Sinha and Parker, 1996; Morris and Williams, 1997; Knighton, 1999; Smith et

al., 2000; Fonstad, 2003; Jain et al., 2006; Goldrick and Bishop, 2007; Barker et al., 2009; Pérez‐

Peña et al., 2009; Gonga‐Saholiariliva et al., 2011; McCleary et al., 2011).

Since the work of Hack (1957, 1973) and Flint (1974), mathematical representations of an ideal concave‐up profile regularly rely on empirical relationships of slope scaled to either river distance (L) or drainage area (Ad). Hack’s (1957) equation for channel slope (S) can be

summarized by

S = k Ln (1)

where k and n are empirically derived constants, and assuming that bed material size is

constant. Hack (1957) found that the case of n = ‐1 provides a useful graded profile index (or

gradient index) for many of the Appalachian rivers he observed. This produces a simplified

version of Eq. (1), where k is simply equal to the product of SL. Assuming that n = ‐1, the

integrated version of Eq. (1) is of the form

H = C ‐ k ln L (2) where H is the channel bed elevation, and k and C are empirically derived constants. This is a

straight line semilog relationship between channel elevation (linear) and channel distance

(logarithmic), which represents an idealized graded profile, and has been branded by some

authors as the Hack profile (Pérez‐Peña et al., 2009; McCleary et al., 2011). For comparison of

river profiles of different lengths, normalization of the semilog profile using the graded river

gradient (K) has shown to be useful for revealing profile anomalies relative to a theoretical SL/K

index (Seeber and Gornitz, 1983; Pérez‐Peña et al., 2009; McCleary et al., 2011). The

normalization factor K being the SL index, but calculated for the entire river profile as

(h h ) (3) K = s f ln Lt

where hs is the elevation of the drainage divide, hf is the elevation of the river outlet, and Lt is

the total length of the entire river.

The primary contribution of Flint (1974) was the equation relating channel slope (S) to drainage

area (Ad) based on an empirically derived power‐law:

–θ S = ks Ad (4)

where ks is known as the steepness index and θ as the concavity index (Whipple, 2004; Gonga‐

Saholiariliva et al., 2011). Values of the concavity index (θ) tend to vary between 0.4 and 1,

with the average often considered to be in the range of 0.6 (Flint, 1974; Whipple, 2004; Gonga‐

Saholiariliva et al., 2011). Assuming that discharge (Q) and drainage area (Ad) increase at roughly an equivalent rate, the lower range value of θ ≈ 0.4 from the Flint equation produces a similar relationship to that proposed by Leopold and Wolman (1957) for the slope‐discharge threshold between meandering and braiding channel patterns:

S* = 0.0125 Q –0.44 (5)

where S* is the meandering–braiding threshold slope (Leopold and Wolman, 1957).

Criticism of the graded river concept tends to focus on the questionable universality of the

smooth concave‐up longitudinal profile as the ultimate state of river evolution (e.g., Phillips and

Lutz, 2008; Phillips, 2011). It is essentially a concern with treating the graded river state as the

normative fluvial condition, in the sense that the geometrically concave‐up profile is not

necessarily the most typical (i.e., exceptions may be more common than the rule) or in terms of

a graded slope profile representing an ideal equilibrium steady state for the entire length of a

fluvial system (but concepts of equilibrium are scale‐dependent in space and time; e.g.,

Schumm and Lichty, 1965; Hickin, 1983; Phillips, 2011). So, at best the geometric graded river

profile represents a deep property of fluvial systems (Smith et al., 2000), or at worst it

represents an arbitrary benchmark (Phillips et al., 2010), from which expected variations in

channel slope can be gauged against a conceptual standard.

The graded river concept from Mackin (1948), and his slope of transportation, seems most in

tune with the ideas of geomorphic equilibrium, whereby slope is a dependent variable and is easily adjusted to spatial and temporal changes in sediment supply and calibre. However, this

idea becomes troublesome if coupled with the geometric notion of smooth concave‐up profiles, particularly when valley fills, sediment inputs, and bed material sizes are spatially variable and

do not always trend downstream in a systematic way. Profile irregularities that may be defined

as knickpoints, knickzones (Whipple, 2004; Phillips and Lutz, 2008; Phillips et al., 2010; Gonga‐

Saholiariliva et al., 2011), or other convex features embedded within a long profile may not be

in disequilibrium if viewed at the reach scale. At reach scales, slope may be considered an

independent variable over periods of years, centuries, and perhaps even millennia (Schumm

and Lichty, 1965; Hickin, 1983). On the other hand, over geologic timescales of the Holocene or

Quaternary, profile evolution and channel slope adjustments may tend toward some ideal form, regardless of whether or not it can ever or will ever be achieved.

The concept of the graded river profile, insofar as it represents an exponential concave‐up

Hack‐type profile, has been demonstrated as a useful model for interpreting landscape diversity

(Hack, 1973), particularly with respect to tectonic, lithological, sedimentological, glacial legacy, and base level controls on fluvial profile evolution. The assumption being that deviations from

the theoretical benchmark profile represent interruptions to a graded condition (idealized by systematic trends in slope, discharge, and sediment size). Thus, the long profile may be used to tease out clues of the underlying environmental controls that complicate fluvial landscape evolution.

2.2.2 Specific stream power approach

In the most general physical terms, power is the rate at which energy is used, or the rate at which work is performed, expressed in units of watts (W). The concept of stream power thus is an expression of the potential for flowing water to perform geomorphic work, specifically in terms of sediment transport rates. As formulated by Bagnold (1966), the potential energy of

water flowing downslope with gravity can provide an expression of the available stream power

in fluvial channels:

 = QS (6)

where  is the specific weight of water (9792 kg∙m∙m‐3∙s‐2 at 20°C; Yang, 1996), Q is the

discharge (m3s‐1), and S is the channel slope. The total stream power () per unit length of

stream (in units of Wm‐1), alternatively referred to as cross‐sectional stream power (Rhoads,

1987), can also be expressed per unit bed area if divided by the channel width (w):

QS = (7) w

where ω is the specific stream power in units of Wm‐2. Expressed as the average cross‐ sectional stream power per unit width of the channel (or stream power per unit bed area, m2),

specific stream power may also be referred to as the mean stream power (Rhoads, 1987).

The introduction of the stream power concept by Bagnold (1966) was explicitly to engage with matters of sediment transport; and in particular, bedload transport was the initial concern.

Mostly proceeding from the work of Shields (1936), an enormous amount of research by engineers and geomorphologists in the last century has focused on the incipient motion of bed material in terms of critical shear stress related to the forces of flowing water acting on the bed

(for recent reviews see: Buffington and Montgomery, 1997; Ferguson, 2005, 2012; Church,

2006; Lamb et al., 2008; Recking, 2009; Parker et al., 2011). However, as reasoned by Eaton and Church (2011), sediment transport rates (or the volume of sediment transported with time) may be more closely related to the bulk transfer of momentum from the fluid to the bed material, which is a stream power phenomenon.

A growing number of recent studies have updated Bagnold’s (1966, 1980) stream power approach, with the notion of critical stream power for bedload sediment transport founded in the parameter of specific stream power (Ferguson 2005, 2012; Petit et al., 2005; Eaton and

Church, 2011; Parker et al., 2011). Parker et al. (2011) proposed a dimensionless form of critical stream power (c) as it relates to grain size (Di):

c c = 3/2 (8) s(gRDi)

‐2 ‐1 ‐1 where c is the critical specific stream power (Wm or N∙m ∙s ); s is the density of the sediment (kgm‐3); g is the gravitational acceleration (9.81 ms‐2); and R is the submerged specific

gravity of the sediment grains (R = s ‐ w)/w; where w is the density of water; Ferguson,

2012). The value of the dimensionless critical stream power parameter is on average c = 0.1

3/2 (Parker et al., 2011). The result that c varies with Di is also supported by empirical field measurements of bedload transport rates reported by Petit et al. (2005).

The spatial properties and downstream trends in stream power have been an important topic of research following Bagnold (1966, 1980), including Graf (1983), Magilligan (1992), Lecce

(1997), Knighton (1999), and Flores et al. (2006). However, rather than the initial focus on

idealized mid‐basin peaks in stream power, Hack’s (1973) view about the spatial arrangement

of channel slope (and thus stream power) as a matter of exploring diversity in the landscape is

perhaps the more productive pursuit. Many studies over the last decade have taken advantage

of advanced automation from geomatic computer mapping applications to investigate the

spatial distribution of stream power in small and large river drainage networks (Finlayson et al.,

2002; Fonstad, 2003; Reinfelds et al., 2004; Jain et al., 2006; Barker et al., 2009; Pérez‐Peña et

al., 2009; Vocal Ferencevic and Ashmore, 2012; Snyder et al., 2013). Particularly in the growing

use of GIS and digital elevation models (DEMs), spatial variations in river channel slope can be

rapidly mapped over large areas and throughout entire drainage networks, a first basic step in

calculating stream power (Eq. 6).

However, some authors have been reluctant to calculate specific stream power (Eq. 7)

presumably because of the difficulties and uncertainties in modeling channel width over large

areas where simple indices such as drainage area or channel length may not explain enough of

the variance (e.g., Jain et al., 2006; Barker et al., 2009; Pérez‐Peña et al., 2009; Vocal Ferencevic

and Ashmore, 2012). Others seem to be willing to accept some uncertainty in width models, including or even focusing on specific stream power calculations in their analysis (e.g.,

Knighton, 1999; Finlayson et al., 2002; Fonstad, 2003; Reinfelds et al., 2004; Flores et al., 2006).

Basic power‐law regime models for width indexed to drainage area or channel length (as proxies for discharge) may in many cases be adequate to explain width variations at the watershed scale and should at least be used for first‐order estimates over catchment scales.

The spatial properties of total stream power (without dividing by width) are strongly dictated by

downstream increases in discharge, so the scaling effects of channel size can mask meaningful

stream power variations if comparing large and small channels.

Acknowledging that channel widths (and thus ω) are a response to total stream power ( ~ QS) and boundary conditions (e.g., bank strength), specific stream power has a more robust theoretical foundation in sediment transport studies (Ferguson, 2005; Eaton and Church, 2011;

Parker et al., 2011), and even channel pattern prediction (Nanson and Croke, 1992; van den

Berg, 1995; Kleinhans and van den Berg, 2011). As such, specific stream power was selected as

the parameter by which to gauge fluvial processes in southern Ontario over geomorphically significant scales in space and time, in relative and in absolute terms.

2.3 Regional Setting

The Laurentian Great Lakes of North America are perched within the low‐relief topography of the Precambrian continental craton (Laurentia), the Paleozoic sedimentary strata of the

intracratonic Michigan basin, and a mantle of Pleistocene glacial drift. The combined watershed of the entire Great Lakes basin covers an area of about 765,990 km2, spanning both

Canada and the United States (Figure 2.1). The mean water surface elevation of each lake

decreases from Lake Superior at 183 m, to Lakes Michigan and Huron at 176 m, Lake Erie at 173

m, and Lake Ontario at 74 m above sea level (asl). From Lake Ontario, the drainage basin

ultimately outlets to the North Atlantic Ocean by way of the St. Lawrence River.

Figure 2.1: Watershed for the Laurentian Great Lakes of North America. Study area of southern Ontario is indicated by (SO). The inset map of southern Ontario in the upper‐right corner shows the locations of the Niagara escarpment (NE) and the Oak Ridges moraine (ORM). (AA) Algonquin Arch of the Precambrian basement bedrock.

The modern topographic impression of the basin is the product of recurring continental

glaciations over the last ~ 2 million years that scoured the Precambrian and Paleozoic bedrock,

while also leaving complex sequences of glacial sediments particularly during deglaciation

about 10 to 15 thousand years ago.

Especially in the southern half of the Laurentian Great Lakes basin, fluvial drainage networks

are carved into the complex architecture of thick glacial overburden, with various formations of

Paleozoic sedimentary bedrock only exposed locally. This study focuses on the area of

peninsular southern Ontario (Figure 2.1), where rivers drain to base levels of the three lower

lakes (Huron, Erie, and Ontario) and have small drainage areas of < 104 km2. The humid

continental climate is characterized by annual rainfall in the range of 800–1200 mm (Paixao et

al., 2011), with annual flooding during spring snowmelt dominating river flow regimes (Gingras

et al., 1994; Javelle et al., 2003). The two most prominent physiographic features that shape

the drainage patterns of the study area as indicated in Figure 2.1 (inset map) are the Niagara 34

escarpment, a ridge of resistant dolomitic limestone from the Silurian Period that arches around the north half of the Michigan basin; and the Oak Ridges moraine, a stratified

interlobated moraine (kame‐type) formed during deglaciation about 12 to 13 thousand years

ago within a complex sequence of subglacial, ice‐contact, and proglacial depositional

environments (Barnett et al., 1998). The highest elevation within the study area is ~ 544 m asl, corresponding to maximum basin relief values of about 368, 371, and 470 m for Lakes Huron,

Erie, and Ontario, respectively.

In addition to the Oak Ridges moraine, southern Ontario is also covered by a succession of

deglacial kame moraines and recessional till moraines that roughly surround the central

uplands (i.e., “Horseshoe moraines”). This area is aptly referred to as the Ontario Island as it was briefly surrounded by glacial ice lobes and proglacial lakes during the early stages of deglaciation just over 13 thousand years ago (Chapman and Putnam, 1951). The moraines, and

associated glaciofluvial outwash valleys, have topographically shaped the modern arrangement

of river catchments and stream drainage networks. Moraine and outwash deposits also

represent the dominant supply of coarse‐grained sediments to fluvial channels and valley fills,

with rivers in many cases cut sharply through the crests of the terminal moraine landforms.

Much of the remaining land surface is characterized by extensive plains or gentle slopes of

glacial till and glaciolacustrine sand, silt, and clay that impart relatively flat channel gradients and fine‐grained alluvial floodplains. The glacial overburden is generally thick, with bedrock

only occasionally exposed near the surface. For southern Ontario, 22 major river catchments were selected to map the spatial properties of specific stream power within the context of the glacial landforms and sediments (Figure 2.2).

Figure 2.2: Map of study area showing 22 major river drainages selected in southern Ontario, including raster image of 10‐m DEM in the background (Ontario Ministry of Natural Resources [OMNR], 2008). Locations of WSC streamflow gauges used for discharge regime models are indicated by black triangles and subregions are labeled I to VIII. Field locations for bankfull width measurements for this study are indicated by white circles. Numbers 1 to 22 identify major river drainages selected for analysis.

2.4 Materials and Methods

To evaluate the spatial properties of specific stream power (Eq. 7) in southern Ontario, channel

slope, discharge, and width were calculated based on digital geomatic mapping and field data

from multiple sources. Specific stream power mapping was then overlaid on glacial landform

and surface geology mapping from published digital datasets. All mapping of the study area was analyzed using standard GIS software, within the Universal Transverse Mercator projection

(UTM, Zone 17N, units of metres) referenced to the North American Datum 1983 (NAD83).

Channel slope (S) was evaluated based on river longitudinal profiles derived from Ontario’s

Provincial DEM Version 2.0.0 (OMNR, 2005, 2008). The provincial DEM is a derivative of

elevation base data acquired through an Ontario base mapping (OBM) program undertaken

between 1976 and 1996. The OBM dataset for southern Ontario was produced at a scale of

1:10,000 and was derived from aerial photogrammetry acquired at a 1:30,000 scale. The absolute positional accuracy of the OBM dataset is 5 m, and the raster cell resolution of the derived DEM is 10 m. The elevation data for the DEM were resolved from a combination of 5‐m

OBM contours, lake shoreline contours, low density spot height elevations, and high density digital terrain models (DTM). The absolute vertical accuracy of the DEM is reported to be 2.5 m

(OMNR, 2008), and elevations are relative to the Canadian Geodetic Vertical Datum 1928

(CGVD28) of the Geodetic Survey Division, measured in metres above sea level (m asl).

The provincial DEM for southern Ontario also includes hydrological enforcement to ensure that hydrologic features such as streams and rivers all decrease with elevation in a downstream direction. Hydrological enforcement was completed by OMNR (2008) using the ANUDEM 5.1

(Australian National University DEM) software package and its associated drainage enforcement algorithm to create a hydrologically conditioned DEM (Hutchinson, 2005). The

hydrologically enforced DEM provided the basis to calculate the drainage areas (Ad) for all raster cells in each drainage network from a flow accumulation raster image (Figure 2.3).

Discharge (Q) was evaluated using streamflow records for gauging stations within the study area available from Water Survey of Canada’s hydrometric monitoring program and corresponding HYDAT database (Environment Canada, 2011). The HYDAT database was

accessed online at http://www.ec.gc.ca/rhc‐wsc.

Figure 2.3: Flow chart of GIS and spreadsheet analysis for DEM slope generalization and stream power mapping. Methods in step 1 are similar to those described by Jain et al. (2006) and Vocal Ferencevic and Ashmore (2012).

Of almost 250 gauging stations within the study area, 210 stations were selected based on a minimum period of record of 6 years. Of the 210 selected gauges, 95% of the stations had > 10

years of record. The average period of record was 33.5 years, with a maximum of 96 years. For

each gauging station, flood frequency analysis was performed using the maximum mean daily

discharge values for each year of record. Although instantaneous maximum discharges are

considered of more geomorphic significance, particularly for smaller watersheds, mean daily

discharges were used for the annual maximum series to provide a larger sample size geographically in terms of the number of stations, as well as to maintain longer periods of

record with fewer gaps in the streamflow data of each station.

Flood frequency analysis of streamflow records was completed using the United States Army

Corps of Engineers Hydrologic Engineering Center Statistical Software Package (HEC‐SSP,

Version 2.0) accessed online at http://www.hec.usace.army.mil. The standard flood frequency

analysis in HEC‐SSP is based on the United States Geological Survey (USGS) Bulletin‐17B,

Guidelines for Determining Flood Flow Frequency (USGS, 1982). As such, HEC‐SSP analysis

assumes that a log‐Pearson type III distribution provides the best fit to define the probabilities

of annual flood series. Flood frequency analysis in HEC‐SSP was used to estimate the flood

discharges at each of the 210 WSC gauging stations, including the conventional 1.5‐year, 2‐year, and 2.33‐year recurrence intervals to represent the approximate bankfull or mean annual flood discharges (Leopold and Wolman, 1957; Leopold et al., 1964).

Channel width (w) data are based on rapid field surveys conducted in 2010 and 2011 on rivers

and streams throughout the study area. Rapid surveys of channel width were completed at a

total of 542 field sites, accessed by road crossings and, in some cases, by inflatable boat. The

bankfull channel width was interpreted based on major bank slope inflections, vegetation

changes, and field evidence of modern floodplain elevations. At each site, the average channel

width was sampled at between 1 and 5 cross sections and was measured using a range finder

and stadia rod. Of the 542 rapid field surveys, a subset of 120 sites (with 15% randomly

selected) also included surveys of channel depth, bed material, floodplain sedimentology,

channel pattern, and channel slope.

Surface geology and landform mapping of southern Ontario were acquired as digital geospatial data from the Ontario Geological Survey (OGS), accessed online at

http://www.geologyontario.mndm.gov.on.ca.

Glacial landform mapping is based on the original work of Chapman and Putnam (1951) titled

The Physiography of Southern Ontario, which was republished and updated in 1966 and 1984,

and was subsequently digitized by the Ontario Geological Survey (Chapman and Putnam, 2007).

Surface geology was also evaluated from digital mapping of bedrock and Quaternary geology from the Ontario Geological Survey (OGS, 1993, 1997). These geospatial datasets collectively provided a range of information on surface geology in southern Ontario, including bedrock types, landforms, depositional environments, material types, and sediment textures.

Accordingly, landform and surface geology maps were cross‐referenced spatially with

topographic information from the 10‐m DEM to classify river reaches by glacial landform type.

2.5 Specific Stream Power Inputs

2.5.1 Discharge regime models

The 2‐year flood (Q2) was selected as the standard flood frequency for mapping of specific

stream power in this study, as it also has widespread relevance in hydrological rainfall–runoff

modeling and river engineering applications (Annable, 1996b; Lecce, 1997; Sherwood et al.,

2005; Mulvihill et al., 2009; Richol and Boley‐Morse, 2009; Annable et al., 2011). Further, the 2‐

year flood discharge has been used in numerous other stream power studies, including

Magilligan (1992), Reinfelds et al. (2004), Jain et al. (2006), Barker et al. (2009), Pérez‐Peña et al. (2009), Vocal Ferencevic and Ashmore (2012), and Snyder et al. (2013). Still others have assumed the near‐equivalence of bankfull discharge with the mean annual flood or its corresponding flood frequency Q2.33 (e.g., Knighton, 1999; Finlayson and Montgomery, 2003;

Phillips and Robert, 2007).

For modeling discharge (Q) continuously downstream, it has been accepted practice in

longitudinal stream power calculations to employ power‐law relationships of the form

 Q =  Ad (9)

2 where Ad is the drainage area in km (derived from DEM analysis; Figure 2.3), with the

coefficient  and exponent  estimated by statistical regression of empirical data (Magilligan,

1992; Knighton, 1999; Fonstad, 2003; Reinfelds et al., 2004; Flores et al., 2006; Jain et al., 2006;

Pérez‐Peña et al., 2009; Vocal Ferencevic and Ashmore, 2012). The normal values for  from

previous studies are in the range of 0.6 to 1.0 (Leopold et al., 1964; Knighton, 1987; Eaton et al.,

2002; Flores et al., 2006), which are consistent with observed values reported for rivers within

the Great Lakes watershed as per Table 2.1. It has also been suggested that  may approach

values closer to 1 with increasing climate humidity (Blöschl and Sivapalan, 1997; Eaton et al.,

2002).

While the strength of this predictive relationship does depend on the relative homogeneity of

surface geology and climate, spatial variations can be mitigated by regional and subregional

calibrations (Gingras et al., 1994; Eaton et al., 2002; Javelle et al., 2003; Flores et al., 2006).

Based on data from a wide range of environments and river sizes, Finlayson and Montgomery

(2003) suggested that the area–discharge proxy is indeed a reasonable model for smaller river

systems (cf. similar conclusions for drainage areas < 104 km2; Eaton et al., 2002), but that the

assumption does begin to break down when significant hydroclimatic variations take effect over

very large drainage areas. Subregional climate variability has been previously investigated in

southern Ontario in terms of snowmelt effects on flood regimes (Gingras et al., 1994; Javelle et

al., 2003) and patterns of heavy rainfall (Paixao et al., 2011).

Previous studies within the Great Lakes watershed demonstrate that drainage area can explain

between 60 and 95% of the variation in the Q2 or bankfull discharge when calibrated to scales

of 104–105 km2, with an average of ~ 80% (Gingras et al., 1994; Annable, 1996a; Pandey, 1998;

Sherwood and Huitger, 2005; Mulvihill et al., 2009; Richol and Boley‐Morse, 2009). For

mapping of specific stream power in southern Ontario, statistically calibrated regime models of

Q2 were tested as a function of drainage area (Ad) over the entire study area and for eight (8)

subregions (Figure 2.2). The Q2 versus Ad regression analysis provided in Figure 2.4 of all

selected 210 stream flow gauges within the study area is statistically significant (p < 0.001), with

an r2 value of 0.86. Regional and subregional results for  and  from Eq. (9) are presented in

Table 2.1.

Figure 2.4: Area–discharge regime model statistical regression of empirical data from this study. Short dashed lines are 95% confidence intervals and long dashed lines are 95% prediction intervals. Standard error is 0.22. Parametric assumptions were tested based on p > 0.05, passing for a Shapiro‐Wilk normality test (p = 0.825) but failing for constant variance (p = 0.017). The bankfull discharge b (Qbf) relation of Annable (1996 , n = 47) is plotted for comparison, with the range of drainage areas from the 1996 dataset between 10 and 1000 km2. Table 2.1: Area–discharge regime models for southern Ontario based on Eq. (9); location of southern Ontario subregions is mapped in Figure 2.2 (I–VIII). Discharge Region Source   r2 n Q Southern Ontarioa This study 2 year 0.248 0.910 0.86 210 Southern Ontario Sub‐Regions South Georgian Bay (I) This study 2 year 0.204 0.863 0.84 25 Toronto and Region (II) This study 2 year 0.283 0.863 0.86 32 Niagara Peninsula (III) This study 2 year 0.156 0.945 0.82 24 Grand River (IV) This study 2 year 0.244 0.937 0.92 38 Lake Erie North Shore (V) This study 2 year 0.094 1.047 0.79 22 Thames St. Clair (VI) This study 2 year 0.401 0.876 0.93 34 South Lake Huron (VII) This study 2 year 0.403 0.901 0.97 22 Saugeen River (VIII) This study 2 year 0.152 1.002 0.87 23 Previous Studies in Canada Ontario and Quebec Gingras et al. (1994) 2 year 0.640 0.734 0.70 183 Lake Ontario, Erie, St. Clair Gingras et al. (1994) 2 year 0.239 0.946 0.86 49 Lake Huron Gingras et al. (1994) 2 year 0.244 0.883 0.88 13 Southern Ontario (all Rosgen Annable (1996b) Bankfull 0.52 0.74 0.64 47 types) Southern Ontario (urban & rural) Annable et al. (2011) Bankfull 0.6 0.73 ‐ 49 Ontario Pandey (1998) Annual Flood ‐ 0.715 0.83 109 Southern Ontario Ribeiro and Rouselle Annual Flood ‐ 0.64 ‐ 109 (1996) Previous Studies in USA New York State Mulvihill et al. (2009) Bankfull 1.067 0.810 0.89 82 ‐ Lake Ontario (New York State) Mulvihill et al. (2009) Bankfull 0.730 0.765 0.94 10 ‐ Lake Erie (New York State) Mulvihill et al. (2009) Bankfull 0.911 0.842 0.90 14 Southern Michigan State Richol and Boley‐ Bankfull 0.073 0.950 0.60 28 Morse (2009) Ohio – Lake Erie Sherwood and Huitger Bankfull 1.951 0.637 0.82 37 (2005) a Note, see Figure 2.4 caption for statistical significance and tests of parametric assumptions.

2.5.2 Bankfull width regime models

Downstream predictions of channel width (w) are well founded in the notions of hydraulic

geometry originally proposed by Leopold and Maddock (1953) based on increasing discharge

(Q), with the conventional power‐law relation often expressed where width tends to vary with

the square‐root of discharge (w ~ Q0.5). For longitudinal specific stream power calculations, it

follows from Eq. (9) that modeling channel width (w) continuously downstream may utilize a

power‐law relationship with drainage area (Ad) of the form

b w = a Ad (10)

with the coefficient a and exponent b estimated by statistical regression of empirical data.

Assuming the exponent  in Eq. (9) is close to a value of 1, expected values for the exponent b

in Eq. (10) will be in the range of 0.4 to 0.5 (Leopold et al., 1964; Anderson et al., 2004). Given

that observations for the lower range of  are about 0.6 to 0.7, it follows that the lowest

observed values for b are typically in the range of 0.3 to 0.35. However, we should

acknowledge that downstream width variations are not only driven by increasing discharge, but actually respond to total stream power (product of discharge and slope) while being mediated

by the relative resistance of the boundary materials and bank strength.

The drainage area–width proxy has been used in previous spatial studies of specific stream

power, either explicitly (Fonstad, 2003) or indirectly from hydraulic geometry assumptions for

width based on discharge (e.g., w ~ Q0.5), where discharge has been modeled using drainage

area as in Eq. (9) (Knighton, 1999; Reinfelds et al., 2004; Flores et al., 2006). Other stream power studies have approached channel width in a more fastidious manner (Magilligan, 1992;

Lecce, 1997; Snyder et al., 2013). Still, many studies within the Great Lakes watershed have

demonstrated reasonable log‐log‐linear predictions of width from drainage area as per Eq. (10) and Table 2.2, with area explaining between 60 and 90% of variance in width at regional scales

(average ~ 82%).

For mapping of specific stream power in southern Ontario, statistically calibrated regime models of channel width (w) were tested as a function of drainage area (Ad) over the entire

study area and for each of the 22 major river catchments mapped for this study (Figure 2.2).

The w versus Ad regression analysis provided in Figure 2.5 of 542 field sites within the study

area is statistically significant (p < 0.001), with an r2 value of 0.88. Regional and catchment

results for a and b from Eq. (10) are presented in Table 2.2. As a reasonable approximation, we

have assumed in this analysis that the predicted bankfull widths (wbf) from the regime models

are representative of the 2‐year flood discharge (Q2), especially for generalized patterns of

stream power at the regional scale (c.f. reference width concept of van den Berg, 1995).

Figure 2.5: Area–width regime model statistical regression of empirical data from this study (bankfull width). Short dashed lines are 95% confidence intervals and long dashed lines are 95% prediction intervals. Censored light‐gray data points were excluded in a second test based on effects of vegetation in headwaters, urban‐ ization, and lake back water. Standard errors are 0.150 and 0.127 for uncensored and censored datasets, respectively. Parametric assumptions are not met in the uncensored dataset for p > 0.05, failing a Shapiro‐Wilk normality test (p = 0.008) and a constant variance test (p = 0.001). The censored dataset does pass for normality (p = 0.066) and for constant variance (p = 0.157). A derived bankfull width (wbf) relation from Annable (1996b, n = 47) is plotted for comparison, with the range of drainage areas from the 1996 dataset between 10 and 1000 km2.

Table 2.2: Area–width regime models (bankfull width) for southern Ontario based on Eq. (10); location of 22 select major river drainages in southern Ontario is mapped in Figure 2.2.

Region Source a b r2 n Southern Ontario (all data)a This study 1.160 0.508 0.87 542 Southern Ontario (censored)a This study 1.201 0.499 0.88 493 Southern Ontario River Drainages Nottawasaga (1), Beaver (22) This study 1.114 0.459 0.79 69 Rouge (2), Humber (3), Credit (4), 16 Mile (5) This study 1.846 0.411 0.88 47 Spencer (6), Twenty Mile (7), Welland (8) This study 1.211 0.580 0.81 35 Grand River (9), Sandusk (10), This study 0.892 0.558 0.93 54 Big (11), Big Otter (12), Catfish (13), Kettle (14) This study 2.051 0.398 0.78 48 Thames (15), Sydenham (16) This study 1.450 0.475 0.93 25 Ausable (17), Bayfield (18) This study 0.991 0.519 0.89 129 Maitland (19) This study 0.732 0.594 0.94 46 Lucknow (20), Saugeen (21) This study 1.317 0.490 0.89 110 Previous Studies in Canada Southern Ontario (all Rosgen types) Annable (1996b) 2.69 0.36 ‐ 47 Previous Studies in USA New York State Mulvihill et al. (2009) 4.256 0.401 0.84 281 ‐ Lake Ontario (New York State) Mulvihill et al. (2009) 2.647 0.458 0.89 33 ‐ Lake Erie (New York State) Mulvihill et al. (2009) 4.220 0.419 0.79 50 Southern Michigan State Richol and Boley‐Morse 2.025 0.440 0.69 28 (2009) Ohio – Lake Erie Sherwood and Huitger 4.632 0.356 ‐ 45 (2005) Conterminous USA Faustini et al. (2009) 2.81 0.24 0.42 1558 ‐ Eastern Highlands, USA Faustini et al. (2009) 2.68 0.38 0.75 275 ‐ Great Lakes, USA Faustini et al. (2009) 2.45 0.33 0.62 53 ‐ Northern Appalachians, USA Faustini et al. (2009) 2.55 0.39 0.72 87 ‐ Upper Midwest, USA Faustini et al. (2009) 1.74 0.39 0.68 70 a Note, see Figure 2.5 caption for explanation of censored dataset, including statistical significance and tests of parametric assumptions.

2.5.3 DEM longitudinal profile extraction and slope generalization

Longitudinal profiles for 22 major watercourses in southern Ontario were extracted from the

hydrologically enforced 10‐m DEM (OMNR, 2005). Methods included GIS and spreadsheet

algorithms to produce generalized longitudinal profiles of elevation, slope, and stream power

using four basic steps: (i) DEM preprocessing, (ii) raw profile extraction, (iii) slope

generalization, and (iv) stream power calculations (Figure 2.3). The initial methods of DEM

preprocessing and profile extraction are similar to those described by others (e.g., Jain et al.,

2006; Vocal Ferencevic and Ashmore, 2012), while the subsequent procedures for slope

generalization are similar to methods from Knighton (1999) and Reinfelds et al. (2004) but are

ultimately unique to this study.

Raw DEM profiles displayed variations in slope over more than five orders of magnitude when calculated between successive 10‐m raster cells (e.g., Figure 2.6). Stream power calculations as envisioned by Bagnold (1966) are to be based on average conditions along a representative

channel length, and thus the high resolution slope data from the 10‐m DEM (or 14.1‐m spacing

for diagonally joined cells) did not meet this criteria (cf. Jain et al., 2006). Real topographic

controls on channel slope included reach scale variations over kilometers as well as subreach

scale variations (e.g., riffle‐pool sequences, dams, geologically controlled knickpoints).

Superimposed over real topographic controls were high frequency variations inherited from

various stages of DEM creation and preprocessing. Such data artifacts may be from original

photogrammetric inaccuracies, multiple contour and DTM data sources, DEM interpolation techniques, hydrologic enforcements, and other preprocessing steps to produce the seamless

DEM mosaic.

Figure 2.6: Example DEM profile from the Ausable River (#17a) showing steps of slope generalization as outlined in Figure 2.3. The raw DEM elevation profile (white line) is presented over the generalized profile (thick black line).

Given multiple scales of variance imposed by real and by artifact sources, a number of slope

generalization techniques were considered to represent reach scale energy gradients (assumed

to be representative of water surface gradients at bankfull discharge flood levels).

Numerous profile smoothing techniques have been suggested to dampen the inherent noise of

DEM‐derived channel slopes and to represent scales meaningful to fluvial processes. These

techniques may be broadly classified as either least‐squares curve‐fitting, profile slicing‐and‐

dicing, or low‐pass moving average filters. For modeling stream power, most profile curve‐

fitting tests have been either classic Hack‐type exponential curves or polynomial curves ranging

from second‐ to fourth‐order polynomial equations (Knighton, 1999; Jain et al., 2006; Barker et

al., 2009; Vocal Ferencevic and Ashmore, 2012); although Phillips and Lutz (2008) also tested

other forms. Profile slicing‐and‐dicing techniques have been the most popular in stream power

studies, including horizontal slicing (Knighton, 1999; Jain et al., 2006; Pérez‐Peña et al., 2009;

Vocal Ferencevic and Ashmore, 2012); horizontal dicing by coarse‐resolution DEMs (Finlayson

et al., 2002; Finlayson and Montgomery, 2003; Barker et al., 2009) or by reach bin‐averages

(McCleary et al., 2011); vertical slicing (Reinfelds et al., 2004; Snyder et al., 2013); and moving window averaging methods (Knighton, 1999; Jenkins, 2007; Snyder et al., 2013).

Each method of profile smoothing was considered to assess the advantages and disadvantages.

While polynomial equations may be appropriate descriptors of some river profiles (e.g., Phillips

and Lutz, 2008), real knickpoints and complex profile irregularities are difficult to maintain in the models (Jain et al., 2006); and modeling of separate polynomial functions for individual tributary branches requires that discrete profiles then be stitched back together (Vocal

Ferencevic and Ashmore, 2012). While horizontal slice methods are appealing, relatively short

slices are required to avoid distorting the location of knickpoints; and the slopes for steeper

sections can be sensitive to the arbitrary start and end points of the successive horizontal slices.

Vertical slices are perhaps most appropriate for maintaining the location and accuracy of slopes

for complex profiles, as the horizontal sampling interval decreases through steeper sections of

the profile. While vertical slice methods may not resolve localized elevation changes less than

the slice thickness over long flat sections of the profile, the approach allows for appropriate

consideration of the origins of the data from which the DEM was derived (Reinfelds et al.,

2004), and specifically the vertical accuracy of the original contours and/or DTM points. Given

that horizontal and vertical slice methods still tend to produce choppy longitudinal plots of the

resultant slope data, moving averages may subsequently be used to further smooth out the

profile and bring out the primary features. Knighton (1999) recognized this by applying a 5‐ point moving average to stream power data that had been derived from 0.5‐km horizontal

profile slices (i.e., a 2‐km moving window).

For this study, a vertical slice method was applied to the raw DEM profiles (Figure 2.3 and

Figure 2.6), using a 2.5‐m slice thickness based on the reported vertical accuracy of the original

DEM (OMNR, 2008). To further smooth out the longitudinal profiles, moving window averages

were subsequently applied. In terms of time series signal processing, the moving average

provides a low‐pass filter to reveal low‐frequency features, fluctuations, and trends in the data

series. Selection of the moving window length was based on a representative reach scale for

southern Ontario rivers. The river reach being defined as a length of channel with relatively

uniform or gradually varying characteristics of channel morphology, slope, discharge, and

boundary materials. Montgomery and Buffington (1997) suggested that reach length is

typically in the range of 10–20 channel widths for mountain drainage basins (cf. 20–30 widths

from Rosgen, 1996). Assuming a factor of 20 widths is representative for channels 10–100 m

wide, most river reaches in southern Ontario are expected to be about 0.2–2 km in length.

Constraining the generalization approach to the upper range, multiple passes of the moving

average were applied to the vertically sliced profiles at increasing window lengths from 1 to 2

km at 500‐m intervals (Figure 2.3). In a few select cases, the moving average window length was further increased with additional passes to emphasize landscape scale patterns in the most complex profiles (e.g., #9 Grand River up to a window length of ~ 5 km).

This study combines vertical slice and moving average approaches to generalize DEM slopes at

multikilometre reach scales, integrating methods used by both Knighton (1999) and Reinfelds et

al. (2004) to calculate downstream stream power trends. However, it is acknowledging that

averaging over 1 to 2 km may appreciably over‐generalize smaller streams and tributaries,

particularly for drainage areas < 102 km2, and higher resolution studies should explore scaling

reach length with channel size. With respect to landscape patterns in stream power,

multikilometre slope generalization (particularly scaled to river reaches draining 102 to 104 km2)

was considered the most appropriate for investigating the geomorphic expression of fluvial

processes operating over geomorphically significant timescales, and specifically over the

Holocene in southern Ontario.

2.6 Results and Discussion

2.6.1 Specific stream power mapping

Based on the modeled inputs for discharge (Q), slope (S), and width (w), longitudinal variations

in specific stream power (ω, Eq. 7) were calculated along the selected river profiles using spreadsheet algorithms. For each watercourse, a contiguous series of extracted watercourse points (corresponding to the UTM locations for the centre of each 10‐m raster cell from the original DEM grid) was then imported back into the GIS environment (Figure 2.3) to create a

stream power map. The specific stream power maps for all selected watercourses within southern Ontario shown in Figure 2.7 were then overlaid on the glacial landform mapping of

Chapman and Putnam (2007).

When viewed at catchment and regional mapping scales, the point series for each watercourse

displays a near‐continuous image of longitudinal patterns in specific stream power. Although values of specific stream power were calculated at discrete intervals, generalized patterns are embedded within the Q, S, and w inputs; and thus low‐frequency spatial patterns visually emerge as conveyed by the colour‐ramp scale in Figure 2.7 (note upper limit of ω ≈ 100 Wm‐2 is roughly equivalent to the grain size boundary between gravel and cobble of D = 64 mm, as per the critical specific stream power approach of Parker et al., 2011).

Figure 2.7: Stream power mapping of select major river drainages in southern Ontario, within context of glacial landform mapping of Chapman and Putnam (2007). The horizontal scale of reach generalization (~ 2 km) is indicated on scale bar for comparison with larger scale fluctuations in specific stream power (5–50 km).

Relative to the 1‐ to 2‐km reach‐scale moving average applied to channel slopes, downstream

fluctuations in stream power are typically of much lower frequency in the range of 5 to 50 km,

arguably corresponding to the spatial scale of primary glacial landforms (cf. glacial macroforms

of Brardinoni and Hassan, 2006).

Spatial associations between stream power and generalized glacial landforms in Figure 2.7 are

visually evident. High values of specific stream power tend to correspond with glacial moraines

(30–100+ Wm‐2) and low values tend to correspond with till and glaciolacustrine plains (< 20

Wm‐2). Till plains are more variable, with some reaches as high as 30–50 Wm‐2. Outwash deposits beyond moraine landforms also exhibit variable results, but stream powers are 52

typically higher than glaciolacustrine plains, in the range of 20–60 Wm‐2. While not directly

associated with glacial landforms, some river reaches are associated with higher stream powers

because of post‐glacial environmental changes and temporally lagged river incision processes.

For example, post‐glacial and Holocene fluctuations in lake levels have resulted in base level

entrenchment at some river outlets. Similarly, tributaries entering incised river valleys often

have higher channel slopes and stream powers as profile adjustments have not kept pace with

the degradation of the main branch.

Other direct glacial landform controls on slope and stream power are apparent, such as

shoreline beach deposits and topographic transitions between glacial landforms (e.g., from till

plain into pro‐glacial basin). Anomalous slopes, and thus higher stream powers, are also occasionally associated with shale and limestone bedrock that is exposed from beneath areas of thin glacial drift. While the spatial pattern of bedrock exposure may not be ostensibly a glacial feature, the variable thickness of the glacial mantle and the underlying bedrock

topography are the product of numerous cycles of continental glaciation over the Quaternary.

The statistical distribution of specific stream power in southern Ontario is strongly right‐skewed with an average condition in the range of ω ≈ 34 Wm‐2 (Figure 2.8A). A sample of 146 reaches from the river catchments presented in Figure 2.7 was selected and classified based on the glacial landform mapping of Chapman and Putnam (2007). Specific stream power results for each reach are plotted versus drainage area in Figure 2.8B. Given the generalized nature of the

glacial landform data, the reach classification process also relied on cross‐referencing of surface

geology maps, DEM topographic maps, and first‐hand field observations. However, in many cases the reach selection process was reversed, whereby particular glacial landforms were identified and then the most representative river reach was selected.

Figure 2.8: (A) Histogram of right‐skewed distribution for specific stream power results in southern Ontario. Mean value of specific stream power is ω ≈ 34 Wm‐2 and the median is 24 Wm‐2. Approximately 4% of the data are in the range of 100 to 200 Wm‐2 and 0.4% of data are above 200 Wm‐2. Percentages are based on ~400,000 points extracted as the centre of each watercourse raster‐cell from the 10‐m DEM grid. (B) Plot of modeled specific stream power versus drainage area for 146 river reaches in southern Ontario. Reaches are classified by dominant glacial landform or other environmental controls, including base level entrenchment in some cases. Shaded area is theoretical stream power domain of meandering river floodplains (10–60 Wm‐2) as per Nanson and Croke (1992).

2.6.2 Profile analysis and SL/K index

Empirical regression analysis of generalized profile data for the selected 22 rivers was applied

to determine the profile index parameters of the Hack and Flint models from Eqs. (2) and (4),

respectively (Table 2.3). Hack’s k and Flint’s ks coefficients are both essentially relative indices

of catchment steepness, while Flint’s θ exponent is a dimensionless index of profile concavity that tends to vary between about 0 and 1. Following work on bedrock and tectonically controlled fluvial systems (Whipple, 2004; Gonga‐Saholiariliva et al., 2011), theoretical ranges of the θ concavity index have been used to interpret the relative concavity of profiles in the context of dominant environmental controls. Specifically, low concavity ranges from 0 to 0.4, moderate concavity ranges from 0.4 to 0.7, and high concavity ranges from 0.7 to 1.0 or greater

(Whipple, 2004; Gonga‐Saholiariliva et al., 2011).

Table 2.3: Profile analysis results for Hack and Flint equations; locations for select major river drainages in southern Ontario are mapped in Figure 2.2. Hack Flint River # k C K SL/K Ratio k θ Drainage s Eq. (2) Eq. (2) Eq. (3) Eq. (8) C/hs Eq. (4) Eq. (4) 1 Nottawasaga 88.1 1211 30.5 2.9 2.2 0.201 0.95 2a Rouge 54.7 697 23.2 2.4 2.2 0.008 0.23 3 Humber 98.2 1264 35.6 2.8 2.6 0.027 0.45 4 Credit 116.5 1492 38.7 3.0 2.8 0.002 ‐0.04 5b Sixteen Mile 106.9 1276 28.7 3.7 3.3 0.001 0.12 6 Spencer 68.6 906 25.1 2.7 2.7 0.002 ‐0.02 7 Twenty Mile 26.8 463 15.7 1.7 1.8 0.001 0.16 8 Welland 27.0 461 11.8 2.3 1.8 0.001 0.01 9 Grand 117.2 1672 29.1 4.0 3.1 0.014 0.41 10 Sandusk 18.5 377 5.4 3.5 1.6 0.002 0.14 11 Big 36.8 607 14.6 2.5 1.8 0.008 0.44 12 Big Otter 36.8 608 13.4 2.7 1.8 0.009 0.46 13 Catfish 28.2 497 11.1 2.5 1.7 0.005 0.31 14 Kettle 35.4 582 12.0 3.0 1.9 0.002 0.10 15 Thames 59.7 925 16.2 3.7 2.5 0.010 0.43 16b East Sydenham 29.8 537 10.5 3.7 1.8 0.010 0.54 17a Ausable 40.9 680 13.9 2.9 2.0 0.011 0.51 18 Bayfield 48.7 784 16.1 3.0 2.2 0.001 ‐0.19 19 Maitland 62.6 1004 23.2 2.7 2.2 0.002 0.05 20 Nine Mile 36.3 620 15.2 2.4 1.8 0.002 ‐0.03 21 Saugeen 116.3 1004 29.6 3.9 3.0 0.003 0.15 22 Beaver 112.7 1422 29.9 3.8 2.8 0.019 0.42 Average 62 870 20 3.0 2.3 0.015 0.25

Concavity results for southern Ontario river systems generally fall within the low concavity to moderate concavity ranges with typical values between 0 and 0.5, although the Nottawasaga

River (#1) does come close to a value of θ = 1. Compared to the interpretations of tectonic uplift and hillslope processes presented by Gonga‐Saholiariliva et al., (2011) for mountainous bedrock landscapes, the low concavities in southern Ontario are instead associated with greater topographic (and/or sedimentologic) conditioning of the river profiles by glacial features, resulting either in large‐scale profile convexity (e.g., #19‐Maitland) or in multiple small‐scale slope anomalies in the middle and lower reaches of the profile (e.g., #4‐Credit, #17a‐Ausable,

#21‐Saugeen).

Pérez‐Peña et al. (2009) and McCleary et al. (2011) used slope deviations from the Hack profile

to investigate landscape controls on fluvial systems in mountainous environments. From this

approach, profile anomalies are identified using the SL/K index, where for this study the

normalization factor K (upper‐case) from Eq. (3) is based on the entire river profile similar to

Pérez‐Peña et al. (2009). By combining Eqs. ((1) and ((3), the SL/K index was calculated along

each point series for selected profiles in southern Ontario using

Sp Lp ln Lt (11) SL/K = (hs hf)

where Sp is the generalized DEM slope at each point, and Lp is the channel distance from the

head of the river to the point location along the profile (cf. lower‐case k of McCleary et al.,

2011).

Longitudinal profiles for eight select rivers in southern Ontario are presented in Figure 2.9, with calculated SL/K curves based on Eq. (11). The exponential Hack profile fit for each river in

Figure 2.9 is shown as a dashed line, with a corresponding Hack curve for the SL/K index that

plots as a nearly horizontal line. Effectively, the Hack curve is a theoretical benchmark for the

SL/K index, which is normalized by river size (Table 2.3). As such, the SL/K curves plot above

the Hack curves for locations where the profiles are oversteepened relative to the Hack profiles,

which often correspond to glacial features such as moraines, bedrock outcrops, or other

topographic and sedimentologic glacial landform transitions. Continuing with this rationale, the

portions of the SL/K curve that fall below the Hack curve may be considered understeepened.

Figure 2.9: Longitudinal profiles for select rivers in southern Ontario with theoretical Hack profiles (dashed). SL/K index curves are based on Eq. (11), and dashed Hack curves plot as a nearly horizontal line. (A) Grand (#9) and Thames (#15) rivers (this page); (B) Nottawasaga (#1), Credit (#4), Big (#11), Ausable (#17a), Maitland (#19), and Saugeen (#21) rivers (next page). Locations of selected rivers identified by numbering are presented in Figure 2.2.

This is particularly apparent for the upper headwater reaches of most profiles where the gentle

slopes of till plains or subtle morainic hillslopes strongly deviate from the expected steep

mountainous topography inherent in an exponential profile model. By fitting the exponential

Hack profile model to rivers in a low‐relief landscape, the theoretical elevation for the drainage

divide for most profiles (i.e., the coefficient C from Eq. (2) tends to be about 2 to 3 times higher

than the actual headwater elevation — hs (i.e., the C/hs ratio in Table 2.3).

Figure 2.9: Continued, see caption on previous page.

2.6.3 Slope–area analysis

The two‐dimensional study of fluvial process domains within slope–area space is a well‐

established concept for mountainous landscapes (e.g., Montgomery and Foufoula‐Georgiou,

1993; Brardinoni and Hassan, 2006) and may essentially be considered a stream power

approach whereby drainage area is a first‐order predictor of discharge. For southern Ontario,

the stream power dimension was formulated within slope–area space by combining Eqs. (7),

(9), and (10) to form

S = A‐0.4 (12) 2100 d

where the relevant range of specific stream power (ω) was used to plot a family of curves

within slope–area space. The exponent ‐0.4 matches Eq. (5) (Leopold and Wolman, 1957) if

0.91 drainage area is substituted for discharge, where for southern Ontario Q ~ Ad based on

Eq. (9), Figure 2.4, and Table 2.1. The slope–area data for 142 reaches with associated glacial

landform classifications are plotted in Figure 2.10A, with constant specific stream power curves

derived from Eq. (12). Following on Figure 2.7 and Figure 2.8B, Figure 2.10A effectively

stratifies reaches based on stream power, particularly between moraine landforms and other glaciogenic plains.

From field and remote sensing observations of channel pattern, single‐channel meandering rivers dominate the study area, while multiple‐channel anabranching or braiding reaches are extremely rare and localized (e.g., braided lower Maitland River where sediment supply and bedload mobility are sufficiently high). Multiple‐channel planforms are likely suppressed in

southern Ontario, as supported by limited field evidence of bedload transport, particularly in

some cobble‐dominated channels (e.g., lack of bed and bar morphology; cf. Snyder et al., 2008).

Figure 2.10: Slope–area plots of 146 river reaches in southern Ontario, with constant specific stream power curves plotted based on Eq. (12). (A) Comparison with reach classifications of dominant glacial landform or other environmental controls, including base level entrenchment in some cases. Shaded area is theoretical stream power domain of meandering river floodplains (10–60 Wm‐2) as per Nanson and Croke (1992). (B) Comparison with reach channel pattern classifications, including reaches with bedrock influence. Shaded area is the theoretical threshold stream power range for multiple‐channel formation of 60–90 Wm‐2 (approximated from multiple sources discussed in text).

Bedload transport rates are most constrained where glacially inherited coarse substrates

coincide with inherited slopes that are not competent to move all of the available bed material.

Further, bedload sediment supply may also be limited within the drainage network further

repressing bar development.

For gravel‐bed rivers, channel pattern transitions between single‐channel meandering and

multiple‐channel braiding have been suggested to be in the range of 30 to 50 Wm‐2 (Ferguson,

1987; Carson, 1984), with meandering river floodplain types generally considered to occupy the range of 10 to 60 Wm‐2 (Nanson and Croke, 1992). However, the onset of well‐developed

gravel‐bed braiding may be more in the range of 150 to 250 Wm‐2, but with distinctive bar

patterns changing at thresholds between about 30 and 90 Wm‐2 (van den Berg, 1995; Kleinhans

and van den Berg, 2011). As such, single‐channel reaches in Figure 2.10B with stream powers

above a theoretical gravel‐bed meandering–braiding threshold (say in the range of 60 to 90

Wm‐2) suggest that high stream powers must be moderated by boundary resistance, in terms of

bed material sediment calibre and bank strength.

Bed material and bank strength

In southern Ontario, systematic decreases in bed material size downstream are frequently interrupted by inherited coarse sediments, disrupting regular fluvial bedload transport patterns. Expected downstream variations in grain size for a theoretical graded profile have been previously described by Hack (1957) based on an empirical power‐law of similar form to

Eq. (4):

0.6 S = 0.006 ( D50 / Ad ) (13)

where D50 is the median grain size of the bed material in mm. Equation (13) was used to plot a

family of curves for expected grain size domains within slope–area space as presented in Figure

2.11A, with observed bed material classifications for each river reach symbolized using increasing circle sizes. Data for southern Ontario appear to reasonably match the Hack (1957)

model (Eq. 13); however, mismatches do suggest that some reaches may exhibit slopes that are

not adjusted to the inherited bed materials (i.e., incomplete or unachievable fluvial slope

adjustments post‐glacially). Snyder et al. (2013) also used slope and drainage area models of

stream power to predict grain size for glacially conditioned watersheds, arguing that their ~

70% success rate indicates channels that are fluvially adjusted in terms of slope and bed materials. However, their model failure rate of ~ 30% may also include some slopes that are not fully adjusted to inherited coarse‐ (or fine‐) grained sediments.

Adjustments in channel width may also be constrained by relative bank strength, further limiting the development of multiple‐channel planforms in southern Ontario. Using the approach of Eaton et al. (2010) in discriminating single‐ and multiple‐channel planforms, increases in relative bank strength are expected to increase the threshold in stream energy

required for multiple‐channel formation such that

S* = 0.40 ’ 1.41 Q* –0.43 (14)

where ’ is dimensionless relative bank strength defined as the ratio of critical shear stress for

entrainment of the channel banks to the channel bed. At low relative bank strengths where the erodibility of the channel bed and banks is roughly equivalent, ’ ≈ 1. The dimensionless

discharge (Q*) is defined by

Q Q* = 2 (15) D50(s‐1)gD50

where s is the specific gravity of the sediment grains, and D50 is the median grain size diameter

in metres (Eaton et al., 2010). The river reach data and planform classifications from southern

Ontario are presented in Figure 2.11B, with slope plotted against dimensionless discharge. The

threshold slopes (S*) for multiple‐channel formation from Eq. (14) are also plotted for

increasing ratios of relative bank strength (’). Horizontal error bars represent the uncertainty

in the D50 estimates used to calculate Q*, with fairly wide ranges in some cases. However, the

analysis does constrain the data in S‐Q* space, suggesting that in many cases single‐channel

meandering rivers in southern Ontario plot above the S* threshold for ’ = 1, particularly for

gravel‐ and sand‐bed channels.

Figure 2.11: (A) Slope–area plot with reach bed material classifications compared to constant grain size curves from Eq. (13) based on Hack (1957). The slope–area domain of gravel is also indicated by shaded area based on Eq. (13). (B) Reach planform classifications compared to a dimensionless threshold for multiple channel formation. The threshold slope (S*) is from Eq. (14) based on the approach of Eaton et al. (2010) with increasing relative bank strength (’). Horizontal error bars reflect uncertainty in estimates of median grain size (D50) used to calculate Q*. The approximate domains of bed material types for southern Ontario data are also indicated.

In other words, the appropriate threshold slope to discriminate multiple‐channel planforms for many gravel‐ and sand‐bed channels should be based on higher relative bank strengths with values of ’ > 1, and perhaps even in the range of 2 to 4. This conclusion is also supported by

field observations of well‐vegetated banks and riparian zones characteristic of the humid

continental climate. Bank strength may be further increased by relatively cohesive floodplain

sediments high in clay, silt, and organic content, particularly in areas of southern Ontario where

fine‐grained glacial till and glaciolacustrine clay are abundant sediment sources.

2.6.4 Glacial conditioning of stream power

As evident in Figure 2.7, Figure 2.8B, and Figure 2.10A, stream power in southern Ontario is

conditioned by glacial features in the landscape. Particularly for river reaches > 102 km2, moraines plot consistently above ~ 30 Wm‐2 with many in the range of 60 to over 100 Wm‐2.

Conversely, till and glaciolacustrine plains tend to plot below 30 Wm‐2, with many under 10

Wm‐2. Reaches characterized by base level entrenchment caused by lake level changes or

lagged tributary incision tend to plot above 30 Wm‐2, with most actually > 60 Wm‐2.

The relative degree to which reach slopes (and stream powers) are inherited may be

interpreted from analysis of the SL/K index (Figure 2.9), whereby oversteepened and

understeepened reaches show evident glacial signals that will influence fluvial processes. As such, the stream power mapping and profile analyses of this study demonstrate that post‐

glacial timescales for fluvial adjustment are spatially variable, from watershed to watershed and longitudinally along river profiles. In terms of Schumm and Lichty’s (1965) ideas for the timescales of slope adjustment, these results suggest that slope in glacially conditioned channels may largely be an independent variable, particularly at large spatial scales and over

the short timescales envisioned by Mackin (1948).

As presented in Figure 2.12, some generalized interpretations of sediment transport in southern Ontario may be drawn from a comparison of modeled specific stream power to critical stream power for bedload transport (Parker et al., 2011) based on observed bed

material size classifications (Figure 2.11A). Self‐forming channels with slopes adjusted to

prevailing fluvial conditions would be expected to have stream powers close to what is required

to transport the existing bed materials, as indicated by a 1:1 line in Figure 2.12.

Figure 2.12: Generalized patterns of bedload transport based on modeled stream power relative to Parker et al. (2011) critical stream power (Eq. 8) for observed sediment classifications. Horizontal arrows indicate the modeled range of stream power within each sediment classification for southern Ontario reaches sampled in this study. The grey area represents an interpreted domain for southern Ontario bedload sediment transport, with the grey scale showing a relative degree of expected bedload transport. The critical stream power boundary between gravel and cobble bedload transport (100 Wm‐2) is also highlighted with a horizontal dashed line.

While many channels in southern Ontario do meet this condition, some reaches are

oversteepened or understeepened based on inherited profile features. Oversteepened reaches

with abundant sources of fine gravel and sand may exhibit higher rates of bedload transport

and/or may be restrained by other forms of boundary resistance (e.g., vegetation, high silt‐clay

content, resistant glacial till). Understeepened reaches, particularly where they coincide with inherited coarse‐grained gravel and cobble valley fills, are expected to exhibit relatively lower rates of bedload transport.

2.7 Conclusions

The interplay between stream energy and channel boundary resistance in fluvial drainage networks of southern Ontario is remarkably complicated by the inherited slope signatures of glacially conditioned topography and sediments. The concept of glacial conditioning (and the degree to which channels are inherited landforms) is essentially the idea that the post‐glacial

timescales for fluvial adjustment may be spatially variable. Adaptation of theoretical

frameworks from mountainous environments to low‐relief glacially conditioned landscapes

such as the southern Laurentian Great Lakes region is critical to extend meaningful

interpretations of fluvial systems at catchment, regional, and supraregional scales.

Study of the spatial properties of stream power over large areas can be facilitated by geomatic

computer mapping applications, and particularly DEMs. However, calculations of specific

stream power must judiciously consider modeling and generalization approaches to inputs of channel discharge, width, and slope that fundamentally depend on the resolution of the source data and the scale of inquiry. Some previous stream power studies using drainage area– discharge models have preferred to not overstate their results by defining the approach as a stream power index (e.g., Flores et al., 2006). Local controls of channel width that are not predicted by drainage area, such as slope and bank strength, do limit local accuracies of specific

stream power; but meaningful patterns are clearly revealed at catchment scales. With similar

intent to van den Berg’s (1995) potential stream power based on floodplain slope (and

reference width), generalization of DEM slopes to reach scales of 1–2 km is considered here to

be representative of fluvial processes operating over Holocene timescales. While the relative

spatial patterns in stream power are enough to interpret glacial signatures within the

landscape, we would also argue that the approach presented in this study does considerably constrain the absolute values of specific stream power that are capable of providing insights into processes of sediment transport and channel pattern formation within the landscape.

The theoretical expectation of a concave‐up graded river profile has been previously

demonstrated as a useful tool for investigating underlying environmental controls on fluvial

profile evolution in mountainous landscapes, and this study has extended the concept to low‐

relief glacially conditioned catchments. Observed slope anomalies relative to the benchmark of

an exponential profile are interpreted as interruptions to a graded condition. Hack’s (1957)

original SL gradient index can be normalized by river size, providing a useful measure of slope

variations — the SL/K index (Pérez‐Peña et al., 2009). From this approach, inherited glacial

signatures appear as deviations in the SL/K index, where river reaches in southern Ontario

frequently appear to be oversteepened or understeepened by glacial landforms and sediments

(c.f. similar interpretations of McCleary et al., 2011, for a mountainous environment).

Slope–area analysis has previously been used to provide insights into fluvial process domains in mountainous catchments (Montgomery and Foufoula‐Georgiou, 1993; Montgomery, 1999;

Brardonini and Hassan, 2006). The slope–area approach is essentially analogous to the Leopold

and Wolman (1957) discrimination of channel patterns based on a critical threshold slope for braiding plotted against discharge, where instead drainage area is used as a surrogate for discharge. Whether defined by fluvial process domains or channel patterns, the morphological products of distinguishable sets of fluvial processes are fundamentally controlled by the

balance between stream power and boundary resistance. Within slope–area space, empirical

data for southern Ontario were used to plot curves of constant specific stream power. The ‐0.4

exponent of the stream power curves matches Eq. (5), the pioneering threshold equation for discrimination of meandering and braiding planforms, if drainage area is substituted for

0.91 discharge (where for southern Ontario Q ~ Ad ). Classification of river reaches based on

glacial landform types shows convincing patterns of higher and lower stream power caused by

glacial conditioning.

Prominent glacial signatures in the spatial patterns of stream power have been demonstrated

in this study by stream power mapping, SL/K analysis, and slope–area analysis. The expected

consequences of glacial landform conditioning on stream power has been explored in terms of

prediction of channel patterns. Channel patterns in southern Ontario are dominated by single‐

channel meandering planforms, and glacially conditioned stream powers for high energy

reaches do not tend to produce multiple‐channel planforms. Field observations support the

idea that bedload transport can be constrained by limited stream competence associated with glacially inherited, coarse‐grained boundary materials in some reaches; and an upward shift of the braiding threshold is expected for larger grain sizes (Ferguson, 1987; Kleinhans and van den

Berg, 2011). Even so, for gravel‐ and sand‐bed channels in southern Ontario, the threshold

slope may still be higher because of greater relative bank strength (Eaton et al., 2010) from

riparian vegetation and from cohesive floodplain materials that promote lateral stability and

limit channel widening. Actually, very few examples of multiple‐channel planforms exist in

southern Ontario, and most anabranching channels are localized and relatively stable. The

lower Maitland River (43°45'17" N., 81°42'20" W.) is perhaps the only genuine location of

braiding in southern Ontario.

Low‐relief glacially conditioned catchments of southern Ontario lack adequate models to

explain the spatial arrangement of river morphologies within the context of complex glacial

legacy effects. A promising approach might be Montgomery’s (1999) concept of lithotopo units,

whereby glacial conditioning of stream energy and channel boundary materials is expected to

produce distinctive fluvial process domains and, thus, river morphologies. Understanding the

spatial arrangement and linkages of distinct morphological groups within a larger landscape

mosaic will lead to a more generalized conceptual framework for interpreting fluvial systems in

low‐relief settings. Slope–area analysis provides a first‐order approach to discriminate between

different suites of dominant geomorphic processes, which is essentially a means of stratifying reach‐scale morphology based on stream power (cf. Flores et al., 2006). However, traditional

channel pattern classifications may be too general to adequately represent reach‐scale variations in channel morphology for southern Ontario. More meaningful groupings will require consideration of the sedimentology and basic stratigraphic architecture of the alluvial floodplains, with possible foundations in the stream power classifications of Nanson and Croke

(1992). The glacial legacy of the southern Laurentian Great Lakes region appears to still be inscribed in the alluvial record of late‐Holocene floodplains.

Chapter 3 Alluvial floodplain classification by multivariate clustering and discriminant analysis for low-relief glacially conditioned river catchments

3 Chapter 3

Abstract: Alluvial floodplains for rivers in low‐relief glacially conditioned catchments of southern Ontario (Canada) are classified and predicted using a sequence of multivariate statistical analyses. An original dataset of 109 floodplain sites is investigated using k‐means clustering, principal component analysis, and discriminant analysis statistical approaches. Four primary floodplain types are proposed representing basic morphological, stratigraphical, and sedimentological characteristics. Classifications are successfully predicted by two principal explanatory dimensions: (1) stream power‐resistance; and (2) floodplain sedimentology. The latter is most effectively represented by the availability of alluvial sand, and specifically a new variable defined as floodplain sand equivalent (FSE). Floodplain types are generally consistent with previous river classifications, however the glacial legacy of cobble bed materials and sand availability does contribute to substantive discrepancies. Representing the residual variability of power‐resistance correlations, a third explanatory dimension of sediment transport is suggested, and may explain some within‐class variability in channel morphology. Balancing the opposing concepts of fluvial process domains and landform continuums, the potential for transitional floodplain types is also explored. The proposed first‐order alluvial floodplain classifications provide a basis from which to further investigate geomorphological diversity within the context of complex glacial legacy effects in low‐relief settings. Revealing the spatial arrangement and linkages of distinct morphological groups within a regional landscape mosaic is expected to provide insights into patterns of post‐glacial fluvial adjustment.

3.1 Introduction

Spatial variations of river and stream landforms are recognized as the product of diverse

landscape controls and local geomorphic processes. Even so, classifications of river

morphologies and floodplain types have been demonstrated as a useful approach to

understanding environmental complexity within fluvial systems relevant to a variety of disciplines in natural and applied science. Over recent decades, rivers and floodplains have

become increasingly important landscape features, particularly within interdisciplinary research

given their role as ecological corridors (Pool, 2002; Nanson and Croke, 2002; Brierley and Fryirs,

2005; Piégay et al., 2005; Dollar et al., 2007; Lóczy et al., 2012). Consequently, river

classifications have the potential to enhance applied science in areas such as stream

restoration, ecological conservation, water resources engineering, and watershed

management; as well as to enhance theoretical linkages between fluvial geomorphology and ecological, biogeochemical, geoarcheological, and paleoenvironmental research. The purpose of this paper is to use multivariate statistical methods to identify and predict natural groupings

of river floodplain types in southern Ontario (Canada), a low‐relief landscape where glacial

legacy effects are pronounced.

3.1.1 Interdisciplinary floodplains

The pioneering river planform morphologies of Leopold and Wolman (1957) have not only given

way to more sophisticated analyses of geomorphic thresholds for river patterns (e.g., van den

Berg, 1995; Eaton et al., 2010; Kleinhans and van den Berg, 2011), but also numerous river classification approaches have been proposed (e.g., Kellerhals et al., 1976; Schumm, 1985;

Church, 1992; Nanson and Croke, 1992; Rosgen, 1994; Montgomery and Bufflington, 1997;

Brierley and Fryirs, 2005; Vaughan et al., 2013). Starting from the meandering‐braiding 71

paradigm, channel planform continuums as presented by previous authors are typically based

on factors such as number of channels, sinuosity, lateral stability, sediment load, and sediment

grain‐size.

The plethora of river classifications includes a mixture of concepts about river planform

morphologies and river floodplain landforms (Lóczy et al., 2012). While the motivations for

river classification may have initially focused on modern channels for engineering purposes

(Hickin, 1993), the development of fluvial facies models by sedimentologists for geological interpretations of floodplain deposits also has a long history (e.g., see synthesis of Miall, 1985).

This has compelled some researchers to propose frameworks for linking river planform

morphology to alluvial floodplain types at the reach‐scale; some notable examples being

Nanson and Croke (1992), Brierley and Fryirs (2005), and Church (2006).

Given its interdisciplinary nature, the term floodplain itself can have different meanings within

a variety of contexts. This was noted by Nanson and Croke (1992) who pointed out the

difference between the hydraulic floodplains of hydrologists and engineers and the alluvial

floodplains of geomorphologists and sedimentary geologists (the latter being more synomous

with their geomorphological definition of genetic floodplains). These physical floodplain

definitions may also be integrated into other disciplines such as fluvial landscape ecology where

floodplains represent important corridors of riparian and aquatic habitat (Pool, 2002; Ward et

al., 2002; Dollar et al., 2007; Lóczy et al., 2012), including important theories such as the river

continuum concept of Vannote et al. (1980) and the hyporheic corridor concept of Stanford and

Ward (1993). From this perspective, geomorphological definitions of floodplains are often

incorporated as the backdrop for ecological systems, highlighting the importance of how physical and biological processes interact.

3.1.2 Genetic floodplains

In translating planform morphology to alluvial floodplain deposits, Nanson and Croke (1992) frame their genetic classification of floodplains, primarily based on energy in terms of specific

stream power and secondarily based on sediment texture. We suggest that their emphasis on

stream power relates back to their definition of genetic floodplains which are “built of sediment

transported” by the river (p. 460). Following the perspective of Church (2006), understanding

alluvial channel morphology requires substantive connections with fluvial sediment transport

processes. Thus, an important measure of river classification schemes lies in the strength of process‐landform linkages; however, processed‐based river classifications have tended to be regionally specific (Simon et al., 2007).

The primary environmental controls on river morphology from numerous sources in the literature (e.g., Church 1992) can be summarized as: (1) the topographic energy gradient or

slope of the channel, (2) the discharge of water and sediment to the channel, and (3) the grain‐ size texture and calibre of transported boundary materials. Specific stream power (ω) is essentially defined by the product of slope (S) and water discharge (Q) as an average per unit

width (w) of channel (ω ~ QSw‐1 in units of Watts per square metre, Wm‐2). Third in our list of

environmental controls, channel boundary materials can also be an important characteristic to

classify alluvial floodplains and may be further organized in terms of depositional processes

(e.g., Miall, 1985). Nanson and Croke (1992) present six main depositional processes for

floodplain accretion, with the most important two in single‐channel rivers being lateral point‐ bar accretion and overbank vertical accretion.

The interconnections between sediment transport, channel morphology, and alluvial floodplain sedimentation are fundamental in fluvial geomorphology. Even so, such essential fluvial process‐landform associations are not so easily reconciled, perhaps in part because of the

practical limitations of defining sediment transport mechanisms. Church (2006) provides an

insightful discussion of this problem, specifically commenting on the role of sand sized

materials that tend to span the definitions of bed load versus suspended load. The relatively

transient nature of sand load transport, whereby it is easily transported and frequently

deposited, has potentially significant implications for fluvial processes, alluvial channel

morphology, and thus floodplain sedimentation. As such, the relative role of sand may in some

cases be essential to understanding how available grain sizes contribute to building alluvial

floodplains.

In this context, glacially conditioned catchments in southern Ontario exhibit significant

variations in the availability and alluvial storage of sand materials which are largely imposed by

the spatial organization of glacial landforms and the associated post‐glacial landscape history

(Phillips and Desloges, 2014). Thus, floodplain sedimentology, particularly in terms of the sand

fraction, is likely important to understanding the variability of river channel morphology and associated fluvial processes.

3.1.3 Hyperdimensional floodplains

While environmental diversity may complicate fluvial landform development in many environments, the variation and architecture of the glacial palimpsest in the southern Great

Lakes region is particularly vast and disorganized with respect to fluvial discontinuities. In this paper, our intention is to develop a conceptual and a statistical framework from which to

organize river landforms in the context of diverse glacial legacy effects, building from a number

of key contributions by previous researchers. Foremost, we include specific stream power in our analysis as per the river classifications of Nanson and Croke (1992), van den Berg (1995), and Vaughan et al. (2013); and given its physical relevance to sediment transport processes

(Parker et al. 2011; Ferguson, 2012; Phillips and Desloges, 2014).

This study also builds on the concept of fluvial process domains as per Montgomery (1999) and

Brardinoni and Hassan (2007) in extending the idea from mountainous catchments to streams

and rivers in low‐relief landscapes. Montgomery (1999) introduces the notion of lithotopo

units as a most basic way of organizing stream types based on channel slope and boundary

materials. A fundamental theme in Montgomery’s (1999) discussion is the competing

abstractions of fluvial systems as physical and biological continuums versus stream reaches as

discrete units with distinct fluvial process domains. Brardinoni and Hassan (2007) not only

extend these ideas to glaciated mountainous catchments, they also approach the problem of stream classification using multivariate statistical approaches.

The river continuum concept is well suited for investigations of natural domains in hyperdimensional space based on multivariate statistical approaches, and the concept has an

established history in both geomorphology (Leopold and Wolman, 1957; Knighton, 1998) and

fluvial ecology (Vannote et al., 1980). This abstract link with multivariate statistical methods

has been embraced by ecologists (e.g., Poff and Ward, 1989; Jeffers, 1998; Guisan and

Zimmermann, 2000), but perhaps less so by fluvial geomorphologists (cf. Brardinoni and

Hassan, 2007; Splinter, 2013; Vaughan et al., 2013). Expressing Montgomery’s (1999) assertion

in statistical terms, fluvial landform continuums are well represented by physical clusters in

multidimensional space based on the theory of reach‐scale fluvial process domains, where

members are more alike within each cluster as compared to the differences between clusters.

The concept of process domains is potentially an important tool for interpreting the spatial

arrangement and linkages of distinct morphological groups within a regional landscape mosaic,

but theories of river landform continuums also emphasize that spatial transitions between

group members (and between multivariate domains) may be more or less distinct (with

interesting applications for hard versus fuzzy statistical clustering methods).

In our analysis of fluvial systems in southern Ontario, the intention is to test a statistical k‐ means clustering (KMC) approach to classify alluvial floodplains in low‐relief glacially conditioned river catchments. We subsequently explore how multiple variables in our floodplain dataset are correlated according to Principal Component Analysis (PCA) in order to

extract and interpret the most important explanatory dimensions of floodplain variability. We

are then able to test our classifications in a multivariate Discriminant Analysis (DA), which

includes fluvial process and floodplain variables, to predict floodplain classes and assess group

membership probabilities.

3.2 Study Area

“Southern Ontario is an area of modest relief” (Chapman and Putnam, 1951, p. vii). An area in

the range of 105 km2, the maximum topographic relief is less than 500 m and the average

catchment gradients are in the range of 0.5 to 5 m per kilometre. The study area of southern

Ontario, Canada (Figure 3.1) is situated within the Laurentian Great Lakes watershed of North

America. The surface geology consists of a diverse range of glacial sediments and landforms, most deposited by the melting Laurentide Ice Sheet 10 to 15 thousand years before present

(Chapman and Putnam, 1951; Larson and Schaetzl, 2001). The two most prominent physiographic features within the study area are a ridge of resistant Paleozoic limestone bedrock known as the Niagara escarpment and a stratified kame moraine complex known as

the Oak Ridges moraine (Barnett et al., 1998).

Figure 3.1: Study area of southern Ontario (left) with glacial landforms of Chapman and Putnam (2007) and (right) with location of 109 floodplain sites from this study. Small inset map in lower right corner shows approximate locations of the Niagara Escarpment (NE) and the Oak Ridges Moraine (ORM). (AA) Algonquin Arch of the Precambrian basement bedrock.

The vast assemblages of glacial landforms have topographically shaped fluvial drainage networks; particularly the succession of deglacial kame moraines, recessional till moraines, and associated glaciofluvial outwash valleys (Figure 3.1). Many areas which were not reworked by

the retreating ice and meltwater remain as extensive plains of glacial till. Large areas of

proglacial lake deposition are also significant features in the landscape (e.g., sand and clay

plains, Figure 3.1). Although the Paleozoic shale and limestone bedrock of the Michigan Basin

is locally exposed, the glacial overburden is generally thick in the range of 10’s of metres (and

greater than 200 m in some areas, Gao et al., 2006).

Streams and rivers of southern Ontario are thus carved into a complex architecture of glacial landforms that have imposed topographic and sedimentological signatures on the river profiles

and alluvial sediments (Phillips and Desloges, 2014). While many river reaches may be

considered self‐adjusted in terms of slope and sediment transport (i.e., in the graded sense of

Mackin, 1948), some reaches may instead be considered oversteepened or understeepened in

the sense that fluvial sediment transport is still in disequilibrium with the dominant boundary

materials (i.e., higher or lower rates than would expected in equilibrium). In addition to

sediment transport, the range of available sediment sizes supplied from tributaries and from within the valley fills will have a significant effect on the potential depositional processes which are responsible for constructing the alluvial floodplains.

For example, cobble and boulders derived from moraine and outwash deposits are more likely

to exhibit lower sediment transport rates and form persistent lag layers within floodplain

deposits. Conversely, sediments supplied from thick glaciolacustrine sand deposits will undergo

frequent sediment transport and deposition potentially contributing to higher rates of

floodplain accretion and turnover. Vast areas of fine‐grained glacial till and glaciolacustrine clay are often associated with fine‐grained clay floodplains dominated by wash load and slack water

sedimentation. Ultimately where a mixture of gravel and sand are available from glacial sources and from fluvial sediment transport, river channels exhibit classic river meandering

processes, including the construction of alluvial floodplains by lateral point‐bar accretion and

vertical over‐bank accretion.

This study is limited to rivers with single‐channel planform morphologies, as very few multiple‐

channel patterns have been observed within the study area. Examples of anabranching

channels tend to be localized and relatively stable (e.g., Croil, 2002). Only one braided river in

southern Ontario has been observed by the authors (lower Maitland River: 43°45'17" N,

81°42'20" W), but this reach has not been included in the present study. We also do not explicitly address the issue of semi‐alluvial channels (Ashmore and Church, 2001), defined as

channels which interact with glaciogenic deposits and/or bedrock along significant proportions

of their boundaries. This issue may have interesting implications for fluvial processes and

floodplain accretion, but from the authors’ experience most channels in the study area are

either fully alluvial or are only in contact with non‐alluvial material to a limited extent (in space

and perhaps also in time!). Still, southern Ontario exhibits a diverse range of single‐channel streams and rivers. In the context of a low‐relief glacially inherited landscape, the considerable topographic and sedimentological discontinuity in the fluvial systems provides a relatively

untested environment to investigate linkages between fluvial processes, channel morphology, and alluvial floodplain landforms.

3.3 Data Collection and Methods

3.3.1 Floodplain dataset

River channel and floodplain data were collected and compiled for 109 representative reaches within the study area of peninsular southern Ontario (Figure 3.1, Appendix A), of which 18 were randomly selected (~16%). Drainage areas for sampled reaches range from about 3 to

6500 km2. The floodplain dataset was derived from a combination of field, laboratory, and geomatic sources. In total, an initial list of 20 variables was assembled into two categories: (1) fluvial process variables and (2) floodplain variables (Table 3.1).

Table 3.1: Statistical transformation and normality tests of 20 reach variables.

Reach Variables Transformationa Normality Fluvial Process Variables (and Environmental Controls) ‐1 d Slope, S (m m ) Log S statistical pass 3 ‐1 d Discharge, 2‐Year Flood, Q2 (m s ) Log Q2 statistical pass Total Stream Power, Ω (Wm‐1) Log Ω statistical passd ‐2 d Specific Stream Power, ω (Wm ) √ω statistical pass ‐1 d Shields Stress, τ/D95 (Nm ) Log τ/D95 statistical pass ‐3 d Specific Stream Power‐to‐D50 Ratio, ω/D50 (Wm ) Log ω/D50 statistical pass b ‐2 e Critical Specific Stream Power for D50, ωcr (Wm ) Log ωcr histogram e Specific Stream Power Ratio, ω/ωcr (dimensionless) Log ω/ωcr histogram

Floodplain Variables (and Channel Metrics) Channel Bankfull Width, w (m) Log w statistical passd Channel Bankfull Depth, d (m) Log d statistical passd Channel Width‐to‐Depth Ratio, w/d (dimensionless) Log w/d statistical passd d Total Floodplain Thickness, Ft (m) Log Ft statistical pass e Fine Alluvial Floodplain Thickness, Ffa (m) Ffa histogram c e Floodplain Sand Equivalent thickness in Ffa, FSE (m) √FSE histogram d Percent Organic Matter in Fine Alluvial Floodplain, OM (%) Log OM statistical pass d Percent Silt‐Clay Fraction in Fine Alluvial Floodplain, MSC (%) MSC statistical pass histogramc Percent Sand Fraction in Fine Alluvial Floodplain, MSa (%) MSa c Bed Material Size, D50 (mm) → Ф50 ‐Log2 D50 histogram c Bed Material Size, D95 (mm) → Ф95 ‐Log2 D95 histogram ‐1 e Relative Roughness, D95/d (m m ) Log D95/d histogram a Logarithmic transformations (base 10) unless otherwise indicated. b Critical specific stream power calculated from equation 17 by Parker et al. (2011). c See discussion in text for definition of Floodplain Sand Equivalent (FSE) thickness. d Non‐normality of the sample distribution is not statistically significant (α = 0.05) according to Shapiro‐Wilk, Anderson‐Darling, Lilliefores, and Jarque‐Bera tests. e Non‐normality of sample distribution is statistically significant for most statistical tests, however quasi‐normality of the population is reasonably assumed based on graphical evaluation of the sample histogram and/or a statistical pass for one test.

For each of the 109 floodplain sites, average reach‐scale estimates of channel slope (S), discharge (Q), and stream power (Ω, ω) were extracted from geomatic modeling using digital elevation models (DEM) and empirical rating curves based on drainage areas for southern

Ontario rivers as described by Phillips and Desloges (2014). Field‐based data collection at each floodplain site included (1) channel surveys of bankfull width (w) and depth (d); (2) observations and sampling of bed material grain‐size (D50, D95); and (3) floodplain sediment

sampling and stratigraphic logs from cut‐banks, hand‐augers, and percussion cores (2.6 cm

diameter).

Basic observations and measurements of floodplain stratigraphic architecture were compiled

for each site from multiple banks and/or boreholes at each site (e.g., fine alluvial thickness <2

mm, Ffa). Total alluvial floodplain thickness (Ft) was normally defined by the depth from the interpreted bankfull elevation to the bottom of the deepest scour pool (as a minimum total thickness). However, in some instances, total alluvial thickness was measured to be less due to exposure of non‐alluvial boundary material and/or due to difficulty in interpreting the bankfull

elevation (e.g., entrenched semi‐alluvial reaches).

For each of the field sites, bulk floodplain sediments were sub‐sampled in the laboratory to

about 100 g; were ignited in a muffle furnace at 550 C for 4 hours to calculate percent organic

matter (OM, Heiri et al., 2001); and baked aggregates were then subsequently ground‐up with a mortar and pestle and wet‐sieved at 63 μm to calculate the percent sand (Msa) versus percent

silt‐clay (MSC) fractions. Prior to sieving, samples were soaked in a deflocculant solution of 0.5%

sodium hexametaphosphate and were blended for 5–10 minutes with an electric mixer. After

sieving, each remaining sample of >63 μm grains was visually inspected under a dissecting

stereo microscope to assess sand purity. Sand separates including >5% silt‐clay aggregates

were reprocessed using the same wet sieving methods.

In most cases only one or a few alluvial floodplain samples were used to represent the

sedimentological character at each field site (MSa, MSC, OM). Given that significant vertical and

lateral variations in alluvial sedimentology were expected within each study reach (Hickin,

1993), a standardized sampling method was implemented where all floodplain sediment samples were collected from immediately above or near the bar‐top surface (or conceptual

equivalent). In other words, variations of representative floodplain texture for this study are

based on the character of the sediments at the base of the fine alluvial deposits for sites where coarse‐lags were present (e.g., gravel and/or cobble), or in close proximity to the low‐flow water surface where sands and finer sediments dominate the channel bed material.

For laterally‐active mixed sand and gravel channels, this vertical zone is intended to be

equivalent to the basal initiation of sandy bar deposits associated with lateral accretion (Miall,

1985). For cobble‐dominated floodplains within the study area, this depositional layer of fine‐ grained material overlying the coarse lag is typically thin (only 10–30 cm thick). For fine‐grained floodplains (e.g., sand or clay), the low‐water surface represents the height above which sediments are most likely to be deposited during the waning stages of a flood, resulting in

floodplain accretion on bars or low‐lying areas of floodplain (note, for clay‐dominated

floodplains in the study area most floodplain accretion may be highly dependent on backwater

deposition during debris‐jams of ice and/or large woody debris, but the long‐term significance

of such accretion processes requires further study).

Thus in terms of floodplain construction, this bar‐top material is deposited at a higher elevation than the bedload in the active channel, but is lower than the vertical accretion of the over‐bank floodplain. For a more general terminology which encompasses all floodplain types within this study, we have designated this vertical portion of the floodplain stratigraphy as the incipient accretionary zone (IAZ) as a representative and consistent sampling position.

It was expected that the variability in floodplain texture would depend to some degree on the relative role of transient sand material within the channels, and in particular its availability

within the river channels. This idea will be discussed further using what we have defined as the

floodplain sand equivalent (FSE), or an estimate of the equivalent vertical depth of sand

contained within the fine alluvial floodplain deposits. For our dataset, an estimate of the FSE

has been calculated based on the product of the average fine alluvial thickness at a representative section (Ffa) and the percentage of sand (MSa) from the IAZ. For example, a

sampled floodplain section with a 1 m Ffa and 50% sand in the IAZ would result in an FSE of 0.5

m. Given that many floodplain sequences fine upwards, it should be acknowledge that our FSE calculations may tend to be over‐estimates, and future detailed studies should be based on reach‐scale averages from more numerous stratigraphic profiles and from multiple weighted sample depths. The approach for this current study was to make first‐order estimates of FSE for a large sample size across the region of southern Ontario (i.e., an FSE index).

3.3.2 Multivariate analysis

The purpose of the study is to use multiple variables to classify, explain, and predict natural

groupings of floodplain types within the study area. Specifically, we have applied a k‐means clustering (KMC) method to statistically group floodplain observations and we have identified

an efficient set of variables to explain and predict floodplain variability using Principal

Component Analysis (PCA) and Discriminant Analysis (DA) statistical approaches.

A common statistical method in exploratory data analysis, k‐means clustering statistically partitions observations into relatively homogeneous groups based on multiple (and potentially

many) variables (Jain et al., 1999). To classify alluvial floodplains, our approach focuses on the

manifest geomorphological and sedimentological landforms themselves, rather than starting

with any explicit consideration of the fluvial processes which produce them. Initially, a set of

four observed floodplain types was identified in the field as preliminary classifications (4

primary types with potentially at least 4 additional transitional types). In statistical terms, k‐

means clustering is considered an unsupervised classification technique (Jain et al., 1999).

Although, preliminary field classifications were used to guide decisions on the number of

meaningful groupings and on the data abstraction to describe group characteristics, the KMC analysis was treated as a relatively objective tool for confirming and refining the floodplain

classifications from the field.

Principal Component Analysis (PCA) was used as an exploration technique to identify the most

important factors controlling the variability of floodplain landforms and to reduce the initial number of 20 variables (Abdi and Williams, 2010). PCA effectively summarizes the variability contained in many (often correlated) dimensions into a smaller number of uncorrelated dimensions (i.e., principal components). As such, PCA is a useful tool to explore variable reduction, providing a basis to discard redundant variables informed by theoretical or statistical criteria (King and Jackson, 1999).

To conduct a discriminant analysis (DA), the floodplain landforms grouped by KMC were treated

as a priori floodplain types. As opposed to KMC analyses, DA is considered to be a supervised

classification technique (McLachlan, 2012). More specifically, DA provides a statistical test to predict a known categorical dependent variable, in this case the KMC floodplain types, by

multiple continuous explanatory variables (McLachlan, 2012). Initially, the DA variable data were selected from the PCA reduced variable set. However, the DA was also preformed iteratively on fewer and fewer variables to assess the relative predictive success of more parsimonious combinations of variables.

It is important to clarify that the KMC floodplain types were based on only floodplain variables,

while the DA tests were based on both floodplain variables and fluvial process variables from

Table 3.1. In other words, a sub‐set of some floodplain variables used in the KMC classifications was also used in the DA tests. While a DA test of floodplain landform classifications based

strictly on fluvial process variables would be a superlative feat, clustering and prediction of

landform groups invariably requires inclusion of landform metrics.

3.4 Results

3.4.1 Field floodplain classifications

Four major floodplain types were interpreted from field observations of channel morphology and floodplain characteristics (Figure 3.2 and Figure 3.3), with a most basic sedimentological classification initially envisioned as cobble‐dominated, gravel‐dominated, sand‐dominated, or

clay‐dominated floodplain types.

Figure 3.2: (left) Schematic cross‐sections of alluvial floodplain types based on field interpretations. Floodplain types are colour coded according to B‐type (red circles); M‐type (orange circles); S‐type (yellow circles); and C‐type (grey circles). Basic alluvial stratigraphic architecture of cross‐sections labeled as: overbank fines (Obf); silt‐clay (SC); sand (Sa); gravel (Gr); and cobble (Cb). Approximate vertical location of incipient accretionary zone (IAZ) is also indicated for each floodplain type. (right) Mapping of floodplain classifications for 109 field sites including field interpretations (inner circles) and KMC grouping results (outer circles). Therefore, misclassifications between field interpretations and KMC results are shown where inner and outer circles are different colours. Black areas on map are glacial moraines of Chapman and Putnam (2007). (For interpretation of colour references for the figure symbols, the reader is referred to the web version of this article).

Cobble‐dominated floodplains were identified not only by coarse bed‐materials in the channel

and prevalent cobble‐lags within the alluvial floodplain architecture, they also exhibited a

distinctively thin layer of fine‐grained alluvial deposits above the coarse‐lag, with overbank

materials containing low percentages of sand. As such, we generally classified cobble‐bed

channels as B‐type floodplains due to the prevalence of the coarse basal‐lags and the bimodal

alluvial grain‐size distributions.

Figure 3.3: Example photographs of channels representing four interpreted floodplain types: (upper left) S‐type floodplain at Big Otter Creek; (upper right) M‐type floodplain at Catfish Creek; (lower left) C‐type floodplain at Deer Creek tributary of Saugeen River; and (lower right) B‐type floodplain at Bayfield River. Inset graphs are schematic interpretations of relative alluvial grain‐size frequency distributions for each floodplain type. Also see Appendix A.

By comparison, gravel‐dominated floodplains were comprised of mixed alluvial grain‐size distributions and were thus designated as M‐type floodplains. Fine‐grained floodplains could often be distinguished in the field as sand‐dominated (S‐type) or clay‐dominated (C‐type)

alluvial floodplains. The thickness of the alluvial floodplains tended to be greatest for the S‐ type and C‐type floodplains, with the M‐type floodplains being of intermediate alluvial thicknesses. Given that alluvial thickness was highly correlated with the bankfull channel depth

(although not always equivalent), the fine‐grained floodplains tended to have the lowest

channel width‐to‐depth (w/d) ratios (i.e., deep and narrow channels), while M‐type and B‐type floodplain channels exhibited increasing w/d ratios respectively (i.e., wider and shallower channels). Schematic cross‐sections of each observed alluvial floodplain type are provided in

Figure 2, including basic stratigraphic archetypes and relative channel dimensions.

3.4.2 K-means clustering analysis

Statistical groupings of floodplain types using k‐means clustering provided a means of testing the preliminary field classifications and objectively assigning floodplain types for further

analysis. For consistency with field interpretations of floodplain landform characteristics, the

KMC analysis was completed using 9 floodplain variables including channel morphology (w/d), alluvial floodplain thickness (Ft, Ffa, FSE); floodplain composition (OM, MSC, MSa); and bed material grain‐size and relative roughness (Ф50, D95/d).

Results of the KMC analysis based on the 9‐variable floodplain dataset are presented in Table

3.2 and Figure 3.2 for the 109 floodplain sites in southern Ontario. Relative to the preliminary

field classifications, the KMC analysis was in agreement at the following percentages for each floodplain type: B‐type (81%); M‐type (78%); S‐type (100%); and C‐Type (81%). However, on average the misclassification was only about 10% if the confounding effects of tributary entrenchment and semi‐alluvial channel boundaries observed in the field were assumed to

explain some of the errors.

From Table 3.2, the centre or mean of each cluster in multivariate space is identified by the class centroid, with the average distances of the class observations from the centroids being less than 1 and the distances between the class centroids being greater than 1. Relative distances between class centroids were also informative, with B‐type floodplains furthest away 88

from S‐type and C‐type floodplains (~4.3). The centroids of the two fine‐grain floodplain types

were relatively close together (1.23) and the M‐type floodplain centroid was intermediate in

terms of distance from the other three types (≥2).

Table 3.2: KMC analysisa results based on 9 selected floodplain variablesb from Table 3.1.

Min Avg Max Distance between class Within class Class Objects distance distance distance centroids variance to centroid to centroid to centroid B M S C B‐type 36 0.93 0.31 0.86 1.88 0 1.98 4.29 4.33 M‐type 37 0.72 0.38 0.78 1.70 1.98 0 2.35 2.56 S‐type 20 0.73 0.44 0.80 1.12 4.29 2.35 0 1.23 C‐type 16 0.89 0.42 0.86 1.57 4.33 2.56 1.23 0 a K‐means clustering based on traditional minimization for the trace of W from the pooled sum of squares and cross products (SSCP) matrix. b 9 floodplain variables include: w/d, Ft, Ffa, FSE, OM, MSC, MSa, Ф50, D95/d.

3.4.3 Principal component analysis

The initial dataset of 20 variables in two categories is listed in Table 3.1. As in many natural systems, several variables may be highly correlated due to representation of a related conceptual phenomenon and/or due to algebraic derivations of factors from physical principles which share common root variables (i.e., spurious correlations). While the latter is undesirable

and should be minimized, the former is a strong basis for PCA applications which seek to

summarize (or in a sense cluster) variables based on uncorrelated dimensions which are the

principal components.

The initial PCA test was conducted on the complete 20‐variable dataset, acknowledging the

existence of several spurious correlations between derived variables. The correlation circle for

the 20‐variable dataset in Figure 3.4A shows the projection of the variables for the first two

principal components (F1 and F2).

Figure 3.4: PCA correlation circles with variable projections for (A) 20‐variable dataset from Table 1; and (B) 12‐variable dataset. Scree‐plots below each corresponding correlation circle to test PCA stopping rules for retaining and interpreting principal components, including latent root and broken stick criteria, are discussed in the text. Grey curves in scree‐plots represent a theoretical broken stick model distribution.

Variables which project in the same direction are highly correlated, and may be more or less

correlated with one or both of the first two principal components (positively or negatively).

Variables which project closer to the outer circle (i.e., closer to a correlation of 1) are more

strongly correlated with one or both of the first two principal components, whereas variables

which are closer to the centre are more correlated with subsequent principal components (e.g.,

F3 or F4, etc).

The first PCA test produced 15 principal components to explain the total variance in the dataset. The latent root and broken stick criteria (Jackson, 1993) were considered as the 90

stopping rules to decide the number of important (or significant) principal components to

retain and interpret. For the latent root criterion (also known as the Kaiser or eigenvalue‐one

criterion) principal components may be retained and interpreted if the eigenvalue is greater

than 1.00 (Kaiser, 1960). For the 20‐variable PCA test, this would suggest retaining the first four

principal components (eigenvalues for F1 to F4 = 9.3, 4.5, 2.7, and 1.3 respectively) that cumulatively explain 88.79% of the data variance. The eigenvalue of 1.0 has been indicated by

a dashed line on the scree‐plot at the bottom of Figure 3.4A.

Another common stopping rule is the broken stick criterion (Jackson, 1993) which indicates a conservative number of principal components to retain based on the point in the scree‐plot where it crosses a broken stick model distribution (Figure 3.4A). For the initial PCA test of 20

variables, the broken stick criterion indicates that only the first three principal components

should be retained and interpreted, accounting for 82.46% of the data variance. Referring back

to the variable projections in the correlation circle of Figure 3.4A, the variables that tend to be

positively or negatively correlated with the first principal component are conceptually related to stream power and/or resistance in terms of roughness and sediment calibre (F1 explains

46.64% of the variance). The second principle component tends to be positively correlated with variables related to channel size (F2 explains 22.38%) and the third principal component

(not shown in Figure 3.4A) tends to be most associated with variables of floodplain

sedimentology (i.e., sediment texture and organic matter, F3 explains 13.44%). The variable

projections for F1 and F2 in Figure 3.4A from our dataset are very similar to those presented by

Vaughan et al. (2013) for a very large dataset in England and Wales (UK).

For the purpose of explaining and predicting fluvial process‐landform associations of southern

Ontario floodplains, we reasoned that channel size was not a desirable dimension to retain in

our analysis. While channel size was clearly important for explaining a significant proportion of

the data variance, our preferred objective was to account for common fluvial processes and

floodplain landforms which occur regardless of channel size. As such, to proceed with a second

PCA test all variables highly correlated with channel size were dropped from further analysis (4

variables removed including Q, w, d, and Ω). Although this variable reduction did remove some

spurious correlations in the PCA dataset, some additional root variables and redundancies were

also removed from further analysis because the variables were considered to have weaker

theoretical value compared to remaining variables (4 more variables removed including S, D50,

D95, ω/ωcr).

A second PCA test was conducted on the remaining 12‐variable dataset (Figure 3.4B) containing

4 fluvial process variables and 8 floodplain variables (from Table 3.1). This PCA test produced

12 principal components (Table 3.3), of which the first three were considered significant to

retain according to the latent root criterion (Figure 3.4B). However, based on the broken stick

criterion only the first two principal components were considered important (F1 and F2

explained 71.39% of the data variance). The 12‐variable PCA test also showed similar variable

projections in the correlation circle (Figure 4B) as compared to the initial PCA test, with the first

component (F1, 48.81%) representing stream power‐resistance (e.g., ω, D95/d, ωcr). The strength of variable correlations with the principal components can also be confirmed in Table

3.4, where the greater the squared cosine of each variable is the stronger the relationship it has with the given principal component.

Table 3.3: PCA eigenvalues of the correlation matrix for 12 variablesa. Variability Cumulative Component Eigenvalue (%) (%) F1 5.86 48.81 48.81 F2 2.71 22.58 71.39 F3 1.44 11.96 83.35 F4 0.58 4.84 88.19 F5 0.48 3.96 92.15 F6 0.36 2.98 95.14 F7 0.32 2.64 97.77 F8 0.14 1.17 98.94 F9 0.07 0.57 99.51 F10 0.04 0.30 99.81 F11 0.01 0.11 99.92 F12 0.01 0.08 100.00 a Bartlett’s sphericity test indicates at least one of the correlations between the variables is significantly different from zero (p < 0.0001). Kaiser‐Meyer‐Oklin measure of sampling adequacy is 0.68.

Table 3.4: PCA squared cosines of the 12 variables for the first four principal componentsa.

Variable F1 F2 F3 F4 √ω 0.569 0.063 0.125 0.148

Log τ/D95 0.439 0.107 0.280 0.001

Log ω/D50 0.284 0.070 0.402 0.157

Log ωcr 0.731 0.151 0.012 0.005 Log w/d 0.502 0.006 0.121 0.117

Log Ft 0.614 0.022 0.183 0.067

Ffa 0.786 0.000 0.129 0.036 √FSE 0.674 0.210 0.032 0.023 Log OM 0.377 0.370 0.003 0.002

MSC 0.037 0.784 0.074 0.024

MSa 0.038 0.836 0.072 0.000

Log D95/d 0.806 0.090 0.001 0.002 a Values in bold correspond for each variable to the factor for which the squared cosine is the largest. For some variables the second largest factor has also been underlined.

With channel size variables removed, the second principal component (F2, 22.58%) in the 12‐ variable PCA test was most correlated with variables of floodplain sedimentology (MSa, MSC,

OM). In Table 3.4, the third principal component (F3, 11.96%) is most correlated with variables of sediment transport (i.e., τ/D95 and ω/D50).

To visualize the meaning of the principal component dimensions in terms of floodplain

landform associations, the four floodplain types grouped in the KMC analysis were plotted on

the first three ordination axes (F1 vs. F2; and F2 vs. F3) as presented in Figure 3.5. Particularly

for the first two principal components (F1 and F2), the four floodplain types show clear separation. In fact, the first component dimension of stream power‐resistance displays a distinct continuum between S‐M‐B floodplain types from left to right. The second component

of floodplain sedimentology appears to be most important for distinguishing between S‐type and C‐type channels. The third component is conceptually appealing by potentially representing variability in sediment transport processes, however the dataset shows poor separation of floodplain types for F3 ordinations in Figure 3.5 and F3 could potentially be discarded based on the broken stick criterion.

The PCA results provided a method of interpreting variable contributions to meaningful data ordinations; for the purposes of variable reduction and for summarizing the variables in terms of a small number of conceptually significant dimensions. In summary, the reduced 12‐variable dataset (Figure 3.5 and Table 3.4) produced two significant dimensions interpreted as (F1)

stream power‐resistance and (F2) floodplain sedimentology; as well as a potential third

dimension (F3) representing sediment transport variability.

Figure 3.5: PCA ordinations for floodplain types based on first three principal components from 12‐variable dataset with floodplain groupings based on KMC results: (left) F1 vs. F2; and (right) F2 vs. F3. Bartlett’s sphericity test indicates at least one of the correlations between the variables is significantly different from zero (p < 0.0001). Kaiser‐Meyer‐Oklin measure of sampling adequacy is 0.68.

3.4.4 Discriminant analysis

A discriminant analysis was used as a supervised statistical technique to test the relative success of predicting a priori floodplain classifications using parsimonious sets of variables to represent explanatory dimensions. As such, the purpose of the DA was to further test floodplain

classification as compared to the results of the KMC and PCA analyses. However, unlike PCA

where components are uncorrelated and thus orthogonal, underlying factors in the DA may be

correlated and thus oblique.

The initial DA test was completed using the 12‐variable PCA reduced dataset. To visualize the

floodplain observations and variable contributions, Figure 3.6 presents the four floodplain

types plotted based on factor scores for the first two DA factors (F1 and F2) that cumulatively

explain 96.58% of the data variance. Variable projections in the correlation circle are similar to

the PCA test with the first factor representing stream power‐resistance and the second factor

representing floodplain sedimentology. Again, well‐defined separation of the S‐M‐B floodplain

types was observed in the first factor, while the second factor distinguishes the floodplain

sedimentology particularly for the S‐type and C‐type floodplains.

The DA computation also provided a membership probability for each observation, where the

group assignment is based on the highest probability. Initially only three observation

misclassifications were computed between KMC groups and DA results, however a subsequent

cross‐validation analysis resulted in 9 classification mismatches. The cross‐validation analysis

recalculated the membership probability of a given observation if it was left out of the analysis

(i.e., leave‐one‐out or jackknifing). Thus the average DA predictive success as presented in the

Table 3.5 confusion matrix for the 12‐variable PCA‐reduced dataset was about 92% compared

to the KMC classes. To explore classification mismatches, membership probabilities of select

examples from the jackknife cross‐validation analysis are presented in Table 3.6. These results

support the idea that fuzzy clustering methods may be appropriate for alluvial floodplain

classifications with individual observations given relative probabilities of being assigned into

one, two, or more groups (i.e., transitional multivariate continuums between floodplain archetypes).

Figure 3.6: (left) First two factors (F1 vs. F2) of DA test for 12‐variables, including KMC floodplain groups, group centroids, and 95% confidence ellipses (dashed grey lines). (right) Correlation circle is plot of variable projections in F1 and F2 factor space.

Table 3.5: KMC‐DA confusion matrix for jackknife cross‐validation of floodplain types based on 12‐ variable dataset.

From/to B M S C Total % Correct B 33 3 0 0 36 91.67 M 2 34 1 0 37 91.89 S 0 0 18 2 20 90.00 C 0 0 1 15 16 93.75 Total 35 37 20 17 109 91.74 Note: Values in bold represent number of KMC‐DA matching classifications for each floodplain type.

Table 3.6: Select examples of floodplain sites from 12‐variable DA jackknife cross‐validation analysis with membership probabilities for KMC‐DA classification mismatches and other transitional floodplain types.

KMC DAa DA Membership Probability River Site Prior Posterior B M S C Ausable River ABR‐5c S C 0.00 0.00 0.08 0.92 Ausable River ABR‐28b S S 0.00 0.23 0.76 0.00 Ausable River ABR‐33 S C 0.00 0.00 0.12 0.88 Ausable River Tributary ABR‐AC‐3 C C 0.00 0.00 0.43 0.57 Bannock Burn River (Bayfield) BFR‐BBR‐4 C C 0.00 0.00 0.50 0.50 Fairchild Creek (Grand) GDR‐FC‐T3‐9 C S 0.00 0.00 0.99 0.01 Speed River (Grand) GDR‐SP‐4 M S 0.00 0.32 0.67 0.01 Nith River (Grand) GDR‐NR‐1b B B 0.92 0.08 0.00 0.00 Nith River (Grand) GDR‐NR‐15 M M 0.05 0.95 0.00 0.00 Nottawasaga River NWR‐9 B M 0.38 0.62 0.00 0.00 Nottawasaga River NWR‐S2 S S 0.00 0.45 0.55 0.00 Snake Creek (Saugeen) SGR‐SKC‐1 B B 0.72 0.28 0.00 0.00 Saugeen River SGR‐9c M B 0.51 0.49 0.00 0.00 Saugeen River SGR‐9d M M 0.33 0.67 0.00 0.00 Saugeen River SGR‐3 B M 0.23 0.77 0.00 0.00 Little Rouge River LRR‐1 C C 0.00 0.17 0.11 0.72 Little Rouge River LRR‐2 M M 0.28 0.72 0.00 0.00 Little Rouge River LRR‐5 B B 0.82 0.18 0.00 0.00 West Duffins Creek WD‐R1 M B 0.89 0.11 0.00 0.00 West Duffins Tributary WD‐WV‐W2 B M 0.47 0.53 0.00 0.00 a Bold types indicate KMC‐DA classification mismatches.

Subsequent discriminant analyses were preformed iteratively on fewer and fewer variables to assess the relative predictive success of more parsimonious combinations of variables (Table

3.7). To start, 6‐variable models were tested using two variables for each of the first three principal components. As presented in Table 3.7, a 96% success rate was found based on ω and

D95/d (stream power‐resistance); FSE and OM (floodplain sedimentology); and τ/D95 and ω/D50

(sediment transport).

Table 3.7: KMC‐DA percent correct classification matches from jackknife cross‐validation results with reduced number of variablesa. Number of Variables 6 3 2

ω, D95/d ω ω Floodplain FSE, OM FSE FSE Type τ/D95, ω/D50 ω/D50 B 97.22 94.44 80.56 M 94.59 97.30 72.97 S 95.00 100 85.00 C 100 93.75 56.25 Average (%) 96.33 96.33 75.23 a For all tests (α = 0.05), Box tests are statistically significant to reject hypothesis that the covariance matrices are equal between groups; and Wilks’ Lambda tests are statistically significant that the means vectors are different.

However, this success rate was equivalent to a simpler 3‐variable model based on ω, FSE, and

ω/D50. In testing various combinations of six variables, most notably FSE and OM tended to provide higher predictive success compared to MSa and MSC. Finally, dropping sediment transport (ω/D50) from the model did significantly reduce the percentage, but the 2‐variable model with just stream power (ω) and floodplain sand equivalent (FSE) still produced successful predictions to more than 75%.

As a final DA test, a 3‐variable ω‐FSE‐ω/D50 model was estimated based on 30% of the most

representative floodplain archetypes as interpreted from field observations (Figure 3.7). As such, for this DA cross‐validation test 70% of the 109 floodplain sites were removed from the

model estimate. Even based on only 3‐variables, the average percentage of correct predictions

from the validation dataset was 97% (i.e., only 2 misclassifications out of 77 sites in validation

sample). By narrowing the multivariate domain of the classifications based on the most

representative floodplain archetypes, the group‐split cross‐validation test in Figure 3.7

demonstrates how transitional floodplain types apply within a multivariate continuum. 99

Figure 3.7: Group‐split cross‐validation DA analysis for 3‐variable model (ω, FSE, and ω/D50), where 70% of dataset is removed and model estimation is by the remaining 30% of floodplain archetypes. Of 109 sites, 32 floodplain archetypes (estimation sample) are represented by large symbols and the remaining 77 floodplain observations (validation sample) are represented by smaller symbols. Average percent prediction success of validation sample is 97%. Transitional floodplain types are indicated between the 95% confidence ellipses for each floodplain archetype from the model estimate.

3.5 Discussion

3.5.1 Floodplain parsimony

The principal of scientific parsimony places value on simpler explanations of natural

phenomena over complex models. Classification is by nature a form of simplification and generalization. Although environmental diversity is certainly not unique to this study, the glacial legacy for river systems of the Laurentian Great Lakes region does impose a myriad of discontinuities on fluvial drainage networks (Phillips and Desloges, 2014). As such, some effort must be made to explain the spatial arrangement of fluvial landforms and process domains in

this context. We present an original empirical dataset for southern Ontario of river channel and floodplain variables and employ multivariate statistical methods to cluster floodplain types and identify controlling factors. While variable selection does require theoretically informed 100

choices, our view is that multivariate statistical methods provide a reasonably objective

approach to develop substantive fluvial process‐landform associations.

The study results have revealed an effective set of underlying variables that identify meaningful groupings for alluvial floodplain landforms in southern Ontario (Figure 3.8). The four floodplain classifications presented are principally governed by the two explanatory dimensions of stream

power‐resistance and floodplain sedimentology. In general terms, these results are consistent

with previous efforts to classify channel and floodplain morphology (e.g., Kellerhals et al., 1976;

Schumm, 1985; Church, 1992; Nanson and Croke, 1992). However, given that stream power (or slope) and resistance (or bed material size) are highly correlated, it is rational to group the

associated variables into one dimension. What follows is that the unexplained variance in the

power‐resistance correlation potentially represents the variability in sediment transport. This idea is significant in that it may provide a link between how variations in sediment transport can be translated into variations in alluvial floodplain construction. The second and third dimensions from our 12‐variable PCA are consistent with this idea, in that they represent floodplain sedimentology (FSE, OM) and bedload sediment transport indices (τ/D95, ω/D50),

which both play a role in governing patterns of alluvial deposition.

Church (2006) points out the intermediary role of sand in sediment transport and alluvial sedimentation, potentially moving in contact with the bed and/or in suspension. In Figure 3.8, we have summarized the second dimension of floodplain sedimentology primarily in terms of a

new variable we propose as the FSE. As defined earlier, conceptually it is the floodplain sand

equivalent depth of the fine alluvial materials and practically it is calculated in this study as the

product of the fine alluvial floodplain thickness (Ffa) and the percent sand (MSa).

101

Figure 3.8: Floodplain classifications for low‐relief glacially conditioned river catchments based on primary factors of stream power‐resistance and floodplain sediment texture interpreted from field observations and multivariate statistical analyses. Secondary factors of width‐to‐ depth ratio, floodplain thickness, floodplain accretion rate, and percent organic matter content (OM) are also presented.

While we acknowledge there is room to refine our estimates of FSE in future studies, the

current dataset is effectively capable of discriminating between the proposed floodplain types.

We propose that the availability of sand in the channel is a substantive control on alluvial

depositional processes, and given its transient nature it is a potential index of floodplain accretion rates. In other words, the turnover of alluvial floodplain material may increase with the amount of sand available. FSE also happens to be negatively correlated with the percent organic matter which accumulates in floodplains with lower accretion rates and less turnover

(Figure 3.8).

As informed by the 2‐variable DA test (Table 3.7), a simple two variable plot of specific stream

power (ω) versus FSE (Figure 3.9) demonstrates reasonably good separations between the four

primary floodplain types. Dashed discrimination lines suggest thresholds between three of the

four floodplain types. 102

Figure 3.9: Four floodplain classifications in bivariate space of specific stream power versus floodplain sand equivalent. The dashed discrimination lines are discussed in the text. Floodplain archetype domains for ω‐FSE are provided in Table 3.8.

A specific stream power of 30 Wm‐2 divides S‐type and C‐type floodplains from B‐type

floodplains (which happens to be average ω for the study area, Phillips and Desloges, 2014).

Further, a floodplain sand equivalent depth of 1 m divides S‐type floodplains from B‐type and C‐ type floodplains. The intermediate ω‐FSE domain is occupied by M‐type floodplains with

stream powers of 20–60 Wm‐2 and floodplain sand equivalent depths of 0.5–2 m.

Despite our discrete classifications, the river landform continuum concept is still relevant. As

demonstrated in Figure 3.5, Figure 3.6, and Figure 3.9, a clear gradient exists between S‐M‐B floodplain types. Extending this idea further, narrower definitions of the four floodplain

archetypes (Figure 3.7) can provide the basis to explore floodplain diversity in southern Ontario

in terms of meaningful transition groups (e.g., B‐M, S‐M, C‐M, or C‐S types). Refining the 3‐

variable ω‐FSE‐ω/D50 domains in terms of narrowly defined floodplain archetypes, general

ranges of each variable are provided in Table 3.8. Separations between these multivariate

domains effectively represent transitional space. 103

Table 3.8: Three‐variable classification domains for primary floodplain archetypesa.

Floodplain ω FSE ω/D50 Archetype (Wm‐2) (m) (Wm‐3) x 10‐3 B 60–100 0.1–0.2 0.5–1 M 25–40 1–1.5 1–1.7 S 10–15 2–3 1.5–3 C 5–10 0.2–0.7 1–2 a Separation between domains represents transitional space.

3.5.2 Floodplain alphabet

River and floodplain classification invariably requires some sort of nomenclature or taxonomic

identification of classes. Our proposed floodplain classification framework is not intended to

contradict other classification schemes, but is rather a complimentary approach to include in the dialog of fluvial landform classification as particularly applied to low‐relief, glacially‐

conditioned environments. As such, some discussion of how our proposed classifications relate

to previous river and floodplain classifications is likely to be informative.

Genetic Floodplains As highlighted earlier, the genetic floodplain classifications of Nanson and Croke (1992) are

aimed at linking previous concepts of channel morphology with alluvial floodplain construction.

The two main criteria used are stream power (ω) and floodplain sediment texture (non‐ cohesive vs. cohesive). Their floodplain classification framework is organized into three primary

orders: Class A, high‐energy non‐cohesive (ω > 300 Wm‐2); Class B, medium‐energy non‐

cohesive (ω = 10–300 Wm‐2); and Class C, low‐energy cohesive (ω < 10 Wm‐2). Each primary

order is then divided into a number of sub‐orders based on distinctive observations of erosion

and depositional processes; channel and floodplain landforms; river planforms; and

environmental contexts. From here forward Nanson and Croke (1992) floodplain classifications

will be written as NC‐class. 104

Comparisons between floodplain types in this study and Nanson and Croke (1992)

classifications are summarized in Table 3.9. Many floodplains in southern Ontario are good

examples of NC‐class B3 (meandering, lateral‐migration) floodplains as per our M‐type

floodplains within the range of ω = 10–60 Wm‐2. However, the degree of lateral activity (with

associated proportions of lateral and vertical accretion deposits) tends to be dependent on the

stream power and the amount of sand available in the system (i.e., the FSE).

Numerous low‐energy cohesive floodplains exist in our study area that are dominated by

vertical accretion processes (and potentially some oblique accretion). These tend to be C‐type

floodplains in southern Ontario and correspond to the single‐channel NC‐class C1. However, this NC‐class does not adequately describe many of our less‐cohesive S‐type floodplains which

are laterally unstable in some cases, but contain very little gravel material and are relatively

low‐energy (e.g., Stewart and Desloges, 2013).

Table 3.9: Comparison of floodplain classifications from this study with previous classifications.

Nanson and Croke Rosgena (1992) (1994, 1996)

Floodplain Match Similar Match Similar Class A C1 B‐type C3 B1, B2 C2, C4

M‐type B3 C4, E4

S‐type C1 B3b,c,d E5

C‐type C1 E6

a For floodplains with channels which are not entrenched.

105

Our S‐type floodplains could potentially be similar to NC‐classes B3b, B3c, or B3d, but they do

not typically have the distinguishing landforms of scroll‐bars or concave benches. Many S‐type

floodplains in southern Ontario are dominated by vertical and oblique accretion, with increasing amounts of lateral accretion in reaches with slightly higher stream powers (e.g., ω =

15–25 Wm‐2).

The B‐type floodplains in this study (about one‐third of those classified) are the most difficult to

place within the Nanson and Croke (1992) classifications. Based on stream power they would

seem to fit into NC‐class B1 or B2 floodplains, but they do not tend to exhibit multiple‐channel planforms (e.g., braiding and wandering, respectively) and the bed materials are typically cobble and boulders (D > 64 mm), with only minor fractions of gravel (D = 2–64 mm). While it is likely that Nanson and Croke (1992) did not intend to exclude cobble channels from their B1

and B2 classes, the apparently low sediment transport rates in our B‐type channels are not

typically capable of producing braiding and wandering channel features, and specifically gravel

bedforms and braid bars tend to be absent.

It is possible that some B‐type floodplains for bedrock channels in southern Ontario might fit

into NC‐class A floodplains (ω > 300 Wm‐2) and some hydraulically confined reaches have been

suggested to be cut‐and‐fill floodplains (e.g., Thornbush and Desloges, 2011), but many B‐type

floodplains in southern Ontario do not exhibit stream powers greater than 200 Wm‐2. In

addition, some B‐type floodplains have strongly cohesive bank materials where sands and

gravels are scarce and cobble transport is exceedingly low (ω << 100 Wm‐2). The dominant

styles of floodplain accretion for B‐type floodplains may not fit well into traditional frameworks

and this issue requires further study. It is likely that vertical accretion and catastrophic

106

stripping of overbank fine material are important processes in some reaches. However, where

bed material transport rates for cobble and gravel are high enough, an important component of

B‐type floodplains might also be a low‐angle style of lateral accretion.

Rosgen Classifications Our comparisons between floodplain types in this study and Rosgen (1994, 1996) stream type

classifications are summarized in Table 3.9. The Rosgen method of stream classification is

established as a descriptive tool for communication between engineers and ecologists, but is

not without its critics in geomorphology (e.g., for recent discussions see Simon et al., 2007;

Roper et al., 2008; Rosgen, 2009; and Buffington et al., 2009). The controversy again highlights

the enigmatic problem of connecting fluvial processes and landforms; as well as the

geomorphic problem of landform equifinality (Simon et al., 2007). This present study has proposed process‐landform associations using multivariate statistical methods to test a large

empirical dataset in a specific environmental context. A similar type of study by Splinter (2013)

in Oklahoma (USA) tested variability in stream reach morphologies using clustering statistics

and comparisons with Rosgen classifications, finding that equivalent Rosgen stream types could

exhibit very different channel morphologies. Nevertheless, some practioners in geomorphology will likely find our brief comparison with Rosgen useful.

For channels which are not entrenched in southern Ontario, three of our four floodplain types can be reasonably classified by Rosgen stream types (Table 3.9), but our M‐type floodplains tend to span Rosgen gravel‐bed E and C stream types based primarily on the w/d ratio

(although given Rosgen’s allowable tolerances, most M‐type floodplains could be Rosgen C4

stream types). Rosgen stream type A may apply to entrenched bedrock reaches of the Niagara

107

escarpment, while Rosgen G, F, B stream types may apply to other entrenched and moderately

entrenched reaches in our study area. Specifically, entrenched channels in southern Ontario tend to be associated with tributary base level effects from lake history and river incision

processes (Phillips and Desloges, 2014) or with more recent land use change (and semi‐alluvial

channel conditions, e.g., Vocal Ferencevic and Ashmore, 2012). Even so, rather than

distinguishing discrete stream types based on entrenchment (and given concerns about

bankfull concepts), we view entrenched channels as modified forms of our core floodplain

classifications, and thus another possible theme to explore in sub‐classification.

3.6 Conclusions

Alluvial floodplain classifications for rivers in low‐relief glacially conditioned catchments have

been presented based on an original floodplain dataset and a sequence of multivariate

statistical analyses (cf. Brardinoni and Hassan, 2007; Splinter, 2013; Vaughan et al., 2013). We

have classified 109 floodplains in southern Ontario into four first‐order groups primarily based

on morphological and sedimentological variables. The four floodplain types are successfully

predicted by two principal explanatory dimensions: (1) stream power‐resistance, which is

effectively represented by specific stream power (ω); and (2) floodplain sedimentology, which

is effectively represented by floodplain sand equivalent (FSE). Motivated by comments of

Church (2006), we propose FSE as an important explanatory variable to predict alluvial

floodplain variations both conceptually and statistically. The stream power approach is similar

to that of Nanson and Croke (1992), but the sedimentological influences of glacial landforms

require refined floodplain classifications for southern Ontario which particularly consider the

effects of inherited cobble bed materials and alluvial sand availability.

108

Variables of stream power and resistance are highly correlated statistically (and theoretically).

Considering environmental controls in statistical terms, correlation between channel slope and bed material grain size indicates that they should be considered a single explanatory dimension to predict variations of alluvial floodplain types. Following this rationale, the residuals of the energy‐resistance correlation potentially represent variations in sediment transport (third principal component, τ/D95, ω/D50). From the perspective of landform continuums, further exploration of floodplain sub‐classifications may initially be based on transitional floodplain types, as well as other environmental contingencies such as channel entrenchment and semi‐ alluvial boundary materials. However, within‐class variations in channel morphology are also likely explained to some degree by variations in sediment transport.

It should be recognized that the four floodplain types presented are generalized, meaning they are not intended to exhaustively represent the diversity of floodplains in southern Ontario.

Even so, the floodplain classifications are meaningful in that they do provide robust first‐order criteria to organize fluvial process‐landform associations and consequently a basis from which to further investigate geomorphological diversity. Classification of alluvial floodplains using a multivariate process‐landform dataset is consistent with the ideas of reach‐scale fluvial process domains and lithotopo units as advocated by Montgomery (1999). Further, revealing the spatial arrangement and linkages of distinct morphological groups within a regional landscape mosaic is expected to provide insights into patterns of post‐glacial fluvial adjustment.

Extending previous research from mountainous glaciated catchments as envisioned by Phillips and Desloges (2014), we propose that adapted models are required to explain the spatial arrangement of river morphologies within the context of complex glacial legacy effects in low‐ relief settings, for southern Ontario and beyond.

109

Chapter 4 Glacial legacy effects on the spatial organization of alluvial floodplain types in the Laurentian Great Lakes region

4 Chapter 4

Abstract: The spatial arrangements of river landforms classified by floodplain types are investigated in the context of glacial legacy effects for three river catchments in southern Ontario, Canada. Patterns of post‐glacial fluvial adjustment in the low‐relief landscape of the Laurentian Great Lakes region are expected to differ from glacially conditioned mountainous environments. The typical Holocene response of southern Ontario rivers is confounded by numerous distinct glacial landform assemblages and dynamic baselevel histories. Mapping of river landforms provides an interpretive framework to explain post‐glacial fluvial adjustments in the context of local sources of glacial sediment and patterns of fluvial sediment flux over the Holocene. A number of general glacial–fluvial landform relationships have been identified and can be broadly clustered into three themes: 1) topographic and sedimentological glacial legacy effects; 2) landforms resulting from isostatic and lake baselevel change, and 3) superimposed patterns of Holocene fluvial sediment supply. Some geochronological evidence is assessed for floodplain landform ages, but radiocarbon methods are limited in many floodplain types due to slow vertical accretion processes. Based on data from this study, laterally active floodplains with point‐bar deposits may be turned over every few centuries in smaller channels and every few millennium in larger channels. However, specific glacial legacy effects can also disrupt alluvial floodplain accretion processes leaving behind older floodplain deposits in some cases. Additional methods such as Optically Stimulated Luminescence (OSL) dating are needed to better assess alluvial floodplain geochronology across all floodplain types in southern Ontario.

110

4.1 Introduction

Since the end of continental scale glaciation in North America at the end of the Pleistocene, vast landscapes of exposed glacial drift have lay open to non‐glacial geomorphic processes.

While deglaciation and the substantial actions of melt‐water also contributed to sculpting the landscape, Holocene fluvial processes have been active to rework glacial debris and to carve

drainage networks into the deglaciated surfaces. Understanding post‐glacial fluvial responses

requires consideration of geomorphic processes at multiple timescales in the context of glacial

landscape features which are expected to control long‐term fluvial landscape evolution. In

particular, the abundance and organization of glacial deposits may impose a paraglacial legacy

on river sediment fluxes (Church and Ryder, 1972) and inherited topographic signatures of

glacial origin may persist in river long profiles (e.g., Brardonini and Hassan, 2006). Effectively, it

is well accepted that much of the Canadian landscape is still responding to the glacial legacy of

the last glacial period, and that fluvial adjustments are ongoing. As dramatically stated by

Church and Slaymaker (1989) for western Canada, “the natural landscape of British Columbia is

imprisoned in its history” (p. 453).

Documenting the spatial organization of river landforms and fluvial process domains can

provide insights into the patterns of post‐glacial fluvial adjustment over the Holocene

(Brardinoni and Hassan, 2006; Collins and Montgomery, 2011; Thayer, 2012). However, glacial

legacy effects on fluvial systems within the low‐relief topography of much of the interior

continent of North America have received considerably less attention as compared to

mountainous environments. As such the purpose of this study is to investigate the spatial

arrangement of river morphologies for catchments in the Laurentian Great Lakes region of

111

North America, a landscape of modest relief and of complex environmental history. More

specifically, the spatial organization of reach‐scale fluvial process domains and alluvial

floodplain types for three select river catchments in southern Ontario are analyzed in the

context of their glacial landforms and their interpreted post‐glacial landscape histories. A

further understanding of post‐glacial fluvial adjustments in such environments is important for

improving our knowledge of fundamental fluvial process‐landform relationships in the context

of complex glacial legacy effects.

4.2 Study Area

The Laurentian Great Lakes watershed is the product of recurring continental glaciations over the Quaternary Period that scoured the underlying cratonic bedrock and deposited extensive sequences of glacial drift, particularly along the southern margins. Within North America, the

Great Lakes watershed may be considered a distinct physiographic region of relatively low‐relief as compared to mountainous areas of the Cordilleran region in the west and the Appalachian

region to the east. While the northern portion of the Great Lakes system is etched into the crystalline Precambrian bedrock of the Canadian Shield, the southern lakes are impounded by a

complex architecture of glacial deposits overlying Paleozoic shales and limestones of the

Michigan sedimentary basin.

The study area of peninsular southern Ontario is bound by lakes Huron, Erie, and Ontario

(Figure 4.1A). Focusing on river catchments which drain to Lake Huron, the three fluvial

drainage networks specifically examined in this study include the Ausable, Saugeen, and

Nottawasaga rivers.

112

Figure 4.1: Study area of southern Ontario. A) Three select study catchments and locations of radiocarbon and OSL samples: 1) Ausable River; 2) Saugeen River; and 3) Nottawasaga River. Location of other field sites with radiocarbon samples also shown: 4) East Humber River (Humber River tributary); 5) Nith River (Grand River tributary); 6) Catfish Creek; and 7) Medway Creek (Thames River tributary). B) Isoline map of southern Ontario for amount of isostatic rebound since 10.6 ka BP (uncalibrated 14C, glacial Lake Algonquin phase) as per analysis of Lewis et al. (2005). Contour interval is 5 m with reference area of zero uplift southwest of Lake Michigan beyond the limit of the last glaciation.

The glacial legacy effects on these river systems can largely be described by the inherited

topographic signatures of the glacial landforms (Chapter 2; Phillips and Desloges, 2014), as well as the inherited sedimentological influences of cobble bed‐material and alluvial sand availability

(Chapter 3). However, post‐glacial fluvial landscape responses in the Great Lakes region have

also been affected by baselevel changes associated with lake water level fluctuations and

broad‐scale patterns of post‐glacial isostatic rebound (Figure 4.1B).

4.2.1 Post-glacial baselevel changes

Rivers tend to respond to changes in baselevel (Schumm, 1993), which in the context of this study baselevel is essentially the water surface in the Lake Huron basin. Further, river profile evolution toward a graded condition (Chapter 2) assumes a constant baselevel elevation for

vertical profile adjustments. However, starting at the river outlet increases in baselevel will 113

impose an aggradational response as stream energy is lost and decreases in baselevel will cause

a degradational response as stream energy is gained. Resulting river landforms may include

low‐gradient backwater floodplain surfaces associated with baselevel rise and river terraces

and valley entrenchment with baselevel lowering. Dynamic baselevel fluctuations can

therefore impose a palimpsest of fluvial landforms with phases of floodplain aggradation and channel incision translated some distance upstream according to the timing and magnitude of the baselevel changes, as well as the antecedent profile conditions. With the evolution of the

Great Lakes, and Lake Huron in particular, the baselevel history for post‐glacial fluvial systems is

confounded by patterns of glacial isostatic rebound and changes in the major hydrological

sources of water (Teller, 1995).

Glacial isostatic rebound

The concept of isostasy accounts for the gravitational equilibrium of the earth’s crust, as it

relates to the relative properties of the lithosphere floating on a viscous upper mantle (Wolf,

1993). Glacial isostasy includes the broad‐scale crustal depression and rebound resulting from

the loading and unloading of continental ice sheets, a recurring dynamic process over glacial periods of the Cenezoic Era (Wolf, 1993; Raymo, 1994). The amount of glacial isostatic depression from the last glacial period generally varies with patterns of continental ice

thickness (Peltier, 2004), which in the Great Lakes region was greatest in the northeast,

gradually decreasing to the southwest (Figure 4.1B). As such, the amount of post‐glacial isostatic rebound varies between the three study catchments, with the greatest amount of

uplift for the Nottawasaga River (#3) and the least amount of uplift for the Ausable River (#1).

114

As was the case on the Nottawasaga River, the greater amount of isostatic depression resulted

in more extensive deposition of proglacial lake deposits during ice retreat.

The increasing amount (and rate) of isostatic rebound from south to north also caused broad‐

scale tilting of the landscape (Clark et al., 1994). Crustal warping was initially recognized in the

tilting of ancient lake shorelines. The Great Lakes region has been particularly rich in paleo‐ shoreline evidence, with early publications on the tilted shorelines by Spencer (1891),

Goldthwait (1910), and Leverett and Taylor (1915). These works were later summarized by

Hough (1958) and then revisited by Karrow and Calkin (1985). Recent papers to revisit paleo‐

shoreline reconstructions in the Lake Huron basin include Schaetzl et al. (2002), Lewis et al.

(2005), Heath and Karrow (2007) and Lewis et al. (2008).

Increasing technical sophistication has allowed researchers to further investigate glacial isostatic rebound using computer models (e.g., Peltier, 1994, 2004). Building on a long history

of measuring water levels in the Great Lakes, Mainville and Craymer (2005) provide analysis

from a network of gauges throughout the Great Lakes which indicates modern uplift rates in the range of 30–50 cm/century. Similar uplift rates have also been reported by others based on

Great Lakes water levels and from a growing network of satellite Global Positioning System

(GPS) sites in North America (Sella et al., 2007; Braun et al., 2008). Satellite monitoring of temporal changes in the earth gravity field is also being used to assess patterns of post‐glacial isostatic adjustment (van der Wall et al., 2008). Even so, late glacial and early Holocene uplift rates are expected to have been much higher, and thus previous work on shorelines and paleo‐ water plains provides important physical constraints to help calibrate models of glacial isostasy.

115

A recent analysis by Lewis et al. (2005) provides a convenient GIS‐based model of glacial

isostatic rebound in the Great Lakes. Specifically, they empirically calibrate glacial isostatic

depression using a large dataset to define paleo‐water plains fit to a simple exponential‐decay

function for post‐glacial uplift as prescribed by Peltier (1998). This uplift expression was used to

determine the amplitude of isostatic depression at any location within the Great Lakes basin

(Lewis et al., 2005). Isostatic depression and rebound of the earth’s surface were reconstructed

(and deconstructed) using a regional digital elevation model (DEM, 30 arc‐second resolution).

This relatively high resolution paleo‐DEM was compared to other geophysical models of glacial

isostasy, including that of Clark et al. (1994) and the ICE‐4G model of Peltier (1995). By comparing the exponential‐decay uplift function with other geophysical models, outlet

chronology elevations showed discrepancies in the order of 5–10 m, likely derived from both empirical error and imperfection of the geophysical models (Lewis et al., 2005).

Post‐glacial Great Lakes

Superimposed on the gradual isostatic uplift and tilting of the Great Lakes were the changing hydrological sources of water following ice retreat, with increasing and decreasing volumes of

water contributing to the rise and fall of the lake levels (Lewis et al., 2005). For example, the

Lake Huron basin initially received direct meltwater inputs from the ice sheet, followed by

indirect meltwater from the drainage of proglacial lakes to the north, and then ultimately

transitioned to strictly climate driven sources of water (Teller, 1995). As presented by Lewis et

al. (2008), four general phases can be used to describe the evolution of the Great Lakes, and

Lake Huron in particular (Figure 4.2): 1) proglacial lakes (PG); 2) extreme lakes (EL); 3) Nipissing

lakes (NLP); and 4) modern lakes (ML).

116

Figure 4.2: Post‐glacial water‐level curve for Lake Huron basin adapted from Lewis et al. (2005), Lewis et al. (2008), and Eschman and Karrow (1985). Lake phases noted as proglacial lakes (PG), glacial Lake Algonquin (GLA), extreme lakes (EL), Nipissing lake phase (NLP), and modern lake (ML). Dashed curve represents approximate Lake Algonquin elevation according to analysis of Lewis et al. (2005).

These phases are generally controlled by complex paleohydrological thresholds (Teller, 1995) which include the dynamic changes in the sources of water and on the physical barriers to flow

(i.e., the locations and elevations of major outlets, or sills). For example, the timing of melt‐

water and climate fluxes would have controlled when sills were over‐topped, the differential

rates of isostatic rebound would have controlled the relative sill elevations, and sill geology would have controlled the rates of incision associated with fluvial overflow processes.

Given its central location in the Great Lakes basin, Lake Huron underwent a complex history of

lake levels and outlet drainage patterns. As a brief overview, the evolution of the Lake Huron

basin is characterized by a shared proglacial lake stage with the Lake Erie basin (PG in Figure

4.2, Glacial Lakes Whittlesley and Warren), a shared proglacial lake stage with the Lake

Michigan basin (GLA, glacial Lake Algonquin), an early Holocene extreme lake phase with

fluctuating low and highstands (EL, Lakes Stanley–Hough and Mattawa, respectively), a rising

Nipissing lake phase (NLP), and ultimately the modern lake (ML, 176–177 m asl) (Eschman and

Karrow, 1985; Lewis et al., 2005; Lewis et al., 2008). 117

Initially most reconstructions of Lake Algonquin (main phase) indicated a water surface

elevation of 184 m asl (Eschman and Karrow, 1985), with similar elevations reported for the

Nipissing lake phase (although Algonquin shorelines are higher in the north due to the

intervening ~5000 years of isostatic uplift). However, based on the shoreline dataset and isostatic reconstructions of Lewis et al. (2005) the main Lake Algonquin elevation may have actually been lower at 150 m asl (Figure 4.2). The early‐Holocene extreme lake phase was due to the compound effects of isostatically depressed outlets to the northeast (lowstands), meltwater pulses from the proglacial lakes to the north (highstands), and possibly fluctuating early‐Holocene climate conditions. The Nipissing lake phase in the mid‐Holocene was

associated with the increasing influence of climate factors and rising lake outlets in the east prior to the redirection of flows to Lakes Erie and Ontario with overflow and incision at the

southern tip of Lake Huron (i.e., Port Huron).

Given the transition to strictly climate driven sources of water in the mid‐Holocene, the general paleoclimate trends for the Great Lakes region can be considered based on data from Edwards

et al. (1996), and more recently from McFadden et al. (2005). Typically, evidence suggests that mid‐Holocene paleoclimate was characterized by higher temperature and moisture conditions

relative to the modern climate, which is consistent with the Nipissing lake phase and subsequent decrease in water levels to the elevation of the modern lake.

4.3 Data and Methods

River reaches were characterized for three major catchments in southern Ontario (Figure 4.1A),

based on field data, geomatic datasets, and previously published information. The glacial

legacies for each drainage network were evaluated based on published mapping and data for

118

glacial landforms, water level history, and isostatic rebound. For alluvial floodplain ages, late

Holocene river deposits were also tested using radiocarbon (14C) and optically stimulated

luminescence (OSL) dating methods (Appendix B).

4.3.1 Field and GIS data

Field observations of river channel and floodplain characteristics were collected for hundreds of

sites across the southern Ontario study area to identify dominant river morphologies and to

document spatial patterns in reach characteristics. As summarized for the three selected

watersheds in Table 4.1, basic field site observations included channel dimensions, substrate type, and valley context (with photographs). For a subset of the field sites, additional measurements were completed to estimate floodplain thickness and alluvial sediment samples were collected to analyze the percentages of sand, silt/clay, and organics (see methods in

Chapter 3).

For each catchment, drainage networks were mapped based on a hydrologically enforced 10‐m

digital elevation model (DEM Version 2.0.0; OMNR, 2005, 2008). Longitudinal river profiles

were extracted from the DEM and generalized as described in Chapter 2. Mapping of river

reaches, including the locations of reach breaks, was assessed based on a synthesis of field

observations (and photographs) in terms of channel morphology and field classifications of

floodplain landforms (Chapter 3). To delineate reaches, field data were also cross‐referenced with secondary sources including river profiles (e.g., changes in slope), surface geology mapping

(e.g., OGS, 1993, 1997; Chapman and Putnam, 2007), and satellite imagery from Google

EarthTM.

119

Table 4.1: Number of field sites per catchment. Catchment Field Sitesa Floodplain Sitesb Ausable River 76 12 Saugeen River 114 16 Nottawasaga River 69 23 a From total of 542 field sites in southern Ontario. b From total of 109 floodplain sites in southern Ontario.

For the final mapping of each drainage network, reaches were assigned an alluvial floodplain type. Classifications were assigned to each reach as per natural groupings interpreted in the field and confirmed using multivariate statistical clustering methods (Chapter 3). Four first‐ order floodplain types were identified (B, M, S, and C types), plus at least four transitional

floodplain types (BM, SM, CS, CM). Floodplain classifications are summarized in Table 4.2.

Transitional floodplain types were initially interpreted based on field observations of hybrid

characteristics of two first‐order floodplain types. For example, laterally migrating rivers with

low‐angle point‐bar accretion and relatively thin alluvial deposits of sand and overbank fine

material (w/d < 15–30) exhibit characteristics of both B and M‐type channels (i.e., BM

floodplains). Thick clay‐rich floodplains (w/d < 10) with 40‐60% sand were considered CS

floodplains. The transitional status of individual sites was subsequently assessed based on the

discriminant cross‐validation analyses discussed in Chapter 3 (Table 3.6 and Figure 3.7).

120

Table 4.2: Summary of floodplain classifications for southern Ontario rivers. Floodplain Stream Power Bed Width‐to‐ Floodplain Floodplain Accretionb Type ω (Wm‐2) Material Depth Ratio Sand FSEa (m) and Bed Morphology B‐type 60–100+ cobble 20–50 0.1–0.2 Strip‐and‐fill, overbank vertical accretion of fines, cobble‐lag in cascade‐run‐pool morphologies, potentially low‐angle lateral accretion where bed mobility is high enough BM 30–90 cobble/ 15–30 0.2–1.0 Transitional from B to M types with gravel more gravel and sand, greater bed mobility, and potentially more lateral accretion compared to B‐types M‐type 25–40 gravel 10–15 1.0–1.5 Lateral point‐bar accretion and overbank vertical accretion, ratios depending on degree of lateral activity, stream power, and sand availability; dynamic riffle‐pools depending on relative mobility of the bed material SM 10–30 sand/ < 10–15 1.5–2.0 Transition from M to S types with gravel greater sand availability and lower stream powers compared to M‐types S‐type 10–15 sand < 10 2.0–3.0 Overbank vertical accretion with increasing lateral accretion given increasing gravel material and stream power; lateral accretion varies from point‐bar accretion to high‐angle oblique accretion processes; limited bed morphology other than sand dunes and occasional fine gravel bars CS 5–15 sand/silt < 10 0.7–2.5 Transition from S to C types with /clay greater sand availability compared to C‐types C‐type < 5–10 silt/clay/ << 10 0.2–0.7 Overbank vertical accretion in sand floodplain and other low‐lying vegetated areas within the bankfull channel; vegetation controls and back‐ water effects from log and ice jams may be important for floodplain accretion processes; limited bed morphology other than occasional silt/sand bars. CM 5–40 silt/clay variable 0.2–1.0 Transitional from C to M types with and (>1.0)* inherited or Holocene gravel/cobble gravel/ lags; lower stream powers and/or sand cobble availability compared to M‐types; common in low‐gradient and vegetation controlled headwater channels (*and some larger channels) a FSE is defined as the equivalent thickness of floodplain sand (see definition Chapter 3). b Floodplain formation processes as per Nanson and Croke (1992), also see Table 1.1 (Chapter 1). 121

4.3.2 Isostatic paleo-DEM and river incision

Modeling of isostatic rebound for each catchment was completed by reconstructing a Late

Pleistocene DEM (paleo‐DEM) based on the analysis and results of Lewis et al. (2005). As

described earlier, the analysis of Lewis et al. (2005) uses a synthesis of empirical data for tilted

shorelines in the Great Lakes basin to calibrate models of glacial isostatic rebound assuming an

exponential‐decay uplift function presented by Peltier (1998). For this study, model results for

the total isostatic rebound since 10.6 ka BP (Lake Algonquin phase, all dates uncalibrated 14C

unless otherwise stated) using a 10‐m resolution raster interpolation method were applied to the DEM of each river catchment to reconstruct a paleo‐DEM. Essentially, as an estimate of the

total isostatic uplift since deglaciation, the amount of rebound was subtracted from the modern

DEM to calculate the paleo‐DEM surface within a 10‐m resolution raster‐GIS data frame.

The resulting paleo‐DEM represents an approximate 10.6 ka BP digital elevation model. River

profiles were extracted from the paleo‐DEM to analyze patterns of isostatic tilting within river

catchments and to assess Holocene temporal trends in relative stream energy. However,

Holocene river incision was not removed from the elevation model so river profiles from the paleo‐DEM are likely to be somewhat lower than the actual Late Pleistocene elevations of each drainage network. Even so, river profiles extracted from the paleo‐DEM were also generalized to 2–5 km of river length so relative slope comparisons between modern DEM and paleo‐DEM

river profiles are expected to be reasonable.

To provide some interpretive reference with respect to spatial patterns of Holocene river

incision, longitudinal profiles of the upland surfaces were collected for comparison with the

modern elevation profiles of the river channels (i.e., to represent post‐glacial pre‐incision

122

surfaces). To reconstruct the upland profiles, valley cross‐sections were extracted from the modern 10‐m DEM along the river length every 2–5 km, with more densely spaced cross‐

sections at glacial landform transitions. Upland surface elevations were then averaged from

each side for the river valley, and the mean upland elevation was assigned an upstream river distance according to the location of the valley cross‐section.

4.3.3 Radiocarbon and OSL

To test interpretations of fluvial landforms relative to Holocene fluvial responses, geochronological evidence of floodplain age was assessed at field sites where organic material was found for radiocarbon (14C) dating. Given the challenges of locating dateable organics and

interpreting the relative age of surrounding alluvial deposits, optically stimulated luminescence

(OSL) dating was also tested at three floodplain sites (Appendix B, 2 OSL samples per site), with

one OSL site in each of the three study catchments (Figure 4.1). In total, 11 radiocarbon

samples were collected from floodplain deposits at 10 field sites in southern Ontario and were

submitted for radiocarbon dating, with 3 of the 10 sites including at least one 14C sample and

two OSL samples (Table 4.3). Radiocarbon samples were submitted to the laboratory of Beta

Analytic Inc. for standard dating methods (Note: 9 radiometric and 2 AMS). OSL samples were

tested at the University of Illinois at Chicago Luminescence Dating Research Laboratory (UIC‐

LDRL).

123

Table 4.3: Radiocarbon samples from southern Ontario alluvial river floodplains for this study. Samplea Dated Host Material Conventional Calibrated Location and Lab ID Materialb and Depth Radiocarbon Age Age (yr BP)c 1. Ausable River 43° 11’ 05” N shell Alluvial sand 5120 ± 40 BP 5750–5840 Main Branch 81° 48’ 57” W (AMS) terrace and (AR‐DSB1) + OSL 3.10 m 5850–5930 Beta 289891 2a. Saugeen River 44° 11’ 12” N wood Alluvial loam on 1870 ± 40 BP 1710–1890 Main Branch 81° 11’ 45” W basal gravel (SR‐B3) 1.80 m Beta 292044 2b. Saugeen River 44° 10’ 58” N wood Alluvial sand 1670 ± 30 BP 1520–1620 Main Branch 81° 11’ 33” W terrace and (SR‐9c‐B1HT) + OSL 2.00 m 1670–1690 Beta 324957 3. Nottawasaga River 44° 12’ 40” N charcoal Alluvial sand on 5920 ± 40 BP 6660–6810 Main Branch 79° 49’ 01” W (AMS) basal gravel and (NR‐B3) + OSL 2.70 m 6810–6850 Beta 289895 3.1. Upper Nottawasaga River 44° 06’ 53” N wood Alluvial sand on 150 ± 30 BP *60–160 Main Branch 79° 52’ 11” W basal gravel and (NWR3‐B1) 1.30 m 170–280 Beta 324956 3.2. Mad River 44° 18’ 19” N wood Alluvial sand on 610 ± 30 BP 540–660 Nottawasaga River Tributary 80° 00’ 11” W basal gravel (NR‐MR1‐B2) 1.20 m Beta 324955 4. East Humber River 43° 50’ 41” N wood Alluvial sand on 580 ± 40 BP 520–660 Humber River Tributary 79° 37’ 13” W basal gravel (EH‐B1) 1.85 m Beta 289893 5.1. Nith River 43° 11’ 12” N wood Alluvial loam on 1540 ± 30 BP 1360–1520 Grand River Tributary 80° 23’ 55” W basal gravel (GR‐NR1b‐B2) 2.10 m Beta 324953 5.2. Nith River 43° 18’ 35” N wood Alluvial sand on 2140 ± 30 BP 2010–2160 Grand River Tributary 80° 29’ 42” W basal gravel and (GR‐NR9‐B1) 2.00 m 2170–2300 Beta 324954 6. Catfish Creek 42° 42’ 18” N wood Alluvial sand on 390 ± 50 BP 310–520 Main Branch 81° 02’ 53” W basal gravel (CF‐B1) 1.90 m Beta 289892 7. Medway Creek 43° 00’ 18” N wood Alluvial sand on 2710 ± 50 BP 2750–2890 Thames River Tributary 81° 17’ 27” W basal gravel and (MW‐BC) 1.75 m 2900–2920 Beta 289894 a Sample numbers as referenced in Figure 4.1; lowercase letters indicate multiple samples within same site; OSL results are presented in Appendix B. b Radiocarbon dating by radiometric analysis unless otherwise indicated as AMS. c Two Sigma Calibrated Results; possible modern samples post 1950 indicated with (*)

124

4.4 Results and Discussion

The spatial organization of alluvial floodplain types was mapped for three select river

catchments in southern Ontario to explore patterns of post‐glacial fluvial adjustment in a

variety of glacially conditioned contexts. The premise being that low‐relief river catchments in

the Laurentian Great Lakes region have been subjected to diverse glacial legacy effects, and represent an environment which has received little systematic attention in previous literature

with respect to Holocene fluvial landscape evolution. Although each river catchment is

essentially unique, identifying broader themes in terms of fluvial process‐landform

relationships will improve our understanding of fundamental geomorphic theories in the

context of low‐relief glacial landscapes and complex environmental histories.

4.4.1 Isostatic and baselevel variations

The general pattern of post‐glacial isostatic uplift in southern Ontario is presented in Figure

4.1B (Lewis et al., 2005). Given the different geographic locations and orientations of each drainage network, the effects of glacial isostatic rebound are not the same for each river

catchment (Figure 4.3). Consistent with the broader pattern of isostatic rebound in the Great

Lakes region, the amount of uplift and the degree of catchment tilting is greatest in the northeast and is least in the southwest. Consequently, catchment‐wide tilting of the

Nottawasaga River has been the most significant with a 35 m difference in the amount of uplift

between the northern and southern parts of the watershed. For comparison, the Ausable River difference is about 10 m and the Saugeen River difference is about 30 m of catchment tilting.

However, as noted the influence of isostatic tilting on river gradients does depend on the

alignment and aspect of the drainage channels.

125

Figure 4.3: Isoline maps of isostatic tilting for each study catchment (right) showing general pattern of river slope change since 10.6 ka BP: (A) Ausable River; (B) Saugeen River; and (C) Nottawasaga River. Contour interval is 5 m and zero contour is referenced to contours from Figure 4.1B by values in brackets. Histograms of percent frequency distributions (right) demonstrate three distinct patterns discussed in the text. Light and dark shading of watercourses and bar graphs reference decreases and increases in river slope (respectively) due to differential isostatic uplift rates since deglaciation.

126

To assess the patterns of isostatic rebound and catchment tilting in terms of river gradients, the

extracted river profiles from the modern‐DEM and paleo‐DEM (10.6 ka BP, Lewis et al., 2005)

were compared. General patterns were mapped for the main branch in each of the three

catchments and percent frequency distributions of changes in slope (ΔS) were plotted (Figure

4.3). Representing long‐term Holocene trends, channel gradients were identified as either

decreasing (light grey shading, where ΔS = ‐0.0003 to ‐0.0001); approximately neutral (no shading, where ΔS = ± 0.0001); or increasing (dark grey shading, where ΔS = 0.0001 to 0.0003).

With reference to the isostatic tilting presented in Figure 4.3, some general observations of

river gradient changes in each catchment can be highlighted. As expected, the Ausable River has experienced the least change due to isostatic rebound. While some subtle increases and decreases in gradient above ± 0.0001 were calculated for local areas in the headwaters and lower reaches (respectively), most of the main branch remained relatively neutral with middle

reaches of the Ausable River showing only slight increases (ΔS = 0 to 0.0001). The main branch

of the Saugeen River has experienced a range of changes in river gradient, with increases most

prevalent in the upper reaches and decreases most prevalent in the lower reaches. The main

branch of the Nottawasaga river has experienced the most marked changes in river gradient,

with decreases of 0.0001 to 0.0003 in the middle and lower reaches that become more

pronounced as the river flows in a northerly direction.

River profiles from the modern DEM were also compared to major post‐glacial highstands in the Lake Huron basin to interpret the potential extent of baselevel change on the lower reaches of the Ausable, Saugeen, and Nottawasaga rivers (Figure 4.4).

127

Figure 4.4: Post‐glacial baselevel highstands in Lake Huron basin relative to modern river profiles for three study catchments: A) Ausable River; B) Saugeen River; C) Nottawasaga River. Highstand water level phases noted as glacial Lake Algonquin (GLA) and Nipissing lake phase (NLP). Relative paleo‐water plains were calculated based on the isostatic depression model of Lewis et al. (2005) for 150 m asl at 10.6 ka BP (GLA) and for 184 m asl at 5 ka BP (NLP). Distance of transgression marked by down arrows.

Specifically, water‐level elevations of glacial Lake Algonquin (GLA) and the mid‐Holocene

Nipissing lake phase (NLP) represent temporary baselevels which would influence the timing of

post‐glacial incision and would potentially translate backwater floodplain aggradation

processes upstream (c.f. Walker et al. 1997). Respectively, the highstands are based on the

modeled paleo‐DEM river profiles (at 10.6 and 5 ka BP) and previously published water level

elevations (150 m and 184 m asl) as per Lewis et al. (2005) and Eschman and Karrow (1985).

Between each of the three catchments, three distinct baselevel relationships are demonstrated.

Comparing the three profiles in Figure 4.4, the relative elevation of the Algonquin paleo‐water surface depends on the amount of glacial isostatic depression. First for the Ausable River, the

paleo‐DEM analysis from the results of Lewis et al. (2005) actually suggests that the main stage

of Algonquin may not have transgressed much beyond the modern day shoreline location (i.e., 128

no GLA noted in Figure 4.4A), but it has also been suggested that the Algonquin and Nipissing

phases were coincident in this area (Karrow, 1980). Even so, proglacial lake deposition in the

lower reaches of the Ausable River and upstream of the Wyoming moraine would have

occurred prior to 10.6 ka BP in the earlier stages of ice retreat (i.e., glacial Lakes Arkona,

Whittlesey, and Warren; Karrow, 1980, Eschman and Karrow, 1985). Second for the Saugeen

River, the amount of upstream transgression that occurred during the Lake Algonquin phase

would depend on the uncertain timing and degree of early post‐glacial fluvial incision.

Algonquin and Nipissing river terraces for Lake Huron drainages have been documented by

Karrow (1986) and the timing of aggradation and incision on the lower Saugeen River to at least

40 km upstream was investigated by Garaci (1998). Based on evidence of terracing, it is actually likely that baselevel signals on the Saugeen River have been transmitted as far as the

Singhampton moraine about 70 km upstream. Third for the Nottawasaga River catchment, the

Lake Algonquin plain is an extensive feature in the basin lowlands, and most of the fluvial

incision of the main branch must have occurred after the glacial lake phase.

Given differences in the inherited glacial landforms, the relative rates of isostatic rebound, and

antecedent profile elevations, the mid‐Holocene transgression of the Nipissing lake phase also

varied considerably between the three catchments. For the Ausable River, the Nipissing

transgression inundated a relatively open embayment (i.e., Thedford Embayment; Karrow,

1980) and backwater would have translated more than 10 kilometres up the narrow valley cut

into the Wyoming moraine (as much as 30 km upstream of modern Lake Huron, Figure 4.4A).

The steep gradient of the lower Saugeen River profile resulted in Nipissing backwater effects to

less than 10 kilometres upstream (Figure 4.4B; Garaci, 1998). In contrast, the Nipissing

129

transgression backwatered the Nottawasaga River about 60 km upstream through the narrow

cut in the Edenvale moraine inundating a wide isolated embayment in the Minesing basin

(Figure 4.4C; Fitzgerald, 1985).

4.4.2 Reach classification and river profiles

To interpret fluvial adjustments, river reaches for the Ausable, Saugeen, and Nottawasaga

rivers, plus major tributaries, were mapped and overlaid with the glacial landform mapping of

Chapman and Putnam (2007) (Figure 4.5, Figure 4.6, and Figure 4.7, respectively). To cross‐ reference reaches, longitudinal profiles for each of the three main branches were also plotted,

including model estimates of glacial isostatic depression (10.6 ka BP) and present‐day profiles of the valley‐uplands which highlight the general patterns of post‐glacial fluvial incision. The approximate post‐glacial water level curve for Lake Huron has also been plotted for reference relative to the modern lake elevation (176 m asl). Figure 4.5, Figure 4.6, and Figure 4.7 provide a basis from which to discuss patterns of post‐glacial fluvial adjustment and associated process‐

landform responses in terms of the proposed floodplain landform classifications (Table 4.2).

Ausable River

The Ausable River catchment drains a series of recessional till moraines, including intervening till, outwash, and glaciolacustrine deposits (Figure 4.5). The largest of these is the Wyoming moraine which separates the upper and lower reaches of the Ausable River. More specifically, the upper and middle reaches flow along the proglacial lee‐side of the moraine, while the lower

reaches are cut into the moraine, flowing through a gorge‐like valley and then entering a widening plain of proglacial and post‐glacial lake deposits. Most of the Ausable River headwaters drain areas south and east of the Wyoming moraine, including till and outwash of

the older Lucan and Seaforth series of recessional moraines (Chapman and Putnam, 1951). 130

Figure 4.5: Reach classification for the Ausable River and select tributaries. (A) Reach mapping overlaid with glacial landforms of Chapman and Putnam (2007). Reach numbers for main branch include prefix for floodplain classifications as outlined in Table 4.2, and based on results in Chapter 3. (B) River profile and reach classifications, with post‐glacial water surface history of Lake Huron basin relative to modern lake elevation (176 m asl; Figure 4.2). Modern river profiles were extracted from the modern‐DEM (Chapter 2) and upland profiles were collected for this study. Modeled profile surface prior to isostatic rebound is for 10.6 ka BP as per results of Lewis et al. (2005) and as per paleo‐DEM analysis described in this paper.

131

As such, headwater source areas of the Ausable River are morainic hillslopes and subtly sloping

till surfaces. Typical reach classifications in the headwaters are CM‐type floodplains (e.g., Reach

CM‐1) due to the abundance of fine‐grained silt and clay, the variable channel gradients, the local sources of gravel and cobble, and the relative lack of sand sized alluvium. The associated

headwater channels tend to be highly vegetation controlled, transitioning from ephemeral to

perennial flow regimes. However, where the main branch (and the Little Ausable tributary) are

cut through the Seaforth moraine, the channel gradient steepens considerably and the

availability of cobble and gravel bed materials observably increases. The alluvial floodplain is

then cobble‐dominated with a thin layer of vertically accreted fine‐grained materials (Reach

B2).

Continuing downstream, the Ausable River main branch enters a relatively flat‐lying till plain surface along the lee‐side front of the Wyoming moraine. Although patchy areas of outwash

may contribute local sand and gravel, the alluvial floodplain is dominantly fine grained (Reaches

CM3 to CM5) and the channel gradient is relatively low (< 0.0003). Poor drainage, annual

flooding, and forested wetlands (e.g., Hay Swamp) are characteristic of this section of the

Ausable River, including associated accumulations of organic matter in the floodplain (Reach

CM4‐Org). Downstream the main branch begins to steepen and transitions into a proglacial lacustrine basin, receiving greater sources of gravel from tributaries such as the Little Ausable

River. The river gradually becomes more deeply incised into the lake plain, thus also increasing

the availability of alluvial sands (Reach SM6).

Within the small proglacial basin and lacustrine deposits along the front of the Wyoming moraine, the Ausable River again has a relatively low gradient (0.0002–0.0004). With increased

132

sand availability from the glaciolacustrine sediments, the S‐type floodplain is highly sinuous

(Reach S7). However, field evidence of lateral activity is sporadic and the width‐to‐depth ratio

is extremely low (w/d < 6). It is unclear as to whether the specific channel morphology of Reach

S7 should be considered entrenched or if long‐term cumulative effects of backwater

sedimentation behind ice and log jams within the reach have resulted in exceptionally high

alluvial banks. Downstream of Reach S7 where the Ausable River makes more frequent contact

with the till and coarse sediments of the Wyoming moraine, the channel steepens with gravel

and cobble supplies influencing the otherwise sand dominated floodplain (Reach SM8).

Through the Wyoming moraine, the Ausable River is sharply cut 40–50 metres below the crest

of the moraine. Although the moraine in this area has been described as a subaqueous

boulder‐clay till deposit (Chapman and Putnam, 1951), the underlying Ordovician shale and limestone bedrock is shallow under portions of the moraine and the Ausable River is cut 10’s of metres into the bedrock. This reach is aptly named the Rock Glen gorge (Reach B9) with

channel gradients as high as 0.005. The main branch and numerous small tributaries continue

to erode weathered bedrock bluffs and knickpoints, thus serving as a significant supply of

cobble and gravel size material to downstream reaches. Over a short 4 km length, the incised

reaches through the moraine transition from B, BM, and SM floodplain types, with reaches

downstream of the gorge continuing to transport the steady supply of gravel.

The lower reaches of the Ausable River are classified as SM‐type floodplains (Reaches SM10 and

SM11), with considerable availability of both sand and gravel. However, the channels in these reaches also have a relatively unique morphology inherited from the Nipissing lake transgression and from the exceptional supply of gravel from upstream. More specifically,

133

these reaches appear to be entrenched into the mid‐Holocene lake deposits of the Thedford

Embayment (Reach SM11) and into the considerably thick backwater deposits which were

translated upstream into the narrow valley that runs through the Wyoming moraine (Reach

SM10). The channels are deep and narrow (w/d = 5–10) with limited hydraulic access to the elevated floodplain surface on an annual basis, particularly in Reach SM10 (except perhaps during backwater associated with ice and/or log jams). Late Holocene annual flooding events in

Reach SM11 have likely become even less frequent following construction of “the cut”, a flood

relief channel creating a short‐cut to Lake Huron built in the 1800s (Reach Cut‐12).

Saugeen River

The Saugeen River catchment drains a variable patchwork of till plains, drumlin fields, kame moraine deposits, and recessional moraines (Figure 4.6). Eastern headwaters originate in the

poorly drained Dundalk till plain with gently sloping till surfaces and open, cedar, or forest

wetlands (e.g., Reaches CM1 and CM2‐Org). Headwaters in the northern and southern parts of

the catchment drain morainic hillslopes, but low‐gradient cedar forest wetlands are also

common in the upper reaches.

Flowing from east to west, the upper reaches of the main branch (and South Saugeen River)

drain through the variable landscape of the Saugeen Kame moraines. Channel bed materials

are dominated by cobble and gravel sourced from the moraines and outwash‐valley fills.

Channel gradients and stream powers are high (S = 0.003–0.006; ω = 80–100 Wm‐2) while sand availability is relatively low resulting in dominantly B‐type alluvial floodplains.

134

Figure 4.6: Reach classification for the Saugeen River and select tributaries. (A) Reach mapping overlaid with glacial landforms of Chapman and Putnam (2007). Reach numbers for main branch include prefix for floodplain classifications as outlined in Table 4.2, and based on results in Chapter 3. (B) River profile and reach classifications, with post‐glacial water surface history of Lake Huron basin relative to modern lake elevation (176 m asl; Figure 4.2). Modern river profiles were extracted from the modern‐DEM (Chapter 2) and upland profiles were collected for this study. Modeled profile surface prior to isostatic rebound is for 10.6 ka BP as per results of Lewis et al. (2005) and as per paleo‐DEM analysis described in this paper.

135

Localized sources of sand contribute to thicker floodplain deposits (e.g., BM floodplain types),

but these are discontinuous and irregular (note overall classification of Reach B3‐BM). Some

sections may also contact shallow bedrock, as is particularly the case for the Rocky Saugeen branch which is semi‐alluvial and bedrock controlled through much of its length.

Where the main branch and South Saugeen come together, gradients are slightly lower and increasing sand availability results in the formation of M and SM floodplain types. However, there is limited field evidence of modern lateral accretion throughout much of Reach M4 due to natural fluvial processes being obscured by a series of low‐head dams between the towns of

Hanover and Walkerton. Reaches M4 and BM5 flow through the proglacial outwash deposits associated with Singhampton moraine, ultimately cutting through the moraine to almost 50 m below the crest of the moraine. Reach 5 is dominated by BM‐type floodplain characteristics,

however occasional impingement against glacial bluffs locally constrain the channel and there is

an abundance of large cobbles and boulders presumably sourced from the moraine. The largest glaciolacustrine and till bluff, with an associated cobble rapids, occurs just downstream

of the Singhampton moraine (44° 10' 29" N, 81° 10' 47" W) and functions as source of sand and

gravel to reaches immediately downstream.

Below the Singhampton moraine the Saugeen River actively meanders through abundant sand and gravel with evidence of M‐type floodplain accretion (Reach M6), but the channel is

somewhat constrained between multiple river terraces. The degree of lateral activity also

declines downstream as the channel gradient decreases with distance from the glacial sources

of gravel. The river valley below the Singhampton moraine is wide, locally over 2 km, and is incised more than 30 m into thick deposits of glaciolacustrine sand. With further reductions in

136

channel gradient downstream and increasing sand supply from within Reach M4, the floodplain

transitions to SM characteristics (e.g., Reach SM7). However, the dominant glacial material also transitions downstream into sources of glaciolacustrine silt and clay. As such, the Saugeen

River floodplain within Reach CM8 can be described as transitional between M and C‐type floodplains with relatively thick alluvial silt‐clay banks and gravel bed materials (also note fully

C‐type floodplain classification for the Deer Creek tributary).

Downstream of its confluences with the Teeswater and North Saugeen rivers, the gradient of the Saugeen River steepens slightly (~0.001) with renewed sources of gravel from the Gibraltar

and Banks moraines. The increased stream power and coarse bedload supply contribute to

dominantly BM‐type floodplains over the last 40 km to Lake Huron (Reaches BM9 to BM11), with multiple river terraces documented with respect to changing baselevels after glaciation

(Garaci, 1998). Reach BM10 is relatively unique given it exhibits local anabranching sections

with semi‐stable islands. While the origin of this channel morphology was not thoroughly

investigated for this study, local supplies of gravel may have contributed to island formation and it is likely that the dynamics of the multi‐thread morphology have diminished in the late

Holocene (c.f. Croil, 2002; possibly influenced by decreasing channel gradients from isostatic

tilting?). A local supply of sand from the Mill Creek tributary (S‐type floodplain) may also

contribute to sustaining a thicker alluvial floodplain in Reach BM11. Before discharging to Lake

Huron, the Saugeen River in Reach B12 is relatively steep (~0.0015) with high stream power

(~100 Wm‐2) and coarse gravel and cobble bed material. The floodplain is locally thin, but is contained within a relatively narrow valley which has been subjected to multiple phases of

aggradation and incision associated with the fluctuating water levels of the Lake Huron basin.

137

Nottawasaga River

The Nottawasaga River catchment drains a diverse landscape of bedrock escarpment, till and

kame moraines, and glaciolacustrine plains (Figure 4.7). Namely, the western tributaries drain

areas of the Niagara escarpment including associated moraines and outwash landforms, with

southern and eastern tributaries draining areas of the Oak Ridges moraine and the Simcoe

uplands, respectively. The central lowlands of the watershed include the Lake Algonquin sand

plain and the wetlands of the Minesing basin. The lower reaches of the Nottawasaga are cut

through the Edenvale moraine and thick aeolian sand‐dune deposits at the head of Georgian

Bay.

In terms of reach classifications, headwater channels of the Nottawasaga river vary somewhat

according to physiographic regions. Specifically the western tributaries start within poorly drained uplands of the Dundalk till plain above the Niagara escarpment in fine‐grained till and

loess, organic‐rich wetlands, and local sources of gravel and cobble materials. The low‐gradient

channels and minimal availability of sands contribute to typical CM‐type floodplains in these

headwaters, dominated by fine‐grained materials with local sources of inherited coarse‐lag

(Reach CM1). Comparatively, tributary headwaters from the southern and eastern parts of the

basin tend to drain subtle morainic hillslopes. With potentially steeper gradients and diverse

source materials compared to the Dundalk till plain, these channels are variable and strongly controlled by vegetation, but still generally fall within the same headwater classification of

CM‐type floodplains.

138

Figure 4.7: Reach classification for the Nottawasaga River and select tributaries. (A) Reach mapping overlaid with glacial landforms of Chapman and Putnam (2007). Reach numbers for main branch include prefix for floodplain classifications as outlined in Table 4.2, and based on results in Chapter 3. (B) River profile and reach classifications, with post‐glacial water surface history of Lake Huron basin relative to modern lake elevation (176 m asl; Figure 4.2). Modern river profiles were extracted from the modern‐DEM (Chapter 2) and upland profiles were collected for this study. Modeled profile surface prior to isostatic rebound is for 10.6 ka BP as per results of Lewis et al. (2005) and as per paleo‐DEM analysis described in this paper.

139

From the western headwaters to the Niagara escarpment, tributaries then drain to semi‐alluvial bedrock reaches of varying extent and gradient, including some abrupt knickpoints (i.e., water‐ falls) and other steep‐channel morphologies including step‐pool channels and boulder cascades. While not expressly B‐type floodplain channels, these high energy semi‐alluvial

bedrock reaches tend to transition to cobble‐dominated alluvial floodplains characterized by

thin overbank deposits (Reach B2). Downstream of the escarpment reaches, the main branch

of the Nottawasaga river is incised into variable morainic materials of the Gibraltar and Oak

Ridges moraines, although much of the incision may be related to glaciofluvial outwash

channels and meltwater during ice retreat. Decreases in channel gradients along the main

branch are met with decreases in bed material grain‐size, with sources of gravel transported

from upstream reaches and derived from local moraine and outwash materials. As such,

alluvial floodplains of the Nottawasaga river transition from BM to M‐type classifications as the

main branch enters the catchment lowlands (Reaches BM3 and M4).

Downstream of its confluence with Innisfil Creek, the Nottawasaga River receives a significant

increase in sediment supply from the tributary and the surrounding glacial Lake Algonquin sand

plain (Reach SM5). The Lake Algonquin plain lies on average about 40–50 m above the modern

elevation of Lake Huron (at 176 m asl). The combined gravel supply from the main branch and

local increases in sand availability contribute to SM‐type floodplain characteristics. As the main

branch transitions deeper into the sand plain and receives additional sediments supplied from

the sand‐dominated lower reaches of the Boyne River, sand‐dominated floodplain characteristics become more prevalent (Reach S6). However, river incision of over 30 m into the Lake Algonquin plain has actually exposed underlying till deposits that contribute significant

140

amounts of cobble and boulder materials locally, particularly within Reach SM7. As such, localized semi‐alluvial conditions and persistent cobble‐boulder rapids tend to complicate the sand‐dominated alluvial floodplain processes for some reaches incised into the Lake Algonquin plain (Reaches S6 and SM7). The semi‐alluvial influences become less significant downstream

and fully sand‐dominated floodplain characteristics are evident in Reach S8.

Entering the Minesing basin, the Nottawasaga River is of significantly low gradient (< 0.0002) with considerable increases in silt, clay, and organic materials available within the basins alluvial and wetland deposits. Even so, significant contributions of sand continue to be delivered from

the Lake Algonquin plain by the main branch and Pine River tributary. The alluvial

characteristics are as such considered to be a CS‐type floodplain, however the special case of the highly organic Minesing swamp requires specific note (Reach CS9‐Org).

The Nottawasaga River outlet from the Minesing basin is confined to a narrow cut through the

Edenvale moraine, incised to at least 30 m below the crest of the moraine. However, there is

less than 3 m of vertical drop between the outlet of the Minesing swamp and modern Lake

Huron water levels (i.e., average slope of less than 0.0001). Despite possible coarse sediment

inputs from the boulder‐clay Edenvale moraine, the floodplain of the lower reaches continues

to be dominated by fine‐grained material (Reach CS10). But with increasing sand supply from

tributaries and isolated valley bluffs associated with thick lake and dune deposits at the head of

Georgian Bay, the floodplain ultimately transitions to sand‐dominated alluvial deposits (Reach

S12). Downstream of the Edenvale moraine and before the sandy reach of S12, a short reach of

SM‐type floodplain is evident (SM11), perhaps due to local increases in gravel supply from

underlying till deposits exposed along the valley walls.

141

Fluctuating water levels in the Lake Huron basin during the Holocene would have had

significant effects on the lower reaches of the Nottawasaga River as far upstream as the

Minesing basin, perhaps further (Figure 4.4C). Following proglacial deposition of the Lake

Algonquin sediments during ice retreat, river incision would have adjusted towards the low‐

water stages in the early Holocene (i.e., Lake Hough in the Georgian Bay basin; Eschman and

Karrow, 1985). Subsequent rising water levels to the mid‐Holocene Nipissing lake stages (7–10

m above modern lake level) would have imposed slack‐water sedimentation in the lower

reaches, with the Edenvale valley and much of the Minesing basin flooded during this period

(Fitzgerald, 1985). This interpretation is further supported by the widespread fine‐grained

alluvial floodplains and low‐gradients of the Nottawasaga River in its lower reaches (e.g., Reach

CS10). Channel gradients and thus sediment transport competence were also decreasing over

the Holocene due to isostatic tilting of the watershed, particularly effecting the lowland reaches of the catchment (e.g., Reaches M4 to S12, Figure 4.3 and Figure 4.7).

4.4.3 Post-glacial fluvial adjustments

The development of late Holocene river landforms in southern Ontario is based on fluvial adjustments that have been complicated by a variety of glacial legacy and post‐glacial environmental factors. Alluvial floodplain types, and thus fluvial process‐landform interactions, are organized by reach‐scale fluvial processes superimposed within the post‐glacial landscape.

As such, some general themes can be extracted from mapping of river landform variations to provide insights into patterns of post‐glacial fluvial adjustment. As summarized in Table 4.4,

factors affecting post‐glacial fluvial adjustment can generally be described in terms of 1) the topographic and sedimentological glacial legacy effects; 2) the landforms resulting from

142

isostatic and lake baselevel change, and 3) the superimposed patterns of Holocene fluvial sediment supply. Although the river profiles are still in disequilibrium and are still responding to glacial legacy effects in the long‐term (see Chapter 1), reach‐scale fluvial landforms may be considered to exhibit a quasi‐equilibrium state in terms of the prevailing fluvial process domains and consequent channel morphologies and floodplain styles. Human impacts may also be an important consideration in some reaches; however, the relative significance of modern process varies spatially and depends on the sensitivity of individual reaches to landuse change.

Table 4.4: Summary of general landform relationships for glacially conditioned catchments. General Theme Landforms Figures 4.5, 4.6, and 4.7 1. Glacial Legacy Glacial Landform Source Reach Examples Coarse alluvium Till and kame moraines Ausable B2, B9 High gradient Local bedrock control Saugeen B3(BM), B12 and BM5, 9–11 Limited sand supply Glaciofluvial outwash Nottawasaga B2, BM3 High sand supply Glaciolacustrine sand plain Ausable S7, SM8, SM10–11 Low gradient (with fluvial incision) Saugeen SM7 Nottawasaga S6, SM7, S8 Silt‐Clay Supply Fine‐grained till plain Ausable CM1, CM3–5 Low gradient Glaciolacustrine clay plain Saugeen CM1–2, CM8 (with fluvial incision) Nottawasaga CS9–10 2. Isostatic and Lake Baselevel Resulting Landforms Reach Examples Baselevel variationsa Entrenched valleys Saugeen Reaches 6–12 River terraces (aggradation, incision) Nipissing transgression Low gradient lake plain Ausable SM10–11 (backwater aggradation) Nottawasaga CS9–10 3. Holocene Fluvial Processes Resulting Floodplain Landformsb Reach Examples Gravel supply from upstream M and SM types Ausable SM6, SM10–11 Saugeen M4, M6, SM7 Nottawasaga M4, SM5 Sand supply from upstream S, SM, BM, and CS types Ausable S7, SM8, SM10–11 (some large channel CM types) Saugeen SM7, BM9–11 Nottawasaga Reaches 5–12 Reduced downstream competence S, SM, M, CS, and CM types Ausable CM3, S7 and transport of gravel and cobble Saugeen M4, SM7, CM8 Nottawasaga Reaches 4–6 and 8–9 Increased downstream competence B and BM types Ausable B9 reducing alluvial storage of sand Saugeen BM5, B12 a Baselevel lowering and valley entrenchment also occur on tributaries due to main branch incision. b Alluvial floodplain classifications in Table 4.2. 143

Holocene floodplain ages

A comprehensive discussion of post‐glacial fluvial adjustments would benefit from

geochronological evidence to compare landform interpretations with the depositional ages of

floodplain alluvium. Radiocarbon dating of organic floodplain materials for this study (Table

4.3) provides a partial picture of late Holocene floodplain ages for comparison with interpreted floodplain classifications. In addition to the challenges of locating dateable organic material, even within thick alluvial deposits, another issue in southern Ontario is the difficulty of using radiocarbon methods in thin alluvial deposits (e.g., B‐type floodplains) and in thick fine‐grained

alluvial deposits that are built by slow vertical accretion processes (e.g., C‐type floodplains). In

both cases, accretion events are typically thin veneers of silt and clay which are subsequently

disturbed by seasonal vegetation growth on an annual basis.

On the other hand some sand‐dominated S‐type floodplains have comparatively thick alluvial

floodplains, but may still lack undisturbed or uncontaminated organics due to the gradual

processes of oblique accretion. S‐type floodplains are also often associated with low‐energy reaches where natural bank exposures are rare, further limiting the potential for radiocarbon sampling. The most suitable floodplain accretion process for preserving dateable organic deposits is lateral point‐bar accretion, and thus M and SM‐type floodplain classifications were

the most likely to be sampled for this study (Figure 4.8A).

From the radiocarbon samples collected at M‐type floodplain sites, three general age

categories were identified varying with channel size (i.e., drainage area, Ad) and specific glacial

legacy effects in some cases (Figure 4.8B).

144

Figure 4.8: Radiocarbon results from alluvial floodplain deposits for this study. A) Relative sampling of 14C datable organic material between four primary floodplain types (Chapter 3) as discriminated by specific stream power (ω) and floodplain sand equivalent (FSE). The shaded area represents the ω‐FSE domain of M‐type floodplains. B) General age categories of floodplain ages for sampled M‐type floodplains varying with channel size indexed by drainage area (Ad). Site numbers are referenced in Table 4.3 and conventional radiocarbon ages BP are uncalibrated 14C dates.

Laterally active M‐type floodplains with the highest alluvial turnover rates and ages less than

1000 14C years BP were most common in the small to intermediate sized channels with drainage areas of 100–500 km2. Intermediate to large channels with drainage areas of 500 to greater than 1000 km2 were more likely to be associated with sampled floodplain ages of about 1000–

3000 14C years BP. However, it is expected that these moderately active channels with M‐type

floodplains have a larger range of floodplain ages compared to the smaller channels, but a

floodplain turnover rate of a few thousand years is perhaps a reasonable average.

The remaining floodplain ages in Figure 4.8B can be explained by other specific environmental

conditions, where for example site 1) on the Ausable River is associated with floodplain

deposits during the Nipissing transgression just over 5000 14C years BP. Further, site 3) on the

Nottawasaga River and site 7) on Medway Creek (Thames River tributary) are associated with

145

semi‐alluvial conditions which complicate floodplain accretion processes in terms of channel

incision into glacial till.

Assuming that feasible alternatives exist for all floodplain types in southern Ontario, further

work is required to investigate geochronological evidence and viable age sampling methods on

thin and fine‐grained alluvial floodplain types in southern Ontario (B, C, and S types).

Preliminary testing of one potential method from this study is based on luminescence dating

(Appendix B).

4.5 Conclusions

River landforms in southern Ontario are strongly controlled by the available sediment grain sizes from local glacial landforms and from fluvial transport of sediments supplied by upstream reaches and tributaries. However with increasing distance from sediment sources, alluvial characteristics may also be moderated downstream by changes in competence (i.e., inherited channel gradients) and by the relative contribution of differing sediment types downstream

(e.g., change from moraine to glaciolacustrine landforms can significantly increase the availability of fine‐grained alluvial material in the floodplains). Channel gradient and stream

power are derived from the inherited signatures of the glacial topography, post‐glacial patterns

of fluvial incision, and the calibre of the available bed materials (where coarser alluvial

materials maintain steeper gradients over the course of incision).

In the sense of lithotopo units as per Montgomery (1999), the spatial organization of fluvial

process domains is strongly controlled by the glacial landforms and sediments. Reach‐scale

mapping of river landforms provides an interpretive framework to explain post‐glacial fluvial

adjustments in the context of local sources of glacial sediment and patterns of fluvial sediment 146

flux over the Holocene. Although the spatial patterns of river adjustment, channel morphology, and floodplain style are complex, a number of general relationships are useful to account for the arrangement of river landforms at the catchment‐scale. These glacial–fluvial landform relationships can broadly be clustered into three themes: 1) topographic and sedimentological

glacial legacy effects; 2) landforms resulting from isostatic and lake baselevel change, and

3) superimposed patterns of Holocene fluvial sediment supply.

The spatial pattern of channel morphologies and floodplain types may be further understood in terms of the timing of post‐glacial fluvial adjustment and late Holocene floodplain development. However, geochronological evidence to interpret floodplain ages and alluvial

turnover rates is limited by the availability of datable organic material. Radiocarbon dating for

this study was practically restricted to M‐type floodplains, and samples were typically collected

within well‐developed lateral point‐bar accretion deposits. The floodplain ages generally vary

with channel size or specific glacial legacy factors, with small to intermediate channels (Ad =

100–500 km2) being the most laterally active in M‐type floodplains. Thin B‐type floodplains and

thick C‐type floodplains are generally not suitable for radiocarbon methods due to the

dominance of fine‐grained vertical accretion processes and yearly bioturbation issues. Optically

stimulated luminescence may be one potential method to improve Holocene floodplain

geochronology in southern Ontario (Appendix B), but OSL methods in this context also have

limitations which require further investigation.

147

Chapter 5 Conclusions 5 Introduction: “The Beginning of the End” Are rivers of southern Ontario unique? Each river network is carved into a relatively distinct assemblage of glacial landforms, with Holocene precipitation draining to the rise and fall of

Great Lake’s water levels after ice retreat. While environmental diversity may complicate fluvial landform development in many environments, the variation and architecture of the

glacial palimpsest in this landscape is particularly vast and disorganized with respect to fluvial

discontinuities. However, countless case studies of individual river histories are only useful in

broader scientific theory to the extent that they contribute to the bigger picture. As such, the

rationale for this research arises from the need to adapt existing ideas about glacial legacy

effects to the low‐relief physiography and river systems of the Laurentian Great Lakes region —

an environment which has received little systematic attention with respect to fluvial landscape

evolution.

As a general terminology, glacial conditioning of fluvial systems represents the lasting

impression of late Pleistocene glaciation, including the paraglacial legacy of sediment

availability and the inherited glacial signatures in river profiles relative to a theoretical self‐

organized equilibrium state. The concept of glacial conditioning (and the degree to which

channels are inherited landforms) is also essentially the idea that the post‐glacial timescales for

fluvial adjustment are spatially variable. Rivers may not completely adjust to the most dramatic

fluvial discontinuities imposed by glacial landforms (and sediments) over typical interglacial

time periods (Brardinoni and Hassan, 2006), but less persistent features may be fully reworked

over periods of a few millennium or less. Spatial scale is also important in that reach‐scale fluvial process‐landform equilibriums (or quasi‐equilibriums) can in theory be nested within the

long‐term disequilibrium state of a river profile. 148

5.1 Thesis Questions

With respect to the low‐relief glacially conditioned river catchments of the southern Laurentian

Great Lakes region, the following thesis questions were proposed.

5.1.1 Thesis question #1: Glacial signatures and river profiles

To what degree are glacial signatures imposed on the longitudinal profiles and stream power distributions of river systems in this low‐relief landscape?

This thesis question is largely addressed in Chapter 2, in which glacial signatures are

documented in river profiles for 22 catchments of southern Ontario. To identify topographic glacial legacy effects, a number of analysis methods were tested including stream power mapping (e.g, Vocal Ferencevic and Ashmore, 2012), stream length–gradient index methods

(SL/K index; e.g., Pérez‐Peña et al., 2009), and slope–area analysis (Montgomery and Foufoula‐

Georgiou, 1993; Brardinoni and Hassan, 2006).

While visual patterns can be recognized from mapping stream power in the context of the

glacial landforms (Chapman and Putnam, 2007), slope–area analysis is a graphical means of

stratifying rivers with different associated glacial landform classifications. This analysis successfully discriminates rivers incised into moraines from those situated within glaciolacustrine and till plains. Slope–area analysis is essentially a stream power domain method, with results from this study matching classic discrimination curves for predicting river

planform types (Leopold and Wolman, 1954). Confirmation of the distinct topographic glacial legacy effects of moraine versus plain landforms can also be demonstrated statistically using

ANOVA difference of means testing for river slope and stream power (Figure 5.1). Although a

theoretical Hack‐type exponential profile model may be somewhat arbitrary (Phillips et al.,

149

2010), the SL/K index provides a relative measure of glacial signatures in terms of the

oversteepening (SL/K = 5–20+) and understeepening (SL/K < 3–5) of river profiles, and thus can

be considered a measure of the degree of topographic glacial conditioning (Figure 5.1C).

Figure 5.1: Distinct glacial legacy signatures of moraines and lake or till plains for rivers in southern Ontario. A) Box‐plots of glacial landform associations and river slope. Note for α = 0.05, the ANOVA difference of the slope means is statistically significant between plains and moraines (p < 0.0001). B) Box‐plots of glacial landform associations and river stream power. Note for α = 0.05, the difference of the stream power means is statistically significant between plains and moraines (p < 0.0001). Stream powers associated with baselevel entrenchment due to lake level regression are statistically significant from all glacial landform effects, but lake level effects are restricted to the lower river reaches where stream powers tend to be higher. C) Generalized model of SL/K profile analysis to identify relative degree of glacial landform influence between moraine and lake/till plains. Anomalies in river profiles (solid lines) are evident relative to a theoretical Hack‐type exponential standard (dashed lines), identifying river reaches that are oversteepened (SL/K curve above Hack line) from those that are understeepened (SL/K curve below Hack line).

150

5.1.2 Thesis question #2: Alluvial floodplain classifications

Are natural groupings of channel morphology, alluvial landforms, and fluvial process domains in this environment consistent with previous river and floodplain classifications from other environments, and in view of scientific parsimony which environmental variables are most effective at explaining and predicting distinct classes?

Four first‐order alluvial floodplain classifications were identified and tested in Chapter 3 based

on a dataset of 109 floodplains sampled in southern Ontario. Field classifications were

confirmed and refined based on natural groupings from k‐means clustering analysis and

predictive variables were explored using principal component and discriminant analysis

methods. The four first‐order classifications are summarized in Table 5.1, along with four

transitional floodplain classes as explored in the discriminant analysis. Floodplain types were

successfully predicted by two principal explanatory dimensions: (1) stream power‐resistance,

which is effectively represented by specific stream power (ω); and (2) floodplain sedimentology, which is effectively represented by floodplain sand equivalent (FSE).

Given that the first principal component is primarily composed of hydraulic factors relating to

the channel (and noting that the hydrologic variables of channel size were removed), this

analysis could also be considered a channel classification. The semantics of the proposed floodplain terminology is intended to extend the natural groupings conceptually to reflect the

sedimentary record of fluvial processes. Even basic sedimentological and stratigraphic

characteristics of the floodplain (e.g., percent sand and floodplain thickness) can represent the

integrated product of three‐dimensional channel morphology in terms of fluvial sediment

transport and alluvial accretion process in the channel. Further, differences in the basic

151

sedimentary architecture of floodplains can account for variations in the style of alluvial record

keeping, which is important for future studies of paleoenvironmental reconstruction.

The proposed floodplain types as natural groupings in the fluvial landscape are generally

consistent with previous single‐thread channel classifications in that they reflect the basic cross‐sectional and hydraulic differences between wide‐and‐shallow channels (e.g., B‐type floodplains, or Rosgen C3 channels) and narrow‐and‐deep channels (e.g., S/C‐type floodplains,

or Rosgen E5,6 channels). Further, M‐type floodplains represent the archetype single‐thread

channel style in fluvial geomorphology of mixed gravel‐bed meandering with lateral point‐bar

accretion deposits (or B3 floodplains according to Nanson and Croke, 1992). However despite

the similarities, the terminology of the proposed floodplain classification allows for a more representative treatment of sedimentological glacial legacy effects in the landscape. In

particular, river classification in southern Ontario must reflect the inheritance of cobble bed

material and the relative availability of sand material, which both vary spatially according to

glacial landforms.

A parsimonious set of explanatory factors were explored to predict the proposed floodplain

types. The focus on stream power is similar to that of Nanson and Croke (1992), where ω

effectively represents the hydraulic factors of the first principal component. Again, removing consideration of hydrology which dictates channel size, the sedimentological characteristics

represent the second most important factor in explaining floodplain variability. A new variable

defined as the floodplain sand equivalent (FSE) is proposed, which effectively summarizes the

floodplain sedimentology. A third factor has also been considered which represents the

variability of sediment transport (e.g., ω/D50), and can be described statistically as the residual

152

variance of the power‐resistance correlation from the first principal component. Although

sediment transport variability does little to discriminate the four first‐order floodplain types, it

is expected that variables such as ω/D50 might be used to explore some within‐class variability

in channel morphology and fluvial processes.

As such, rivers in southern Ontario can be efficiently classified and predicted based on models

of ω, FSE, and ω/D50. However, these first‐order classifications only represent the most

fundamental differences in channel hydraulics and alluvial floodplain characteristics at the landscape scale. Further variability has been initially explored in terms of transitional floodplain types (Table 5.1), but additional research is required to investigate geomorphic diversity in terms of sediment transport, channel entrenchment, and semi‐alluvial boundary conditions.

Table 5.1: Summary of floodplain classifications for southern Ontario rivers.

Floodplain Stream Power Dominant Width‐to‐ Floodplain Other Main Identifying Features Type ω (Wm‐2) Material Depth Ratio Sand FSEa (m) B‐type 60–100+ cobble 20–50 0.1–0.2 Thin alluvial floodplain Cobble‐lag and vertical accretion

BM 30–90 15–30 0.2–1.0  

M‐type 25–40 gravel 10–15 1.0–1.5 Point‐bars and cut‐banks Lateral accretion

SM 10–30 < 10–15 1.5–2.0  

S‐type 10–15 sand < 10 2.0–3.0 Thick sand‐dominated floodplain Oblique accretion and sand bars

CS 5–15 < 10 0.7–2.5  

C‐type < 5–10 silt‐clay << 10 0.2–0.7 Thick silt/clay‐dominated floodplain Vertical accretion, vegetation control

CM 10–30 mixed variable 0.2–1.0 Relatively low sand availability (>1.0)b Vegetation control (headwaters) a FSE is defined as the equivalent thickness of floodplain sand (see definition Chapter 3). b FSE > 1.0 m possible for CM floodplains associated with large channels (e.g., width > 10–50 m). 153

5.1.3 Thesis question #3: Glacial legacy and landform organization

How are the spatial arrangements of river landforms and process domains spatially organized in the context of glacial landforms and post‐glacial landscape histories, and how have the fluvial systems responded to the glacial legacy over the Holocene?

Chapter 4 investigates how reach‐scale floodplain classifications can be used to interpret post‐

glacial fluvial landscape development. Applying Montgomery’s (1999) concept of lithotopo

units to document glacial legacy effects in terms of inherited sediments and slopes, a number

of general landform relationships account for the spatial arrangement of river landforms (Table

5.2). These glacial–fluvial landform relationships can broadly be clustered into three themes: 1)

topographic and sedimentological glacial legacy effects; 2) landforms resulting from isostatic

and lake baselevel change, and 3) superimposed patterns of Holocene fluvial sediment supply.

Returning to the idea that the post‐glacial timescales for fluvial adjustment are spatially

variable, river systems in southern Ontario contain varying proportions of glacial and fluvial

signatures. These fluvial systems have thus responded over the Holocene by developing quasi‐ equilibrium landforms governed by both glacial inheritance and patterns of sediment transport within the drainage network (also see Thayer, 2012). While human impacts on modern river landforms are evident in southern Ontario, particularly in some reaches, the glacial features

persist and clearly remain dominant environmental controls in the fluvial landscape.

Geochronological evidence is important for paleoenvironmental reconstruction of post‐glacial river history, temporal and spatial patterns of river incision, and rates of alluvial floodplain

turnover. However, studies of the geomorphology and history of fluvial environments in the

Great Lakes watershed have been relatively few and site specific (e.g., Karrow, 1984; Croil,

2002; Kiesel and Mickelson, 2005; Karrow et al., 2007; Arbogast et al., 2008; Thornbush and

154

Desloges, 2011; Stewart and Desloges, 2013), with radiocarbon dating being the primary technique for constraining post‐glacial river history in the region.

Radiocarbon techniques rely on locating and sampling datable organic matter within an appropriate alluvial context, also assuming that it has not been contaminated by old or new carbon since uptake and deposition. Attempts in this study to sample organic matter from a wide range of locations and floodplain types suggest that radiocarbon methods are unsuitable or may not be practical for many alluvial deposits in southern Ontario. As such, paleoenvironmental research on the full range of fluvial environments in the region will require development of additional geochronological methods (e.g., OSL dating, Appendix B).

Table 5.2: Glacial–fluvial landform relationships for southern Ontario river catchments.

Landscape Feature Inherited Features Holocene Fluvial Processes Floodplain Landformsc Morainesa Incised valleys Minor incision B and BM type Steeper gradients (>0.001) Gravel to downstream (to M and SM types Cobble, gravel, minor sand downstream) Outwash deposits Spillway channels/valleys Similar to moraines Similar to moraines Variable gradients but variable but variable Cobble, gravel, minor sand Till plains Variable gradients Minor incision CM and BM types Silt‐clay source material Concentration of coarse lag Silt‐clay storage in floodplains Glaciolacustrine Low gradients (<0.001) Moderate incision C,CS, and CM types Clay plains Silt‐clay source material Silt‐clay storage in floodplains Silt to downstream Glaciolacustrine Low gradients (<0.001) Incision S and SM types Sand plains Sand‐silt source material Sand‐silt storage in floodplains (to SM, CS, CM, and BM Sand‐silt to downstream types downstream) Baselevel regressionb Incised valleys Incision Variable based on other River terraces Channel entrenchment landscape features Steeper gradients Nipissing transgression Low gradients Late Holocene channel SM and CS types Sand‐silt‐clay alluvium entrenchment in some cases (may override other landscape features) a Legacy effects of exposed bedrock channels are similar to moraines. b Baselevel lowering and valley entrenchment also occur on tributaries due to main branch incision. c Alluvial floodplain classifications in Table 5.1.

155

5.2 Adapted Fluvial Landscape Model

An idealized river system is conventionally envisioned with a smooth concave‐up profile, where

gradual downstream increases in drainage area and discharge are matched by complimentary

decreases in river gradient and bed material grain‐size. Whether interrupted by tectonics,

geological heterogeneity, baselevel history, or climate change, few fluvial landscapes fully

adhere to this standard. Geomorphological diversity in fluvial landscapes is thus a matter of scientific inquiry in most environments.

The low‐relief physiography and glacial geology of the southern Laurentian Great Lakes are distinct. Within North America, the river catchments of this region are genuinely different from those in mountainous environments around the continent. At the source, these river catchments lack steep headwaters and are not coupled with hillslope sediment production. At

the outlets, the river channels have experienced dramatic changes in baselevel over a relatively

short geologic timescale associated with post‐glacial isostatic and hydrological adjustments.

However, recent research on glacial legacy effects has largely focused on mountainous landscapes (e.g., Ballantyne, 2002; Brardinoni and Hassan, 2006; Collins and Montgomery,

2011) and no previous effort has been made to systematically generalize fluvial landscape

development in low‐relief glacially conditioned catchments.

This thesis provides a basis to explore an adapted fluvial landscape model for low‐relief glacially

conditioned rivers (Figure 5.2A). At the core is the idea that sediment cascades in these river

catchments are truncated. Headwaters are not necessarily steep or the dominant source of

alluvial material and sea‐level is not baselevel, at least over the Holocene timescale.

156

Figure 5.2: A) Adapted fluvial landscape model for low‐relief glacially conditioned river catchments of southern Ontario, this study (dashed curves represent theoretical equilibrium graded profile surface for mountainous catchments). B) River profile models for southern Ontario divided into three primary segments: proglacial lake lowlands, Horseshoe moraines, and upland drainages. C) Representative profile model examples for southern Ontario rivers.

Instead the glacial landforms and sediments largely govern patterns of fluvial sediment supply.

Lake level fluctuations have reworked river outlet deposits, followed by a gradual lowering of water levels most recently over the late Holocene. Thus, a general understanding of the associations between glacial and river landforms provides a starting point to explain river geomorphology. The glacial–fluvial landform relationships developed from this study are presented in Table 5.2.

However, these quasi‐equilibrium river landforms are superimposed on long‐term disequilibrium trends in river profile adjustment. The theory is that river profiles will gradually

adjust toward a concave‐up graded form, regardless of whether or not it can ever or will ever

be achieved. As such, in documenting the diversity of river profiles in the region, it is useful to

identify recurring forms which indicate common downstream patterns in river slope, and

perhaps river landform sequences. 157

A family of generalized river profile models is presented in Figure 5.2B, including the names of

example river catchments from southern Ontario. Each model profile is split into lowland and

upland segments, divided in most cases by the series of “Horseshoe moraines” which surround

the central uplands of peninsular southern Ontario (i.e., the “Ontario Island” uplands of

Chapman and Putnam, 1951). Most rivers within the study area can thus be organized by

matching characteristic upland and lowland profile segments. For example, both the Maitland

and Bayfield rivers exhibit abrupt baselevel entrenchment (L3) upstream of Lake Huron and

drain through moderate gradients associated with till plains and outwash channels (U3)

upstream of the Wyoming moraine (e.g., L3‐U3 in table for Figure 5.2C). Given the

considerable geomorphic diversity in the landscape, some river profiles remain relatively

unique (e.g., Saugeen River, L2‐U2). A few river profiles in the region might be considered

graded (e.g., East Duffins, L0xU0) and rivers on the north shore of Lake Ontario lack the central

dominant Horseshoe Moraine (e.g., Oak Ridges and Credit types, Figure 5.2C).

Rivers in southern Ontario are geomorphologically diverse, but some generalized relationships

are possible to improve our knowledge of fundamental fluvial process‐landform interactions in

the context of complex glacial legacy effects. The results of this thesis provide some foundation for an adapted fluvial landscape model which begins to reflect the geomorphological diversity of rivers in low‐relief glacially conditioned landscapes. The conceptual model presented in

Figure 5.2A — including low‐energy headwaters, glacial landform controls, and dynamic

baselevels — is expected to be relevant to other areas of the southern Great Lakes region. The

general conceptual ideas presented may also be relevant to low‐relief sub‐catchments beyond

the foothills of mountainous drainage basins in other previously glaciated regions.

158

5.3 Research Significance

Fluvial geomorphology is well situated as an interdisciplinary river science. Established theoretical frameworks from mountainous river catchments and non‐glaciated environments can inform fundamental fluvial process‐landform relationships. Specifically, concepts of

equilibrium and self‐organization are important themes in geomorphology, especially for

understanding “disturbed” landscapes (Slaymaker, 2009). A major contention of this thesis is

that existing ideas about glacial legacy effects on river systems must be adapted to reflect the

distinct low‐relief physiography of the Laurentian Great Lake region.

An adapted fluvial landscape model and a set of generalized glacial–fluvial landform

associations are presented as a basis from which to further investigate geomorphological

diversity within the context of complex glacial legacy effects in low‐relief settings. Particularly

with respect to documenting variations in alluvial architecture and sedimentary accretion process, distinct floodplain types provide a first‐order framework to group fluvial process domains in glacially conditioned river systems. For example, the geometry and proportion of lateral versus vertical accretion deposits may vary according to the relative availability of sand

and increasing stream power (between S, M, and B‐type floodplains, respectively). On the

other hand, low‐energy channels with silt‐clay boundaries (C‐type floodplains) are expected to

be dominated by overbank vertical accretion processes which are strongly controlled by

vegetation and the backwater effects of ice and woody debris jams. Additional research is

needed to investigate the detailed accretionary processes associated with each of the

floodplain types proposed in this thesis. Building on the classic facies model for meandering

159

mixed‐bed channels (e.g., Miall 1985, 2010), such research will contribute to the interpretation

of modern and ancient fluvial sedimentary records for single‐channel rivers.

In summary, the landform relationships presented in this thesis will be relevant to other low‐ relief glacially conditioned landscapes beyond southern Ontario, contributing to the broader

discourse regarding the effect of past glaciations on river systems. The proposed conceptual

framework also has the potential to enhance interdisciplinary and applied science in the areas

of biogeochemistry, geoarcheology, conservation ecology, and environmental water resources

management.

5.3.1 Interdisciplinary significance

Biogeochemistry

Understanding spatial patterns of floodplain sedimentological and stratigraphic architecture

can provide a working template on which to interpret other natural and physical sciences. For

example, at the relevant timescales of hydrological and biogeochemical processes, the alluvial

floodplain is essentially a hydrogeological template, which controls the patterns of runoff,

infiltration, groundwater movement, and biological activity. In this case, the texture,

distribution, and geometry of the floodplain sedimentary architecture will significantly

influence the patterns and rates of hydrological and biogeochemical processes (e.g., Devito et

al., 2000; Vidon and Hill, 2006). These previous studies suggest that the thickness of permeable

floodplain sediments (e.g., sands), the groundwater flow path(s), and the location of organic‐

rich subsurface deposits (e.g., abandoned channels, buried organic debris) are of significant

importance to the denitrification potential of riparian zones (i.e., floodplains).

160

Geoarcheology

Paleoenvironmental reconstruction of floodplains is an important component of Quaternary

science, including its contributions to geoarcheology. Floodplains in the southern Great Lakes

region are seen as important landscape features for early shifts to horticulture by ancient

indigenous societies (Walker et al., 1997; Ellis et al., In Preparation). As such, knowledge of

floodplain development can help to elucidate the local scale morphology‐physiography of the past river environment and the preservation potential of archeological evidence in what are referred to as “deeply buried” cultural sites in geoarcheology (Stewart et al., 2008).

Conservation ecology

Research on the spatial patterns of river processes and landforms can also inform disciplines such as fluvial landscape ecology, with channels and floodplains representing important corridors of aquatic and riparian habitat (Pool, 2002; Ward et al., 2002; Dollar et al., 2007; Lóczy et al., 2012), including important theories such as the river continuum concept of Vannote et al.

(1980) and the hyporheic corridor concept of Stanford and Ward (1993). Of growing interest in

southern Ontario is interdisciplinary research to investigate critical habitat criteria for sensitive

and endangered species at risk. For example, variations in sediment supply and hydraulic

factors of sediment transport are important to aquatic ecologists studying critical habitat for endangered freshwater mussels such as Snuffbox (Epioblasma triquetra) on the lower Ausable

River (Di Miao and Corkum, 1995; Wilkins et al., 2009; Upsdell et al., 2010; Appendix C.1).

Work has also begun recently by the Canadian Wildlife Service to investigate the spatial

distribution of Bank Swallow (Riparia riparia) in Ontario, a songbird species in decline that

commonly nests in sandy river banks (Cadman and Lebrun‐Southcott, 2013). This project is

161

using stream power mapping and surface geology (as per Chapter 2; Phillips and Desloges,

2014) to select river reaches for Bank Swallow monitoring. Understanding the spatial

relationships of fluvial landforms and processes is important in conservation ecology,

particularly given that the geomorphological diversity of southern Ontario rivers is so strongly

governed by the glacial features in the landscape. For example, it is important to predict the locations of eroding banks in thick alluvial deposits for Banks Swallows or the availability and stability of gravel bed material for other aquatic fish species (e.g., Redside Dace, Clinostomus elongatus; Parish Geomorphic Ltd., 2004; OMNR, 2009; Aquafor Beech Ltd. [ABL], 2010).

Environmental water resources management

River research in southern Ontario has important implications for environmental water

resources management, and specifically stream restoration, river engineering, and land use

planning. Stream restoration initiatives are very active in southern Ontario to protect private property and infrastructure; to provide storm water and local ecological services for new developments; and/or to enhance degraded aquatic and riparian habitat due to the historical effects of agricultural practices or urban development (Appendix C.2). Erosion mitigation

based on standard river restoration practices often ignores issues of sediment supply in the

long‐term sustainability of channel morphology and bank stability (Appendix C.3). In municipal planning, assessment and mitigation of long‐term river hazards based on designated erodible corridors (also known as “meander‐belts”) is a practice of growing importance which relies on assessment of past and future river behaviour (Piégay et al., 2005; Appendix C.4).

New storm water management practices may increasingly be required for local applications in

floodplains to manage road runoff at bridge and culvert crossings. For example, bioretention

162

units are a new storm water management technology in a growing suite of what are termed

‘Low Impact Development’ practices (LIDs) (Bradford and Denich, 2007; ABL, 2010). Vegetated

bioretention units consist of landscaped depressions that are strategically engineered with

subsurface media (e.g., engineered sand and organic material) to provide a designated pathway for infiltration and biogeochemical processing of pollutants and contaminants (which effectively treats urban runoff). The potential use of bioretention units or other LID technologies in floodplain areas may be necessary in some cases at road crossings of greater environmental sensitivity (e.g., OMR‐ESA regulations for Redside Dace; ABL, 2010). Floodplain sedimentology

and stratigraphy are important design components in assessing infiltration capacities and shallow groundwater flow paths for such initiatives.

5.3.2 Contribution to applied geoscience

Fluvial geomorphologists explore the natural science of rivers. This thesis has been equally motivated by aspirations to inform applied geoscience, particularly with respect to fluvial geomorphology and its relevance in environmental water resources management and

conservation ecology (see above Section 5.3.1). It is hoped that professional practice in fluvial

geomorphology will benefit from the development of a more consistent framework to explain

spatial patterns of river and stream morphology in low‐relief glacially conditioned settings.

This research has specifically focused on southern Ontario.

Natural science and applied geoscience in fluvial geomorphology can contribute new perspectives to many interdisciplinary initiatives in environmental research, management, and planning. This thesis is offered as a contribution to our understanding of the ways of rivers.

163

References

Abdi, H., Williams, L.J. 2010. Principal component analysis. WIREs Computational Statistics, 2: 433–459. DOI: 10.1002/wics.101

Anderson, R.J., Bledsoe, B.P., Hession, W.C. 2004. Width of streams and rivers in response to vegetation, bank material, and other factors. Journal of the American Water Resources Association, 40: 1159–1172. DOI: 10.1111/j.1752‐1688.2004.tb01576.x

Anderson, T.W., Lewis, C.F.M. 1985. Postglacial water‐level history of the Great Lake Ontario basin. In: Karrow, P.F., Calkin, P.E. (eds.), Quaternary Evolution of the Great Lakes. Geological Association of Canada Special Paper 30, St. John’s, Newfoundland: 231–253.

Annable, W.K. 1996a. Database of morphologic characteristics of watercourses in southern Ontario. Ontario Ministry of Natural Resources, Peterborough, Ontario.

Annable, W.K. 1996b. Morphological relationships of rural water courses in southwestern ontario and selected field methods in fluvial geomorphology. Ontario Ministry of Natural Resources, Peterborough, Ontario.

Annable, W.K., Lounder, V.G., Watson, C.C. 2011. Estimating channel‐forming discharge in urban watercourses. River Research and Applications, 27: 738–753. DOI: 10.1002/rra.1391

Aquafor Beech Ltd. 2010. Stream restoration for crossings of Fourteen Mile Creek (Upper Middle Road widening) requiring Ministry of Natural Resources Endangered Species Act permitting for Redside Dace, including storm water treatment bioretention facility and overall net benefit habitat compensation works. Prepared for Halton Region. ABL Reference: 64363.

Arbogast, A.F., Bookout, J.R., Schrotenboer, B.R., Lansdale, A., Rust, G.L., Bato, V.A. 2008. Post‐glacial fluvial response and landform development in the upper Muskegon River valley in North‐Central Lower Michigan, U.S.A. Geomorphology, 102: 615–623. DOI: 10.1016/j.geomorph.2008.06.008

164

Ashmore, P., Church, M. 2001. The impact of climate change on rivers and river processes in Canada. Geological Survey of Canada Bulletin 555: Ottawa, Canada.

Bagnold, R.A. 1966. An approach to the sediment transport problem from general physics. United States Geological Survey Professional Paper 422‐I: Washington, DC.

Bagnold, R.A. 1980. An empirical correlation of bedload transport rates in flumes and natural rivers. In: Thorne, C.R., MacArthur, R.C., Bradley, J.B. (eds.), The Physics of Sediment Transport by Wind and Water. American Society of Civil Engineers, New York: 323–345.

Ballantyne, C. 2002. Paraglacial geomorphology. Quaternary Science Reviews, 21: 1935–2017. DOI: http://dx.doi.org/10.1016/S0277‐3791(02)00005‐7

Barker, D.M., Lawler, D.M., Knight, D.W., Morris, D.G., Davies, H.N., Stewart, E.J. 2009. Longitudinal distributions of river flood power: the combined automated flood, elevation and stream power (CAFES) methodology. Earth Surface Processes and Landforms, 34: 280– 290. DOI: 10.1002/esp.1723

Barnett, P.J. 1985. Glacial retreat and lake levels, north‐central Lake Erie Basin, Ontario. In: Karrow, P.F., Calkin, P.E. (eds.), Quaternary Evolution of the Great Lakes. Geological Association of Canada Special Paper 30, St. John’s, Newfoundland: 185–194.

Barnett, P.J., Sharpe, D.R., Russell, H.A.J., Brannand, T.A., Gorrell, G., Kenny, F., Pugin, A. 1998. On the origin of the Oak Ridges moraine. Canadian Journal of Earth Sciences, 35: 1152– 1167. DOI: 10.1139/cjes‐35‐10‐1152

Blöschl, G., Sivapalan, M. 1997. Process controls on regional flood frequency: coefficients of variation and basin scale. Water Resources Research, 33: 2967–2980. DOI: 10.1029/97WR00568

Bradford, A., Denich, C. 2007. Rainwater management to mitigate the effects of development on the urban hydrologic cycle. Journal of Green Building, 2: 37–52. DOI: http://dx.doi.org/10.3992/jgb.2.1.37

165

Brardinoni, F., Hassan, M.A. 2006. Glacial erosion, evolution of river long profiles and the organization of process domains in mountain drainage basins of coastal British Columbia. Journal of Geophysical Research, 111: F01013. DOI: 10.1029/2005JF000358

Brardinoni, F., Hassan, M.A. 2007. Glacially induced organization of channel‐reach morphology in mountain streams. Journal of Geophysical Research, 112: F03013. DOI: 10.1029/2006JF000741

Braun, A., Kuo, C.Y., Shum, C.K., Wu, P., van der Wal, W., Fotopoulos, G. 2008. Glacial isostatic adjustment at the Laurentide Ice Sheet margin: Models and observations in the Great Lakes region. Journal of Geodynamics, 46: 165–173. DOI: 10.1016/j.jog.2008.03.005

Brierley, G.J., Fryirs, K. 2005. Geomorphology and River Management: Applications of the River Styles Framework. Blackwell, Oxford. DOI: 10.1002/9780470751367

Buffington, J.M., Montgomery, D.R. 1997. A systematic analysis of eight decades of incipient motion studies, with special reference to gravel‐bedded rivers. Water Resources Research, 33: 1993–2029. DOI: 10.1029/96WR03190

Buffington, J.M., Roper, B.B., Archer, E., Moyer, C. 2009. Reply to Discussion–“The role of observer variation in determining Rosgen stream types in northeastern Oregon mountain streams” by David L. Rosgen. Journal of the American Water Resources Association, 45: 1298–1312. DOI: 10.1111/j.1752‐1688.2009.00363.x

Burchsted, D., Daniels, M., Wohl, E.E. 2014. Introduction to the special issue on discontinuity of fluvial systems. Geomorphology, 205: 1–4. DOI: 10.1016/j.geomorph.2013.04.004

Burt, T., Allison, R. 2010. Sediment Cascades: An Integrated Approach. John Wiley & Sons, Oxford. DOI: 10.1002/9780470682876

Cadman, M., Lebrun‐Southcott, Z. 2013. Bank Swallow colonies along the Saugeen River, 2009‐ 2013. Ontario Birds, 31: 136–147.

166

Carson, M.A. 1984. The meandering‐braided river threshold: A reappraisal. Journal of Hydrology, 73: 315–334. DOI: 10.1016/0022‐1694(84)90006‐4

Chapman, L.J., Putnam, D.F. 1951. Physiography of Southern Ontario. University of Toronto Press: Toronto, Ontario, Canada.

Chapman, L.J., Putnam, D.F. 2007. Physiography of Southern Ontario [computer file]. Miscellaneous Release‐Data 228. Ontario Geological Survey: Sudbury, Ontario, Canada.

Charlton, R. 2009. Fundamentals of Fluvial Geomorphology. Routledge: London.

Chorley, R.J. 2000. Classics in physical geography revisited. Progress in Physical Geography, 24: 563–578. DOI: 10.1177/030913330002400405

Church, M. 1983. Pattern of instability in a wandering gravel bed channel. In: Collinson, J.D., Lewin, J. (eds.), Modern and Ancient Fluvial Systems. International Association of Sedimentologists Special Publication 6, Blackwell Publishing Ltd, Oxford, UK: 169–180. DOI: 10.1002/9781444303773.ch13

Church, M. 1992. Channel morphology an typology. In: Calow, P., Petts, G.E. (eds.), The River Handbook: Volume 1. Blackwell Scientific Publications, Oxford: 126–143.

Church, M. 2006. Bed material transport and the morphology of alluvial river channels. The Annual Review of Earth and Planetary Science, 34: 325–354. DOI: 10.1146/annurev.earth.33.092203.122721

Church, M., Ryder, J.M. 1972. Paraglacial sedimentation, a consideration of fluvial processes conditioned by glaciation. Geological Society of America Bulletin, 83: 3059–3071. DOI: 10.1130/0016‐7606(1972)83[3059:PSACOF]2.0.CO;2

Church, M., Slaymaker, O. 1989. Disequilibrium of Holocene sediment yield in glaciated British Columbia. Nature, 337: 452–454. DOI: 10.1038/337452a0

167

Clark, J. A., Hendriks, M., Timmermans, T. J., Struck, C., Hilverda, K. J. 1994. Glacial isostatic deformation of the Great‐Lakes region. Geological Society of America Bulletin, 106: 19–31. DOI: 10.1130/0016‐7606(1994)1062.3.CO;2

Collins, B.D., Montgomery, D.R. 2011. The legacy of Pleistocene glaciation and the organization of lowland alluvial process domains in the Puget Sound region. Geomorphology, 126: 174– 185. DOI: 10.1016/j.geomorph.2010.11.002

Croil, J.S. 2002. Channel pattern alternation, anabranch development and floodplain evolution of the lower Grand River, Ontario. Unpublished Masters Thesis, University of Toronto, Toronto, Ontario.

Desloges, J.R., Church, M. 1989. Canadian landform examples; 13, Wandering gravel‐bed rivers. The Canadian geographer, 33: 360–364. DOI: 10.1111/j.1541‐0064.1989.tb00922.x

Devito, K.J., Fitzgerald, D., Hill, A.R., Aravena, R. 2000. Nitrate dynamics in relation to lithology and hydrologic flow path in a river riparian zone. Journal of Environmental Quality, 29: 1075–1084. DOI: 10.2134/jeq2000.00472425002900040007x

Di Maio, J., Corkum, L.D. 1995. Relationship between the spatial distribution of freshwater mussels (Bivalvia: Unionidae) and the hydrological variability of rivers. Canadian Journal of Zoology, 73: 663–671. DOI: 10.1139/z95‐078

Dollar, E.S.J., James, C.S., Rogers, K.H., Thoms, M.C. 2007. A framework for interdisciplinary understanding of rivers as ecosystems. Geomorphology, 89: 147–162. DOI: 10.1016/j.geomorph.2006.07.022

Dyke, A.S., Moore, A., Robertson, L. 2003. Deglaciation of North America. Geological Survey of Canada Open File 1574: Ottawa, Canada. DOI: 10.4095/214399

Eaton, B.C., Church, M. 2011. A rational sediment transport scaling relation based on dimensionless stream power. Earth Surface Processes and Landforms, 36: 901–910. DOI: 10.1002/esp.2120

168

Eaton, B.C., Church, M., Ham, D. 2002. Scaling and regionalization of flood flows in British Columbia, Canada. Hydrological Processes, 16: 3245–3263. DOI: 10.1002/hyp.1100

Eaton, B.C., Millar, R.G., Davidson, S. 2010. Channel patterns: Braided, anabranching, and single‐thread. Geomorphology, 120: 353–364. DOI: 10.1016/j.geomorph.2010.04.010

Edwards, T.W.D., Wolfe, B.B., MacDonald, G.M. 1996. Influence of changing atmospheric circulation on precipitation δ18O‐temperature relations in Canada during the Holocene. Quaternary Research, 46: 211–218. DOI: 10.1006/qres.1996.0061

Ellis, C.J., Keron, J.R., Menzies, J.H., Monckton, S.G., Stewart, A. In Preparation. For immediate occupancy: Cosy 3000 year old heritage winter house with river view near Lake Huron. Apply to Terminal Archaic Realty. In: Redmond, B.G., Genheimer, R.A. (eds.), Building the Past: Studies of Prehistoric Wooden Post Architecture in the Ohio Valley‐Great Lakes Region. University Presses of Florida, Gainesville.

Environment Canada. 2011. Water Survey of Canada (WSC) Hydrometric Monitoring Program ‐ HYDAT Database. Environment Canada, Ottawa. Accessed online: http://www.ec.gc.ca/rhc‐ wsc [May, 2011].

Eschman, D.F., Karrow, P.F. 1985. Huron Basin glacial lakes: A Review. In: Karrow, P.F., Calkin, P.E. (eds.), Quaternary Evolution of the Great Lakes. Geological Association of Canada Special Paper 30, St. John’s, Newfoundland: 79–94.

Eyles, N. 2002. Ontario Rocks: Three billion years of environmental change. Fitzhenry & Whiteside: Markham, Ontario.

Eyles, N., Kocsis, S. 1989. Reply to M. Church and J.M. Ryder’s discussion of ‘Sedimentology and clast fabrics of subaerial debris flow facies in a glacially influenced alluvial fan’. Sedimentary Geology, 59: 15–28. DOI: 10.1016/0037‐0738(89)90017‐1

Faustini, J.M., Kaufmann, P.R., Herlihy, A.T. 2009. Downstream variation in bankfull width of wadeable streams across conterminous United States. Geomorphology, 108: 292–311. DOI: 10.1016/j.geomorph.2009.02.005 169

Ferguson, R.I. 1981. Channel form and channel changes. In: Lewin, J. (ed.), British Rivers. Allen & Unwin, London: 90–125.

Ferguson, R.I. 1987. Hydraulic and sedimentary controls of channel pattern. In: Richards, K. (ed.), River Channels – Environment and Processes. Blackwell, Oxford, UK: 129–158.

Ferguson, R.I. 2005. Estimating critical stream power for bedload transport calculations in gravel bed rivers. Geomorphology, 70; 33–41. DOI: 10.1016/j.geomorph.2005.03.009

Ferguson, R.I. 2012. River channel slope, flow resistance, and gravel entrainment thresholds. Water Resources Research, 48: W05517. DOI: 10.1029/2011WR010850

Finlayson, D.P., Montgomery, D.R. 2003. Modelling large‐scale fluvial erosion in geographic information system. Geomorphology, 53: 147–164. DOI: 10.1016/S0169‐555X(02)00351‐3

Finlayson, D.P., Montgomery, D.R., Hallet, B. 2002. Spatial coincidence of rapid inferred erosion with young metamorphic massifs in the Himalayas. Geology, 30: 219–222. DOI: 10.1130/0091‐7613(2002)030<0219:SCORIE>2.0.CO;2

Fitzgerald, W.D. 1985. Postglacial history of the Minesing Basin, Ontario. In: Karrow, P.F., Calkin, P.E. (eds.), Quaternary Evolution of the Great Lakes. Geological Association of Canada Special Paper 30, St. John’s, Newfoundland: 133–146.

Flint, J.J. 1974. Stream gradient as a function of order, magnitude and discharge. Water Resources Research, 10: 969–973. DOI: 10.1029/WR010i005p00969

Flores, A.N., Bledsoe, B.P., Cuhaciyan, C.O., Wohl, E.E. 2006. Channel‐reach morphology dependence on energy, scale, and hydroclimatic processes with implications for prediction using geospatial data. Water Resources Research, 42: W06412. DOI: 10.1029/2005WR004226

Fonstad, M.A. 2003. Spatial variation in the power of mountain streams in the Sangre de Cristo Mountains, New Mexico. Geomorphology, 55: 75–96. DOI: 10.1016/S0169‐555X(03)00133‐8

170

Gao, C., Shirota, J., Kelly, R.I., Brunton, F.R., Van Haaften, S. 2006. Bedrock topography and overburden thickness mapping, southern Ontario. Miscellaneous Release—Data 207. Ontario Geological Survey: Sudbury, Ontario, Canada.

Garaci, M.C. 1998. River terrace development of the lower Saugeen River valley, southern Ontario. Unpublished Masters Thesis, University of Toronto, Toronto, Ontario.

Gingras, D., Adamowski, K., Pilon, P.J., 1994. Regional flood equations for the provinces of Ontario and Quebec. Journal of the American Water Resources Association, 30: 55–67.

Goff, J.R., Ashmore, P. 1994. Gravel transport and morphological change in braided Sunwapta River, Alberta, Canada. Earth Surface Processes and Landforms, 19: 195–212. DOI: 10.1002/esp.3290190302

Goldrick, G., Bishop, P. 2007. Regional analysis of bedrock stream long profiles: Evaluation of Hack’s SL form, and formulation and assessment of an alternative (the DS form). Earth Surface Processes and Landforms, 32: 649–671. DOI: 10.1002/esp.1413

Goldthwait, J.W. 1910. Isobases of the Algonquin and Iroquois beaches and their significance. Geological Society of America Bulletin, 21: 227–248.

Gonga‐Saholiariliva, N., Gunnell, Y., Harbor, D., Mering, C. 2011. An automated method for producing synoptic regional maps of river gradient variation: Procedure, accuracy tests, and comparison with other knickpoint mapping methods. Geomorphology, 134: 394–407. DOI: 10.1016/j.geomorph.2011.07.013

Graf, W.L. 1983. Downstream changes in stream power in the Henry Mountains, Utah. Annals of the Association of American Geographers, 73: 373–387. DOI: 10.1111/j.1467‐8306.1983.tb01423.x

Guisan, A., Zimmermann, N.E. 2000. Predictive habitat distribution models in ecology. Ecological Modeling, 135: 147–186. DOI: 10.1016/S0304‐3800(00)00354‐9

171

Hack, J.T. 1957. Studies of longitudinal profiles in Virginia and Maryland. United States Geological Survey Professional Paper 294‐B, Washington, DC: 45–97.

Hack, J.T. 1973. Stream‐profile analysis and stream‐gradient index. Journal of Research of the United States Geological Survey, 1: 421–429.

Heath, A.J., Karrow, P.F. 2007. Northernmost (?) glacial lake Algonquin series shorelines, Sudbury basin, Ontario. Journal of Great Lakes Research, 33: 264–278. DOI: 10.3394/0380‐1330(2007)33[264:NGLASS]2.0.CO;2

Heiri, O., Lotter, A.F., Lemcke, G. 2001. Loss on ignition as a method for estimating organic and carbonate content in sediments: Reproducibility and comparability of results. Journal of Paleolimnology, 25: 101–110. DOI: 10.1023/a:1008119611481

Hickin, E.J. 1983. River channel changes: Retrospect and prospect. In: Collinson, J.D., Lewin, J. (eds.), Modern and Ancient Fluvial Systems. International Association of Sedimentologists Special Publication 6, Blackwell Publishing Ltd, Oxford, UK: 61–83. DOI: 10.1002/9781444303773

Hickin, E.J. 1993. Fluvial facies models: A review of Canadian research. Progress in Physical Geography, 17: 205–222. DOI: 10.1177/030913339301700207

Hough, J.L. 1958. Geology of the Great Lakes. University of Illinois Press, Urbana, Illinois.

Hutchinson, M.F. 2005. ANUDEM 5.1 User’s Guide, 2005. Centre for Resources and Environmental Studies, Australian National University, Canaberra, Australia.

Jackson, D.A. 1993. Stopping rules in principal component analysis: A comparison of heuristical and statistical approaches. Ecology, 74: 2204–2214. DOI: 10.2307/1939574

Jain, A.K., Murty, M.N., Flynn, P.J. 1999. Data clustering: A review. ACM Computing Surveys, 31: 264–323. DOI: 10.1145/331499.331504

172

Jain, V., Preston, N., Fryirs, K., Brierley, G. 2006. Comparative assessment of three approaches for deriving stream power plots along long profiles in the upper Hunter River catchment, New South Wales, Australia. Geomorphology, 74: 297–317. DOI: 10.1016/j.geomorph.2005.08.012

Javelle, P., Ouarda, T.B.M.J., Bobée, B. 2003. Spring flood analysis using the flood‐duration‐ frequency approach: Application to the provinces of Quebec and Ontario, Canada. Hydrological Processes, 17: 3717–3736. DOI: 10.1002/hyp.1349

Jeffers, J.N.R. 1998. Characterization of river habitats and prediction of habitat features using ordination techniques. Aquatic Conservation: Marine and Freshwater Ecosystems, 8: 529– 540. DOI: 10.1002/(SICI)1099‐0755(199807/08)8:4<529::AID‐AQC301>3.0.CO;2‐9

Jenkins, P.A. 2007. Map‐based tests on controls of anabranching river character on the lower Yellowstone River. Unpublished M.Sc. Thesis, Department of Earth Sciences, Montana State University, Bozeman, MT.

Kaiser, H.F. 1960. The application of electronic computers to factor analysis. Educational and Psychological Measurement, 20: 141–151. DOI: 10.1177/001316446002000116

Karrow, P.F. 1980. The Nipissing transgression around southern Lake Huron. Canadian Journal of Earth Science, 17: 1271–1274. DOI: 10.1139/e80‐132

Karrow, P.F. 1986. Valley terraces and Huron basin water levels, southwestern Ontario. Geological Society of America Bulletin, 97: 1089–1097. DOI: 10.1130/0016‐7606(1986)97<1089:VTAHBW>2.0.CO;2

Karrow, P.F., Appleyard, E.C., Endres, A.L. 2007. Geological and geophysical evidence for pre‐ Nipissing (>5,000 years BP) transgression inﬁlled valleys in the Lake Huron basin, Ontario. Journal of Paleolimnology, 37: 419–434. DOI: 10.1007/s10933‐006‐9048‐z

Karrow, P.F., Calkin P.E. 1985. Quaternary evolution of the Great Lakes, Special Paper 30, St. John’s, Newfoundland.

173

Kellerhals, R., Church, M., Bray, D.I. 1976. Classification and analysis of river processes. American Society of Civil Engineers Journal of the Hydraulics Division, 102: 813–829.

Kiesel, D.S., Mickelson, D.M. 2005. Influence of Lake Michigan Holocene lake levels on fluvial sediments of the lower Pigeon River, Wisconsin. Journal of Great Lakes Research, 31: 343– 359. DOI: 10.1016/S0380‐1330(05)70265‐3

King, J.R., Jackson, D.A. 1999. Variable selection in large environmental datasets using principal component analysis. Environmetrics, 10: 67–77. DOI: 10.1002/(SICI)1099‐095X(199901/02)10:1<67::AID‐ENV336>3.0.CO;2‐0

Kleinhans, M.G., van den Berg, J.H. 2011. River channel and bar patterns explained and predicted by an empirical and physics‐based method. Earth Surface Processes and Landforms, 36: 721–738. DOI: 10.1002/esp.2090

Knighton, A.D. 1987. River channel adjustment – the downstream dimension. In: Richards, K., (ed.), River Channels – Environment and Process. Basil Blackwell Ltd, Oxford, UK: 95–128.

Knighton, A.D. 1998. Fluvial Forms and Processes: A New Perspective. Arnold: London. DOI: 10.1002/(SICI)1096‐9837(199904)24:4<378::AID‐ESP958>3.0

Knighton, A.D. 1999. Downstream variation in stream power. Geomorphology, 29: 293–306. DOI: 10.1016/S0169‐555X(99)00015‐X

Lamb, M.P., Dietrich, W.E., Venditti, J.G. 2008. Is the critical Shields stress for incipient sediment motion dependent on channel‐bed slope? Journal of Geophysical Research Earth Surface, 113: F02008. DOI: 10.1029/2007JF000831

Larson, G., Schaetzl, R. 2001. Origin and evolution of the great lakes. Journal of Great Lakes Research, 27: 518–546. DOI: 10.1016/S0380‐1330(01)70665‐X

Lecce, S.A. 1997. Nonlinear downstream changes in stream power on Wisconsin’s Blue River. Annals of the Association of American Geographers, 87: 471–486. DOI: 10.1111/1467‐8306.00064

174

Leopold, L.B., Maddock, Jr., T. 1953. The hydraulic geometry of stream channels and some physiographic implications. United States Geological Survey Professional Paper 252: Washington, DC.

Leopold, L.B., Wolman, M.G. 1957. River channel patterns: Braided, meandering, and straight. United States Geological Survey Professional Paper 282‐B: Washington, DC.

Leopold, L.P., Wolman, M.G., Miller, J.P., 1964. Fluvial Processes in Geomorphology. W.H. Freeman: San Francisco, CA.

Leverett, F., Taylor, F. B. 1915. The Pleistocene of Indiana and Michigan and the history of the Great Lakes. United States Geological Survey Monograph 53: Washington, DC.

Lewis, C.F.M., Blasco, S.M., Gareau, P.L. 2005. Glacial isostatic adjustments of the Laurentian Great Lakes basin: Using empirical record of strandline deformation for reconstruction of early Holocene paleo‐lakes and discovery of a hydrologically closed phase. Géographie Physique et Quaternaire, 59: 187–210. DOI: 10.7202/014754ar

Lewis, C.F.M., Karrow, P.F., Blasco, S.M., McCarthy, F.M.G., King, J.W., Moore, T.C., Rea, D.K. 2008. Evolution of lakes in the Huron basin: Deglaciation to present. Aquatic Ecosystem Health & Management, 11: 127–136. DOI: 10.1080/14634980802095263

Lewis, C.F.M., Teller, J.T. 2007. North American late‐Quaternary meltwater and floods to the oceans: Evidence and impact. Palaeogeography Palaeoclimatology Palaeoecology, 246: 1–7. DOI: 10.1016/j.palaeo.2006.10.023

Lóczy, D., Pirkhoffer, E., Gyenizse, P. 2012. Geomorphometric floodplain classification in a hill region of Hungary. Geomorphology, 147–149: 61–72. DOI: 10.1016/j.geomorph.2011.06.040

Mackin, J.H. 1948. Concept of the graded river. Geological Society of America Bulletin, 59: 463– 511. DOI: 10.1130/0016‐7606(1948)59[463:COTGR]2.0.CO;2

175

Magilligan, F.J. 1992. Thresholds and the spatial variability of flood power during extreme floods. Geomorphology, 5: 373–390. DOI: 10.1016/0169‐555X(92)90014‐F

Mainville, A., Craymer, M. R. 2005. Present‐day tilting of the Great Lakes region based on water level gauges. Geological Society of America Bulletin, 117: 1070–1080. DOI: 10.1130/B25392.1

McCleary, R.J., Hassan, M.A., Miller, D., Moore, R.D., 2011. Spatial organization of process domains in headwater drainage basins of the glaciated foothills region with complex longitudinal profiles. Water Resources Research, 47: W05505. DOI: 10.1029/2010WR009800

McFadden, M.A., Patterson, W.P., Mullins, H.T., Anderson, W.T. 2005. Multi‐proxy approach to long‐ and short‐term Holocene climate‐change: Evidence from eastern Lake Ontario. Journal of Paleolimnology, 33: 371–391. DOI: 10.1007/s10933‐004‐7634‐5

McLachlan, G.J. 2012. Discriminant analysis. WIREs Computational Statistics, 4: 421–431. DOI: 10.1002/wics.1219

Miall, A.D. 1985. Architectural‐element analysis: A new method of facies analysis applied to fluvial deposits. Earth Science Reviews, 22: 261–308. DOI: 10.1016/0012‐8252(85)90001‐7

Miall, A.D. 2010. Alluvial deposits. In: James, N.P., Dalrymple, R.W. (eds.), Facies Models 4. Geological Association of Canada, St. John’s, Newfoundland: 105–138.

Montgomery, D.R. 1999. Process domains and the river continuum. Journal of the American Water Resources Association, 35: 397–410. DOI: 10.1111/j.1752‐1688.1999.tb03598.x

Montgomery, D.R., Buffington, J.M. 1997. Channel‐reach morphology in mountain drainage basins. Geological Society of America Bulletin, 109: 596–611. DOI: 10.1130/0016‐7606(1997)109<0596:CRMIMD>2.3.CO;2 v.

Montgomery, D.R., Foufoula‐Georgiou, E. 1993. Channel network source representation using digital elevation models. Water Resources Research, 29: 3925–3934. DOI: 10.1029/93WR02463

176

Morris, P.H., Williams, D.J. 1997. Exponential longitudinal profiles of streams. Earth Surface Processes and Landforms, 22: 143–163. DOI: 10.1002/(SICI)1096‐9837(199702)22:2<143::AID‐ESP681>3.0.CO;2‐Z

Mulvihill, C.I., Baldigo, B.P., Miller, S.J., DeKoskie, D., DuBois, J. 2009. Bankfull discharge and channel characteristics of streams in New York state. United States Geological Survey Scientific Investigations Report 2009–5144: Reston, VA.

Nanson, G.C., Croke, J.C. 1992. A genetic classification of floodplains. Geomorphology, 4: 459– 486. DOI: 10.1016/0169‐555X(92)90039‐Q

Nanson, G.C., Croke, J.C. 2002. Emerging issues in flood plain research. In: Dyer, F.J., Thoms, M.C., Olley, J.M. (eds.), The Structure, Function and Management Implications of Fluvial Sedimentary Systems, IAHS Publication No 276, International Association of Hydrological Sciences Press, Wallingford: 271–278.

Nanson, G.C., Knighton, A.D. 1996. Anabranching rivers: Their cause, character and classification. Earth Surface Processes and Landforms, 21: 217–239. DOI: 10.1002/(SICI)1096‐9837(199603)21:3<217::AID‐ESP611>3.0.CO;2‐U

Ontario Geological Survey. 1993. Bedrock geology, seamless coverage of the province of Ontario [computer file]. ERLIS Dataset 6 (released March 2000), Ontario Geological Survey, Sudbury, Ontario, Canada.

Ontario Geological Survey. 1997. Quaternary geology, seamless coverage of the province of Ontario [computer file]. ERLIS Dataset 14 (released March 2000), Ontario Geological Survey, Sudbury, Ontario, Canada.

Ontario Ministry of Natural Resources. 2005. Provincial digital elevation model—Tiled dataset, version 2.0.0 [computer file]. Ontario Ministry of Natural Resources, Peterborough, Ontario, Canada.

177

Ontario Ministry of Natural Resources. 2008. Provincial digital elevation model (DEM) version 2.0.0. Product Specification – Edition 1.1. Water Resources Information Program, Geographic Information Branch, Ontario Ministry of Natural Resources, Peterborough, Ontario, Canada.

Ontario Ministry of Natural Resources, Redside Dace Recovery Team. 2009. DRAFT Recovery strategy for Redside Dace (Clinostomus elongatus) in Ontario. Ontario Recovery Strategy Series. Prepared for the Ontario Ministry of Natural Resources (OMNR), Peterborough, Ontario.

Paixao, E., Auld, H., Qader Mirza, M.M., Klaassen, J., Shephard, M.W. 2011. Regionalization of heavy rainfall to improve climatic design values for infrastructure: case study in southern Ontario, Canada. Hydrological Sciences Journal, 56: 1067–1089. DOI: 10.1080/02626667.2011.608069

Pandey, G.R. 1998. Assessment of scaling behaviour of regional floods. Journal of Hydrologic Engineering, 3: 169–173. DOI: 10.1061/(ASCE)1084‐0699(1998)3:3(169)

Parish Geomorphic Ltd. 2004. Redside Dace recovery strategy fluvial geomorphology study. Mississauga, Ontario. Prepared for OMNR Redside Dace Recovery Team.

Parker, C., Clifford, N.J., Thorne, C.R. 2011. Understanding the influence of slope on the threshold of coarse grain motion: Revisiting critical stream power. Geomorphology, 126: 51–65. DOI: 10.1016/j.geomorph.2010.10.027

Peltier, W.R. 1994. Ice‐age paleotopography. Science, 265: 195–201. DOI: 10.1126/science.265.5169.195

Peltier, W.R. 1995. VLBI base‐line variations from the ICE‐4G model of postglacial rebound. Geophysical Research Letters, 22: 465–468. DOI: 10.1029/94GL03213

Peltier, W.R. 1998. Postglacial variations in the level of the sea: Implications for climate dynamics and solid‐earth geophysics. Reviews of Geophysics, 36: 603–689. DOI: 10.1029/98RG02638 178

Peltier, W.R. 2004. Global glacial isostasy and the surface of the ice‐age earth: The ICE‐5G (VM2) model and GRACE. Annual Review of Earth and Planetary Sciences, 32: 111–149. DOI: 10.1146/annurev.earth.32.082503.144359

Pengelly, J.W., Tinkler, K.J., Parkins, W.G., McCarthy, F.M. 1997. 12600 years of lake level changes, changing sills, ephemeral lakes and Niagara gorge erosion in the Niagara peninsula and eastern Lake Erie basin. Journal of Paleolimnology, 17: 377–402. DOI: 10.1023/A:1007946401036

Pérez‐Peña, J.V., Azañon, J.M., Azor, A., Delgado, J., González‐Lodeiro, F. 2009. Spatial analysis of stream power using GIS: SLk anomaly maps. Earth Surface Processes and Landforms, 34: 16–25. DOI: 10.1002/esp.1684

Petit, F., Gob, F., Houbrechts, G., Assani, A.A. 2005. Critical specific stream power in gravel‐bed rivers. Geomorphology, 69: 92–101. DOI: 10.1016/j.geomorph.2004.12.004

Phillips, J.D. 2011. Emergence and pseudo‐equilibrium in geomorphology. Geomorphology, 132: 319–326. DOI: 10.1016/j.geomorph.2011.05.017

Phillips, J.D., Lutz, J.D. 2008. Profile convexities in bedrock and alluvial streams. Geomorphology, 102; 554–566. DOI: 10.1016/j.geomorph.2008.05.042

Phillips, J.D., McCormack, S., Duan, J., Russo, J.P., Schumacher, A.M., Tripathi, G.N., Brockman, R.B., Mays, A.B., Pulugurtha, S.P. 2010. Origin and interpretation of knickpoints in the Big South Fork River basin, Kentucky‐Tennessee. Geomorphology, 114: 188–198. DOI: 10.1016/j.geomorph.2009.06.023

Phillips, R.T.J., Desloges, J.R. 2014. Glacially conditioned specific stream powers in low‐relief river catchments of the southern Laurentian Great Lakes. Geomorphology, 206: 271–287. DOI: 10.1016/j.geomorph.2013.09.030

Phillips, R.T.J., Robert, A. 2007. Hydrological control of waveforms on small meandering rivers. Earth Surface Processes and Landforms, 32: 1533–1546. DOI: 10.1002/esp.1483

179

Piégay, H., Darby, S.E., Mosselman, E., Surian, N. 2005. A review of techniques available for delimiting the erodible river corridor: A sustainable approach to managing bank erosion. River Research and Applications, 21: 773–789. DOI: 10.1002/rra.881

Poff, N.L., Ward, J.V. 1989. Implications of streamflow variability and predictability for lotic community structure: A regional analysis of streamflow patterns. Canadian Journal of Fisheries and Aquatic Sciences, 46: 1805–1818. DOI: 10.1139/f89‐228

Poole, G.C. 2002. Fluvial landscape ecology: Addressing uniqueness within the river discontinuum. Freshwater Biology, 47: 641–660. DOI: 10.1046/j.1365‐2427.2002.00922.x

Rachol, C.M., Boley‐Morse, K. 2009. Estimated bankfull discharge for selected Michigan rivers and regional hydraulic geometry curves for estimating bankfull characteristics in southern Michigan rivers. United States Geological Survey Scientific Investigations Report 2009–5133: Reston, VA.

Raymo, M.E. 1994. The initiation of northern‐hemisphere glaciation. Annual Review of Earth and Planetary Sciences, 22: 353–383. DOI: 10.1146/annurev.ea.22.050194.002033

Recking, A. 2009. Theoretical development on the effects of changing flow hydraulics on incipient bed motion. Water Resources Research, 45: W04401. DOI: 10.1029/2008WR006826

Reinfelds, I., Cohen, T., Batten, P., Brierley, G. 2004. Assessment of downstream trends in channel gradient, total and specific stream power: A GIS approach. Geomorphology, 60: 403–416. DOI: 10.1016/j.geomorph.2003.10.003

Rhoads, B.L. 1987. Stream power terminology. The Professional Geographer, 39: 189–195. DOI: 10.1111/j.0033‐0124.1987.00189.x

Ribeiro, J., Rousselle, J. 1996. Robust simple scaling analysis of flood peaks series. Canadian Journal of Civil Engineering, 23: 1139–1145. DOI: 10.1139/l96‐923

180

Roper, B.B., Buffington, J.M., Archer, E., Moyer, C., Ward, M. 2008. The role of observer variation in determining Rosgen stream types in northeastern Oregon mountain streams. Journal of the American Water Resources Association, 44: 417–427. DOI: 10.1111/j.1752‐1688.2008.00171.x

Rosgen, D.L. 1994. A classification of natural rivers. Catena, 22: 169–199. DOI: 10.1016/0341‐8162(94)90001‐9

Rosgen, D.L. 1996. Applied River Morphology. Wildland Hydrology, Pagosa Springs, CO.

Rosgen, D.L. 2009. Discussion–“The role of observer variation in determining Rosgen stream types in northeastern Oregon mountain streams” by Brett B. Roper, John M. Buffington, Eric Archer, Chris Moyer, and Mike Ward. Journal of the American Water Resources Association, 45: 1290–1297. DOI: 10.1111/j.1752‐1688.2009.00332.x

Schaetzl, R.J., Drzyzga, S.A., Weisenborn, B.N., Kincare, K.A., Lepczyk, X.C., Shein, K., Dowd, C. M., Linker, J. 2002. Measurement, correlation, and mapping of glacial lake Algonquin shorelines in northern Michigan. Annals of the Association of American Geographers, 92: 399–415. DOI: 10.1111/1467‐8306.00296

Schumm, S.A. 1985. Patterns of alluvial rivers. Annual Review of Earth and Planetary Sciences, 13: 5–27. DOI: 10.1146/annurev.ea.13.050185.000253

Schumm, S.A. 1993. River response to baselevel change: Implications for sequence stratigraphy. The Journal of Geology, 101: 279–294. DOI: 10.1086/648221

Schumm, S.A., Lichty, R.W. 1965. Time, space and causality in geomorphology. American Journal of Science, 263: 110–119. DOI: 10.2475/ajs.263.2.110

Seeber, L., Gornitz, V. 1983. River profiles along the Himalayan arc as indicators of active tectonics. Tectonophysics, 92: 335–367. DOI: 10.1016/0040‐1951(83)90201‐9

181

Sella, G.F., Stein, S., Dixon, T.H., Craymer, M., James, T.S., Mazzotti, S., Dokka, R.K. 2007. Observation of glacial isostatic adjustment in "stable" North America with GPS. Geophysical Research Letters, 34: 1–6. DOI: 10.1029/2006GL027081

Sherwood, J.M., Huitger, C.A., 2005. Bankfull characteristics of Ohio streams and their relation to peak stream‐flows. United States Geological Survey Scientific Investigations Report 2005– 5153: Reston, VA.

Shields, A.F. 1936. Application of similarity principles and turbulence research to bed‐load movement. Mitteilungen der Preussischen Versuchsanstalt für Wasserbau und Schiffbau. Berlin, Germany, 26: 5–24.

Simon, A., Doyle, M., Kondolf, M., Shields, F.D., Rhoads, B., McPhillips, M. 2007. Critical evaluation of how the Rosgen classification and associated “Natural Channel Design” methods fail to integrate and quantify fluvial processes and channel response. Journal of the American Water Resources Association, 43: 1117–1131. DOI: 10.1111/j.1752‐1688.2007.00091.x

Sinha, S.K., Parker, G. 1996. Causes of concavity in longitudinal profiles of rivers. Water Resources Research, 32: 1417–1428. DOI: 10.1029/95WR03819

Smith, T.R., Merchant, G.E., Birnir, B. 2000. Transient attractors: Towards a theory of the graded stream for alluvial and bedrock channels. Computers and Geosciences, 26: 541–580. DOI: 10.1016/S0098‐3004(99)00128‐4

Snyder, N.P., Castele, M.R., Wright, J.R. 2008. Bedload entrainment in low‐gradient paraglacial coastal rivers of Maine, U.S.A.: Implications for habitat restoration. Geomorphology, 102: 430–446. DOI: 10.1016/j.geomorph.2008.07.013

Snyder, N.P., Nesheim, A.O., Wilkins, B.C., Edmonds, D.A. 2013. Predicting grain size in gravel‐ bedded rivers using digital elevation models: Application to three Maine watersheds. Geological Society of America Bulletin, 125: 148–163. DOI: 10.1130/B30694.1

182

Spencer, J.W. 1891. Deformation of the Algonquin beach and the birth of Lake Huron. American Journal of Science, 41: 12–21. DOI: 10.2475/ajs.s3‐41.241.12

Splinter, D.K. 2013. Clustering and classifying channel morphology in eastern Oklahoma ecoregions using dissimilarity coefficients. Physical Geography, 34: 512–528. DOI: 10.1080/02723646.2013.817028

Stanford, J.A., Ward, J.V. 1993. An ecosystem perspective of alluvial rivers: Connectivity and the hyporheic corridor. Journal of North American Benthological Society, 12: 48–60. DOI: 10.2307/1467685

Stewart, A.M., Desloges, J.R. 2013. A 9000‐year record of vertical and lateral accretion on the floodplain of the lower Thames River, southwestern Ontario, Canada, and implications for archaeological research. Quaternary International. DOI: 10.1016/j.quaint.2013.10.002

Stewart, A.M., Desloges, J., Ellis, C., Smith, D. 2008. What lies beneath? A survey of the evidence for buried sites in southern Ontario. Paper presented to the Canadian Archaeological Association meeting in “Identifying Contexts for Deeply Buried Sites,” May 9, 2008: Peterborough, Ontario.

Tabata, K.K., Hickin, E.J. 2003. Interchannel hydraulic geometry and hydraulic efficiency of the anastomosing Columbia River, southeastern British Columbia, Canada. Earth Surfaces Processes and Landforms, 28: 837–852. DOI: 10.1002/esp.497

Teller, J.T. 1995. History and drainage of large ice‐dammed lakes along the Laurentide Ice‐ Sheet. Quaternary International, 28: 83–92. DOI: 10.1016/1040‐6182(95)00050‐S

Thayer, J.B. 2012. Downstream variability of fluvial form, process, and character in a small deglaciated watershed, southern Ontario. Unpublished Masters Thesis, University of Toronto, Toronto, Ontario.

Thorn, C.E., Welford, M.R. 1994. The equilibrium concept in geomorphology. Annals of the Association of American Geographers, 84: 666–696. DOI: 10.1111/j.1467‐8306.1994.tb01882.x 183

Thornbush, M.J., Desloges, J.R. 2011. Environmental change and evidence for Archaic and Woodland floodplain occupation along the lower Nottawasaga River, southern Ontario, Canada. In: Brown, A.G., Basell, L.S., Butzer, K.W. (eds.), Geoarchaeology, Climate Change, Sustainability, Geological Society of America Special Paper 476. Geological Society of America: Boulder, Colorado: 105–116. DOI: 10.1130/2011.2476(09)

United States Geological Survey. 1982. Guidelines for determining food flow frequency. Hydrology Subcommittee, United States Department of the Interior, Geological Survey, Office of Water Data Coordination, Bulletin‐17B, Reston, VA: 14–35.

Upsdell, B., Veliz, M., Staton, S., Jean, K. 2010. Ausable River geomorphology and the distribution of freshwater mussel species at risk. DRAFT report prepared for Ausable River Recovery Team, Ausable Bayfield Conservation Authority (ABCA). van den Berg, J.H., 1995. Prediction of alluvial channel pattern of perennial rivers. Geomorphology, 12: 259–279. DOI: 10.1016/0169‐555X(95)00014‐V van der Wal, W., Wu, P., Sideris, M. G. , Shum, C. K. 2008. Use of GRACE determined secular gravity rates for glacial isostatic adjustment studies in North‐America. Journal of Geodynamics, 46: 144–154. DOI: 10.1016/j.jog.2008.03.007

Vannote, R.L., Minshall, G.W., Cummins, K.W., Sedell, J.R., Cushing, C.E. 1980. The river continuum concept. Canadian Journal of Fisheries and Aquatic Sciences, 37: 130–137. DOI: 10.1139/f80‐017

Vaughan, I.P., Merrix‐Jones, F.L., Constantine, J.A. 2013. Successful predictions of river characteristics across England and Wales based on ordination. Geomorphology, 194: 121– 131. DOI: 10.1016/j.geomorph.2013.03.036

Vidon, P., Hill, A.R. 2006. A landscape‐based approach to estimate riparian hydrological and nitrate removal functions. Journal of the American Water Resources Association, 42: 1099– 1112. DOI: 10.1111/j.1752‐1688.2006.tb04516.x

184

Vocal Ferencevic, M., Ashmore, P. 2012. Creating and evaluating digital elevation model‐based stream‐power map as a stream assessment tool. River Research and Applications, 28: 1394– 1416. DOI: 10.1002/rra.1523

Walker, I.J., Desloges, J.R., Crawford, G.W., Smith, D.G. 1997. Floodplain formation processes and archaeological implications at the Grand Banks site, lower Grand River, southern Ontario. Geoarchaeology, 12: 865–887. DOI: 10.1002/(SICI)1520‐6548(199712)12:8<865::AID‐GEA3>3.0.CO;2‐3

Ward, J.V., Tockner, K., Arscott, D.B., Claret, C. 2002. Riverine landscape diversity. Freshwater Biology, 47: 517–539. DOI: 10.1046/j.1365‐2427.2002.00893.x

Whipple, K.X. 2004. Bedrock rivers and the geomorphology of active orogens. Annual Review of Earth and Planetary Sciences, 32: 151–185. DOI: 10.1146/annurev.earth.32.101802.120356

Wilkins, B., Siddiqui, A., Parish, J. 2009. Ausable River mussel study – Geomorphic assessment (Draft Report). Mississauga, Ontario: Parish Geomorphic Ltd.

Wolf, D. 1993. The changing‐role of the lithosphere in models of glacial isostasy: A historical review. Global and Planetary Change, 8: 95–106. DOI: 10.1016/0921‐8181(93)90017‐I

Yang, C.T. 1996. Sediment Transport: Theory and Practice. The McGraw‐Hill, New York, NY.

185

Appendix A Raw Floodplain Dataset and Example Photographs

186

Appendix A ‐ Raw Floodplain Dataset Fluvial Process Variables Page 1 of 2

Easting Northing A d Q 2 Ω ω τ /D 95 ω /D 50 ω cr ID# Field Site K‐Means Sample Date S ω/ω cr (m) (m) (km2) (m3s‐1) (Wm‐1) (Wm‐2) (Nm‐3) (Wm‐3) (Wm‐2) 1EHR Bank 1.1 (M, SM) M 25/05/2010 610785 4855585 137 0.0023 19.8 452 32.4 616 1.4 21.2 1.5 1.4 EHR Bank 1.4 (M, SM) M 27/07/2010 612035 4854255 140 0.0022 20.2 434 30.7 580 1.4 21.2 1.5 1.7 EHR Bank 7 (M, SM) M 26/08/2010 613025 4852935 179 0.0027 24.9 658 42.2 788 1.9 21.2 2.0 2MWC‐1 BX (M) M 30/06/2010 476215 4761535 201 0.0024 41.8 984 54.7 608 2.4 21.2 2.6 3MWC‐2 B1 (M) M 24/06/2010 475415 4766005 182 0.0018 38.3 690 40.2 339 2.5 12.6 3.2 4FL Bank2 (M) M 29/07/2010 603275 4830775 38 0.0050 6.6 325 39.3 777 1.7 21.2 1.9 4FL Bank1 (M) M 29/07/2010 603295 4830695 38 0.0053 6.6 338 41.0 810 1.8 21.2 1.9 5.1 NWR‐S1 (S) S 04/08/2010 594775 4892335 1092 0.0007 85.4 589 21.3 215 1.9 7.5 2.9 5.2 NWR‐S2 (S) S 04/08/2010 595385 4893325 1094 0.0004 85.6 354 12.8 129 1.1 7.5 1.7 6.2 SG‐9b S2 (M)‐X B 11/08/2010 485115 4891535 2285 0.0013 352.7 4619 78.4 131 0.4 478.9 0.2 6.3 SG‐9d Stop3 (M) M 11/08/2010 484315 4892595 2293 0.0008 353.9 2848 48.3 121 1.1 59.9 0.8 7CF YM Bank1 (M) M 17/08/2010 496105 4728125 350 0.0012 43.3 508 24.0 323 1.1 21.2 1.1 8AR‐Bar‐15 DS (S, SM) S 18/08/2010 433655 4781645 1086 0.0002 219.2 401 10.7 531 0.9 7.5 1.4 9GR Cayuga (C)‐X M 24/08/2010 592875 4755615 6477 0.0002 909.2 1824 15.2 142 1.0 12.6 1.2 10 AR‐5c (S, SM) S 21/06/2011 433575 4773045 1049 0.0005 212.3 1127 30.7 79 4.1 4.0 7.7 11 AR‐28b (S, SM) S 21/06/2011 434485 4773925 1056 0.0005 213.5 1038 28.2 364 2.3 8.3 3.4 12 AR‐33 (S, SM) S 22/06/2011 434185 4777925 1069 0.0002 215.9 502 13.5 446 2.2 2.9 4.6 101 NWR‐S3 (M, SM) M 04/08/2010 594495 4895985 1107 0.0010 86.4 831 29.9 57 0.7 59.9 0.5 102 ISC 3b (S, SM) S 10/06/2011 597595 4883575 86 0.0011 969.6 102 11.9 2671 151.5 444.4 272.7 103 ICST2 (C, CM) C 10/06/2011 595925 4879525 77 0.0025 8.7 210 25.6 5772 3.2 4.4 5.8 104 AR‐5b (M) M 20/06/2011 433595 4770825 1037 0.0018 210.1 3731 102.3 443 1.1 169.3 0.6 105 AR‐AC‐3 (C) C 19/09/2011 440275 4764295 36 0.0011 10.1 105 16.6 2405 4.2 1.6 10.6 106 AR‐AC‐4 (C, CM) C 19/09/2011 439775 4768215 65 0.0013 17.2 220 25.5 2550 6.4 1.6 16.2 108 AR‐UT4‐1 (M)‐X C 19/09/2011 434615 4772525 3 0.0083 1.2 95 51.6 808 8.5 2.9 17.6 109 BFR‐1 (B) B 20/09/2011 452395 4822145 455 0.0027 100.0 2642 111.1 37 0.9 284.8 0.4 110 BBR‐4 (CM)‐X C 20/09/2011 454155 4819425 157 0.0015 38.4 578 42.2 195 5.3 4.4 9.5 111 BFR‐TC‐1 (C, BM) C 20/09/2011 452815 4826565 20 0.0031 6.1 183 38.8 287 4.8 4.4 8.7 112 BFR‐4 (B) B 20/09/2011 460345 4825175 183 0.0012 44.1 521 35.2 24 0.5 100.7 0.3 114 AR‐3 (B) B 26/09/2011 458705 4801155 111 0.0028 28.1 759 66.3 50 0.7 169.3 0.4 116 PHC‐5 (M, SM) M 27/09/2011 451635 4782695 84 0.0016 21.8 341 34.5 179 1.5 21.2 1.6 117 LAR‐7b (B) B 03/10/2011 462755 4780705 143 0.0034 35.4 1165 89.2 110 1.4 100.7 0.9 118 PHC‐6b (C, CM) C 03/10/2011 450015 4781215 107 0.0016 27.3 434 38.6 775 6.4 2.9 13.3 119 AR‐13 (S) S 04/10/2011 451715 4771555 76 0.0003 159.4 451 14.5 1424 3.6 1.6 9.2 120 AR‐18 (S) S 04/10/2011 442075 4769185 913 0.0004 187.3 642 18.8 1542 4.7 1.6 12.0 121 NWR‐13 (B) B 11/10/2011 573155 4867345 30 0.0184 3.9 700 131.2 300 0.7 478.9 0.3 122 NWR9 (M)‐X B 11/11/2010 581365 4873945 164 0.0050 16.6 821 70.9 246 1.1 100.7 0.7 123 NWR‐3 (M) M 03/08/2010 590465 4885275 328 0.0017 30.3 512 32.2 776 1.4 21.2 1.5 124 NR‐1R (SM)‐X M 11/10/2011 595075 4887935 827 0.0013 67.2 857 35.3 131 1.6 21.2 1.7 125 ISC‐1R (S, SM)‐X M 11/10/2011 599285 4886365 214 0.0010 20.9 204 15.6 78 0.7 21.2 0.7 126 NR‐10 (S) S 11/10/2011 592185 4904735 1259 0.0002 96.6 218 747.4 1059 121.2 292.9 262.6 127 BYR‐2 (B) B 17/10/2011 569035 4883295 91 0.0100 10.0 977 110.9 293 0.6 478.9 0.2 128 BYR‐3 (B) B 17/10/2011 574365 4889865 122 0.0109 12.9 1390 137.3 143 1.1 284.8 0.5 129 BYR‐7 (S) S 17/10/2011 593525 4890615 239 0.0014 23.1 307 22.3 3980 3.7 2.9 7.7 130 PC‐1 (B)‐X M 17/10/2011 583225 4897215 197 0.0039 19.4 742 59.1 295 1.8 35.6 1.7 131 WC‐2 (S) S 18/10/2011 596815 4922385 135 0.0013 14.0 173 16.3 3067 2.7 2.9 5.7 132 WC‐3 (M) M 18/10/2011 601045 4922065 128 0.0044 13.4 576 55.9 243 2.5 21.2 2.6 133 NR‐14 (S, CS) S 05/08/2010 585755 4925855 2642 0.0002 183.2 298 7.2 679 1.8 1.6 4.6 134 NR‐18 (M) M 18/10/2011 575895 4924695 2747 0.0004 189.4 687 16.3 18 0.5 35.6 0.5 135 MR‐1 (M) M 18/10/2011 579615 4906255 304 0.0030 28.4 833 54.2 293 2.4 21.2 2.6 136 PR‐4 (M, BM) M 18/10/2011 574785 4895745 147 0.0027 15.1 399 36.3 263 1.6 21.2 1.7 137 PR‐5b (B) B 24/10/2011 567575 4891025 70 0.0162 7.9 1266 162.0 191 0.6 805.5 0.2 138 MR‐2 (B) B 24/10/2011 568255 4908225 226 0.0068 21.9 1465 109.2 144 0.6 478.9 0.2 139 SGR3 (M)‐X B 12/08/2010 474065 4922615 3912 0.0008 604.6 4434 57.8 51 0.9 100.7 0.6 140 SR‐SKC‐1 (M, BM)‐X B 25/10/2011 472905 4918375 178 0.0026 27.3 688 40.9 67 0.6 100.7 0.4 141 NSR‐10 (B) B 25/10/2011 480345 4907315 241 0.0045 37.1 1647 84.2 160 0.9 169.3 0.5 143 GR‐NR‐1b (B, BM) B 31/10/2011 548815 4781715 1111 0.0021 171.3 3613 80.6 226 1.3 100.7 0.8 144 NR5 (S, SM) S 29/10/2010 538845 4787915 954 0.0008 146.9 1224 29.7 482 3.7 4.4 6.7 145 SR‐15d (B) B 01/11/2011 522805 4897585 278 0.0051 42.7 2134 101.9 200 1.1 169.3 0.6 146 RSR‐1 (B) B 01/11/2011 518765 4901055 224 0.0039 34.4 1324 70.2 90 0.5 284.8 0.2 147 SR‐13c (B, BM) B 01/11/2011 507755 4895225 655 0.0027 100.9 2622 82.1 127 0.9 169.3 0.5 148 SR‐_DC‐3 (C) C 02/11/2011 489035 4899565 81 0.0011 12.5 130 11.4 2150 282.8 161.6 727.2 149 SR‐DC‐4 (CM)‐X B 02/11/2011 484835 4900895 152 0.0014 23.3 323 20.8 104 0.3 100.7 0.2 150 TW‐0 (B) B 02/11/2011 477945 4902585 674 0.0015 103.8 1532 47.3 61 0.5 169.3 0.3 151 TW2 (B) B 12/08/2010 478805 4897575 592 0.0034 91.2 3033 99.8 117 1.6 100.7 1.0 152 SR‐9c (M) M 02/11/2011 484485 4892295 2291 0.0009 353.7 3283 55.7 139 1.2 59.9 0.9 153 SSG4 (B) B 10/08/2010 506425 4874215 553 0.0029 85.1 2420 82.4 114 0.9 169.3 0.5 154 SSG6 (S, SM)‐X M 10/08/2010 497705 4886755 1064 0.0007 164.0 1076 26.6 56 0.8 35.6 0.7 155 TW5 (M, SM) M 12/08/2010 473005 4879585 300 0.0009 46.1 407 18.7 132 0.8 21.2 0.9 156 LNR‐1a (B)‐X M 08/11/2011 448565 4856845 226 0.0041 34.7 1398 73.9 565 3.3 21.2 3.5 157 MR‐3 (B) B 08/11/2011 455495 4845175 1836 0.0013 351.7 4396 69.1 80 0.8 169.3 0.4 158 MMR‐1a (B) B 08/11/2011 474715 4846705 603 0.0009 128.9 1186 36.1 46 0.6 100.7 0.4 159 MR‐6 (B) B 08/11/2011 493155 4858325 308 0.0011 70.4 774 35.2 132 0.5 100.7 0.3 NTS‐UTM Coordinates Zone 17N Appendix A ‐ Raw Floodplain Dataset Fluvial Process Variables Page 2 of 2

Easting Northing A d Q 2 Ω ω τ /D 95 ω /D 50 ω cr ID# Field Site K‐Means Sample Date S ω/ω cr (m) (m) (km2) (m3s‐1) (Wm‐1) (Wm‐2) (Nm‐3) (Wm‐3) (Wm‐2) 160 GR‐SPR‐UT‐1 (B) B 15/11/2011 556405 4838105 29 0.0053 5.7 303 51.8 194 0.8 100.7 0.5 161 GR‐SPR‐3 (B) B 15/11/2011 558575 4826305 236 0.0023 40.8 931 49.4 99 0.8 100.7 0.5 162 GR‐SPR‐4 (M, SM) M 15/11/2011 558035 4814835 635 0.0013 103.1 1272 38.8 279 2.4 12.6 3.1 163 GR‐NR‐9 (M) M 18/11/2011 540975 4795365 817 0.0009 130.7 1158 30.7 71 1.4 21.2 1.4 164 GR‐NR‐15 (M) M 18/11/2011 515025 4812455 328 0.0009 55.5 493 21.8 77 0.5 59.9 0.4 165 GR‐3 (B) B 18/11/2011 558475 4856725 658 0.0020 106.7 2032 60.8 45 0.7 169.3 0.4 166 GR‐4 (M, BM)‐X B 18/11/2011 542235 4830985 1136 0.0017 177.9 2886 63.6 120 0.7 169.3 0.4 167 TM‐N‐4b (C) C 05/12/2011 608425 4774935 117 0.0005 14.1 72 3.8 867 0.9 1.6 2.4 168 WL‐4b (C) C 05/12/2011 612595 4764155 222 0.0003 25.7 78 2.8 1062 0.7 1.6 1.8 169 LET1‐4 (C)‐X M 05/12/2011 577865 4745145 100 0.0013 11.7 151 11.8 595 0.7 12.6 0.9 170 GR‐FC‐3 (C) C 09/12/2011 567335 4777675 383 0.0004 43.1 271 10.9 1477 1.8 2.9 3.8 171 GR‐FC‐T3‐9 (C, CS) C 09/12/2011 564975 4786315 130 0.0002 15.5 44 3.2 335 0.8 1.6 2.1 172 STR‐2 (B) B 12/12/2011 525045 4784785 148 0.0018 31.9 551 35.4 121 0.6 100.7 0.4 173 STR‐3 (S) S 12/12/2011 518055 4775365 271 0.0012 54.3 614 29.6 2495 4.9 2.9 10.2 174 NSYD‐2 (C) C 12/12/2011 393225 4740505 530 0.0002 97.7 190 6.7 585 1.7 1.6 4.2 175 NSYD‐3 (C) C 12/12/2011 400365 4736265 201 0.0002 41.7 101 5.6 605 1.4 1.6 3.6 176 TR‐4 (S, SM) S 13/12/2011 442285 4721405 4028 0.0001 577.0 480 6.4 138 0.8 4.4 1.4 177 TR‐6 (M, BM) M 13/12/2011 465345 4750805 3373 0.0007 494.0 3377 49.2 98 0.8 100.7 0.5 178 TR‐DC (M) M 13/12/2011 471295 4753505 148 0.0015 31.9 469 30.1 118 131.3 21.2 141.4 179 BGC‐0 (S, CS) S 19/12/2011 545265 4759535 161 0.0013 19.2 251 16.2 3133 2.7 2.9 5.6 180 BGC‐1A (S, SM) S 19/12/2011 541915 4749065 265 0.0009 32.4 300 15.9 455 2.6 2.9 5.5 181 BGC‐2 (S, SM)‐X M 19/12/2011 540205 4740985 382 0.0017 47.5 793 36.2 392 1.6 21.2 1.7 182 BGC‐5 (S, SM) S 19/12/2011 537895 4722835 589 0.0003 74.8 237 9.1 891 1.5 2.9 3.2 183 KTC‐6 (M) M 20/12/2011 480695 4729995 366 0.0009 45.4 404 18.7 321 0.8 21.2 0.9 184 CFC‐6 (M, SM) M 20/12/2011 494915 4734335 247 0.0018 30.0 529 28.7 282 1.3 21.2 1.4 185 BOC‐5b (S) S 20/12/2011 511845 4731105 526 0.0006 66.3 374 15.0 1299 2.5 2.9 5.2 186 WD‐WV‐W2 (B, BM) B 26/04/2013 649375 4859985 5 0.0297 1.1 334 92.8 1454 2.1 59.9 1.6 201 JBT‐LR‐1 (C, CM) C 2011 638665 4868895 17 0.0061 3.3 195 32.9 725 4.1 4.4 7.4 202 JBT‐LR‐2 (M) M 2011 638945 4867705 22 0.0076 4.0 298 45.6 304 1.3 42.5 1.1 203 JBT‐LR‐3 (C) C 2011 641365 4863215 72 0.0009 11.4 103 9.6 262 2.4 1.6 6.1 204 JBT‐LR‐4u (BM, M)‐X B 2011 644115 4861035 87 0.0028 13.3 364 31.5 141 0.5 96.0 0.3 205 JBT‐LR‐4d (B, BM) B 2011 644305 4860925 87 0.0030 13.3 385 33.3 159 0.6 82.4 0.4 206 JBT‐LR‐5 (B) B 2011 646515 4854575 107 0.0085 16.0 1328 105.1 310 1.9 78.0 1.3 207 JBT‐LR‐6 (M) M 2011 649415 4852385 114 0.0035 16.9 577 44.6 446 1.4 33.9 1.3 208 TH‐WD‐1+2 (M) M 2013 654875 4857275 133 0.0028 19.3 531 38.4 235 1.3 32.3 1.2 209 TH‐WD‐R1 (M) M 2013 652195 4858105 129 0.0067 18.8 1234 90.5 590 2.0 59.9 1.5 210 TH‐ED‐R1 (B, BM) B 2013 655265 4863145 84 0.0045 13.0 573 50.1 119 1.6 35.6 1.4 NTS‐UTM Coordinates Zone 17N Appendix A ‐ Raw Floodplain Dataset Floodplain Variables 1 of 2

ID# Field Site K‐Means w (m) d (m) w/d F t (m) F fa (m) FSE (m) OM (%) M SC (%) M Sa (%) D 50 (m) D 95 (m) D 95 /d

1 EHR Bank 1.1 (M, SM) M 12.3 1.6 7.6 2.5 1.8 1.35 1.3% 24% 74% 0.02 0.06 0.037 1.4 EHR Bank 1.4 (M, SM) M 12.3 1.6 7.6 2.9 2.9 1.61 1.3% 43% 55% 0.02 0.06 0.037 1.7 EHR Bank 7 (M, SM) M 15.7 1.8 8.8 2.1 1.7 1.50 0.8% 11% 88% 0.02 0.06 0.034 2 MWC‐1 BX (M) M 24.0 1.8 13.3 1.5 1.0 0.67 2.1% 31% 67% 0.02 0.07 0.039 3 MWC‐2 B1 (M) M 16.9 1.5 11.3 2.7 1.4 1.06 1.7% 20% 79% 0.02 0.08 0.053 4FL Bank2 (M) M 11.5 1.1 10.5 1.1 0.9 0.53 1.8% 39% 59% 0.02 0.07 0.064 4FL Bank1 (M) M 9.7 1.1 8.8 1.0 0.8 0.60 1.4% 24% 74% 0.02 0.07 0.064 5.1 NWR‐S1 (S) S 25.3 2.5 10.1 3.0 2.8 2.05 3.6% 23% 73% 0.01 0.08 0.032 5.2 NWR‐S2 (S) S 27.5 2.5 11.0 3.0 2.8 2.18 1.1% 21% 78% 0.01 0.08 0.032 6.2 SG‐9b S2 (M)‐X B 59.0 3.0 19.7 3.5 2.3 1.87 1.2% 17% 81% 0.18 0.30 0.100 6.3 SG‐9d Stop3 (M) M 55.0 3.0 18.3 2.9 1.8 1.10 2.4% 36% 61% 0.05 0.20 0.067 7CF YM Bank1 (M) M 21.8 2.2 9.9 2.2 1.9 0.96 2.2% 47% 51% 0.02 0.08 0.036 8AR‐Bar‐15 DS (S, SM) S 37.2 5.8 6.4 4.2 4.2 2.65 2.1% 35% 63% 0.01 0.02 0.003 9GR Cayuga (C)‐X M 163.0 3.5 47.1 4.0 3.4 0.91 2.7% 71% 27% 0.02 0.05 0.014 10 AR‐5c (S, SM) S 32.0 3.0 10.7 5.0 4.0 1.98 2.1% 48% 49% 0.01 0.20 0.067 11 AR‐28b (S, SM) S 34.0 3.0 11.3 5.0 4.0 2.06 2.1% 46% 51% 0.01 0.04 0.013 12 AR‐33 (S, SM) S 26.0 5.0 5.2 5.0 4.0 1.63 2.0% 57% 41% 0.01 0.03 0.005 101 NWR‐S3 (M, SM) M 33.0 3.0 11.0 4.8 3.0 2.35 0.5% 21% 78% 0.05 0.50 0.167 102 ISC 3b (S, SM) S 9.0 2.5 3.6 2.6 2.5 2.45 0.5% 1% 98% 0.01 0.01 0.004 103 ICST2 (C, CM) C 8.0 2.4 3.3 2.5 2.4 0.41 4.7% 78% 17% 0.01 0.01 0.004 104 AR‐5b (M) M 38.0 3.0 12.7 4.0 1.8 1.22 1.8% 28% 70% 0.09 0.12 0.040 105 AR‐AC‐3 (C) C 9.5 2.3 4.1 2.3 2.3 1.14 4.9% 44% 51% 0.00 0.01 0.004 106 AR‐AC‐4 (C, CM) C 11.1 2.0 5.6 2.0 2.0 0.62 4.9% 64% 31% 0.00 0.01 0.005 108 AR‐UT4‐1 (M)‐X C 3.4 1.0 3.4 1.3 1.0 0.64 2.6% 30% 68% 0.01 0.10 0.100 109 BFR‐1 (B) B 15.3 0.6 25.5 0.8 0.4 0.20 4.2% 38% 58% 0.13 0.42 0.693 110 BBR‐4 (CM)‐X C 13.5 1.3 10.4 1.5 1.2 0.76 4.7% 32% 63% 0.01 0.10 0.077 111 BFR‐TC‐1 (C, BM) C 6.6 1.0 6.9 1.0 0.3 0.19 9.5% 28% 63% 0.01 0.10 0.105 112 BFR‐4 (B) B 12.5 0.7 18.1 0.8 0.7 0.12 7.8% 74% 18% 0.06 0.35 0.500 114 AR‐3 (B) B 8.9 0.7 12.7 0.7 0.5 0.25 5.5% 39% 55% 0.09 0.38 0.536 116 PHC‐5 (M, SM) M 12.2 2.3 5.4 1.5 1.3 1.05 2.0% 17% 81% 0.02 0.20 0.089 117 LAR‐7b (B) B 15.3 1.0 15.3 1.2 0.7 0.26 5.9% 57% 37% 0.06 0.30 0.300 118 PHC‐6b (C, CM) C 12.5 2.5 5.0 2.5 2.3 0.78 3.4% 62% 35% 0.01 0.05 0.020 119 AR‐13 (S) S 28.5 5.1 5.6 5.1 5.1 3.31 2.8% 32% 66% 0.00 0.01 0.002 120 AR‐18 (S) S 27.0 4.5 6.0 4.5 4.5 2.41 1.9% 45% 54% 0.00 0.01 0.002 121 NWR‐13 (B) B 9.5 1.0 9.5 1.0 0.3 0.15 15.5% 33% 51% 0.18 0.60 0.600 122 NWR9 (M)‐X B 12.0 1.5 8.0 1.5 1.0 0.86 2.9% 11% 86% 0.06 0.30 0.200 123 NWR‐3 (M) M 14.5 2.3 6.3 2.7 2.0 1.94 0.8% 2% 97% 0.02 0.05 0.022 124 NR‐1R (SM)‐X M 21.5 3.1 6.9 3.6 3.0 2.08 3.4% 27% 69% 0.02 0.30 0.097 125 ISC‐1R (S, SM)‐X M 7.5 2.4 3.1 2.4 2.0 1.56 1.2% 21% 78% 0.02 0.30 0.125 126 NR‐10 (S) S 32.0 4.7 6.8 4.7 4.7 4.47 0.5% 4% 95% 0.01 0.01 0.002 127 BYR‐2 (B) B 10.9 1.5 7.3 1.6 0.3 0.25 3.1% 13% 84% 0.18 0.50 0.333 128 BYR‐3 (B) B 12.8 0.8 16.0 0.9 0.4 0.26 5.4% 29% 66% 0.13 0.60 0.750 129 BYR‐7 (S) S 16.5 3.0 5.5 3.0 3.0 2.68 1.3% 9% 89% 0.01 0.01 0.003 130 PC‐1 (B)‐X M 15.6 1.6 10.1 1.7 0.5 0.38 1.9% 22% 76% 0.03 0.20 0.129 131 WC‐2 (S) S 11.5 2.5 4.6 2.5 2.5 2.19 1.1% 11% 88% 0.01 0.01 0.004 132 WC‐3 (M) M 11.1 1.7 6.5 2.0 1.0 0.95 0.7% 4% 95% 0.02 0.30 0.176 133 NR‐14 (S, CS) S 38.4 4.2 9.1 4.2 4.2 2.68 3.0% 33% 64% 0.00 0.01 0.002 134 NR‐18 (M) M 40.0 2.0 20.0 1.7 0.8 0.75 0.8% 5% 94% 0.03 0.40 0.200 135 MR‐1 (M) M 22.0 2.0 11.0 2.3 1.0 0.74 1.4% 25% 74% 0.02 0.20 0.100 136 PR‐4 (M, BM) M 13.2 1.0 13.2 1.3 1.0 0.60 4.9% 32% 64% 0.02 0.10 0.100 137 PR‐5b (B) B 9.8 0.9 10.9 0.9 0.1 0.06 16.9% 23% 60% 0.26 0.75 0.833 138 MR‐2 (B) B 17.5 1.3 13.5 1.3 0.4 0.20 5.6% 37% 58% 0.18 0.60 0.462 139 SGR3 (M)‐X B 73.3 2.8 26.2 2.8 2.3 1.75 1.9% 20% 78% 0.06 0.40 0.143 140 SR‐SKC‐1 (M, BM)‐X B 14.2 0.8 17.8 1.0 0.8 0.65 3.3% 16% 81% 0.06 0.30 0.375 141 NSR‐10 (B) B 16.8 0.9 18.7 0.9 0.8 0.31 5.2% 56% 38% 0.09 0.25 0.278 143 GR‐NR‐1b (B, BM) B 39.7 2.7 14.7 2.7 1.3 0.60 1.6% 52% 46% 0.06 0.25 0.093 144 NR5 (S, SM) S 42.6 3.0 14.4 3.0 1.7 1.37 1.5% 18% 81% 0.01 0.05 0.017 145 SR‐15d (B) B 24.1 0.8 30.1 0.9 0.2 0.17 4.0% 13% 83% 0.09 0.20 0.250 146 RSR‐1 (B) B 22.0 0.7 31.4 0.7 0.3 0.03 49.3% 40% 10% 0.13 0.30 0.429 147 SR‐13c (B, BM) B 25.7 2.0 13.2 2.0 1.2 0.91 3.0% 21% 76% 0.09 0.40 0.205 148 SR‐_DC‐3 (C) C 12.1 2.1 5.9 2.1 2.0 0.80 5.4% 54% 40% 0.00 0.01 0.005 149 SR‐DC‐4 (CM)‐X B 14.8 1.5 9.9 1.8 0.9 0.62 3.5% 28% 69% 0.06 0.20 0.133 150 TW‐0 (B) B 39.5 1.7 23.9 1.7 0.9 0.19 9.1% 70% 21% 0.09 0.40 0.242 151 TW2 (B) B 38.0 1.1 36.2 1.1 0.2 0.08 9.5% 51% 40% 0.06 0.30 0.286 152 SR‐9c (M) M 53.5 3.0 17.8 3.0 1.0 0.73 1.5% 25% 73% 0.05 0.20 0.067 153 SSG4 (B) B 27.1 1.2 22.6 1.2 0.2 0.13 4.5% 31% 65% 0.09 0.30 0.250 154 SSG6 (S, SM)‐X M 28.1 3.4 8.3 3.4 3.3 1.70 2.0% 47% 51% 0.03 0.40 0.118 Appendix A ‐ Raw Floodplain Dataset Floodplain Variables 2 of 2

ID# Field Site K‐Means w (m) d (m) w/d F t (m) F fa (m) FSE (m) OM (%) M SC (%) M Sa (%) D 50 (m) D 95 (m) D 95 /d

155 TW5 (M, SM) M 21.5 1.5 14.3 1.5 1.3 1.10 1.3% 11% 88% 0.02 0.10 0.067 156 LNR‐1a (B)‐X M 19.3 1.4 13.8 1.4 0.4 0.25 3.8% 33% 63% 0.02 0.10 0.071 157 MR‐3 (B) B 78.9 1.6 49.3 1.6 0.3 0.18 10.4% 29% 61% 0.09 0.25 0.156 158 MMR‐1a (B) B 26.4 1.5 17.6 1.7 0.3 0.11 8.2% 54% 37% 0.06 0.30 0.200 159 MR‐6 (B) B 22.5 1.2 18.8 1.4 0.3 0.14 7.1% 37% 55% 0.06 0.10 0.083 160 GR‐SPR‐UT‐1 (B) B 8.5 1.3 6.5 1.3 0.5 0.29 7.6% 28% 64% 0.06 0.35 0.269 161 GR‐SPR‐3 (B) B 26.3 1.3 20.2 1.3 0.6 0.42 5.7% 24% 71% 0.06 0.30 0.231 162 GR‐SPR‐4 (M, SM) M 34.2 2.3 15.2 2.3 2.0 1.25 8.3% 29% 62% 0.02 0.10 0.044 163 GR‐NR‐9 (M) M 38.7 2.4 16.1 3.0 2.0 1.52 1.6% 23% 76% 0.02 0.30 0.125 164 GR‐NR‐15 (M) M 24.8 2.6 9.5 2.6 2.2 1.77 1.3% 18% 80% 0.05 0.30 0.115 165 GR‐3 (B) B 34.2 0.7 48.9 0.7 0.2 0.08 6.7% 40% 54% 0.09 0.30 0.429 166 GR‐4 (M, BM)‐X B 53.5 2.2 24.3 2.3 1.5 0.93 4.4% 34% 62% 0.09 0.30 0.136 167 TM‐N‐4b (C) C 16.0 1.7 9.4 1.7 1.7 0.12 7.8% 85% 7% 0.00 0.01 0.006 168 WL‐4b (C) C 27.5 3.5 7.9 3.5 3.5 0.53 5.3% 80% 15% 0.00 0.01 0.003 169 LET1‐4 (C)‐X M 21.7 2.3 9.4 2.3 0.9 0.18 4.4% 75% 20% 0.02 0.05 0.022 170 GR‐FC‐3 (C) C 18.9 3.5 5.4 3.5 3.4 0.21 3.2% 91% 6% 0.01 0.01 0.003 171 GR‐FC‐T3‐9 (C, CS) C 14.1 1.8 7.8 1.8 1.8 1.02 1.4% 42% 56% 0.00 0.01 0.006 172 STR‐2 (B) B 16.6 1.4 11.9 1.4 0.6 0.20 5.9% 60% 34% 0.06 0.20 0.143 173 STR‐3 (S) S 22.1 2.2 10.0 2.3 2.2 1.81 1.5% 16% 82% 0.01 0.01 0.005 174 NSYD‐2 (C) C 29.5 3.0 9.8 3.0 3.0 0.80 4.3% 69% 27% 0.00 0.01 0.003 175 NSYD‐3 (C) C 15.5 2.5 6.2 2.5 2.5 0.30 5.1% 83% 12% 0.00 0.01 0.004 176 TR‐4 (S, SM) S 61.0 8.3 7.3 8.3 4.3 2.71 1.7% 35% 63% 0.01 0.05 0.006 177 TR‐6 (M, BM) M 62.0 5.0 12.4 5.0 3.5 3.21 1.1% 7% 92% 0.06 0.35 0.070 178 TR‐DC (M) M 16.2 1.6 10.1 2.0 1.3 0.90 2.3% 28% 70% 0.02 0.20 0.125 179 BGC‐0 (S, CS) S 16.6 2.4 6.9 2.4 2.4 1.93 4.2% 15% 80% 0.01 0.01 0.004 180 BGC‐1A (S, SM) S 16.1 2.5 6.6 2.5 2.4 2.31 0.6% 3% 96% 0.01 0.05 0.020 181 BGC‐2 (S, SM)‐X M 16.6 2.4 7.1 2.4 2.3 1.91 1.6% 15% 83% 0.02 0.10 0.043 182 BGC‐5 (S, SM) S 19.5 2.8 7.0 2.8 2.8 2.07 0.8% 25% 74% 0.01 0.01 0.004 183 KTC‐6 (M) M 26.9 3.6 7.5 1.5 1.0 0.88 1.2% 10% 88% 0.02 0.10 0.028 184 CFC‐6 (M, SM) M 19.9 3.2 6.2 3.2 2.9 2.74 0.9% 5% 94% 0.02 0.20 0.063 185 BOC‐5b (S) S 20.9 4.6 4.5 4.6 4.6 3.42 0.9% 25% 74% 0.01 0.02 0.004 186 WD‐WV‐W2 (B, BM) B 5.0 1.0 5.0 1.0 0.4 0.17 4.0% 52% 44% 0.05 0.20 0.200 201 JBT‐LR‐1 (C, CM) C 5.6 0.7 8.3 1.0 1.0 0.52 6.5% 41% 52% 0.01 0.06 0.082 202 JBT‐LR‐2 (M) M 9.4 0.7 14.5 1.0 0.8 0.43 3.0% 45% 52% 0.04 0.16 0.243 203 JBT‐LR‐3 (C) C 9.1 1.0 8.8 2.5 2.5 0.58 2.6% 75% 23% 0.00 0.04 0.035 204 JBT‐LR‐4u (BM, M)‐X B 13.4 0.7 18.3 1.7 0.6 0.40 1.8% 35% 63% 0.06 0.14 0.195 205 JBT‐LR‐4d (B, BM) B 13.6 0.7 18.7 1.8 0.6 0.36 2.3% 41% 57% 0.06 0.13 0.182 206 JBT‐LR‐5 (B) B 17.5 0.7 24.9 1.4 0.2 0.12 1.6% 39% 60% 0.05 0.19 0.267 207 JBT‐LR‐6 (M) M 20.5 0.8 25.3 2.4 2.4 1.18 1.6% 49% 49% 0.03 0.06 0.077 208 TH‐WD‐1+2 (M) M 16.2 0.9 18.9 1.6 1.6 0.51 3.8% 63% 33% 0.03 0.10 0.117 209 TH‐WD‐R1 (M) M 12.4 0.9 13.8 0.9 0.9 0.52 2.0% 38% 60% 0.05 0.10 0.111 210 TH‐ED‐R1 (B, BM) B 8.5 0.3 31.5 0.3 0.3 0.05 9.0% 72% 19% 0.03 0.10 0.370 South Saugeen River 44° 01' 15" N, 80° 55' 09" W River South Saugeen ID#: 153 43° 08' 54" N, 81° 11' 32" W Thames River B B B-Type Floodplains B-Type Appendix A: Boyne River (Nottawasaga) 44° 09' 30" N, 80° 04' 12" W (Nottawasaga) Boyne River ID#: 128 Grand River 43° 51' 09" N, 80° 16' 53" W River Grand ID#: 165 B B

Appendix A 191 Nottawasaga River 44° 06' 54" N, 79° 52' 10" W River Nottawasaga ID#: 123 Catfish Creek 42° 44' 49" N, 81° 03' 29" W Creek Catfish ID#: 7 M M M-Type Floodplains M-Type Appendix A: Mad River (Nottawasaga) 44° 18' 19" N, 80° 00' 11" W (Nottawasaga) Mad River ID#: 135 Saugeen River 44° 10' 59" N, 81° 11' 36" W River Saugeen ID#: 152 M M

Appendix A 192 Nottawasaga River 44° 17' 23" N, 79° 50' 37" W River Nottawasaga ID#: 126 Fairchild Creek (Grand) 43° 09' 30" N, 80° 10' 12" W (Grand) Creek Fairchild ID#: 170 S C S/C-Type Floodplains S/C-Type Appendix A: Innisfil Creek (Nottawasaga) 44° 05' 55" N, 79° 46' 51" W (Nottawasaga) Creek Innisfil ID#: 102 Sydenham River 42° 48' 35" N, 82° 18' 23" W River Sydenham ID#: 174 S C

Appendix A 193

Appendix B OSL Testing

194

Appendix B Optically Stimulated Luminescence Dating Potential of Quartz for Holocene Alluvial Deposits in the Southern Laurentian Great Lakes Glaciated Region

B Introduction

Optically stimulated luminescence (OSL) dating methods using quartz (and feldspar) sands have emerged in Quaternary geochronology as an effective alternative to traditional radiocarbon

techniques. However, use of quartz‐OSL dating for alluvial floodplains, and fluvial systems in

general, requires special attention to the issues of: 1) partial solar resetting; and 2) the

occurrence of dim‐quartz. To test the viability of quartz‐OSL in southern Ontario, six (6) OSL core samples were collected from three (3) river banks in southern Ontario for comparison with radiocarbon ages, including sites on the Ausable, Nottawasaga, and Saugeen rivers (Figure B.1).

Figure B.1: Map of southern Ontario study area shown the locations of three OSL sampling sites in association with radiocarbon ages.

Appendix B 195 B.1 Luminescence Dating

Luminescence dating of sediments relies on the build‐up of a luminescence signal within the

crystalline mineral lattice of geologic materials such as quartz and feldspar from natural

irradiation sources (Figure B.2); the luminescence signal being photon emissions yielded by

energy release from electron traps when stimulated by sunlight. The longer the sediment is buried and exposed to low‐level background radiation (which recharges election traps), the

stronger the luminescence signal will be when stimulated in the laboratory, providing a

measure of the length of time since it was naturally stimulated (or “reset”) by sunlight. Thus,

the luminescence signal can be used as a dating method for the time since “last” burial of

fluvially deposited sediments (as well as other geomorphic and depositional environments).

Laboratory luminescence dating techniques typically use optical light (OSL) or infrared (IRSL) to

stimulate the luminescence signal in quartz and feldspar sand grains, respectively.

Figure B.2: Cycle of luminescence signal in sediments during fluvial erosion, transport, deposition, and burial with recurring sunlight exposure. i) Solar resetting and build up of luminescence signal after burial. ii) Process of solar resetting for luminescence signal of sediments in a fluvial system – Saugeen River image (Google EarthTM).

Appendix B 196 B.2 Field Sites – OSL Core Samples

Six (6) sediment cores were collected from three (3) field sites (Figures B.1 and B.3) for quartz‐

OSL dating at the UIC Luminescence Dating Research Laboratory. Samples were collected in the field in a “light‐tight” container (2 inch dia PVC pipe) for shipment to the UIC lab. For each field

site, two (2) OSL core samples were collected, with sample ‘A’ being older than sample ‘B’

based on stratigraphic superposition (Sites 1 and 3) or river terracing (Site 2). At least one

radiocarbon age was also obtained for each field site from shell, wood, or charred organic

material within each alluvial sequence.

Figure B.3: Photographs for three river bank field sites showing locations of six OSL samples collected on the Ausable, Saugeen, and Nottawasaga rivers.

Appendix B 197 B.3 Laboratory Methods

Each sediment sample was pretreated to isolate quartz fractions (Figure B.4). Steps included magnetic separation, HCL acid wash for organics, sieving to 150–250 microns, density separation using the heavy liquid Na polytungstate, and 40‐min immersion in HF acid to etch outer 10+ microns of quartz grains. Quartz separates were then evaluated by petrographic inspection, and samples showing > 1% non‐quartz minerals were retreated with HF. Single‐ aliquot regenerative (SAR) protocols (Murray and Wintle, 2000) were used to estimate the equivalent dose (De) of the 150–250 micron quartz separates with an automated Risø TL/OSL‐

DA‐15 system (Bøtter‐Jensen et al., 2000). The dose rate (DR) for natural exposure of the

sediment to ionizing radiation is typically from U and Th decay series, 40K, and cosmic sources

(see notes for Table B.1). The luminescence age of the sample is determined by the equivalent dose (De) divided by the dose rate (DR).

Figure B.4: Quartz separates.

Appendix B 198 B.4 Quartz-OSL Results

Six OSL samples were collected at three field sites where radiocarbon ages were also obtained

(Table B.1): 1) Ausable River; 2) Saugeen River; and 3) Nottawasaga River. For each field site,

two OSL core samples were collected with sample ‘A’ being older than sample ‘B’ based on

stratigraphic superposition (Sites 1 and 3) or river terracing (Site 2). Dating for three of the six

quartz‐OSL samples (1A, 1B, 3B) showed reasonable concordance with late Holocene

radiocarbon ages (Table B.1). However, partial solar resetting was confirmed as a major

limitation to the dating accuracy of quartz‐OSL independent of radiocarbon results. Specifically, over‐dispersion values of >100% were observed (with < 20% being optimal), despite small aliquot sizes (10–20 grains) and sampling of 50–70 aliquots for each OSL core (Figure B.5‐1B).

The remaining three quartz‐OSL samples (2A, 2B, 3A) did not provide reliable equivalent dose

estimates from blue‐light stimulation due to low photon emissions (< 400 counts), and thus fit

the description of “dim” quartz as described by others for this region (Lamarche et al., 2007).

Luminescence was observed in one quartz sample (Figure B.5‐3A) using infrared stimulation

(instead of blue‐light), which is hypothesized to be due to microscopic “feldspatic” inclusions

contained within the quartz sand separates (Huntley et al., 1993). However, the infrared

luminescence age of this sample was not concordant with the radiocarbon date or the paired

OSL date (3B) based on stratigraphic superposition (Table B.1).

Appendix B 199

Figure B.5: OSL/IRSL radial plots. 1B) Ausable River quartz‐OSL age with >100% overdispersion. 3A) Nottawasaga River quartz‐IRSL age (inclusions) with 33% overdispersion.

Table B.1: OSL ages compared to 14C dating results

Appendix B 200 While successful quartz‐IRSL dating of feldspar inclusions has been previously demonstrated

(Huntley et al., 1993), explanations of the luminescence and mineralogical properties of dim‐ quartz, and consequently insights on its spatial distribution, are not well documented in literature. A leading theory may be that dim‐quartz is commonly associated with glaciogenic

deposits and recently deglaciated areas (Rhodes, 2000; Duller, 2006; Rhodes, 2011), possibly due to relatively low levels of sand maturity in these areas.

B.5 Elemental Analysis of Quartz Separates

To test the possibility of a mineralogical explanation for dim quartz (e.g., feldspatic inclusions and crystal lattice imperfections), a sub‐sample of the Quartz separates (% SiO2 ~ 90 ‐ 95%) for

each OSL sample was submitted for elemental analysis by ICP‐MS. Dim‐quartz from Sample 3A

showed distinctly higher percentages of feldspar analytes, including Al2O3, Na2O, and K2O

(Figure B.6), as compared to the bright‐quartz samples (1A, 1B, 3B). However, subsequent

elemental analysis of samples 2A and 2B suggests that these results are inconclusive and the

issue of limited solar resetting cycles of “young” sands from recently eroded bedrock in glacial

landscapes may be another major reason for low Quartz sensitivity (Forman, S., personal communication).

Figure B.6: Percent content of select analytes for quartz separates from ICP‐MS.

Appendix B 201 B.6 OSL Conclusions

Previous quartz‐OSL dating in the continental region of the Laurentian Great Lakes has been demonstrated with some success (Lepper et al., 2007; Schaetzl and Forman, 2008) and some rejection (Lamarche et al., 2007). The results of this study indicate that quartz‐OSL dating methods for Holocene alluvial deposits within the study area are possible; however,

independent credible ages require special consideration of:

1) Partial solar resetting: smaller aliquot sizes (<10 grains) and potentially 100–1000 aliquot samples may be required to statistically overcome the issue of partial solar resetting. Single quartz‐grain aliquot approaches may be required (Rittenour, 2008; Rhodes, 2011).

2) Dim‐quartz: while the occurrence of dim‐quartz was observed in half of the samples tested, more research is required to assess the spatial distribution, mineralogical characteristics, and luminescence properties of this phenomenon.

The occurrence of dim‐quartz luminescence may be associated with deglaciated areas, and thus

Holocene alluvial deposits of the Laurentian Great Lakes region are likely to consist of recently

reworked glaciogenic deposits where the viability of quartz‐OSL approaches may be limited.

However, successful quartz‐OSL dating of half the samples in this study indicates that some

sediments may contain quartz which is bright enough to allow for quartz‐OSL dating.

Uncertainties related to the viability of quartz‐IRSL ages from feldspar inclusions and the role (if

any) of feldspatic inclusions in the occurrence of dim‐quartz remain unresolved.

Appendix B 202 References Bøtter‐Jensen, L., Bulur, E., Duller, G.A.T., Murray, A.S. 2000. Advances in luminescence instrument systems. Radiation Measurements, 32: 523–528.

Duller, G.A.T. 2006. Single grain optical dating of glacigenic deposits. Quaternary Geochronology, 1: 296–304. DOI: 10.1016/j.quageo.2006.05.018

Galbraith, R.F., Roberts, R.G., Laslett, G.M., Yoshida, H., and Olley, J.M. 1999. Optical dating of single and multiple grains of quartz from Jinmium rock shelter, northern Australia, part 1, Experimental design and statistical models. Archaeometry, 41: 339–364.

Huntley, D.J., Hutton, J.T., Prescott, J.R. (1993). Optical dating using inclusions within quartz grains. Geology, 21: 1087–1090.

Lamarche, L., Bondue, V., Lemelin, M.J., Lamothe, M., and Roy, A.G. 2007. Deciphering the Holocene evolution of the St. Lawrence River drainage system using luminescence and radiocarbon dating. Quaternary Geochronology, 2: 155–161.

Lepper, K., Fisher, T.G., Hajdas, I., and Lowell, T.V. 2007. Ages of the Big Stone Moraine and the oldest beaches of glacial Lake Agassiz: Implications for deglaciation chronology. Geology, 35: 667–670.

Moore, T.C., Jr., Rea, D.K., Godsey, H. 1998. Regional variation in modern radiocarbon ages and the hardwater effects in Lakes Michigan and Huron. Journal of Paleolimnology, 20: 347– 351.

Murray, A.S., and Wintle, A.G. 2003. The single aliquot regenerative dose protocol: potential for improvements in reliability. Radiation Measurements, 37: 377–381.

Prescott, J.R., and Hutton, J.T. 1994. Cosmic ray contributions to dose rates for luminescence and ESR dating: large depths and long‐term time variations. Radiation Measurements, 23: 497–500.

Rhodes, E.J. 2011. Optically simulated luminescence dating of sediment over the past 200,000 years. Annual Review of Earth and Planetary Sciences, 39: 461–488.

Rhodes, E.J. 2000. Observations of thermal transfer OSL signals in glacigenic quartz. Radiation Measurements, 32: 595–602. Appendix B 203 Rittenour, T.M. 2008. Luminescence dating of fluvial deposits: applications to geomorphology palaeoseismic and archaeological research. Boreas, 37: 613–635.

Schaetzl, R.J. and Forman, S.L. 2008. OSL ages on glaciofluvial sediment in northern Lower Michigan constrain expansion of the Laurentide ice sheet. Quaternary Research, 70: 81– 90.

Acknowledgements

This research was supported by the University of Toronto Centre for Global Change Science (CGCS) with a graduate student research grant.

Luminescence dating was by, and in collaboration with, Dr. Steve L. Forman at the University of Illinois at Chicago (UIC) Luminescence Dating Research Laboratory.

Appendix B 204

Appendix C Interdisciplinary Conference Presentations 2009‐2012

Appendix C.1 Presentation Latornell Conference November 15, 2012

205

2. 4. Appendix C.1 Presentation Latornell Conference November 15th, 2012 1. 3. Appendix C.1 206 5. 6.

7. 8. Appendix C.1 207 9. 10.

11. 12. Appendix C.1 208 13. 14.

15. 16. Appendix C.1 209 17. 18. Appendix C.1 210

Appendix C.2 Presentation Natural Channel Systems Conference September 27–28, 2010

211

Appendix C.2 Presentation Natural Channel Systems Conference September 27 - 28, 2010 1. 2.

3. 4.

5. 6.

Appendix C.2 212 7. 8.

9. 10.

11. 12.

Appendix C.2 213 13. 14.

15. 16.

17. 18.

Appendix C.2 214 19. 20.

21. 22.

23.

Appendix C.2 215

Appendix C.3 Presentation CAG‐ONT Conference October 12–13, 2012

216

2. 4. Appendix C.3 Presentation CAG-ONT October 12 - 13, 2012 1. 3. Appendix C.3 217 5. 6.

7. 8. Appendix C.3 218 9. 10.

11. 12. Appendix C.3 219 13. 14.

15. 16. Appendix C.3 220

Appendix C.4 Presentation AGU Conference May 24–27, 2009

221

2. 4. Appendix C.4 Presentation AGU Conference May 24 - 27, 2009 1. 3. Appendix C.4 222 5. 6.

7. 8. Appendix C.4 223 9. 10.

11. 12. Appendix C.4 224