UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Watershed responses to timber harvesting disturbance Campbell, David Andrew 2005

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2005-0172.pdf [ 19.06MB ]
Metadata
JSON: 831-1.0091815.json
JSON-LD: 831-1.0091815-ld.json
RDF/XML (Pretty): 831-1.0091815-rdf.xml
RDF/JSON: 831-1.0091815-rdf.json
Turtle: 831-1.0091815-turtle.txt
N-Triples: 831-1.0091815-rdf-ntriples.txt
Original Record: 831-1.0091815-source.json
Full Text
831-1.0091815-fulltext.txt
Citation
831-1.0091815.ris

Full Text

WATERSHED RESPONSES TO TIMBER HARVESTING DISTURBANCE by D A V I D A N D R E W C A M P B E L L B.Sc. (Hons.), The University of British Columbia, 1999 A THESIS SUBMITTED IN PARTIAL F U L F I L L M E N T OF THE REQUIRMENTS FOR THE DEGREE OF M A S T E R OF SCIENCE in THE F A C U L T Y OF G R A D U A T E STUDIES (Department of Geography) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH C O L U M B I A March 2005 © David Andrew Campbell, 2005 A b s t r a c t In mountainous watersheds, such as those in the Coast Mountains of British Columbia, sediment mobilized on hillslopes can be delivered directly to stream channels. In coupled watersheds timber harvesting operations increases landslide sediment production on hillslopes, and decreases channel stability where riparian logging affects bank strength or sediment supply is increased due to landslide delivery. This report documents an investigation of hillslope and stream channel responses to timber harvesting in 119 reaches of 7 study watersheds in the Coast and Cascade Mountains of British Columbia. Air photos and GIS were used to document landslide and stream channel changes over a 40 to 50 year period within each study basin. Response times, response magnification, and relaxation times were calculated for hillslope and stream channel responses. Five response regimes are proposed to explain watershed response mechanisms; no response, riparian-driven responses, landslide-driven responses, propagated response, and compound responses. Channel change was considered to be significant if width changes exceeded 5.0 m and 7% of the initial width. The 46 reaches that did not exceed these criteria are considered to have no significant response. Riparian-driven responses were observed in 40 reaches, and average channel widening was 94% of initial channel width. Landslide-driven responses were observed in 16 reaches, and average channel widening was 86%. Propagated responses were minor (8 reaches), though they exhibited the largest amount of channel width increase (179%). Channels that exhibited compound responses had characteristics similar to the other three response types, and therefore the mechanism of response was difficult to discern. Overall 17 reaches exhibited compound response, and the average channel width increase was 120%. Average response times ranged from 17.4 to 22.5 years for all channel response types, and response time was 16.4 years for landslide responses Relaxation times were 13.8 years for landslides, 14.2 years for landslide-driven channel responses, and 28.4 years for riparian-driven responses. Confined channels are less prone to significant channel responses than channels with developed alluvial floodplains. An important issue for land management is the potential for disturbances to propagate to undisturbed reaches downstream. Table of Contents Abstract ii Table of Contents iv List of Figures vii List of Tables x List of Appendices xi Acknowledgements xii 1 Introduction 1 1.2 Context 1 1.2.1 Definitions .• 1 1.2.2 The Coupled Watershed as a Geomorphic System 2 1.2.3 Disturbance Mechanisms... 3 1.2.4 Timber Harvesting and Watershed Responses 6 1.3 Theoretical Framework 9 1.4 Objectives 11 2 Study Sites 14 2.1 Introduction 14 2.2 Selection Criteria 14 2.3 Study Site Locations and Watershed Characteristics 16 2.4 Reach Characteristics 26 3 Methods 31. 3.1 Introduction 31 iv 3.2 Aerial Photograph Imaging 32 3.3 Hillslope Sediment Production Measurements 35 3.4 Stream Channel Measurements 38 3.5 Field Methods 41 4 Error Analysis 44 4.1 Introduction 44 4.2 Sources of Error 45 4.3 Digital Image Processing and Geo-referencing 46 4.4 Measurement Precision 50 4.5 Discussion 55 5 Results.. 57 5.1 Introduction 57 5.2 Landslide Responses 63 5.3 Hydroclimatic Factors 68 5.4 Discussion 68 6 Channel Response Regimes 71 6.1 Introduction 71 6.2 No Significant Channel Changes (Type I) 71 6.3 Riparian Disturbance Response (Type II) 76 6.4 Landslide Coupling Disturbance Response (Type III) 82 6.5 Disturbance Propagation Response (Type IV) 8 7 6.6 Compound Responses (Type V) 93 6.7 Discussion 98 v 7 Conclusion . .108 7.1 Introduction . . .108 7.2 Methods and Error: Applying GIS for long-term monitoring of landslides and stream channels 108 7.3 Watershed responses to timber harvesting disturbance .." 110 7.4 Future research directions 112 8 References . .115 vi List of Figures Figure 1.1 Theoretical response of a geomorphic system to disturbance 10 Figure 2.1. Overview map of British Columbia and the location of study sites 18 Figure 2.2 Map of Cascade Creek and location of study reaches 20 Figure 2.3 Map of Chapman Creek and locations of study reaches 21 Figure 2.4 Map of Cedarflat and Dewdney Creeks and location of study reaches. .22 Figure 2.5 Map of Norrish Creek and locations of study creeks 23 Figure 2.6 Map of Slesse Creek and locations of study reaches 24 Figure 2.7 Map of Theodosia River and the locations of study reaches 25 Figure 2.8 Distribution of channel morphology within study sites 27 Figure 2.9 Initial channel width to drainage basin area. 28 Figure 2.10. Distribution of disturbance regimes within study basins 30 Figure 3.1. Summary of aerial photograph coverage of study sites. 32 Figure 3.2 An example of geo-referencing procedure 34 Figure 3.3 Measurements made for landslide areas 36 Figure 3.4. Landslide width and depth measurements for Centre Creek, Chapman Creek, Nesaquatch Creek, and Theodosia River 38 Figure 3.5 Measurements made for stream channel assessment 39 Figure 3.6 Field observations of channel aggradation and landslide delivery 42 Figure 4.1 Distribution of the number of control points used for image geo-referencing .47 Figure 4.2 Sampling distribution of geo-referenced image pixel resolution ....47 Figure 4.3 Standard error of estimate of reach width 52 vii Figure 4.4 Fractional width error of estimate of reach width 53 Figure 4.5. Fractional standard error of estimate for replicate landslide area measurements 55 Figure 5.1 Annual Maximum Daily Precipitation and Annual Peak Discharge for stations at or near Study Sites 61 Figure 5.2 Coupled landslide frequency for study watersheds 64 Figure 6.1. Variable distributions for Type I response reaches 72 Figure 6.2 Type I Response-Chapman Reach 10 73 Figure 6.3 Airphotos (1957-1998) of Chapman Reach 10 74 Figure 6.4. Variable distributions for Type II response reaches 77 Figure 6.5 Channel widening in Type II response reaches 78 Figure 6.6 Response Type II- Theodosia River Reach 4 79 Figure 6.7 Air photo coverage for Theodosia Reach 4 (1947-1996) 80 Figure 6.8. Variable distributions for Type III response reaches 83 Figure 6.9 Channel Widening in Type III Response Reaches 84 Figure 6.10 Response Type III- Cascade Creek Reach 14 85 Figure 6.11 Air photo coverage for Cascade Creek Reach 14 (1954-1993) 86 Figure 6.12. Variable distributions for Type IV response reaches 89 Figure 6.13 Type IV Propagation Response-Dewdney Creek Reaches 3 to 6.. 90 Figure 6.14 Air photo coverage for Dewdney Creek Reaches 3 to 6 (1954-1996). .91 Figure 6.15 Variable distributions for Type V response reaches 94 Figure 6.16 Type V Compound Response-Chapman Reach 20 95 Figure 6.17 Air photo coverage for Chapman Creek Reach 20 (1957-1998) 96 viii Figure 6.18 Distribution of disturbance types and channel morphology 100 Figure 6.19 Channel width magnification for all reaches 101 Figure 6.20 Riparian disturbance response variable distributions 104 Figure 6.21 Landslide disturbance and channel responses 105 ix List of Tables Table 2-1 Summary of study watershed characteristics 19 Table 5-1 Summary of landslide response and relaxation times for study basins 67 Table 5-2 Environment Canada Climate Station Location Summary 58 Table 5-3 Environment Canada Streamflow Gauging Station Locations 60 Table 5-4 Timing of extreme precipitation and discharge events for study sites 69 Table 6-1 Summary of Response Characteristics 102 x List of Appendices Appendix 1. Summary of reach characteristics 124 Appendix 2. Response curves for study reaches 131 xi Acknowledgements This thesis would not have been possible without the influence and support of many individuals. Sincere thanks must be first given to Michael Church his supervision of this project, as well as his recognition and support of my academic potential. I would also like to recognize the other members of my committee; Olav Slaymaker for thoughtful comments and review of the preliminary manuscript, and Man/van Hassan for his ongoing enthusiasm and encouragement. I would like to thank Bruce Thompson and Mike Miles for their assistance in locating potential study sites, as well as for providing air photos for use in analysis. I must extend a further appreciation to both gentlemen for their sincere encouragement and support they provided. Useful discussions with Tom Millard and Richard Guthrie were also helpful for identifying study sites and for project development. I would like to thank Kevin Tabata, Jon Tunnicliffe, and Andre Zimmermann for their assistance in the field. I would also like to thank Jon for his technological expertise throughout the project. Extended thanks go to Erica Ellis for her collegial support throughout the project. xii 1 Introduction This thesis represents an attempt to understand the linkages between hillslope and stream channel processes, and to examine how these linkages affect long term responses of watersheds to timber harvesting. The current chapter provides the background information of the research. It begins with a review of previous research and identifies research gaps in order to develop research goals. The review includes a background of geomorphic processes in steep, forested watersheds, and how timber harvesting can constitute a disturbance to these systems. From knowledge gained from previous work, a conceptual framework is developed. 1.2 Context 1.2.1 Definitions Chorley and Kennedy (1971) define a system as "a structured set of objects and/or attributes," which, "consist of components or variables that exhibit discernible relationships with one another and operate together as a complex whole". A well defined geomorphic system, for example, can be the collection of physical matter included within that system, such as water and sediment, and the collection of / forces acting on that matter, such as gravity. Geomorphic systems are governed by the physics which balance the driving and resisting forces within a landscape system. Changes to the driving or resisting forces can result in changes to the type or character of the geomorphic processes. Any change to the driving or resisting forces within a given system can be considered to be a disturbance. The effects of 1 a disturbance on a system depend on the nature, magnitude, and timing of the disturbance. The changes to the system as a result of the disturbance are the response. Depending on the sensitivity of the system and the nature of the disturbance, the response may result in a permanent change to the driving and resisting forces of the system, and may therefore cause the system to conform to a new state. The sensitivity of a system refers to how susceptible the system is to be affected by imposed forcings. If the changes are temporary, then the system may recover to a condition similar to its pre-disturbance state. The capacity for the system to be able to recover from disturbance events is the resilience. 1.2.2 The Coupled Watershed as a Geomorphic System A watershed can be usefully defined as a geomorphic system. The system consists of the collection of geomorphic processes and landform features that occur within the watershed. Within a watershed, sediment is moved from one point on the landscape to another. An understanding of the watershed as a geomorphic system requires an understanding of processes within the watershed that are responsible for sediment production (Dietrich et al., 1982). Sediment production is the mobilization of sediment on the landscape, and sediment transfer is the movement of sediment from one location to another. In the Pacific Northwest, the headwater portions of drainage basins are typically steep areas with high relief, and sediment transfers are dominated by hillslope processes. In headwater areas, hillslope processes can be coupled to stream systems (Church, 1983), meaning that sediment transfer can occur directly from hillslopes to the stream system. Higher 2 order streams are distanced from hillslopes, and valley-flats act to buffer sediment transfer from hillslopes. Here the stream system is "de-coupled" from hillslope processes (Ibid.). In headwater systems of mountainous watersheds, then, it is important to understand the interactions between hillslope and stream channel sediment transfers in order to understand the watershed system. Within a given region, the processes and rates of sediment production are a function of the geological (and geomorphic) history, the hydroclimate and the nature of the local disturbance regime (Church et al., 1999). The geologic history dictates the distribution and character of rock material in an area. Similarly, the geomorphic history determines the shape and nature of topography, as well as the distribution of surficial materials. The hydroclimatic regime determines the distribution of water in both space and time. This in turn determines the amount of water available to contribute to erosional processes and to influence mass wasting events. There are also feedbacks amongst these driving forces. For instance, topography has a strong influence on microclimate, and hence erosional processes (Gimbarzevsky, 1988). 1.2.3 Disturbance Mechan i sms The governing conditions of hillslope stability can be examined in a factor of safety (F) approach, such as the Coulomb equation: _ resisting forces c' + (y - myw )z cos2 /? tan <j) driving forces yz sin /3 cos /3 where c' is the effective soil cohesion, y is the unit weight of soil, yw the unit weight of water, z is the soil depth, m is the fraction of the soil depth which is saturated, /? is 3 the slope angle, and </>' is the internal friction angle of the soil. Hillslope instability is driven by gravity and pore water pressure, and therefore slope angle and water content are key components of the hillslope system. Depth of soils is based on thickness of unconsolidated soil material above a highly consolidated material or bedrock. This is generally determined by the amount of pedogenic development or by the history of surficial material deposition. On forested hillslopes, strength from root networks increases the shear strength of soils (Sidle, 1991). Soil properties, such as grain size distribution, consolidation, and silt and clay content, as well as the material friction angle, can affect the cohesion of soils. Mountainous watersheds in western British Columbia have topography with steep relief, variable spatial distribution of glacial drift material, and perhumid climate. All of these factors help define the geomorphic system within this area. Studies of erosional processes in the Western Cordillera suggest that, due to these factors, debris slides and debris flows are the important processes of hillslope sediment production and transfer (Slaymaker and McPherson, 1977; Dietrich and Dunne, 1978; Roberts and Church, 1986; Martin et al., 2002; Guthrie and Evans, 2004). In the Coast Mountains of British Columbia, tectonic uplift has created major mountain relief. Repeated glaciations have dramatically shaped the landscape, and have both created steep topography and deposited unstable drift material (Clague et al., 1989). Thin drift material deposited on top of competent material, such as bedrock or basal till, can create a plane of weakness on which unconsolidated material fails (Martin et al., 2002). Slopes can be over-steepened due to glaciation, and increased landslides can occur due to de-buttressing (Holm era/., 2004). Weak 4 bedrock, such as is found in volcanic areas, can also contribute to hillslope instability (Bovis and Jakob, 2000; Friele and Clague, 2004). The hydroclimate of coastal British Columbia is characterized by infrequent intense storm events (Hogan and Schwab, 1990; Septer and Schwab, 1995) and these events, combined with antecedent soil moisture conditions, may be a primary control over landslide initiation (Sidle era/., 1985; Schwab, 1998; Bovis and Jakob, 2000; Jakob etal., 2000; Martin et al., 2002; Guthrie and Evans, 2004). Stream channel systems function in a manner similar to hillslope systems. Processes that occur reflect the balance of driving and resisting forces. Stream channel dynamics are affected by discharge, sediment caliber and character, slope, bank strength and form resistance due to channel morphology and bed structures (Millar and Quick, 1993; Church etal., 1998; Eaton etal., 2004). Headwater streams in mountainous regions tend to have high gradients. In forested stream channels root strength can provide added bank strength (Millar, 2000) and increase channel stability. In steep watersheds, hillslope sediment production can deliver sediment directly to stream channels (Benda and Dunne, 1987; Benda and Dunne, 1997; Nakamura etal., 2000; Martin etal., 2002) and, where coupling occurs, hillslope sediment delivery can affect stream dynamics. Sediment delivery to fluvial systems can affect the quantity and size distribution of channel sediment (Rice and Church, 1996; Buffington, 1999;). Increased sediment loads due to hillslope delivery can also have strong control over channel pattern and channel stability (Miller and Benda, 2000; Sutherland et al., 2002; Clague et al., 2003). 5 At a watershed scale, geomorphic processes occur both on hillslopes and in stream channels, and the intersection of these processes occurs in coupled regions. In a study from the Deer Creek Basin, northwest Washington, Eide (1990) suggested that landslides (debris slides and debris flows) account for 74% of the total sediment production within the basin. Campbell and Church (2003) found that hillslope processes dominated the contemporary and Holocene sediment budgets for the Lynn Valley watershed in the Coast Mountains of British Columbia. Stream bank erosion is also a major process of sediment production. Roberts and Church (1986) suggested that landsliding events and stream bank erosion together are the dominant processes of sediment production in watersheds in the Queen Charlotte Islands. Other processes of sediment production can include soil creep (Dietrich et al., 1982; Roberts and Church, 1986), other mass movements such as earth flows (Eide, 1990) and snow avalanches, surface erosion of landslides scars (Hudson, 2001) and tree-throw (Hudson and D'Anjou, 2001). Local variation of controlling factors, such as the distribution and extent of glacial deposits, or local geography, can determine the relative importance of certain processes (e.g. Eide, 1990; Friele and Clague, 2004). 1.2.4 Timber Harvesting and Watershed Responses Timber harvesting activities have the potential to affect watershed geomorphic processes. Road construction, timber removal operations, and loss of timber from slopes can affect hillslopes and stream channels. Forest management operations in British Columbia have impacted local geomorphic processes by 6 affecting the governing conditions of these geomorphic systems (Slaymaker, 2000). The sensitivity of hillslopes and stream channels to timber harvesting disturbances is determined by the degree to which the governing conditions are affected, and how susceptible the system is to respond to those changes. On hillslopes, road networks change local slope gradient and structure, and can re-route drainage. On steep terrain these factors have led to dramatic increases in landslide frequency (Swanson and Dyrness, 1975; Rood, 1984; Guthrie, 2002). Roads also introduce surface erosion to the sediment production and can be a significant source of sediment (Reid etal., 1981; Wemple etal., 2001; Hudson, 2001). Similarly, changes to hydrologic routing and loss of root strength due to root decay have led to increases in landsliding on logged slopes (Jakob, 2000; Brardinoni etal., 2003; Roberts etal., 2004). Regionally focused studies of logging impacts have examined sediment production in the Queen Charlotte Islands (Roberts and Church, 1986; Rollerson, 1992; Martin etal., 2002), Vancouver Island (Jakob, 2000; Guthrie, 2002), the Coast Mountains of British Columbia (Millard, 2002; Brardinoni etal., 2002, northwest Washington (Eide, 1990; Paulson, 1997) and Oregon (Swanson and Dyrness, 1975). Although there is variation between studies, local landsliding rates in general appear to be roughly on the order of 10-102 m3/km2/yrfor natural systems, and 102-103 m3/km2/yr for logged terrain. Logging appears to accelerate landsliding rates by about an order of magnitude (Rood, 1984; Jakob, 2000; Guthrie, 2002; Brardinoni etal., 2002). Stream channel responses are the result of cumulative changes in their lateral and upstream condition. Timber harvesting can affect the governing 7 conditions of stream systems by increasing sediment supply from hillslopes, changing the hydrologic regime of the stream catchment, and changing the composition of bank vegetation (Sullivan et al., 1987). Removal of riparian vegetation reduces bank strength and can cause channel instability and channel widening (Millar and Quick, 1993; Millar, 2000). Increased sediment loads from logging disturbances can affect channel morphology (Wood-Smith and Buffington, 1996; Hogan et al., 1998) and can result in channel instability and aggradation (Roberts and Church, 1986; Kondolf et al., 2002). This issue is particularly important in coupled reaches where the impact of logging on hillslopes can influence stream channel responses through increased sediment delivery. The timing of hillslope responses depends on the timing of significant precipitation events and the time period over which hydrological re-routing and loss of root strength occur. The recovery of the hillslope system occurs as new vegetation growth provides a new source of root strength on slopes (Sidle, 1992; Sakals and Sidle, 2004) or appropriate hydrologic pathways are re-established. Theoretical root reinforcement through re-vegetation occurs on the order of 10 to 20 years (Sidle et al., 1985). Similarly, Brardinoni etal. (2002) found landslide frequencies in logged watersheds of south-west British Columbia were elevated for about 20 years before reduction to pre-logging rates. Grant and Wolff (1991) found that sediment transport in disturbed watersheds was increased for a 10 to 15 year period. Channel response to disturbance is the overall change to the character of the stream channel. The recovery of the system occurs as increased sediment inputs 8 are reduced, banks re-vegetate, and active channel widths narrow (Friedman etal., 1996; Scott etal., 1996; Liebault and Piegay, 2002). Miller and Benda (2000) and Sutherland etal. (2002) both examined how stream systems respond to discrete sediment delivery events, which either translate or disperse sediment downstream overtime. Research that has tried to examine stream channel recovery has drawn from observations of stream channels in different stages of recovery (Hogan, 1987), and studies that have systematically examined channel response through time have focused on fairly large river systems (Liebault and Piegay, 2002). In coastal British Columbia, long-term monitoring of channel recovery processes has been limited (e.g. Carnation Creek, cf. Hogan et al. 1998). 1.3 Theoretical Framework A theoretical framework can be used to understand watershed responses (Figure 1.1). Disturbances change the character of the governing conditions of a system, and this may be reflected in the change of some variable that characterizes the state of the system (i.e. the response variable). The change to the response variable may lag the disturbance event. This time lag is called the response time. The relative magnitude of change can be described as the response. The recovery of the system occurs if there is a trajectory towards initial pre-disturbance conditions. The time from maximum disturbance to recovery is called the relaxation time. 9 Figure 1.1 Theoretical response of a geomorphic system to disturbance (modified from (Church, 2002)) A initial condition of response variable : >-Time •< response time —>\< relaxation time >• In watersheds, disturbance from logging can affect hillslopes and stream channels. The response time of the hillslope system is the result of the timing of changes to root strength and drainage patterns relative to the timing of geomorphically significant precipitation events. The recovery of hillslopes is facilitated by the re-establishment of stable hydrologic pathways and root strength from vegetation re-generation, and the relaxation time should reflect the rates of these processes. The response of stream channels is driven by changes in sediment character and riparian vegetation. The response times reflect the timing of these changes combined with the timing of geomorphically significant water discharge events. In coupled regions, the timing of sediment delivery from hillslopes can play a significant role in the timing and nature of the stream channel response. Lagged stream channel responses may occur as hillslopes respond to logging 10 through increased landsliding, and then stream channels respond to increased sediment delivery from those landslides. Hillslope recovery allows for decreased sediment delivery to channels but channel recovery may also depend on the succession of riparian vegetation. 1.4 Objectives Efforts to understand hillslope responses have typically focused on the initial response magnitude (Rood, 1984; Howes, 1987; Gimbarzevsky, 1988; Rollerson, 1992; Jakob, 2000; Millard etal., 2002). In a few recent cases attempts have been made to understand the timing of response and relaxation of landsliding activity (Chang and Slaymaker, 2002; Guthrie, 2002; Brardinoni etal., 2002). In stream channels, efforts have often examined the magnitude of responses due to logging disturbance (Roberts and Church 1986; Millar, 2000). Efforts have been made to quantify watershed response to disturbances (Roberts and Church, 1986; Schnackenberg and MacDonald, 1998; Beaudry and Gottesfeld, 2001). There appears to be little focus on the timing of stream channel responses, or the potential for stream channel recovery. Researchers examining catchment sediment processes have acknowledged the need for more detailed research examining hillslope and stream channel linkages (Wainwright et al., 2002), but there appears to be little examination of how, in coupled regions, hillslope responses affect stream channel responses. Given the need to investigate these issues, the objectives of this thesis are: 11 1. ) to examine hillslope and stream channel responses to timber harvesting 2. ) to determine the interaction between hillslope and stream channel responses 3. ) to examine the recovery of hillslope and stream channel systems after responses to timber harvesting. The aim is to better understand the timing of hillslope and stream channel responses, their potential magnitude, and the process and timing of recovery. The approach is to examine system changes in logged coupled watersheds in south-west British Columbia using a time sequence of air photographs, and to investigate how coupling affects watershed responses to logging disturbance. The sensitivity of hillslopes and stream channels is implied through the examination of how responses vary between channels and hillslopes with varied characteristics. The thesis is divided into seven chapters to describe the development, implementation, results and conclusions of the research. The first chapter discussed the background for this study. This included a review of previous research and development of a conceptual framework for research, and then identified the main objectives of this study. Chapter 2 describes the study locations, including background information on geographic location, watershed characteristics, and details on data and information coverage. Chapter 3 highlights the methods used for data preparation and collection. Chapter 4 provides an identification of the sources of error and an analysis of measurement error. Chapter 5 gives the results of the investigation of stream channel and hillslope disturbances. Chapter 6 discusses the implications of the results, and suggests a framework to understand disturbance 12 regimes. Chapter 7 summarizes the research and results, and suggests further directions for research into watershed disturbance regimes. 13 2 S t u d y S i t e s 2.1 Introduction This chapter describes the study sites of this research. First, a framework for the selection criteria is discussed. This is followed by introducing the chosen study sites, with a brief discussion of the study site characteristics. 2.2 Selection Criteria As the primary focus of this research is to examine the interaction of stream channels with hillslopes, coupling is a prerequisite for selected watersheds. As well, a regional focus on watersheds in south-western British Columbia was also a constraint. Suitable watersheds for this study must have a history of logging and, as well, possess characteristics that make them susceptible to response. In general, watershed characteristics should be conducive to both hillslope and stream channel responses to logging. Suitable characteristics include hillslopes which are steep and have surficial or lithological materials which are prone to mass wasting. The hydroclimatic regime in the Coast Mountains is typically wet, with episodic intense storm events, and this can be a primary driving force of landsliding activity (Church, 1998; Schwab, 1998). Stream channels should be alluvial, so that channel widening is feasible; these channels are the most susceptible to channel changes (Wood-Smith and Buffington, 1996; Hogan, 2001). To examine the linkages between hillslope and stream channel processes, watersheds should have a high degree of hillslope coupling such that changes to hillslope sediment transfers will 14 changes in sediment delivery to stream channels. These criteria typically limit drainage basin size to roughly the 10-100 km2 range for mountainous watersheds in coastal British Columbia. Aside from physical characteristics, logistical constraints helped to identify potential watersheds. While the Province of British Columbia maintains a substantial aerial photograph record library, obtaining photos for analysis can be expensive. Availability of photos from external sources influenced study site selection. The timing of photo coverage was also a consideration, and this included both the overall time period of coverage, and the time period between sequential photos. Further, in order to reduce the costs of field work, study sites were selected that were fairly accessible from Vancouver. Given the outlined criteria, site selection was clearly not randomly prescribed. Efforts were made to select sites which had a known history of watershed response in order to examine the mechanisms and patterns of those responses. This may bias results in that watershed responses to logging operations observed in these basins may not be indicative of typical responses to logging. In basins which have similar geomorphic, geologic, and hydro-climatic conditions as those in this study, however, similar patterns of response may be expected to occur. Furthermore, while there has been no effort to define these basins as "typical" for the coastal region of British Columbia, they are not obviously atypical (cf. Cheong 1992, and Trainor and Church, 2003 for a further description of watershed and channel comparisons). 15 2.3 Study Site Locations and Watershed Characteristics Potential study sites.were identified based on the criteria listed above. An initial selection of 15 watersheds was identified. A preliminary air photo reconnaissance reduced this number to 7 watersheds in southwestern British Columbia (Figure 2.1, Table 2-1). Study sites are broadly located throughout the Coast and Cascade Mountains (Holland, 1976). Chapman, Cascade, and Norrish Creeks and Theodosia River are on the western edge of the Coast Mountains, and on the eastern edge of Georgia Strait. Slesse Creek is located in the Cascade Mountains, and is more centrally located within the span of the Western Cordillera. Cedarflat and Dewdney Creeks are located towards the eastern edge of the Cascade Mountains. Average area of the study basins is 76 km2, the largest being Slesse Creek (162 km2), and the smallest being Cascade (19.5 km2) (Table 2-1). Watersheds have a high proportion of steep (>30°) topography and contain regions of hillslope coupling. Study sites have a history of hillslope or stream channel disturbance post-logging. Norrish and Cascade Creeks were investigated as part of an early case study in British Columbia examining the effects of logging on landsliding (Howes, 1987). Since it has been 15 years since this initial study, there is the opportunity to examine what the longer term trends of watershed responses are in these basins. Similarly, Chapman Creek has a history of documented hillslope response (Brardinoni et al., 2003). Slesse Creek has been used as a study site for examining stream channel responses to riparian vegetation removal (Millar, 2000). Similarly, Theodosia River, Cedarflat Creek and Dewdney Creek all have a history of channel 16 changes post-logging (M.J. Miles, personal communication, 2003). While there have been previous histories of research in these basins, the work tended to focus on specific research goals (e.g. effects of bank riparian removal on stream channels (Millar, 2000), or the degree of landslide frequency increase post-logging (Rood, 1984)). Re-visiting some of these sites within the framework of this research allows for an examination of all of the components of watershed response. This includes extending the time frame of study to capture the longer term patterns of response, as well as extending the focus on hillslopes to include the stream channel component, and vice versa. 17 ce Is isrv/7 A/orf/i Chapman Theodosia s Williams Lake Cedarflat Cascade Norrish Dewdney Slesse Figure 2.1. Overview map of British Columbia and the location of study sites (Maps reproduced from the Ministry of Sustainable Resource Management, Government of British Columbia) 18 Table 2-1 Summary of study watershed characteristics . . . . . . . Area Elevation Watershed Location / f 2» n / » (km) Range(m) Geology Slesse Norrish Cascade Chapman Theodosia Cedarflat Dewdney Chilliwack Valley Lower Fraser Valley Lower Fraser Valley Sunshine Coast Sunshine Coast Coquihalla Coquihalla 161.8 110.8 19.5 66.8 73.0 21.2 76.6 280-2400 20-1280 120-1340 0-1440 170-1620 550-1950 340-1950 felsic intrusives, volcanics, argilitte, conglomerates (Monger 1966) felsic intrusives (incl. quartz monzonite) (Roddick 1965 in Howes 1987) felsic intrusives(incl. quartz monzonite) (Roddick 1965 in Howes 1987) volcanics, volcanoclastics, conglomerates, sandstones (Journeay and Monger 1994) quartz diorites, granodiorites (Roddick et al. 1976) basalts, pelites, siltstones, argilites, sandstones, conglomerates (Journeay and Monger 1994) basalts, pelites, siltstones, argilites, sandstones, conglomerates (Journeay and Monger 1994) 19 Cascade Creek Figure 2.2. Map of Cascade Creek and locations of study reaches. 20 Figure 2.3. Map of Chapman Creek and locations of study reaches. 21 Cedarflat Creek Figure 2.4. M a p of Cedarflat and Dewdney Creeks and locations of study reaches. 2 2 N Legend Slesse Creek 0 0.5 1 2 3 i Reach Location Watershed Boundary Main Drainage Contours (20 m) Wt (Tl«t IS U. S.A (out of study area coverage) Figure 2.6. Map of S lesse Creek and locations of study reaches. 24 Theodosia River 1:100 000 0 0.5 1 2 3 4. i Kilometers Figure 2.7. Map of Theodosia River and the locations of study reaches 25 2.4 Reach Characteristics Study watersheds were divided into reaches in order to examine reach scale dynamics (Appendix 1). This is important since governing conditions, such as slope, sediment character, and morphology can be significantly different between reaches. Reach breaks were defined based on significant changes in morphology or channel characteristics from adjacent reaches. Reach morphology was classified in a manner following Church (1992). Reach morphology was classified as confined (cf), single-thread (st), or multi-thread (mt). Confined channels fell in between the Church (1992) classification of "intermediate" and "large" channels. Reaches that were defined as confined included channels that had little or no alluvial floodplain, or were constrained by adjacent hillslopes. Confined channels were typically straight or sinuous, and had characteristic morphologies such as cascades, step-pools, or in some cases riffle-pools (Ibid.). The caliber of the largest clasts in the channel bed and bank was typically on the same order of magnitude as peak discharge flow depth. Channels classified as single-thread had developed alluvial floodplains, but only one channel and no in-channel bars. These channels typically had a riffle-pool morphology and typical grain size diameters smaller than flow depths. Single-thread channels were also sinuous (Ibid.). Multi-thread channels were defined based on the presence of frequent mid-channel bars, and multiple channels. These included wandering and anastomosed channels, as well as three braided reaches and one fan reach. Overall 119 individual reaches were identified, and the distribution of channel morphology types is given in Figure 2.8. Confined and single-thread reaches are the dominant reach morphologies, with a lesser 26 number of multi-thread channels. This is reasonable given that channels tend to be headwater systems that are steep and coupled. Channels that are more strictly alluvial or multi-thread would be expected to be dominant in more distal reaches of a drainage basin (Church, 1992). Most watersheds contain the range of morphology types. Cedarflat Creek and Chapman Creek have only confined or single-thread channels. Slesse Creek and Theodosia River contain all reach morphologies, but have a lower proportion of confined reaches than the other watersheds. 25 20 w o> •c u S 15 o -Q E 2: 10 Distribution of Channel Morphology within Study Basins B MT a ST OCF Increasing drainage area 4<f <r Watershed Figure 2.8 Distribution of channel morphology within study sites 27 Initial channel widths (W0) vary from 8.3 to 131.4 m, and have contributing drainage areas (Ad) from 9.0 to 157.0 km2. A wide range of W 0 / A d relationships exist (Figure 2.9). In general, Wo/ A d relationships are stratified by morphology. Confined channels tend to have narrower widths, for a given watershed area, than multi-thread and single-thread channels. The average length of study reaches is 54 W 0 , with a range of 7 to150 W 0 . Some shorter reach lengths were chosen where channel characteristics were significantly different over a fairly short distance along the thalweg length. Trainor and Church (2003) suggest that 50-70 W b is a conservative reach length for riffle-pool channels, and 30-50 W b is a conservative reach length for smaller forested channels. Width to Drainage Area Relations 1000 o st • cf A mt E A 100 A A 10 A 0.1 1.0 10.0 100.0 1000.0 Catchment Area (km2) Figure 2.9 Initial channel width to drainage basin area. 28 Disturbance regimes were identified for each reach (Appendix 1). Disturbance types forms a 2x2 matrix which identifies the presence or absence of riparian logging, and the presence or absence of hillslope coupling. The four disturbance types are no coupling and no riparian disturbance (NL/NC), coupling but no riparian disturbance (NL/C), riparian disturbance but no coupling (L/NC), and coupling and riparian disturbance (L/C)(). Most reaches have some degree of riparian logging disturbance (89 of the 119 reaches). This reflects logging practices during the pre-Forest Practices Code era to log trees up to channel banks. There is the same number of each of the not disturbed and the not riparian logged but coupled reaches (15 each). NL L NC 15 30 C 15 59 Table 2-2 Distribution of disturbance type. Within each study watershed, there is a range of disturbance types (Figure 2.10). All basins have numerous reaches that have been subject to the combination of riparian logging and landslide coupling. Chapman Creek and Dewdney Creek have proportionately more reaches with no disturbances than other basins. Conversely, Cascade Creek and Slesse Creek have some disturbance in all reaches, and for Slesse Creek, all reaches had some degree of riparian logging. 29 25 20 w 0) -c o ro o •Q I 15 10 Distribution of Disturbance Regimes within Study Basins BL/C ®NUC • L/NC • NUNC Increasing drainage area Watershed Figure 2.10. Distribution of disturbance regimes within study basins. 30 -3 Methods 3.1 Introduction Response times for logging disturbances on hillslopes in the Pacific Northwest are on the order of 20 or more years (Sidle et al., 1985; Brardinoni et al., 2003), and may be longer for stream channels. Assessing long term changes to hillslope and stream channel patterns, then, requires methods that can capture long time scales. Ground surveys are beneficial because they can incorporate measurements of features that are not feasibly measured through remote sensing. This can include detailed measurement of sediment texture, bank and bed conditions, and channel width, depth and slope. Ground surveys can be difficult to conduct due to cost, logistics of maintaining long-term monitoring, and the long time periods required to gain detailed results. They are also limited in that they do not allow for retrospective analysis. Aerial photographs provide an instantaneous perspective on the landscape. Chrono-sequential air photos of the same location can be used to examine long term changes to the landscape. In British Columbia, many regions have air photo coverage extending back for 50 or 60 years. This record creates the opportunity for long-term retrospective studies of landscape changes. Aerial photographs can be scanned into digital images for use with Geographic Information Systems (GIS) computer applications. This purpose of this chapter is to detail the methods used for the analysis of landscape changes using chrono-sequenced aerial photographs and GIS. 31 3.2 Aerial Photograph Imaging A detailed inventory of the aerial photograph coverage of study basins was conducted to identify the available resources. Photos were generally taken by the Province of British Columbia, but some were taken by private forest companies. Photos were taken roughly once every 10 years, with an overall photo coverage time frame from 40 to 50 years. A summary of the air photo coverage for each study site is given in Figure 3.1. Figure 3.1. Summary of aerial photograph coverage of study sites. Theodosia Slesse Norrish Dewdney Chapman Cedarflat Cascade -TJ C-• • • T> • D-• Year of Air Photo ~\ r -0 0- -o • -TJ TJ • • -O-D-i r 1940 1950 1960 1970 1980 1990 2000 2010 32 Aerial photograph prints (152 mm) were scanned into a digital image using a HP Scanjet™ flatbed scanner. Images were scanned at a resolution between 150 and 300 dpi, and digital image files were created and managed using HP Scan™ software. Images were formatted as Tagged Image Format Files (TIFF), since TIFF files are most easily manipulated during geo-referencing procedures. The pixel resolution of TIFF images is dependent on the scale of the original photo. The raster image of a photo with a scale of 1:20000, scanned at 150 dpi, will have a pixel width of approximately 3.4 meters. While precision can be increased by increasing the scan resolution, there is a trade-off with file size for each image. To increase the resolution by two requires a 2 2, or four, times increase in file size. Raw image files were geo-referenced in order to link images with a real-space co-ordinate system. This is important for establishing scale on the images, as well as using the photo images for comparison of the same locations through time. Geo-referencing involved identification of control points on air photo images and corresponding points on a fixed map base (Figure 3.2). Digital files of the Province of British Columbia TRIM (Terrain Resource Information Program) maps were used as a base map to which control points were referenced. TRIM maps are accurate to within 10 m for position and 5m for elevation, with a confidence of a=0.9 (Anonymous, 1997). Control points were objects that could be easily identified on both the digital image and the base TRIM maps, and would not change position through time. Objects used for control points were typically road junctions or obvious bends in roads. TRIM maps are projected using the North American 1983 Datum. 33 Figure 3.2 An example of geo-referencing procedure. Control point locations included road intersections (A), bridges (B), and identifiable features such as river bends (C). Geo-referencing was facilitated through ArcGIS™ software. Geo-referencing is a two-dimensional "best-fit" procedure. Control points from the digital image are referenced to known real space coordinates, based on an X , Y coordinate system. Generally 10-20 control points were identified on each image for geo-referencing. Geo-referencing uses a polynomial regression to create a least squares fit of the X , Y coordinates of the control points on the digital image relative to map X , Y coordinates. The digital image is stretched to fit the regression of the control points. The geo-referencing produces a TIFF World file, which is a TIFF file that has an X , Y position attached to each pixel in the raster file. A positional error based on the geo-34 referencing regression is given as a root mean square (RMS) error. Pixels are stretched to fit the polynomial regression, and therefore pixel size can be different than the original scan resolution. Overall, 256 air photos were scanned into digital format and geo-referenced. A formal discussion of the error associated with this procedure is given in Chapter 4. 3.3 Hillslope Sediment Production Measurements Photogrammetric analysis of landslides has become a standard method for examining sediment production from hillslopes (Rood, 1984; Reid and Dunne, 1996; Jakob, 2000). Methodology for aerial photo analysis of sediment production from landslides has been well-established (cf. Rood, 1984) and a similar methodology is used for this study. The term "landslide" is used in this study to functionally refer to mass movements on hillslopes that lead to identifiable scars on the landscape. These include debris slides and gullied debris flows. Landslide scars are visible on aerial photographs down to some threshold size. Digital images of air photos are used for landslide identification. In ArcGIS™, a shapefile layer is created that highlights the area of individual landslide scars (Figure 3.3). Vegetation obscures landslide scars, and identification of landslides on air photos becomes difficult after 20 years due to new growth (Jakob, 2000). Smaller landslides (<500 m2) can be difficult to identify due to vegetation cover (Brardinoni et al., 2003). Given the typical time frame between air photos (10 years), the visibility of new slides should not be affected by vegetation re-growth. 35 Figure 3.3 Measurements made for landslide areas, a.) Original image without landslide polygons identified b.) Image after landslide polygon construction. The total coupled landslide polygon area is calculated for each reach. The area of each landslide (A) is measured. In order to assess hillslope coupling, only landslides that deliver sediment to a channels were recorded. The overall area of landslides affecting stream channels over the time span between air photos is given as: AT = ^  A + A + A3 +... + An Total landslide area measurements do not necessarily reflect the total amount of sediment mobilized between the time periods of sequential air photos. The same landslide may be identified on sequential air photos. The overall slide area is included in the inventory for each year, and therefore new sediment may not be mobilized during that period. Similarly, the area of a slide does not necessarily represent the amount of sediment being delivered to channels. Landside depths vary, therefore it can be difficult to ascertain the volume of sediment mobilized from a landslide based on landslide area. While Rood (1984) and Guthrie and Evans (2004) argue that area-volume relationships exist for landslides in coastal British Columbia, the high degree of variability in landslide depths measured in the field does not suggest that strong relationships exist between landslide width and depth (Figure 3.4). Average depth of landslides measured in the field is 2.1 m (S.D. = ±1.2 m). 37 Figure 3.4. Landslide width and depth measurements for Centre Creek, Chapman Creek, Nesaquatch Creek, and Theodosia River. 5 10 Landslide Width (m) 15 3.4 Stream Channel Measurements Air photo analysis has been a valuable tool to evaluate changes in the sedimentary regimes of stream channels (Lyons and Beschta, 1983; Ham and Church, 2000; Mount et al., 2003). The dynamic nature of rivers adds an element of complexity to the investigation of channel responses to logging disturbance. Changes in the overall character of the stream are important indicators of channel responses. These indicators include changes to channel width, depth, morphology, slope, riparian vegetation, and sediment grain size distribution. In the context of a retrospective survey, many of these factors cannot easily be quantified. For this study, changes in stream channel width will be measured to assess historical 38 channel changes since it is a component of channel change that can be monitored from air photos. Figure 3.5 Measurements made for stream channel assessment. Original Image TO DC Average Width = ( A c h - A i ) reach boundary reach boundary island area (Aj) thalweg length (Lt) channel area(Ach)\ GIS Layers As with landslide measurements, stream channel measurements were made on digital geo-referenced air photo images. Previous research has focused on using GIS applications to examine channel morphodynamics (Ham and Church, 2000), or channel cross-sectional width (Mount et al., 2003). However, due to large within-reach variability of channel width (Trainor and Church, 2003), temporal comparison of channel width can be difficult. For example, the standard error of mean channel width in a reach, based on a number of width measurements, will 39 likely exceed temporal variability resulting from actual changes in width (a full discussion of error is given in Chapter 4). Reach locations were fixed in space for all time periods. For a given reach, an average width measurement based on (Figure 3.5): w k „ - 4 ) where Wavg is the reach average width, Acn is the reach channel area, A, is the reach area of vegetated islands, and Lt is the reach thalweg length. Shapefile layers were constructed in ArcGIS™ for thalweg length, channel area, and island area. Thalweg location was assumed to be along the location of flow in the channel. Where there were multiple channels, thalweg was assumed to follow the path along the widest channel. Where there was no obvious thalweg location, it was assumed to be in the location of flow in the middle of the channel. Channel areas were identified based on the edge of the visible bankfull channel and bank edge was estimated in regions where vegetation obscured the channel edge location. A further investigation of the bias of width measurements associated with bank vegetation is discussed in Chapter 4. Channel islands were identified based on the presence of vegetation. Shapefiles were constructed for each time period for the entire channel study area. Layers were then clipped to the fixed reach breaks to create individual reach channel area, island area, and thalweg length. 40 3.5 Field Methods Field work was conducted in several of the study watersheds. The purpose of the fieldwork was to undertake a reconnaissance of the study sites and to collect information in order to compare air photo data with conditions on the ground. Field work was conducted in July and August, 2003, in Slesse Creek, Nesequatch Creek, Centre Creek, Norrish Creek, Chapman Creek and Theodosia River. Nesequatch and Centre Creeks were initially considered as potential study watersheds, but were not included in the final analysis of this research. These sites do, however, provide valuable information on the comparison of field measurements to air photo measurements, and were therefore included for measurement comparisons. Since many of the sites were logged over 20 years previous to inspection, access to sites was difficult. Logging roads are typically de-activated after use, and motor-vehicle access was not possible. Bicycles were used to access remote study sites where vehicle access was impractical or not possible. Boat and bicycles were also required to access sites on the Theodosia River. In the field, general observations were made of stream channel conditions. This included observations of grain size (estimated D 5 0 and Dmax), channel morphology, channel aggradation (Figure 3.6a), locations of bank erosion (Figure 3.6b), and sediment delivery from landslides (Figure 3.6c). Measurements of channel width were made in several reaches. Widths were measured as active bank-full width, and were measured at a spacing of 2 mean bankfull widths (Wb). A total of 10 width measurements was made per reach, and therefore surveys included a channel distance of approximately 20 W b . 41 All accessible landslides at field sites were surveyed. Surveys consisted of measurements of landslide width and landslide depth at a number of locations along the landslide track. Landslides were followed to stream channels to assess channel coupling (Figure 3.6d). (Note: caption for figure that follows) Figure 3.6 a.) Field observation of channel aggradation (Theodosia River). Flow is from the left of the photo to right. Aggradation is occurring and a sediment wedge has developed behind a log jam (right). Aggradation in channels indicates a high sediment supply, and highlights that channel widening is not the only potential channel response. b.) Field observations of bank erosion (Theodosia River). This photo highlights some of the mechanisms of channel width response. The field assistant is sitting on top of a logged tree stump. To the right is the old active channel, and the logged stump would have been at or near the pre-logging channel bank. Bank erosion has occurred to the left of the stump. Channel widening has increased channel width by approximately 100%. Channel recovery can be seen occurring through the development of pioneer tree species mid-channel. The caliber of bed material is large cobble to boulder size. c.) Field observation of hillslope coupling through delivery of landslide derived sediment to stream channels (Chapman Creek). This view looks down from a road-derived landslide. Stream channel (Chapman Creek) is below and almost the entire landslide volume was delivered to the channel. Aggradation is visible in the channel. This landslide is approximately 10 m wide and 2 m deep. d.) Field survey of landslide tracks. This landslide is approximately 15 m wide and 4 m deep. Note the distribution of sediment size, including boulder size clasts. Note the distribution of sediment size, including boulder size clasts. 42 4 Error Analysis 4.1 Introduction In order to assess changes in channel widths or landsliding incidence through time, there must be an understanding of the precision and accuracy of the measurements that are made. The precision refers to the error associated with any given measurement, while the accuracy indicates how close measurements are to their true value. Quantification of the precision of measurements is necessary for rigorous comparison of channel and landslide changes between two time periods. Determining the accuracy of measurements will allow for an understanding of how measurement values compare with the true quantities that are being estimated. To assess temporal changes, analysis of measurement error needs to encompass both the precision and accuracy. To assess temporal changes to parameters measured on sequential aerial photographs, the amount of change between successive photos must be greater than the amount of error associated with those measurements (e.g. Downward et al., 1994; Mount etal., 2003). Therefore the measured precision must be smaller than the magnitude of the changes one is attempting to observe. Several factors, the result of image processing and limitations of the person making observations, can affect measurement precision. The purpose of this chapter is to examine the sources of error of measurements made in this study, as well to try to quantify those errors in order to validate the following analysis. 44 4.2 Sources of Error Errors occur when an operator makes length and area measurements on a digitally produced image. Errors can be the result of the optical properties of aerial photographs, of photo digital imaging and geo-referencing, or of the measurement process. Aerial photograph scale is related to the camera's focal length and the distance the camera is above the landscape. Variation in topographic relief causes scale heterogeneity within aerial photographs. Uncorrected two-dimensional measurements made on aerial photographs will be affected by changes in scale. Scale distortion can be addressed through ortho-rectification of the image. The scale of the photos will determine the precision with which measurement can be made. Creating digital images from analog images affects error. As discussed in the methodology, analog images are digitally scanned using a flatbed scanner, and the resulting digital image is a rasterized version of the original image. The size of the pixels of the raster depends on the resolution with which the scan is made. In turn, the size of the pixelation affects the precision of measurements made on the digital image. Images are then geo-referenced to a map projection. Depending on the topography of the landscape contained in the image, errors associated with scale distortion can be reduced or enhanced. The geo-referencing procedure also affects the accuracy with which positional measurements can be made on the image. Once the digital images are geo-referenced, length and area measurements are made of stream channels and landslides. These measurements are subject to 45 error based on the precision with which the operator can identify such features. The accuracy of these measurements can be affected by how visible the feature is. Vegetation, aspect, and bank overhang can introduce bias into measurements made from aerial images. 4.3 Digital Image Processing and Geo-referencing Geo-referencing was facilitated through ArcGIS™ software. Geo-referencing is a two-dimensional "best-fit" procedure. Control points from the digital image are referenced to known real space coordinates, based on an X,Y coordinate system. Generally 10-20 control points (mode of 13) (Figure 4.1) were identified on each image for geo-referencing. Geo-referencing uses a polynomial regression to create a least squares fit of the X,Y coordinates of the control points on the digital image relative to map X,Y coordinates. A positional error based on the geo-referencing regression is given as a root mean square (RMS) error. Pixels are stretched to fit the polynomial regression, and therefore pixel size can be different than the original scan resolution. Overall, 256 air photos were scanned into digital format and geo-referenced, though data were recorded for only 173 photos. For recorded images, the average pixel width is 2.9 m (Figure 4.2), and the average positional RMS error is ±13.7 m. 46 Figure 4.1 Distribution of the number of control points used for image geo-referencing n=173 "0.3 10" 3 -0.2 I o a Tl 0.1 I p 0.0 5 10 15 20 Niarter of Geo-Referendng Control Points Figure 4.2 Sampling distribution of geo-referenced image pixel resolution. O -*—' o JC D_ < o d z 2 3 4 Pixel Width (m) 47 Geo-referencing is limited in the sense that positional error is reduced only in two-dimensions. While this can reduce scale disparities with elevation from original images, it does not account for or correct for these errors. In order to address topographic effects on scale, ortho-rectification needs to be done. Ortho-rectification involves image correction using control points in the X,Y,Z co-ordinate system, and can be manually done using a stereo-plotter. Software is available for ortho-rectification but was unavailable for this project. Mount et al. (2003) report little difference in the accuracy of ortho-rectified images compared with un-rectified images. However, they caution that their results were obtained for photo analysis of low gradient flood-plains, and do not apply to complex terrain. In order to reduce the error associated with elevation, geo-referencing control points were generally selected near valley bottoms. This reduced RMS errors associated with geo-referencing, but it should be noted that this reduction in error is applicable only to valley bottom measurements. Biasing the geo-referencing procedure in this way is advantageous for the precision of stream channel measurements, but affects the precision and accuracy of landslide measurements. Measurement accuracy can be assessed based on how close measurements made on digital images are to the same measurements made in the field. Measurement accuracy can be affected by vegetation cover, bank overhang, and aerial photograph perspective. To assess measurement accuracy, field measurements of stream channel widths and landslide widths were made during the summer of 2003. Channel widths were measured at a total of 12 reaches at the Centre, Norrish, Chapman, and Theodosia sites. At each reach, 10 width 48 measurements were made at a spacing of approximately two average channel bankfull widths (a total of n=128 channel widths were measured in the field). Width measurements were also taken on geo-referenced air photo digital images at the same channel cross-section locations as those where width measurements were taken in the field. Field measured widths were compared with remote image measurements using one-way ANOVA with measurements paired for each cross-section. The hypotheses were: H 0 :p i - U2 = 0, the difference between means is zero H A : Mr M2 5*0, the difference between means is not zero The difference in a paired measurement, d,, is the difference between the field measured width, X,, and the remote image measured width, Y-,. The mean difference, d, is 3.48 m (S.E.=±0.945 m). For a two-tailed test, a=0.05 and df=127, Ho is rejected. Therefore, the field and remote image measurements of width are not equivalent. The strong positive difference of paired measurements indicates a bias in the channel width measurements made on remote images. Specifically, widths measurements made on images tend to under-estimate the field measurements of the same channel cross sectional width. This seems reasonable since riparian vegetation cover and bank overhang should act to obscure the visibility of channels on digital images. Since riparian logging is no longer practiced along larger channels, it was not possible to examine whether this bias exists for width measurements on riparian logged channels. It may be reasonable to suppose that GIS measurements of channel width could increase by an amount similar to this 49 bias when logging occurs (i.e. channels may appear to increase in size after logging as a result of increased visibility from reduced bank cover). In order to detect significant changes in channel width, the change, d, should significantly exceed this bias. For a 95% confidence (one-tailed), the difference between two channel measurements needs to exceed d +1.65 x S.E.- = 5.0 m. d 4.4 Measurement Precision Large positional error can result from the geo-referencing of images of complex terrain. On average, RMS Error for geo-referenced images was 13.7 m. If channel changes of small to medium sized streams (say less than 50m wide) are examined based on fixed cross-sections, there will a high degree of uncertainty whether or not widths are being measured at the same location on the stream channel through time. An object-based measurement system can help to overcome positional errors. For example, measuring channel area and channel length based on a defined reach, the positional error is proportionally smaller than it would be for a single channel width measurement. The positional error is associated with the delineation of what segment of the channel constitutes the reach being measured. The precision of measurements will be affected by the pixel resolution, positional error, and human-operator errors. Calculating average width based on area and length allows for an integrated measure of width. Due to high variability in channel width within a reach (Trainor and Church, 2003), the variation in channel 50 widths within a reach likely exceeds the variation in width that might occur between time periods. Therefore, unless there are a large number of width measurements (n) used to calculate the mean reach width, the error in mean width (S.E.- = -yjs2/n ), could be large because the variance of width measurement (s2) is large. Comparison of object measurements, such as area and length, depend less on positional precision and are more affected by error based on scale variations between images. The standard error of estimate of a length measurement, L, can be given as 6L, and the standard error of estimate of an area measurement A, can be given as 5A. The fractional standard error of estimate, 5W/W, can then be On portions of photos with overlap with adjacent photos, replicate measurements can be made on each photo and compared. Individual photos have inherent error associated with measurements as a result of the scanning and geo-referencing procedures as discussed. The error between replicate measurements on different images is an estimate of the effect of geo-referencing error on the precision of those measurements. To examine the precision of reach length, area, and average width measurements, replicate measurements were made on reaches that had overlapped photo coverage. Efforts were made to not make replicate measurement too close to photo edges, where photo distortion will have a greater effect on precision. A total of 45 replicate measurements of reach length and area were made across 8 different reaches with photo overlap. For a given measurement pair, the standard error of reach length measurements is calculated as 51 SL = ^sL2 / 2 , and reach area measurements is SA = ^sA2/2 , where in each case n = 2 since it is a paired measurement. From the 45 replicates, the average standard error of estimate for reach width (SW) is 1.2 m and the standard deviation of SW is 1.2 m (Figure 4.3). The range of SW values is 0.1 to 4.4 m. From Figure 4.3 it is apparent that the magnitude of error increases with increasing reach channel width, however, fractional standard error of estimate (SWIW) decreases with increasing channel width (Figure 4.4). The average fractional standard error of estimate for reach width is SWIW=0.0^ (S.D =± 0.021). Therefore, in order for channel width measurement comparisons to be significant (one-tailed test, a=0.05), the difference must be greater than; SW IW +1.65 x s.D.—— = 0.067, or approximately 7% . Figure 4.3 Standard error of estimate of reach width. —. 3 2 Width Standard Error of Estimate • All Basins n=42 20 40 60 W{m) 80 100 52 Figure 4.4 Fractional width error of estimate of reach width. Width Fractional Standard Error of Estimate 0.12 0.1 0.08 g 0.06 0.04 0.02 • All Basins n=42 20 40 60 W (m) 80 100 Similar precision issues affect landslide area measurements. Air photo scale distortions are greater in regions where the photo captures a wide range of elevation. On watershed hillslopes, elevation may vary on the order of 1000 m or more. For non-rectified images, this introduces error that can be on the order of 20-30%. Geo-referencing can help to reduce errors by "averaging" distortion across wide elevation ranges. This is most effective when control points for geo-referencing are selected from a wide range of elevations. This may produce a "standard" scale for the referenced images, but will lead to large R.M.S. error as well as leading to an under-estimation of the lower elevations on the image. Since high precision was desired for stream channel measurements, geo-referencing was biased towards 53 selection of control points in the lower elevations surrounding the stream channel and valley bottom. The precision of landslide area measurements was evaluated in a manner similar to stream channel widths. Individual landslides that were visible on two photos of the same photo series were identified. Landslide area (A) was measured from both images, and then those areas were compared. Overall, 46 landslides were identified and measured. The average fractional standard error of estimate for landslide area is 0.090 (S.D.=±0.080) (Figure 4.5). In order to detect significant (a=0.05) differences in landslide areas between time periods, the change in landslide area needs to exceed 0.22, or 22%. This is a fairly significant error, and is the result of two factors. First, scale changes with elevation will cause larger error for measurements made over a wide elevation range. Scale introduced error can be corrected by ortho-rectification, but this procedure was not conducted in this study. Geo-referencing can help to reduce elevation-related scale distortion by establishing average scales over varied elevation ranges; scale will be over-estimated in low elevation and under-estimated in higher elevations. However, this is effective only if the post-geo-referencing scale is similar to the average elevation scale for the air photo. In the way geo-referencing was applied in this study, the second way in which the error of landslide measurements is affected is through this geo-referencing related scale distortion. Geo-referencing procedures were biased to increase the precision of channel measurements, and therefore geo-referencing control points were selected primarily in lower elevation locations near the stream channel. Therefore, scale distortion is expected to be higher for hillslope 54 measurements, and the variability in this distortion will also be high. This is reflected in the high measurement error for landslides (22%). Interpretation of results must take this error into consideration. However, given study objectives and the need for higher precision measurements for stream channel features, the sacrifice of measurement precision for hillslope features seems reasonable. iu <0 c 45 c .o o 2 Landslide Area Fractional Standard Error 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 • All Basins n=46 10000 20000 30000 40000 50000 Landslide Area (m2) Figure 4.5. Fractional standard error of estimate for replicate landslide area measurements. 4.5 Discussion In order to assess whether changes observed between air photos are real, the magnitude of change needs to exceed the error inherent in measurements. Traditionally, error estimates have largely been based on either R.M.S. errors from 55 geo-referencing, or on pixel resolution size (Mount et al., 2003). Single cross-sectional measurements of channel width can have a large amount of error due to positional uncertainty and channel width variability. This error can be reduced by width measurements based on length and area since this provides an estimate of error averaged over the entire feature. Measurement accuracy can be compromised due to vegetation and bank overhang affecting visibility of channels. This can be important when riparian vegetation is logged because it can cause an apparent increase is channel width. While positional errors are fairly large (average of ±13.7 m), and pixel width is typically large (2.9 m), overall error in width measurements is reduced (±1.2 m or a fraction of 0.031) since the error in area and length measurements is smaller. On average, measurements on digital images under-estimate channel width by 3.5 m (S.E. = ±0.95). When riparian vegetation is removed, a bias may be introduced that causes an apparent width increase because of an unobstructed view of the channel. Therefore, in order to detect width changes in channels which have undergone riparian logging, the magnitude of change needs to equal at least both the amount of bias and the measurement precision. As previously discussed, the difference between successive channel widths needs to exceed 5.0 m to be significant (a=0.05). Similarly, channel changes need to exceed 0.07 (7%) of channel width in order to be significant (a=0.05). For landslides, change in landslide area needs to exceed 0.22 (22%) of landslide area to be significant (a=0.05). These are the criteria that will be used to assess watershed changes in this study. 56 5 R e s u l t s 5.1 Introduction This chapter discusses the results obtained from this research. First, hydroclimatic data are discussed. A discussion of landslide responses follows, and will highlight logging and hydroclimatic histories. 5.2 Hydroclimatic Factors While logging affects hillslope or channel stability through loss of root strength, the response of these systems is driven by hydroclimatic forcing. On hillslopes, precipitation events drive landslide initiation (Schwab, 1998; Bovis and Jakob, 2000; Jakob etal., 2000). Important driving conditions can include antecedent soil moisture (Jakob etal., 2000), rain-on-snow events, and short-term, high intensity precipitation (Lynn and Jang, 2000). Stream channel instability is driven by stream discharge acting on channel conditions. Watersheds in this study have extreme discharge events typically in response to extreme precipitation events during the fall and winter, though some peak flows occur during spring run-off. In order to assess the timing of extreme hydroclimatic events, the history of precipitation and streamflow events is reviewed. Precipitation data were obtained from Environment Canada (Table 5-1). Climate station networks have limited spatial coverage, and stations were not generally located within study basins. It is, however, useful to examine precipitation trends at locations near to study sites. Stations were selected based on their 57 proximity and their similarity, in terms of elevation and physiographic setting, to study watersheds, and the time period of record. Study Site Nearest Weather Station Latitude Longitude Elevation (m) Period of Record Chilliwack Hatchery* 49° 5' N 121° 42'W 213 1961-2002 Hatzic Lake* 49° 10' N 122° 15'W 25 1959-2000 Stave Falls* 49° 10' N 122° 15'W 110 1909-2002 Powell River 49° 53' N 124° 33'W 52 1924-2002 Sechelt* 49° 28' N 123° 46'W 23 1937-1968 Sechelt West 49° 28' N 123° 48'W 61 1989-2002 Hope 49° 22' N 121° 29'W 39 1934-1995 Slesse Creek Norrish Creek Cascade Creek Theodosia River Chapman Creek Dewdney/ Cedarflat Creeks Table 5-1 Environment Canada Climate Station Location Summary. * reflects weather stations that are less than 5 km from study sites. Some climate stations are less than 5 km away from study watersheds (Table 5-1). However, Cedarflat Creek, Dewdney Creek and Theodosia River do not have nearby climate stations with long term precipitation records. Powell River, approximately 25 km away, is the closest station to the Theodosia River. The station at Powell River located in a similar physiographic setting to the mouth of Theodosia River. Similarly, the Hope climate station is approximately 21 km away from Dewdney and Cedarflat Creeks. However, the Hope station is situated on the western, windward side of the Cascade Mountains, whereas Cedarflat and Dewdney Creeks are situated more centrally in the mountains, and are also at a higher elevation. Climate data from Hope are not likely to closely reflect conditions at Dewdney and Cedarflat Creeks, but records are included to examine temporal trends in extreme precipitation events that may have affected the region. Another 58 factor that will affect the congruity between gauged precipitation from the station of record and actual precipitation conditions in study watersheds is elevation. Precipitation stations are at lower elevations (Table 5-1) than the majority of the terrain within study watersheds. Orographic effects within basins will likely result in much higher precipitation rates and more intense events at the higher elevations. It is likely that precipitation within study basins, therefore, exceeds values measured at the local weather stations. As well, low elevation stations may not be able to predict whether precipitation in upper elevations falls as rain or as snow. This can be particularly important, since it may be difficult to identify rain-on-snow events. Precipitation data (Figure 5.1) include daily precipitation (mm), where available, or daily rainfall (mm) if overall precipitation was not available. Daily records represent the finest temporal resolution available. However, landslide inducing precipitation events may be of high intensity but short duration. Daily precipitation records do not necessarily reflect the occurrence of short, high intensity events. It can still be useful to examine precipitation records to identify the occurrence of extreme events which may coincide with significant watershed change. Historical streamflow data were provided by Environment Canada through the HYDAT hydrometric data archive. Streamflow gauging stations are established on some study streams; Norrish Creek, Chapman Creek, Slesse Creek, and Theodosia River all have long-term stream flow gauging stations (Table 5-2). Records are not necessarily complete, nor do they overlap completely the time periods of air photo coverage of the same watersheds. Cascade Creek, Cedarflat 59 Creek and Dewdney Creek do not have gauging stations. For local comparison, Cascade Creek may have temporal patterns of streamflow similar to nearby Norrish Creek. For comparison with Cedarflat Creek and Dewdney Creek, Tulameen River is included. Tulameen River is a larger watershed than Cedarflat Creek or Dewdney Creek. Tulameen Creek drains eastward and is positioned slightly more leeward within the Cordillera. Table 5-2 Environment Canada Streamflow Gauging Station Locations Gauge Location Station # Drainage Area (km2) Record Norrish Creek near Dewdney 08MH058 117 1959-2001 Chapman Creek above Sechelt Diversion 08GA060 64.5 1970-1988 Chapman Creek below Sechelt Diversion 08GA078 N/A 1993-2001 Chapman Creek near Wilson Creek 08GA046 71.5 1959-1970 Slesse Creek near Vedder Crossing 08MH056 162 1957-2001 Theodosia River near Bliss Landing 08GC004 140 1953-1993 Tulameen River below Vuich Creek 08NL071 256 1974-2003 Annual maximum daily average discharges are given for Norrish Creek, Slesse Creek, Theodosia River, Chapman Creek near Wilson Creek, Chapman Creek above Sechelt Diversion, Chapman Creek below Sechelt Diversion, and at Tulameen River (Figure 5.1). Peak instantaneous discharge will exceed the daily average, and the duration and magnitude of these events can cause significant geomorphic change. A daily average discharge measurement may not identify these important events. In some cases, stream flow records include peak instantaneous discharge; however, the records have sporadic temporal coverage, and are not included since year to year comparisons would not be meaningful due to the incomplete record. 60 Figure 5.1 Annual Maximum Daily Precipitation and Annual Peak Discharge for stations at or near Study Sites (data from Environment Canada). Slesse Creek Norrish Creek 1950 1960 1970 1980 1990 2000 2010 Year 1950 1960 1970 1980 1990 2000 2010 Year Theodosia River Chapman Creek — i 1 1 r Powell River J I I L_ 1950 1960 1970 1980 1990 2000 2010 Year 70i 1 "c 60 s a 75 « i t 50 Q- 40 30h 200 5-150 g §>100| £ a e> Q 50 Sechelt Near Wilson Creek Above Diversion Below Diversion 1950 1960 1970 1980 1990 2000 2010 Year 61. 200 Cedarflat/Dewdney Creeks If s .s-§ s < 150 100 50 150| E. 100 o o £ 50 a Hope Tulameen River 0| i i i i _ 1950 1960 1970 1980 1990 2000 2010 Year Cascade Creek 1501 1 1 1 1 r i f s .& < 100 50 250 J O 200 E a i5oi (0 a 50 Stave Falls i i 1_ Norrish Creek J l l l_ 1950 1960 1970 1980 1990 2000 2010 Year Issues with the ability to interpret stream flow records are similar to issues with precipitation records. In many cases, records have missing data and records are incomplete. Annual maximum daily average discharges are not given in years with incomplete records since extreme events may have occurred during the unrecorded period. For Chapman Creek, gauging stations have been moved several times during the record. While extreme events can still be identified, comparison between sites is difficult because of the diversion of water to Sechelt, and there are no overlapping time frames between stations in order to establish correlations between them. 62 5.3 Landslide Responses All study watersheds exhibit hillslope responses after logging of steep slopes. Patterns of landslide responses varied between watersheds (Figure 5.2). The history of logging in basins can affect the timing of responses in basins. In most cases, logging takes place in watersheds over a wide period of time. Generally, logging occurs first in the lower reaches within watersheds, and development proceeds upstream into the headwater regions. Steeper slopes tend to be logged later in the development process, after more easily accessible timber from valley bottoms has been logged. The extended history of logging leads to extended cumulative responses on hillslopes 63 , Figure 5.2 Coupled landslide frequency for study watersheds. Logging increments are displayed as linear increases between observed periods, and landslides are grouped between dates. Dashed lines indicate years of extreme precipitation events. Cascade Creek 0 J 1 E T3 C CD 40000 30000h 20000 looooh 1940 1950 1960 1970 1980 Year 1990 2000 Chapman Creek 1940 1950 1960 1970 1980 1990 2000 Year 64 Dewdney Creek 1940 1950 1960 1970 1980 1990 2000 2010 Year 65 10000 <= ?r 8000 (1) C Q = CD 6000 1 E « " 4000 JS 2000 1940 1950 1960 1970 1980 1990 2000 2010 Year Most watersheds exhibit increases in landslide activity after logging. An exception is Theodosia River, where the watershed was subject to high landslide activity prior to logging (Figure 5.2). While the pre-logging landslides occurred in a few individual debris flow paths in one region of the basin, their occurrence does highlight the potential for natural variability of landslide activity. Other basins, such as Norrish Creek, have also been subject to multiple episodes of landslide activity. Again, this reflects the logging history within the basin, where major logging activity took place in stages. A summary of landslide response characteristics is given in Table 5-3. Response times are calculated as the time between logging at a particular location until the time of maximum landslide area, and therefore are calculated on an individual landslide basis. Since logging occurs at different locations within a watershed, response times are not calculated at a basin scale (e.g. Figure 5.2). Relaxation times are calculated as the time from peak landslide area until the region has no landslides. In regions where landslide persist, relaxation had not occurred, 66 and therefore those regions were excluded from basin average relaxation time calculations. Response Time Relaxation Time* Total Response Watershed (years) (years) (years) AtreSp S.E. Atreiax S.E. Attotai S.E. Cascade 23.8 6.0 13.5 5.5 37.3 8.1 Cedarflat 0.0 8.0 13.7 8.0 13.7 11.3 Chapman 14.4 5.0 11.0 5.5 25.4 7.4 Dewdney 11.4 8.0 20.0 6.3 31.4 10.1 Norrish 22.0 6.7 12.1 7.9 34.1 10.3 Slesse 18.6 7.6 14.5 7.3 33.1 10.5 Theodosia 8.3 6.4 15.9 9.0 24.2 10.9 All Basins 16.4 6.5 13.8 6.6 30.2 9.3 Table 5-3 Summary of landslide response and relaxation times for study basins. * Note: Relaxation times calculated only from regions where hillslopes appeared to stabilize. Overall, the average response time for all basins is 16.4 years, the average relaxation time is 13.8 years, and the average total response (response time plus relaxation time) is 30.2 years. Standard error is calculated JV ( S i s ) 2 —— r e t t C h—. There is considerable variability in response times ^reaches between basins. Average response times range from 0 to 23.8 years. The low values in Cedarflat Creek are the result of a small sample of landslides in that basin, and all landslides occurred, within the temporal resolution, at the same time as logging. The nature of responses also varies between watersheds, some basins having longer response times compared to their relaxation times, and other 67 channels have the reverse pattern. The combined overall responses, however, are fairly similar between watersheds, with an average of about 30 years. This is roughly similar to the time frame suggested by Brardinoni etal. (2002) (>20 years) for hillslope response. 5.4 Discussion Study sites have variable hydroclimatic histories. Long-term trends are apparent in precipitation and discharge, and likely reflect climatic variability within the region (Moore and McKendry, 1996; Eaton etal., 2002). For example, most study sites appear to have increased daily maximum precipitation during the past 20 years. Increases in long-term average peak discharge, however, do not appear to occur in most basins (with the exception of perhaps Slesse Creek). A derived history of extreme precipitation and discharge events is given in Table 5-4. The years of extreme precipitation are not necessarily the same as years of extreme streamflow. However, extreme peak flows occur in the fall and winter period (except for the 5 t h largest peak flow in Tulameen River). Peak discharges are associated with rain, or rain-on-snow, events but are not necessarily related to extreme precipitation. Antecedent conditions may be important for the timing of fall extreme stream discharge events, for example they may occur after several days of heavy rain, rather than a one-day event. 68 Table 5-4 Timing of extreme precipitation and discharge events for study sites3. Study Site Years of Maximum Precipitation (ranked) Magnitude of Largest Event (Maximum/A verage) Years of Maximum Discharge (ranked) Magnitude of Largest Event (Maximum/Average) Cascade 1990, 2002, 1955, 1951,1980 1.63 1963, 1990, 1961, 1997, 1980 1.93 Chapman 1968, 1945, 1963, 1995, 1953 1.68 1983, 1981, 1962, 1968, 1975 -Cedarflat/ Dewdney 1990, 1975, 1974, 1984,1979 2.25 1995, 1989, 1990, 1980, 1997 1.95 Norrish 1990, 1986, 1972, 1971,1995 1.64 1963, 1990; 1961, 1997, 1980 1.93 Slesse 1979, 1990, 1986, 1972,1989 1.69 1995, 1980, 1999, 1997, 1994 1.96 Theodosia 1997, 1947, 1951, 1998, 1949 1.77 1965, 1963, 1975, 1955, 1974 1.89 a information is based on the precipitation and discharge stations discussed in text. Some issues arise with the timing of extreme precipitation and discharge events for study sites. In general, the range of the magnitudes of extreme events is similar between study sites. One exception is the largest precipitation event in Hope (applied to Cedarflat Creek and Dewdney Creek) in 1990. This event was relatively large (over twice as large as the average annual maximum daily precipitation) in comparison to other events recorded at that site. While extreme precipitation does not necessarily coincide with the timing of extreme discharge events, there is some overlap. In Cascade Creek, extreme precipitation and discharge events occurred in 1990 and in 1980. Also in 1990 there were extreme precipitation and discharge events in Cedarflat Creek, Dewdney Creek and Norrish Creek. In some cases, the timing of extreme precipitation or extreme discharge events was grouped in time. In Dewdney and Cedarflat Creeks (from Hope data), four of the five most extreme precipitation events from 1934 to 1995 occurred during the decade from 1974 to 69 1984. At Theodosia River (Powell River) another three (of the five most extreme) events occurred in a four year period from 1947 to 1951. Similar trends occur for patterns of extreme streamflow. In Slesse Creek, four of the five most extreme events from 1957 to 2001 occurred from 1994 to 1999. With the exception of the temporal trends in extreme hydroclimatic events discussed, there does not appear to be any extraordinary pattern in precipitation or streamflow. 70 6 Channel Response Regimes 6.1 Introduction Patterns are evident in the nature of watershed responses. The purpose of this chapter is to discuss these patterns. Five types of watershed response regimes are identified and these regimes are defined based on the timing and evident mechanisms of response. They include watersheds with no significant channel changes (Type I), riparian disturbance response (Type II), landslide coupling disturbance response (Type III), disturbance propagation response (Type IV) and compound response (Type V). In this chapter, each of these response regimes is presented, and observed watershed responses in study sites are discussed in relation to these response regimes. 6.2 No Significant Channel Changes (Type I) Many stream channels exhibited no significant channel changes. This is the result of either no actual change to channel width, or the variability in width not exceeding the potential error in width measurements (either 0.07W0 or 5.0 m as discussed in Chapter 4). Channels with no significant changes to width are classified as Type I response. Thirty-six of the total 119 reaches exhibited no significant channel change. Typically, these channels tended to be smaller in size. The average W0 of Type I disturbance is 15.7 m (S.D. = ±6.3 m) and average maximum width, Wmax (maximum width observed in a reach over the time period), is 17.7 m (S.D. = ±6.5 71 m)(Figure 6.1). The average change in width, AW, is 1.9 m (S.D. = ±1.5 m), and average fractional width change, %AW, is 10%(S.D.= ±13 %). While these channels do exhibit change in width throughout the study period, the width changes are not significantly larger than the error associated with the measurements. The summary statistics reflect this; average A W is much smaller than the threshold of 5 m. Figure 6.1. Variable distributions for Type I response reaches (n=36). a.) Initial reach width (W0) b.) Maximum reach width (Wm a x) c.) Width magnification. CO CD •s E b.) 20 Wo(m) W m a x (m) 2 15r 10h 0.1 0.2 0.3 Magnification 0.4 0.5 Channels which exhibited Type I response included channels with both landslide coupling and riparian disturbance. Nineteen of the Type I channels (52%) had some amount of hillslope coupling, and 19 channels (52%) had riparian logging. The watersheds of Cedarflat Creek, Chapman Creek, Dewdney Creek, and Norrish Creek contain reaches with Type I responses. As an example, results 72 (Figure 6.2) and images (Figure 6.3) are shown from Chapman Reach 10. As is apparent from Figure 6.2, the reach is coupled and undergoes landslide delivery of sediment during the study period. While there is some variability in channel width, changes are smaller than potential measurement errors. Figure 6.2 Type I Response-Chapman Reach 10. Error bars for channel width represent average measurement error. The timing of extreme discharge events are indicated with dashed line. c c (0 6 0 0 ) f Channel Width • Landslides 5000 4000 Q . 3000 » 2000 1000 1950 1960 1970 1980 1990 2000 CD D) Year 73 Figure 6.3 Airphotos (1957-1998) of Chapman Reach 10. 74 75 6.3 Riparian Disturbance Response (Type II) In many reaches that undergo logging of riparian vegetation, channel instability occurs after the loss of bank strength. In coupled reaches, the maximum changes to channel width occur prior to the timing of maximum landslide delivery of sediment to the channel. This suggests that changes in the riparian vegetation are what is driving channel response rather than the delivery of increased sediment loads from landslides. Type II disturbances were defined based on having AW>5.0 m and %AW>7% and, if the channels were coupled, the timing of peak response is not evidently affected by timing of landslide inputs. Overall, 40 reaches exhibited Type II response. The average W0 of reaches with Type II responses is 16.9 m (S.D. = ±9.1 m), while the average maximum width, Wmax, is 31.6 m (S.D. = ±16.1 m) (Figure 6.4). The average channel response magnification for Type II responses is %AW= 94% (S.D. = ±60%). The average response time, the time from logging disturbance to maximum width response, is 17.4 years (S.D. = ±14.2 years). There is a wide range of the timing of response times. This likely reflects variability in the timing of extreme peak discharges and variability in the timing of the loss of root strength, as well as low precision in time measurements. Channels, through time, do appear to return to pre-disturbance widths. The average relaxation time is 28.4 years (S.D. = ±11.0 years). Relaxation times were calculated as the time period over which the maximum width returns to a width that is similar to (i.e. within the error of measurements) the initial width. Not all channels returned to a width similar to initial 76 conditions, and only channels which exhibited significant narrowing (n = 22) were included in relaxation time calculations. The process of narrowing may be ongoing in reaches where it was not observed, however there is neither adequate time elapsed nor adequate time coverage of air photos to assess the relaxation. Figure 6.4. Variable distributions for Type II response reaches (n=40). a.) Initial reach width (W0) b.)Maximum reach width (Wm a x) c.) Response time (T res) d.) Relaxation time (Tre!) e.) Width magnification a.) c.) <D 1 20 1 n s 1 2 i CD CD X> £ 8 6 4-2" 10 10 20 30 40 Wo (m) 20 30 Tres (years) 50 40 60 50 b.) 8 1 5 i CD CD OH CO £ 3 10 T Z Z L 0' 10 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0 Wmax (m) d.) OH 8" 7 6" 5 4 3 2 1 0' T 1 1 r 10 20 30 40 50 Trel (years) e.) o co CD or E Z 8 7-6 5 4! 3h 2 1 0.5 1.0 1.5 Magnification There is a wide range of magnitudes of channel widening amongst Type II response reaches. There are some patterns of channel response amongst channel 77 morphology types (Figure 6.5). The greatest amount of channel widening (up to %AW=330%) occurred in multi-thread channels, though all reach types all show significant width increases in response to riparian disturbance. Figure 6 . 5 Channel widening in Type II response reaches. 5 < 350% 300% 250% 200% 150% 100% 50% 0% Disturbance II • • A • Single-Thread • Confined A Multi-Thread 0.0 10.0 . 20.0 30.0 40.0 50.0 60.0 70.0 Wo Type II responses were observed in reaches in all study watersheds, and in general was the most common type of channel response observed. As an example of Type II responses, results (Figure 6.6) and images (Figure 6.7) are shown from Reach 4 of Theodosia River. This reach illustrates the time lag between logging 78 riparian disturbance and channel width response (Figure 6.6). Within the study time period, recovery has not occurred in this reach. Figure 6.6 Response Type II- Theodosia River Reach 4. Error bars for channel width represent average measurement error. The timing of extreme discharge events are indicated with dashed line. 60 E $ Channel Width 4 5 " O 01 i I L _ U L A U i i I 1940 1950 1960 1970 1980 1990 2000 Year 79 Figure 6.7 Air photo coverage for Theodosia Reach 4 (1947-1996). (Images courtesy of M.J. Miles) 80 81 6.4 Landslide Coupling Disturbance Response (Type III) Some reaches exhibited responses to landslide delivery of sediment to channels. The timing of channel widening coincided with or lagged the delivery of sediment from landslides. Reaches were coupled, and had either riparian logging or no riparian logging. In riparian logged reaches where the timing of peak landslide response and channel response coincide, it is difficult to attribute the cause of response. The timing of hillslope response may precede channel response; however it may not cause the channel response. In coupled reaches where no riparian logging occurs, significant channel widening at the onset or after increased sediment delivery is likely caused by the increased sediment loads derived from landsliding. Landslide coupling disturbance responses are classified as Type III responses. Overall, 16 of the study reaches had Type III responses. The average W0 of reaches with Type III responses is 19.3 m (range of 8.3 to 88.2 m), while the average maximum width, Wmax, is 32.8 m (range of 18.0 to 107.5 m) (Figure 6.8). The average width increase for Type III responses is AW= 13.5 m (S.D. = ±7.1 m), and average magnification is %AW= 86% (S.D. = ±63%). The average response time for Type III regimes is 21.9 years (S.D. = ±15.2 years). The average lag time between peak landslide delivery and maximum channel width is 1.3 ± 3.1 years. Average relaxation time (based on observed relaxation in 9 reaches) of Type III responses is 14.2 years (S.D. = ±5.7 years), which is much shorter than the Type II relaxation times. This likely reflects a more extended time period of sediment 82 delivery from bank erosion as compared to hillslopes, which might deliver a shot impulse of sediment in a shorter period of time. As well, this may reflect the longer period of time that is required for channels to re-establish bank stability from vegetation regeneration. Figure 6.8. Variable distributions for Type III response reaches (n=16). a.) Initial reach width (W0) b.) Maximum reach width (Wm a x) c.) Response time (T res) d.) Lag time between peak landslide response and peak channel width response time (T d ei) e.) Width magnification f.) Relaxation time (Tre!). 15r 10r _i_ 0 10 20 30 40 50 60 70 80 90 Wo(m) b.) 9i 1 1 1 1 1 1 1 r 8-7" 6-5 " 4 -3 " 2 -1 -10 20 30 40 50 60 70 80 90 100 110 Wmax (m) 20 30 Tres (years) 1 1 1 r- r ] I I 2 4 6 8 10 12 Tdei (years) 1 2 Magnification 9 "<& 8 t 7 a) 6 on 10 20 30 T r e i (years) There is a wide range of magnitudes of channel widening amongst Type III response reaches (Figure 6.9). The greatest amount of channel widening was 83 %AW= 245%. In general, smaller channels (<20 m) exhibited the greatest amount of proportional channel widening due to landslide delivery. This seems reasonable, since larger stream systems have a greater ability to convey sediment that is delivered to them, and therefore need increasingly large amounts of sediment delivered to them to cause major channel responses. Figure 6.9 Channel Widening in Type III Response Reaches 300% 250% 200% < 150% 100% 50% 0% Disturbance • Single-Thread • Confined • Multi-Thread 0.0 20.0 40.0 60.0 80.0 100.0 W 0 Type III response regimes were observed in reaches in all seven study watersheds, however it was most prevalent at Cascade Creek sites (a total of 9 reaches out of the 16 reaches studied in Cascade Creek had Type III responses). As an example, responses (Figure 6.10) and images (Figure 6.11) are given for 84 Cascade reach 14. In this example the reach receives large amounts of landslide delivered sediment after logging. Channel aggradation is visible on air photos, and coincides with the timing of landslide sediment delivery. Channel recovery processes are starting to take place. After major channel widening and increased landslide frequency during the 1970s and 1980s, by 1993 the hillslopes were beginning to stabilize and the channel was starting to narrow. Figure 6.10 Response Type III- Cascade Creek Reach 14. Error bars for channel width represent average measurement error. The timing of extreme discharge events are indicated with dashed line. c c CO 6 50 40 30 20 10 1 \ Channel Width • Landslides Riparian Logg\ng 1 1950 1960 1970 1980 Year - . 70000 - 60000 §" - 50000 | - 40000 £ - 30000 £ - 20000 j& - 10000 % 1990 2000 85 86 6.5 Disturbance Propagation Response (Type IV) Some reaches underwent major width changes after logging within watersheds, without having any direct riparian or landslide changes of input. Channel responses are, therefore, not a result of changes to bank condition or hillslope sediment delivery. Channel responses coincide with or lag channel responses upstream. In these cases, disturbance appears to propagate downstream. These disturbance regimes are classified as a disturbance propagation response (Type IV). Overall, 8 reaches exhibited Type IV responses. Seven of these reaches had no riparian logging. The reach that had riparian logging did not have significant changes to channel width until nearly 60 years post-logging, and the channel widening was synchronized with significant channel changes in adjacent upstream 87 reaches. One of the reaches had minor hillslope coupling, and the other eight reaches had no landslide delivery during the study period. ' Type IV responses occurred in Cedarflat Creek, Chapman Creek, Dewdney Creek, Slesse Creek, and Theodosia River, and were observed only in alluvial or multi-thread morphologies (in one instance, disturbance propagated through a confined reach and affected downstream reaches without significant changes to the confined reach). Type IV disturbances tended to occur in the lower, downstream portions of watersheds. This reflects the mechanism of disturbance, but can also be the result of the increasingly alluvial nature of channels further downstream in watersheds. Average Type IV initial width, W0, is 19.7 m (range of 11.1 to 40.3 m), and average Wmax is 54.2 m (range of 12.8 to 93.2 m) (Figure 6.12). The average width increase for Type IV responses is AW= 34.5 m (S.D. = ±22.0 m), and average magnification is %AW= 179% (S.D. = ±104%). Average response time for Type IV responses is 22.5 years (S.D. = ±16.3 years). Compared with Type II and Type III responses, propagation type responses, have much larger magnifications. 88 Figure 6.12. Variable distributions for Type IV response reaches (n=8). a.) Initial reach width (W0) b.) Maximum reach width (Wm a x) c.) Width magnification d.) Response time (T r e s). a.) 30 40 Wo(m) 1 2 3 Magnification b.) or 2 CP 11 10 20 30 40 50 60 70 80 90 100 W m a x (m) d.) 5 sL 2 4 Z 3 1 2 2 1 0 L 10 20 30 40 T r e 5 (years) 50 60 As an example of propagation type responses (Type IV) Dewdney Creek reaches 3 to 6 are given; reach 3 is furthest downstream, and reach 6 is upstream. Reaches 3 to 5 have no riparian disturbance or delivery of sediment from landslides; at reach 6 there was some riparian logging during the 1967-1983 period. All reaches (except one) upstream of reach 6 have been subjected to riparian logging, and some have had sediment delivery from landsliding. Response curves for Dewdney Creek reaches 3 to 6 are given in Figure 6.13, and images for these reaches are given in Figure 6.14. 89 Figure 6.13 Type IV Propagation Response-Dewdney Creek Reaches 3 to 6. The timing of extreme discharge events are indicated with dashed line. 80.0 1950 1960 1970 1980 1990 2000 Year . / The maximum observed width in all reaches was at the last photo image in 1996. Given the temporal resolution provided by the air photos, it is not possible to examine the dynamics of how these disturbances propagate from upper reaches, but the observations appear to reflect a cumulative effect of changes in up-stream channel stability. Upstream reaches (Reaches 6 to 10) exhibit riparian disturbance responses (Type II), and the timing of maximum widths coincides with the timing of maximum widths observed in Reaches 3 to 5. Response times in Type IV regimes are dependent on the response times of upstream reaches. Over fairly long stretches (several kilometers) along channels, disturbances propagate quickly or 90 may occur concurrently with upstream responses. In the case of Dewdney Creek, it appears that there is the potential for disturbances to magnify as they move downstream. Reach 3, the furthest downstream, has been subject to the greatest increases in channel width (%AW= 315%). A similar pattern is observed in Slesse Creek, where Reach 2 exhibits a Type IV response, and has experienced more extreme channel changes (%AW= 294%) than most of the reaches in the upper portion of the watershed. Several large peak flows occur over the same period of time as channel changes (Figure 6.13) and these discharge events are likely driving these changes. However, channel widths in disturbed reaches (upstream) widen prior to these extreme events and this suggests that channels are sensitive to change as a result of riparian disturbance. Channel instability is facilitated by extreme discharge events, and this instability is then propagated to undisturbed reaches downstream. Figure 6.14 Air photo coverage for Dewdney Creek Reaches 3 to 6 (1954-1996). (Images courtesy of M.J. Miles) 91 92 6.6 Compound Response (Type V) Some channels experienced responses that had components of Type II and Type III response types. These tended to exhibit a response to riparian disturbance, and later to experience increased sediment delivery from landslides; the response of the channels is multi-phased. Other reaches underwent simultaneous landslide and channel width response, such that the mechanism of response is not clearly distinguishable. Both of these situations are classified as Type V responses, and they reflect compound response mechanisms. Overall, 17 reaches exhibited compound response. Initial widths, W 0 , range from 10.9 to 66.4 m, with an average of 21.0 m (Figure 6.15). Post-disturbance widths, W m ax, range from 18.7 to 120.1 m, with an average W m a x of 46.7 m. Average magnification of Type V responses is %AW = 120% (S.D. = ±90%). The timing of width changes varies greatly, and the average response time is 18.7 years (S.D. = ±19.0 years). Magnification varied with initial channel width, and smaller channels generally had larger channel widening. 93 Figure 6.15 Variable distributions for Type V response reaches, a.) Initial Reach Width (W0) b.) Maximum Reach Width (Wm a x) c.) Width Magnification d.) Response time (T r e s). i 1 1 1 1 1 1 r 20 30 40 50 60 70 80 90 100 W m a x (m) 1 1 1 1 1 20 30 40 50 60 Tres (years) As an example of a compound response regime (Type V), response curves (Figure 6.16) and air photo images (Figure 6.17) are given for Chapman Reach 20. In Chapman reach 20, the channel initially widened after riparian logging. About 15 years after logging (riparian and hillslope), hillslopes were subject to an increase in mass wasting with sediment delivery to the channel. Following sediment delivery, channels continued to exhibit instability and widening. Images from 1990 () indicate that hillslope were recovering (little signs of landslide scars), yet the channel showed continued widening. Channels were beginning to show narrowing by 1998. a.) o CO CD CC CD E 30 40 50 W 0 (m) b.) CO CD CC 0' 1 d.) o CO CD CC CD E 1 2 3 4 Magnification § 8 € 7 S 6 K 5 2 4 2 3 I 2 z 1 °1 94 Figure 6.16 Type V Compound Response-Chapman Reach 20. Error bars for channel width represent average measurement error. The timing of extreme discharge events are indicated with dashed line. 25 20 15 10 5 4> 0 I Riparian Logging O P I 35000 -30000 » $ ^25000 | O - 20000 £ —115000 > 10000 » I Channel Width • Landslides 5000 0 3 1950 1960 1970 1980 Year 1990 2000 95 Figure 6.17 Air photo coverage for Chapman Creek Reach 20 (1957-1998). 97 6.7 Discussion Disturbance mechanisms varied among study reaches. Most channels exhibited distinct response patterns (Type I to Type IV regimes). However, some reaches exhibited response patterns that had characteristics of more than one response type (Type V). The observation of response regimes with multiple disturbance mechanisms highlights the variability in the mechanisms responsible for disturbance; it is likely that a variety of mechanisms of disturbance can affect the responses in all channels, and across all of the described response types. Disturbance regimes are given to highlight observed patterns of responses, and to suggest probable modes of disturbance. Due to variability in hydroclimatic forcing, watershed geomorphology, and spatial and temporal impacts of timber harvesting, watershed response is inherently complex (Figure 6.18). A / 2 -goodness-of-fit test was used to examine whether the distribution of channel reach morphologies within each disturbance type is significantly different than the overall sample distribution. The x2 -test statistic is given as %fv) = — , for a given morphology /' (total of k=3), where O, is the observed frequency, and E, is the expected frequency based on the overall sample morphology distribution. With degrees of freedom of v=2 and a=0.05, the critical x 2 is 5.99. The distributions of channel morphology types for different disturbance regimes are not significantly different then the sampling distribution for Type II and III regimes. However, Types I, IV and V response regimes all have a significantly different distribution of channel 98 morphologies than the overall sample 0f 2 > 5.99). Type I responses are over-represented by confined channels (64%), and under-represented by multi-thread channels (3%). For confined channels, this seems reasonable since stable boundary conditions in these systems will make them more resistant to instability. Multi-thread channels appear less likely to be subject to no significant response. This might suggest that these channels are less stable, and therefore more likely to be affected by changes in bank stability or sediment supply. Both Type IV and Type V responses are under-represented by confined channels and over-represented by multi-thread channels. This seems reasonable given channel dynamics. Multi-thread channels have high energy and, in general, have lower stability compared to confined channels. Confined channels typically have poorly developed or no alluvial floodplains over which the channel can move laterally, and therefore are less sensitive to channel disturbance. Confined channels are constrained from lateral movement, which may be the result of high bank strength either from reinforcement from footslopes, from coarse sediment, or from bedrock. 99 Figure 6.18 Distribution of disturbance types and channel morphology. T3 > o 1 o , <D -Q E 3 II C 45 40 35 30 25 20 15 10 5 0 Distribution of Disturbance Types and Morphology M Multi-Thread • Single-Thread S Confined II III IV Disturbance Type V Trends appear in the distribution of morphological types for other disturbance regimes, however they are not statistically significant. Type I responses appear to be over-represented by confined channels, and under-represented by multi-thread channels. Another trend is the over-representation of confined channels in Type III (landslide disturbance) responses. Increased sediment loads do have the ability to affect channel stability in confined systems. Confined channels have smaller channel widths, and therefore can be more easily overwhelmed by increases in sediment load. A summary of channel magnification is given for all channels (Figure 6.19). In general, confined channels are smaller, and are subject to lower magnitude 100 disturbances. Large magnitude channel disturbance occur in both single- and multi-thread channels. It appears as if disturbance affects medium-sized channels (20-30 m width) more than "larger" channels (>50 m). It is difficult to assess whether the lower magnitude disturbances observed in larger channels is the result of lower sensitivity of these systems; a larger sample of "large" channels is required. Figure 6.19 Channel width magnification for all reaches. 400% 350% 300% 250% % 200% 150% 100% 50% 0% A * • # A Confined • Single-thread • Multi-thread 0.0 25.0 50.0 Wo(m) 75.0 100.0 A summary of the characteristics of each response type is given in Table 6-1. Statistically, the response times for all disturbance types are similar, and range from 17.4 to 22.5 years. Riparian disturbances have the shortest response times, and propagation and landslide responses have the longest response times. Given the lagged effect of sediment delivery or transfer of disturbances down through the 101 watershed, longer response times would be expected in these systems. There is a high degree of variability in response times for a given disturbance type. This appearance is affected by poor temporal resolution using air photo records. Variability also arises from variability in the timing of hydroclimatic events in each basin, reach characteristics, and the timing of disturbances in adjacent reaches and on hillslopes. Table 6-1 Summary of Response Characteristics. Disturbance Type Disturbance Mechanism Response Time (years) Magnification (%) Relaxation Time (years) S.D. %AW S.D. Atreiax S.D. Type I No Significant Changes N/A N/A 10 13 N/A N/A Type II Riparian Disturbance 17.4 14.2 94 60 28.4 11.0 Type III Landslide Coupling 21.9 15.2 86 63 14.2 5.7 Type IV Propagation 22.5 16.3 179 104 - -Type V Compound 18.7 19.0 120 90 - -Landslides Hillslope Disturbance 16.4 6.5 - - 13.8 6.6 Propagation disturbances have the largest average magnifications (%AW = 179%). From a management perspective, this is of particular concern. While riparian management, such as riparian buffer zones, for a downstream reach may be appropriate, disturbances generated upstream can cause disturbance to propagate to downstream reaches. As well, the magnitude of the disturbance can magnify channel response as it progresses downstream (e.g. Figure 6.13). Timber harvest planning should consider the downstream transfer of disturbance in the 102 planning of riparian management. Protection of the local riparian zone may not be sufficient to protect riparian and channel resources. Riparian and landslide disturbance regimes have similar magnitude channel responses. However, landslide disturbances tend to affect smaller channels than riparian disturbance, for reasons previously discussed. Relaxation times vary among disturbance regimes. For many reaches, recovery was not observed during the time period of record. Average relaxation times were longer for riparian disturbances (28 years) compared with landslide disturbances (14 years). In landslide disturbed channels, instability in the channels is driven largely by increased sediment loads. The relaxation of the system occurs through conveyance of that material, or incision into aggraded material, so that the channel can re-equilibrate with discharge and sediment characteristics. Given fairly high frequency competent discharges, there is the potential for this process to occur on a fairly short time scale. In riparian disturbance situations, channel instability is largely driven by changes in bank strength and stability. For the system to stabilize, riparian vegetation needs to be re-introduced to disturbed areas. Stabilization may take a longer period of time, since the channels may continue to be active without the riparian vegetation. It may be difficult for vegetation to re-establish on active channel margins and, therefore, relaxation of these systems may be delayed. Overall responses (response and relaxation times combined) are 46 years for riparian disturbances, and 36 years for landslide disturbance. As expected, these time periods are similar to or longer than overall responses that have been observed for hillslope responses to logging (Brardinoni etal., 2002). 103 While four disturbance regimes were identified (not riparian logged and not coupled, riparian logged and not coupled, not riparian logged and coupled, and logged and coupled), it is not possible to predict response type based on disturbance regime (with the exception of Type II responses which are defined based on the presence of riparian logging, and landslide responses which are defined based on hillslope coupling). However, the effects of both riparian and landslide disturbance can be examined through an investigation of the characteristic responses to each of these disturbance types. Figure 6.20 Riparian disturbance response variable distributions. a.)lnitial Width (WO) b.) Maximum Width (Wmax) c.) Width Magnification d.) Response Time (Tres) e.) Recovery Time (Tree) f.) Response Regime Distribution. a.) » 80 J2 70 ro 60 G* 50h 0 40 jjj 30 E 20 1 10 0. —i 1 1 1 1 1 1 r n=89 1 1 1 0 10 20 30 40 50 60 70 80 90 Wo(m) c.) » 40 o 3 30 -O 10h n=89 Magnification 50 100 W m a x (m) d.) 1 1 1 r n=85 10 20 30 40 50 60 e.) » 16| 2 8| CO a I 4 n=54 10 20 30 40 50 rv v Response Type 104 A summary of response variables for all riparian disturbed reaches is given in Figure 6.20. Average initial width (W0) is 18.3 m (S.D. = ±12.2 m), average maximum width {Wmax) is 33.6 m (S.D. = ±23.1 m), and average magnification is 87% (S.D. = ±74%). Response and relaxation times were quite varied. Riparian responses were the most prevalent response type for channels which had riparian disturbance. However, this is essential given riparian disturbance is a prerequisite for channel responses that were defined as Type II. A similar number of channels had compound responses, or no significant response. Figure 6.21 Landslide disturbance and channel responses. Open symbols represent reaches without riparian logging, and closed symbols represent reaches with riparian logging. Landslide Disturbance 350% 20000 40000 60000 80000 100000 Landslide Area (m2) 105 Landslide disturbance effects also varied among reaches (Figure 6.21). Due to the low number of reaches which were subject only to landside disturbance (n=15), it is difficult to assess the impact of landslide sediment inputs alone for most channels. The average channel magnification for reaches is 48% (S.D. = ±50%). There does not appear to be a relationship between the area of coupled landslides in a reach, and the overall channel response magnification. Overall response is confounded by presence of both riparian and landslide disturbance in many of the reaches. One important issue is that there is an apparent sensitivity to disturbance in the lower sediment supplied (<10 000 m2) reaches. Most of these reaches, however, also have some degree of riparian logging. Given that there does not appear to be a strong increase in channel width responses with increased sediment delivery, it is reasonable to suppose that riparian disturbance exhorts strong controls on channel response. Another issue that cannot be addressed is the influence of down-stream effects. The apparent sensitivity of non-riparian logged reaches with low quantities of landslide sediment delivery can be the result of disturbances upstream. Propagated disturbances were hard to identify, particularly for reaches where local disturbances confound the overall disturbance mechanism regime. Hydroclimate is a confounding factor in understanding channel responses. Generally, extreme discharge events are scattered in time throughout the study periods. In some cases, extreme events are clustered in shorter time periods (less than a decade). The timing of major channel changes, however, does not generally occur concurrent with extreme discharge, rather the timing of channel changes lags 106 riparian and landslide disturbance. This pattern is ubiquitous, and suggests that changes are caused by the changes to the stream system through logging disturbance. Extreme events that occur before disturbance and after recovery (e.g. Figure 6.10) do not coincide with major changes to stream channel widths. Extreme discharge may, however, help drive the response in systems which are disturbed. In Dewdney Creek (Figure 6.13) significant channel changes coincide with a series of extreme streamflow events over the 1980s and early 1990s. Streamflow may act as the driving force that causes instability in systems that are more sensitive due to changes that have occurred to bank stability or sediment load due to other disturbances. 107 7 Conclusion 7.7 Introduction This report addresses issues surrounding the interactions between hillslopes and stream channels, and how those processes can be affected by timber harvesting activities. The purpose of this chapter is to highlight the key results of this research and to suggest future research directions. 7.2 Methods and Error: Applying GIS for long-term monitoring of landslides and stream channels This research applies a combination of traditional and new methods for analyzing landslide and stream channel changes. Emphasis has been placed on examining the interactions between hillslopes and stream channels in medium-sized mountainous watersheds (10-100 km2). While field studies can yield detailed information about watershed changes, long-term monitoring is challenging to implement due to logistics and cost. Air photo records can be beneficial for long-term monitoring because they allow for an investigation of processes that occur on a multi-decade time scale. Air photos were scanned into digital format to accommodate GIS analysis. The use of GIS technology allows for higher precision measurements. Images were geo-referenced to establish spatial control. Errors associated with image processing result from original image scale, resolution with which images are scanned and the quality of geo-referencing. Errors can be associated with the 108 precision of location or scale. Due to the large positional error of geo-referenced images, channel width measurements using fixed cross-sections would have error that is proportional to the width variability within a region of a channel, since there is high uncertainty associated with the location of measurements between images. Field measurements of channel width suggested that width variability for a given reach can vary by as much as 50% within a given reach. Therefore, given the uncertainty of measuring width at the same location from image to image, width measurement using fixed cross-sections would likely not be precise enough to detect temporal changes in width. Reach average width measurements calculated from reach length and area provide higher precision since positional error is integrated over the whole reach. Replicate measurements were made to assess the precision of reach average width calculations. Precision of average width measurements varies with channel width, and averaged 1.2 m (S.D. = ±1.2 m), or 3.1 % (S.D. = ±2.1 %). The accuracy of channel measurements can be affected by the channel being obscured due to vegetation or bank overhang, and for this study, GIS measurement of width tended to under-estimate field measured width by 3.5 m. This can affect channel comparisons after riparian logging since vegetation may then no longer obscure channel banks. In order to detect changes in channel width from one time period to another, the magnitude of the change needs to be greater than the error inherent in channel measurements. For this study, changes in channel width that were less than 5.0 m and less than 7% of previous measurements were considered to be insignificant. 109 7.3 Watershed responses to timber harvesting disturbance Timber harvesting can affect geomorphic processes in watersheds. Road construction and tree removal on hillslopes can cause increases in the frequency of landslides. Changes to riparian vegetation can decrease channel stability and, in coupled watersheds, sediment supply from hillslopes can also cause channel instability. This study examined hillslope and stream channel responses in seven steep, coupled watersheds in the Coast and Cascade Mountains of British Columbia. The purpose of this investigation was to examine how hillslopes and stream channels respond to logging, and to determine how hillslope responses influence stream channel response. Watershed changes after logging were ubiquitous. Channel responses are classified into five patterns which specify processes evidently responsible. These response types are riparian response, landslide-driven response, propagated response, compound response, and no significant response. Channels which did not undergo detectable changes during the study period were classified as having no significant response. Riparian responses appeared to be characterized by changes in channel stability as the result of loss of bank strength. Landslide driven responses had channel width increases which coincided or lagged major sediment delivery from hillslopes. Propagated disturbance occurred in reaches which had no local timber harvesting, but were affected by upstream channel destabilization. Compound responses exhibited a combination of characteristics from riparian and landslide-driven response regimes. 110 Response patterns varied slightly among disturbance regimes. In general, channel responses occurred about 20 years after logging disturbance. Channels appear to recover to pre-disturbance widths, and this process takes approximately 15 to 30 years. Channels with riparian responses appear to take longer to recover than areas with landslide-induced response. Recovery was not observed in propagation and complex type responses, though this is likely due to an insufficient length of record rather than an inability to recover. The greatest impact on channel widths occurred in propagated responses (average width increases of 179%). This is an important issue for forest management, since it reflects the possibility for channels to be impacted by distant operations rather than just by local operation conditions. Therefore, management of upstream regions must consider the potential effect on downstream reaches. Riparian management, such as riparian buffers, may be insufficient to protect aquatic environments if upstream management affects channel stability. Propagation responses appear to affect alluvial and multi-thread channel types more severely than confined channels. Riparian disturbance regimes produced the second greatest impacts. On average, channel widening of riparian disturbance responses was 94%. Riparian disturbance regimes took longer to recover than landslide induced responses (28 years on average, compared with 14 years for landslide-induced responses). Landslide-induced responses also produced significant channel changes. Confined channels were affected more often by landslides; however this could be 111 the result of a higher degree of coupling in regions with confined channels rather than higher sensitivity to this kind of disturbance. In general, the greatest amount of proportional width changes occurred in smaller channels (10 to 15 m). Landslide-driven responses tended to affect smaller channels (less than 30 m), which may indicate increased capacity of larger channels to convey sediment or for adjacent terrain to buffer landslide sediment delivery. Wider channels tended to be affected more by riparian disturbance regimes, representing higher sensitivity of these systems to changes in bank strength. This seems reasonable since wider and more completely alluvial channels will likely have lower bank strength than confined channels. Propagated responses also tend to affect wider streams. 7.4 Future research directions i Several recommendations stem from this research. This report addresses some of the aspects of watershed responses; it also highlights further questions that need to be investigated. A primary concern that should be addressed is the poor temporal resolution in this research. In general, air photos were taken every 10 years, which provides much too coarse a resolution to properly describe hillslope and channel responses. Similarly, a total of 4 to 7 photo time periods were available for each study basin. This allowed for an investigation of the patterns of watershed change over an approximately 50 year period. Further research into watershed changes would benefit from increased temporal resolution. Increased temporal resolution would 112 allow for the identification of more complex relationships between response characteristics. Increased temporal resolution may be significantly improved through the use of satellite imagery. At present the cost of obtaining long time series coverage from satellite images is prohibitively expensive. Another benefit that could be added to this study would be increased investigation into the relations between channel pattern, watershed conditions, logging "treatment", and degree of hillslope coupling. A multivariate approach to watershed responses would require greater experimental control over these factors. This study is largely a case study into watershed response, and therefore has not dealt with watershed response in an experimental way. To properly relate the interactions between hillslopes and stream channels, better methodology is required, since there need not be relations between the area of a landslide and the amount of sediment actually delivered to a stream channel. Further research is required to properly quantify how much of the sediment that is mobilized in mass wasting events is actually delivered to streams. A final direction would be to investigate other types of disturbance regimes. This study focused on the effects of timber harvesting on watershed processes. This problem could be extended to strictly examining the response of stream channels to discrete sediment inputs, in order to investigate thresholds for hillslope-channel interactions. Investigating the dynamics of propagation disturbance in greater detail is also required. In terms of forest management, the potential for disturbances to propagate downstream needs to be investigated to assess how to incorporate downstream effects into riparian management practice. Similarly, 113 severe forest fires can dramatically affect root strength in watersheds, and may have effects similar to timber harvesting on channel stability. Patterns of response and recovery may be unique. 114 8 References Anonymous, 1997. Policies and specifications for TRIM II (Version 2). Geographic Data BC, BC Ministry of Environment, Lands and Parks. 83pp. Beaudry, P.G. and A. Gottesfeld, 2001. Effects of forest-harvest rates on stream channel changes in the central interior of British Columbia. In: D.A.A. Toews and S. Chatwin (Editors), Watershed assessment in the southern interior of British Columbia. Research Branch, BC Ministry of Forests, Victoria. Benda, L. and T. Dunne, 1987. Sediment routing by debris flows. In: R.L. Beschta, T. Blinn, G.E. Grant, F.J. Swanson and G.G. Ice (Editors), Erosion and Sedimentation in the Pacific Rim. IAHS Publication, pp. 213-223. Benda, L.E. and T. Dunne. 1997. Stochastic forcing of sediment supply to channel networks from landsliding and debris flow. Water Resources Research, 33: 2849-2863. Bovis, M.J. and M. Jakob. 2000. The July 29, 1998, debris flow and landslide dam at Capricorn Creek, Mount Meager Volcanic Complex, southern Coast Mountains, British Columbia. Canadian Journal of Earth Sciences, 37(10): 1321-1334. Brardinoni, F. and M. Church. 2004. Representing the landslide magnitude-frequency relation: Capilano River basin, British Columbia. Earth Surface Processes and Landforms{29): 115-124. Brardinoni, F., O. Slaymaker, and M.Hassan, 2003. Landslide inventory in a rugged forested watershed: a comparison between air-photo and field survey data. Geomorphology, 54(3-4): 179-196. Brardinoni, F., M.A. Hassan and H.O. Slaymaker. 2002. Complex mass wasting response of drainage basins to forest management in coastal British Columbia. Geomorphology, 49: 109-124. Buffington, J.M., and Montgomery, D.R. 1999. Effects of sediment supply on surface textures of gravel-bed rivers. Water Resources Research, 35(11): 3523-3530. 115 Campbell, D. and M. Church. 2003. Reconnaissance sediment budgets for Lynn Valley, British Columbia: Holocene and contemporary time scales. Canadian Journal of Earth Sciences, 40: 701 -713. Chang, J.C. and O. Slaymaker, 2002. Frequency and spatial distribution of landslides in a mountainous drainage basin: Western Foothills, Taiwan. Catena, 46(4): 385-307. Cheong, A.L., 1992. Quantifying drainage basin comparisons within a knowledge-based system framework. MSc Thesis, University of British Columbia. 133pp. Chorley, R.J., and B.A. Kennedy. 1971. Physical Geography: A Systems Approach. Prentice-Hall International Inc., London. 370pp. Church, M. 1983. Concepts of sediment transfer and transport on the Queen Charlotte Islands. Fish/Forestry Interaction Program. Working Paper 2/83. 30pp. Church, M. 1992. Channel morphology and topology. In: P. Calow and G.E. Petts (Editors), The River Handbook. Blackwell, pp. 126-143. Church, M., 1998. The Landscape of the Pacific Northwest. In: D.L. Hogan, J.J. Tschaplinski and S. Chatwin (Editors), Carnation Creek and Queen Charlotte Islands Fish/Forestry Workshop: Applying 20 Years of Coastal Research to Management Solutions. BC Ministry of Forests, pp. 13-21. Church, M., 2002. Fluvial sediment transfer in cold regions. In: K.E.A. Hewitt (Editor). Kluwer Academic, Netherlands, pp. 93-117. Church, M., D. Ham, M. Hassan and O. Slaymaker. 1999. Fluvial clastic sediment yield in Canada: scaled analysis. Canadian Journal of Earth Sciences, 36: 1278-1280. Church, M., M. Hassan and J. Wolcott. 1998. Stabilizing self-organized structures in gravel-bed stream channels: Field and experimental observations. Water Resources Research, 34(11): 3169-3179. Clague, J . , J.M. Ryder, W.H. Mathews et al., 1989. Quaternary Geology of the Canadian Cordillera. In: R.J. Fulton (Editor), The Geology of North America. Geological Survey of Canada, pp. 15-96. 116 Clague, J.J., R. Turner, and A. Reyes. 2003. Record of recent river channel instability, Cheakamus Valley, British Columbia. Geomorphology, 53(3-4): 317-332. Dietrich, W.E. and T. Dunne. 1978. Sediment budget for a small catchment in mountainous terrain. Z. Geomprph. N.F., Suppl. Bd. 29: 191-206. Dietrich, W.E., T. Dunne, N.F. Humphrey and L.M. Reid, 1982. Construction of Sediment Budgets for drainage basins. In: F.J. Janda, T. Dunne and D.N. Swanson (Editors), Sediment Budgets and Routing in Forest Drainage Basins. USDA Forest Services, pp. 5-23. Downward, S.R., A.M. Gurnell and A. Brookes, 1994. A methodology for quantifying river channel changes using GIS. In: L.J. Oliva, R.J. Loughran and J.A. Kesby (Editors), Variability in stream erosion and sediment transport. International Association of Hydrological Sciences, Wallingford, UK, Volume 224, pp. 449-456. Eaton, B.C., M.Church and D.Ham. 2002. Scaling and regionalization of flood flows in British Columbia, Canada. Hydrological Processes, 16(16): 3245-3263. Eaton, B.C., M. Church and R.G. Millar. 2004. Rational regime model of alluvial channel morphology and response. Earth Surface Processes and Landforms, 29(4): 511-529. Eide, J . , 1990. A 48-year sediment budget (1942-1989) for Deer Creek Basin, Washington. MSc Thesis, University of Western Washington. 122 pp. Friedman, J.M., W.R. Osterkamp and J . Lewis, W.M. 1996. The role of vegetation and bed-level fluctuations in the process of channel narrowing. Geomorphology, 14: 341-351. Friele, P. and J . Clague. 2004. Large Holocene landslides from Pylon Peak, southwestern British Columbia, Canadian Journal of Earth Sciences, 41(2): 165-182. Gimbarzevsky, P., 1988. Mass wasting on the Queen Charlotte Islands: A regional inventory. No. Land Management Report Number 29, Ministry of Forests and Lands, Victoria, pp. 96. 117 Grant, G.E. and A.L. Wolff, 1991. Long-term patterns of sediment transport after timber harvest, western Cascade Mountains, Oregon, USA, Sediment and Stream Water Quality in a Changing Environment: Trends and Explanation. IAHS, Vienna, p. 31-40. Guthrie, R. and S. Evans. 2004. Magnitude and frequency of landslides triggered by a storm event, Loughborough Inlet, British Columbia. Natural Hazards and Earth System Sciences, 4(3): 475-483. Guthrie, R.H. 2002. The effects of logging on frequency and distribution of landslides in three watersheds on Vancouver Island, B.C. Geomorphology, 43: 273-293. Halwas, K.L. and M. Church. 2002. Channel units in small, high gradient streams on Vancouver Island, British Columbia. Geomorphology, 43: 243-256. Ham, D.G. and M. Church. 2000. Bed material transport estimated from channel morphodynamics: Chilliwack River, British Columbia. Earth Surface Processes and Land forms, 25: 1123-1142. Hogan, D.L., 1987. The influence of large organic debris on channel recovery in the Queen Charlotte Islands, British Columbia, Canada, Erosion and Sedimentation in the Pacific Rim (Proceedings of the Corvallis Symposium, August 1987). IAHS, pp. 343-353. Hogan, D.L., 2001. Stream channel assessment in the interior of British Columbia. In: D.A. Toews and S. Chatwin (Editors), Watershed Assessment in the Southern Interior of British Columbia: Workshop Proceedings. Research Branch, B.C. Ministry of Forests, pp. 112-133. Hogan, D.L., S.A. Bird and M.A. Hassan, 1998. Spatial and temporal evolution of small coastal gravel-bed streams: influence of forest management on channel morphology and fish habitats. In: P.C. Klingeman, R.L. Beschta, P.D. Komar and J.B. Bradley (Editors), Gravel-Bed Rivers in the Environment. Water Resources Publications, pp. 365-392. Hogan, D.L., J.J. Tschaplinski and S. Chatwin (Editors), 1998. Carnation Creek and Queen Charlotte Islands Fish/Forestry Workshop: Applying 20 Years of Coastal Research to Management Solutions. BC Ministry of Forests. 118 Hogan, D.L. and J.W. Schwab. 1990. Precipitation and Runoff Characteristics, Queen Charlotte Islands. Land Management Report No. 40. BC Ministry of Forests, 43 pp. Holland, S.S., 1976. Landforms of British Columbia: A Physiographic Outline. Bulletin, No. 48, British Columbia Department of Mines and Petroleum Resources, Government of the Province of British Columbia, pp. 138. Holm, K., M.J. Bovis and M. Jakob. 2004. The landslide response of alpine basins to post-Little Ice Age glacial thinning and retreat in southwestern British Columbia. Geomorphology, 57(3-4): 201-216. Howes, D.E., 1987. A terrain evaluation method for predicting terrain susceptible to post-logging landslide activity: A case study from the southern Coast Mountains of British Columbia. Technical Report, No. 28, Ministry of Environment, Lands and Parks, pp. 35. Hudson, R., 2001. Storm-based sediment budgets in a partially harvested watershed in Coastal British Columbia. Forest Research Technical Report, No. TR-009, BC Forest Service, pp. 40. Hudson, R. and B. D'Anjou, 2001. Roberts Creek Study Forest: the effects of shelterwood harvesting and blowdown on sediment production in a small zero-order creek. Forest Research Extension Note, No. EN-004, B.C. Forest Service, pp. 7. Jakob, M. 2000. The impact of logging on landslide activity at Clayoquot Sound, British Columbia. Catena, 38: 279-300. Jakob, M., D. Anderson, T. Fuller, O. Hungr and D. Ayotte. 2000. An unusually large debris flow at Hummingbird Creek, Mara Lake, British Columbia. Canadian Geotechnical Journal, 37(5): 1109-1125. Kerr, D. and W.G. Kendrew. 1955. The Climate of British Columbia and the Yukon Territory. Queen's Printer: Ottawa, 222 pp. Kondolf, G.M., H. Piegay and N. Landon. 2002. Channel response to increased and decreased bedload supply from land-use changes: contrasts between two catchments. Geomorphology, 45: 35-51. 119 Liebault, F. and H. Piegay. 2002. Causes of 20th century channel narrowing in mountain and piedmont rivers of southeastern France. Earth Surface Processes and Landforms, 27: 425-444. Lynn, M.L, and F.S. Jang. 2002. Characteristics of hazards induced by extremely heavy rainfall in Central Taiwan - Typhoon Herb. Engineering Geology, 58(2): 191-207. Lyons, J.K. and R.L. Beschta. 1983. Land use, floods, and channel changes: Upper Middle Fork Willamette River, Oregon (1936-1980). Water Resources Research, 19(2): 463-471. Martin, Y., K. Rood, J.W. Schwab and M. Church. 2002. Sediment transfer by shallow landsliding in the Queen Charlotte Islands, British Columbia. Canadian Journal of Earth Sciences, 39: 189-205. Millar, G.R. 2000. Influence of bank vegetation on alluvial channel patterns. Water Resources Research, 36: 1109-1118. Millar, R. and M.C. Quick. 1993. Effect of bank stability on geometry of gravel bed rivers. Journal of Hydraulic Engineering, 119(12): 1343-1363. Millard, T., Rollerson, T.P., and Thomson, B., 2002. Post-logging landslide rates in the Cascade Mountains, southwestern British Columbia. Technical Report, No. TR-023, Forest Research, Nanaimo, Canada, 18 pp. Miller, D.J. and L.E. Benda. 2000. Effects of punctuated sediment supply on valley-floor landforms and sediment transport. Geological Society of American, Bull., 112: 1814-1824. Montgomery, D.R. and J.M. Buffington. 1997. Channel-reach morphology in mountain drainage basins. Geological Society of America, Bulletin, 109(5): 596-611. Moore, R.D., and I.G.McKendry. 1996. Spring snowpack anomaly patterns and winter climatic variability, British Columbia, Canada. Water Resources Research 32(3): 623-632. Mount, N.J., J . Louis, R.M. Teeuw, P.M. Zukowskyj and T. Stott. 2003. Estimation of error in bankfull width comparisons from temporally sequenced raw and corrected aerial photographs. Geomorphology, 56: 65-67. 120 Nakamura, F., F.J. Swanson, and S. Wondzell, 2000. Disturbance regimes of stream and riparian systems-a disturbance-cascade perspective. Hydrological Processes, 14, 2849-2860. Paulson, K.M., 1997. Estimating changes in sediment supply due to forest practices: A sediment budget approach applied to the Skagit River basin in Northwestern Washington. MSc Thesis, University of Washington. 156 pp. Reid, L.M. and T. Dunne. 1996. Rapid Evaluation of Sediment Budgets. GeoEcology Paperback, 164 pp. Reid, L.M., T. Dunne and C.J. Cedarholm. 1981. Application of Sediment Budget studies to the evaluation of logging road impact. Journal of Hydrology (New Zealand): 49-62. Rice, S. and M. Church. 1996. Bed material texture in low order streams on the Queen Charlotte Islands, British Columbia. Earth Surface Processes and Landforms, 21(1): 1-18. Roberts, B., B. Ward and T. Rollerson. 2004. A comparison of landslide rates following helicopter and conventional cable-based clear-cut logging operations in the Southwest Coast Mountains of British Columbia. Geomorphology, 61(3-4): 337-346. Roberts, R.G. and M. Church. 1986. The sediment budget in severely disturbed watersheds, Queen Charlotte Ranges, British Columbia. Canadian Journal of Forest Research, 16(5): 1092-1106. Rollerson, T.P., 1992. Relationships between landscape attributes and landslide frequencies after logging: Skidegate Plateau, Queen Charlotte Islands. Final Report, BC Ministry of Forests. Rood, K.M. 1984. An Aerial Photograph Inventory of the Frequency and Yield of Mass Wasting on the Queen Charlotte Islands, British Columbia. Land Management Report No. 34. BC Ministry of Forests. Sakals, M. and R. Sidle. 2004. A spatial and temporal model of root cohesion in forest soils. Canadian Journal of Forest Research, 34: 950-958. 121 Schnackenberg, E.S. and L.H. MacDonald. 1998. Detecting cumulative effects on headwater streams in the Routt National Forest, Colorado. Journal of the American Water Resources Association, 34(5): 1163-1177. Schwab, J.W., 1998. Landslides on the Queen Charlotte Islands: Processes, Rates, and Climatic Events. In: D.L. Hogan, J.J. Tschaplinski and S. Chatwin (Editors), Carnation Creek and Queen Charlotte Islands Fish/Forestry Workshop: Applying 20 Years of Coastal Research to Management Solutions. BC Ministry of Forests. Scott, M.L., J.M. Friedman and G.T. Auble. 1996. Fluvial processes and the establishment of bottomland species. Geomorphology, 14: 327-339. Septer, D. and J.W. Schwab, 1995. Rainstorms and flood damage: Northwestern British Columbia, 1891-1991. Land Management Handbook, No. 31, BC Ministry of Forests. Sidle, R. 1991. A conceptual-model of changes in root cohesion in response to vegetation management. Journal of Environmental Quality, 20(1): 43-52. Sidle, R. 1992. A theoretical-model of the effect of timber harvesting on slope stability. Water Resources Research, 28(7): 1897-1910. Sidle, R.C., A.J. Pearce and C.L. O'Loughlin. 1985. Hillsope Stability and Land Use. Water Resources Monograph Series 11. American Geophysical Union, 140 PP-Slaymaker, O., and H.J. McPherson. 1977. An overview of geomorphic processes in the Canadian Cordillera. Zietschrift fur Geomorphologie, 21:169-186. Slaymaker, O. 2000. Assessment of the geomorphic impacts of forestry in British Columbia. Ambio, 29: 381-387. Sullivan, K., T.E. Lisle, C.A. Dollof, G.E. Grant and L.M. Reid, 1987. Stream channels: The link between forests and fishes. In: E.O. Salo and T.W. Cundy (Editors), Streamside Management: Forestry and Fisheries Interactions. University of Washington, Seattle, pp. 39-97. Sutherland, D.G., Ball, M.H., Hilton, S.J., and Lisle, T.F. 2002. Evolution of a landslide-induced sediment wave in the Navarro River, California. Geological Society of American, Bull., 114: 1036-1048. 122 Swanson, F.J. and L.T. Dyrness. 1975. Impact of clear-cutting and road construction on soil erosion by landslides in the western Cascade Range, Oregon. Geology, 3: 393-396. Trainor, K. and M. Church. 2003. Quantifying variability in stream channel morphology. Water Resources Research, 39(9); Article 1248. Wainwright, J., A.C. Cases, J. Puigdefabregas and K. Michaelides. 2002. Linkages of hillslope erosion to sediment transport and storage in river and floodplain systems (Editorial). Earth Surface Processes and Landforms, 27: 1363-1364. Wemple, B.C., F.J. Swanson and J.A. Jones. 2001. Forest Roads and Geomorphic Process Interactions, Cascade Range, Oregon. Earth Surface Processes and Landforms, 26: 191-204. Wood-Smith, R.D. and J.M. Buffington. 1996. Multi-variate geomorphic analysis of forest streams: implications for assessment of land use impacts on channel condition. Earth Surface Processes and Landforms, 21: 377-393. 123 Appendix 1 Summary of reach characteristics. Creek Reach Area (km2) Length (m) Wo (m) Morphology mpanan Logged Year Coupled* Cascade 1 19.5 962 31.5 mt 1948 N Cascade 2 18.6 395 10.6 cf 1948 N Cascade 3 18.0 497 14.4 al 1948 N Cascade 4 17.1 619 9.3 cf NA Y Cascade 5 10.6 1129 23.0 mt 1952 Y Cascade 6 9.4 734 17.1 al 1952 Y Cascade 7 6.7 722 15.9 al 1952 Y Cascade 8 4.5 996 12.3 al 1952 Y Cascade 9 2.7 1180 9.2 cf 1952 Y Cascade 10 1.2 1039 9.9 cf 1952 Y Cascade 11 1.4 559 11.5 fn 1952 Y Cascade 12 0.4 944 9.0 cf 1968 Y Cascade 13 4.5 728 12.6 cf 1952 Y Cascade 14 2.9 893 11.2 al 1968 Y Cascade 15 1.2 1141 10.5 cf NA Y Cascade 16 0.5 704 8.3 cf NA Y 124 Creek Reach Area (km2) Length (m) Cedarflat 1 21.2 1102 Cedarflat 2 19.9 941 Cedarflat 3 18.3 875 Cedarflat 4 16.4 657 Cedarflat 5 15.7 700 Cedarflat 6 14.5 . 768 Cedarflat 7 13.6 547 Cedarflat 8 12.1 532 Cedarflat 9 11.2 513 Cedarflat 10 7.4 980 Riparian Morphology Logged Coupled* Year 10.4 al NA N 12.2 al NA N 10.3 al 1983 N 9.9 cf NA Y 11.5 al 1983 Y 15.2 al 1983 Y 11.1 cf 1983 N 11.2 cf 1983 N 9.3 al 1983 N 9.3 cf 1983 Y 125 Creek Reach Area (km2) Length (m) Wo (m) Morphology t\i/jai laii Logged Year Coupled* Chapman 1 66.8 913 52.6 al 1946 N Chapman 2 66.3 527 9.8 cf 1946 N Chapman 3 64.8 1068 36.0 cf 1946 N Chapman 4 64.2 879 18.4 cf NA N Chapman 5 63.6 625 10.2 cf NA N Chapman . 6 62.9 943 12.5 cf 1946 N Chapman 7 61.4 1333 12.2 cf 1946 N Chapman 8 60.5 1036 13.7 al NA N Chapman 9 59.2 615 15.8 cf 1946 N Chapman 10 58.1 830 14.2 al NA Y Chapman 11 56.9 634 14.4 cf NA Y Chapman 12 54.4 721 22.2 cf 1946 Y Chapman 13 52.9 886 24.9 al 1946 N Chapman 14 51.7 646 28.3 al 1946 Y Chapman 15 49.1 707 12.7 cf 1946 N Chapman 16 45.1 1509 28.0 al 1946 Y Chapman 17 43.0 509 19.9 cf 1946 Y Chapman 18 42.0 628 10.7 cf 1967 Y Chapman 19 40.7 571 11.2 cf 1967 Y Chapman 20 36.8 734 12.1 cf 1967 Y Chapman 21 27.5 957 11.3 al 1967 Y Chapman 22 26.3 741 11.4 cf 1967 Y Chapman 23 22.8 629 12.0 cf NA N 126 Creek Reach Area (km2) Length (m) Wo (m) Morphology Kipanan Logged Year Coupled* Dewdney 1 76.6 453 11.1 al NA N Dewdney 2 76.4 143 10.7 cf NA N Dewdney 3 75.4 1027 18.2 mt NA N Dewdney 4 73.1 1542 18.2 mt NA Y Dewdney 5 70.0 1350 20.2 mt NA N Dewdney 6 68.4 771 17.3 mt 1983 N Dewdney 7 45.1 452 16.6 mt 1983 N Dewdney 8 35.1 1252 11.2 al 1983 Y Dewdney 9 33.9 682 10.9 al 1983 Y Dewdney 10 30.6 778 12.9 al 1983 Y Dewdney 11 29.6 488 12.6 al 1983 Y Dewdney 12 27.3 668 12.2 al NA Y Dewdney 13 24.0 1036 12.9 a l 1983 Y Dewdney 14 19.8 1020 14.7 al 1983 Y Dewdney 15 17.4 595 11.1 cf 1983 Y Dewdney 16 11.1 757 11.7 cf 1983 Y Dewdney 17 9.8 477 9.5 cf 1983 Y Dewdney 18 8.8 567 12.3 cf 1983 Y Dewdney 19 3.8 1371 8.8 cf NA Y 127 Creek Reach Area (km2) Length (m) Wo (m) Morphology Riparian Logged Year Coupled* Norrish 1 110.8 1722 131.4 mt NA N Norrish 2 104.3 1013 24.5 cf NA Y Norrish 3 102.7 748 22.1 cf NA Y Norrish 4 102.0 410 18.9 cf NA Y Norrish 5 99.3 1042 20.2 cf NA Y Norrish 6 92.9 1175 15.9 cf 1968 Y Norrish 7 89.2 696 31.7 al 1954 Y Norrish 8 75.2 1175 21.2 al 1954 Y Norrish 9 72.0 1068 22.4 cf 1954 Y Norrish 10 66.6 1038 35.4 al 1954 Y Norrish 11 65.2 635 31.0 al 1954 N Norrish 12 57.7 676 20.5 al NA Y Norrish 13 56.7 1052 31.4 mt 1954 N Norrish 14 54.0 691 88.2 mt 1954 N Norrish 15 48.7 717 66.4 mt 1954 Y Norrish 16 45.9 820 58.3 mt 1954 N Norrish 17 45.0 764 29.1 mt 1954 N Norrish 18 32.1 935 22.7 mt 1954 N Norrish 19 13.1 544 18.5 al NA N Norrish 20 15.8 1156 19.5 al NA N Norrish 21 13.0 1485 16.2 al 1954 N 128 Creek Reach  A r e a L e n g t h creek Reach ( k n j 2 ) ( m ) Slesse 1 161.8 423 Slesse 2 157.0 770 Slesse 3 156.3 364 Slesse 4 155.3 731 Slesse 5 152.8 1010 Slesse 6 149.7 1202 Slesse 7 144.4 1366 Slesse 8 132.6 1263 Slesse 9 124.6 1050 Slesse 10 123.2 863 Slesse 11 116.3 862 Slesse 12 108.8 986 Slesse 13 107.6 926 Slesse 14 103.2 1041 Riparian Morphology Logged Coupled* Year 23.2 al 1940 N 23.7 al 1940 N 14.2 cf 1940 N 29.7 al . 1940 Y 31.9 br 1950 Y 26.6 mt 1950 Y 27.7 br 1950 Y 24.6 al 1966 Y 17.3 cf 1990 Y 15.3 mt 1990 N 13.6 cf 1966 Y 14.3 cf 1966 Y 14.1 cf 1966 N 19.2 mt 1966 Y 129 Creek Reach Area (km2) Length (m) Wo (m) Morphology Logged Year Coupled* Theodosia 1 73.0 273 40.3 mt NA N Theodosia 2 71.2 1163 37.5 mt 1947 N Theodosia 3 69.3 782 23.5 mt 1947 N Theodosia 4 66.5 900 18.9 al 1978 N Theodosia 5 61.5 1079 17.0 al 1965 Y Theodosia 6 40.9 1397 11.3 al 1978 Y Theodosia 7 39.9 701 11.1 al 1978 Y Theodosia 8 39.4 540 11.6 al 1978 N Theodosia 9 24.6 631 22.9 mt 1960 Y Theodosia 10 21.4 1151 13.0 al 1960 Y Theodosia 11 19.0 605 14.6 al 1960 Y Theodosia 12 17.5 595 16.3 al 1960 Y Theodosia 13 15.1 763 20.6 al 1965 Y Theodosia 14 5.0 946 18.2 al NA Y Theodosia 15 4.2 434 26.6 cf 1978 Y Theodosia 16 3.0 653 13.7 cf 1978 Y 130 Appendix 2. Response Curves for study reaches. Cedarflat Reach 1 1970 19 Year Cedarflat Reach 2 1950 1960 1970 1980 1990 2000 Year Cedarflat Reach 3 _ 15 - Channel Writh 1950 1970 1980 Year Cedarflat Reach 4 _ 15 E § 10 - Channel Width -Landslkles 1950 1960 1970 1980 Year 1990 2000 Cedarflat Reach 5 20 I 1 5 s S 10 to c c m JZ <-> 5 Cedarflat Reach 6 13/ Cedarflat Reach 7 _ 15 -1970 1980 Year Cedarflat Reach 8 S 10-- Channel Width -Landslides 1960 1970 1980 Year Cedarflat Reach 9 - Channel Width — Landslides 1970 1980 Year Cedarflat Reach 10 -Channel Width — Landslkles 1970 1980 Year Dewdney Reach 1 - Channel Width —Landstides 1970 1980 . Year Dewdney Reach 2 - Channel Width —Landslkles 1970 1980 Year 132 Dewdney Reach 3 - C h a n n e l Width -Lands l ides 1970 1980 Year Dewdney Reach 4 1970 1980 Year 1990 2000 Dewdney Reach 5 1970 1980 Year Dewdney Reach 6 -Channe l Width | E, TJ 3 20 1970 1980 Year Dewdney Reach 7 - C h a n n e l Width -Landsl ides 1970 19 Year Dewdney Reach 8 - C h a n n e l Width -Lands l ides 1950 1960 1970 1980 Year 2000 133 Dewdney Reach 9 20 -- C h a n n e l Width — Landslides 1950 1960 1970 1980 Year 1990 2000 Dewdney Reach 10 - Channel Width -Landsl ides 1970 1980 Year Dewdney Reach 11 10 H - Channel Width -Landsl ides 1970 1980 Year Dewdney Reach 12 _ 15 E - C h a n n e l Width — Landslides 1500 1950 1960 1970 1980 1990 2000 Year Dewdney Reach 13 35 30 25 H 20 15 10 5 - C h a n n e l Width -Landsl ides 1970 1980 Year Dewdney Reach 14 20 | i 10 1970 1980 Year 134 Dewdney Reach 15 20 _ 15 E _ 15 E Dewdney Reach 17 1970 1980 Year Dewdney Reach 18 1970 19 Year 15 • / • 2000 • 1800 — / : 1600 idth (m / : • 1400 - 1200 s / • 1000 Channel / • 800 - 600 Channel Widthl / - 400 Landslides / • 200 o - i i i ' i • 0 Cascade Reach 2 - Channel Wdth 1970 1980 Year 135 Cascade Reach 3 1970 1980 Year Cascade Reach 4 5 1 0 -- Channel Wkdth — Landslides 1970 1980 Year Cascade Reach 5 40 -35 • 30 -25 20 -15 10 -5 -- Channel Width -Landslktes 70000 60000 50000 40000 30000 20000 10000 1950 1960 1970 1980 Year 1990 1950 Cascade Reach 6 1960 1970 1980 Year 1990 2000 Cascade Reach 7 - 80000 70000 - 60000 - 50000 40000 - 30000 - 20000 - 10000 - 0 1970 1980 Year Cascade Reach 8 250000 1960 1970 1980 19 Year 136 Cascade Reach 9 3 15 H I = 10 -1970 1980 Year Cascade Reach 10 ? 1 5 -1970 1980 Year Cascade Reach 11 1970 1980 Year 14000 12000 + 10000 8000 6000 4000 2000 20000 18000 16000 14000 12000 10000 8000 6000 4000 2000 Cascade Reach 12 ? 1 5 - Channel Width -Landslkjes 70000 60000 50000 40000 30000 20000 10000 1970 1980 Year 2000 Cascade Reach 13 5 1 5 i 01 i 10 n o - Channel Width -Landslides 1950 1960 1970 1980 19 Year Cascade Reach 14 1950 1960 1970 1980 Year 137 Cascade Reach 15 7000 -f 6000 5000 4000 3000 2000 1000 1970 1980 Year Cascade Reach 16 E 10 - Channel Width -Lendslkles 25000 20000 15000 10000 5000 1990 2000 Slesse Reach 1 45.0 40.0 H . 35.0 30.0 25.0 20.0 15.0 10.0 -Channel Width -Landslkies 1 0.9 0.8 0.7 0.6 0.5 0.4 - 0.3 - 0.2 0.1 1950 1970 1980 Year 1990 2000 Slesse Reach 2 1960 1970 1980 Year 1990 Slesse Reach 2 100.0 90.0 H 80.0 70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0 - Channel Wklth -Land: Slesse Reach 3 15.0 • 1 O.S + 0.£ • 0. - 0. + 0. - 0 - 0. • 0 • 0. 7 _ 6 ^ (A 0) 5 2 vt 4 1 CO 3 J .2 1 1950 1960 1970 1980 1990 Year 2000 138 Slesse Reach 4 90. 80 70 .§. 60 | 50 1 40 c J 30 O 20 10 0 - Channel Wkjth •Landslkles 1950 Slesse Reach 5 1970 1980 Year Slesse Reach 6 6000 1990 2000 1500 4 45000 40000 35000 30000 ~ ~ 25000 S TJ 20000 « TJ C 15000 ™ 10000 + 5000 Slesse Reach 7 150.0 125.0 £ 100.0 TJ S 75.0 o> c c 50.0 o 25.0 0.0 - Channel Wklth •Landslkjes 25000 15000 — 1950 1960 1970 1980 1990 2000 Year Slesse Reach 8 70.0 60.0 • 50.0 40.0 30.0 20.0 10.0 0.0 -Channel Writh •Landslides 1970 1980 19 Year Slesse Reach 9 40.0 1950 1960 1970 1980 Year 35000 30000 25000 _ \ 20000 i» v TJ 15000 -3 c m 10000 5000 • 0 2000 1600 1400 1200 1000 "i. (A d> 800 2 "» 600 1 _ i 400 200 - 0 2000 139 Slesse Reach 10 20.0 ? £ 15.0 •D 3 0) £ 10.0 -Channel Widih| "Landslides ! 1000 900 800 700 600 500 400 300 200 100 1950 1960 1970 1980 1990 2000 Year Slesse Reach 11 1950 1960 1970 1980 Year 1990 2000 Slesse Reach 12 30.0 25.0 .§. 20.0 S 15.0 o> c 1 10.0 0.0 -Channel Width "Landslides-5000 3000 — 2000 1950 1960 1970 1980 Year Slesse Reach 13 20.0 £ 15.0 •o s 01 i io.o 0.0 - Channel Width "Landslkies 5000 3000 — 1950 1960 1970 19 Year 1990 2000 Slesse Reach 14 20.0 £ 15.0 | £ 10.0 -Channel Wklth — Landslkies 30000 25000 20000 ~ ~ to 15000 | <A TJ C 10000 ™ 5000 1950 1960 1970 19 Year 1990 2000 140 Theodosia Reach 1 80 • 70 • .c 1 50 H s g 40-c rs 5 30-20 H 10 -- Channel Wkith 1970 Year Theodosia Reach 2 70 -60 -- 5 0 -3 40 -o 1970 Year Theodosia Reach 3 50 -3 30 -" 20-- Channel Wklth 1970 Year Theodosia Reach 4 30 i 1970 Year Theodosia Reach 5 3 30 " 20 1940 1950 1960 1970 1980 1990 2000 Year Theodosia Reach 6 - 1 5 141 Theodosia Reach 7 5 20 1970 1980 1990 2000 Year Theodosia Reach 8 20 — 15 E S 10 - Channel Width — Landslkles 1940 1950 1960 1970 1980 1990 2000 Year Theodosia Reach 9 20 1940 1950 1960 1970 Year 1980 1990 2000 Theodosia Reach 10 - Channel Wklth — Landslides 20000 1940 1950 1960 1970 1980 19 Year Theodosia Reach 11 1940 1950 19 1970 1980 1990 2000 Year Theodosia Reach 12 S 15 1940 1950 1970 1980 1990 2000 Year 142 Theodosia Reach 13 40 35 I 30 S 25 v c I 20 o 15 10 18000 16000 14000 12000 10000 8000 6000 4000 2000 1940 1950 1970 Year 1980 1990 2000 Theodosia Reach 14 Theodosia Reach 16 - Channel Wdth — Landslkies 1940 1950 1960 1970 19 Year 1990 2000 Norrish Reach 1 143 Norrish Reach 3 25 ? .c 5 20 d> c c (1) 5 15 1950 1960 1970 1980 1990 2000 2010 Year Norrish Reach 4 S. 20 s - Channel Wklth -LandslWes 1950 1960 3000 2500 2000 cT-1500 | f 1000 5 500 1980 1990 2000 2010 Year Norrish Reach 5 2500 2000 *T in 1500 | w •a c 1000 ™ 1950 1960 1970 1980 1990 2000 2010 Year Norrish Reach 6 - Channel Width •Landslides 90000 80000 70000 60000 50000 8 •o 40000 5 •a 30000 n _i 20000 10000 2000 2010 Norrish Reach 7 1950 1960 1970 1980 1990 2000 2010 Year Norrish Reach 8 1990 2000 144 Norrish Reach 9 Norrish Reach 10 70000 60000 50000 _ ™E 40000 *Z o> T l 30000 1 c n 20000 - 1 10000 1950 1960 1980 1990 2000 2010 Year Norrish Reach 11 1980 Year Norrish Reach 12 Norrish Reach 13 - Channel Width • Landslides 50000 40000 tn 30000 | tn TJ 20000 J 2000 2010 Norrish Reach 14 - Channel Writh "Landslides 60000 50000 40000 ' 30000 20000 10000 1950 1960 1970 1980 1990 2000 2010 Year 145 Norrish Reach 15 100 90 80 (m) 70 T3 60 S •3 50 c c .c 40 0 30 20 10 30000 25000 •f 20000 «~ in 15000 J "O 10000 J3 5000 Norrish Reach 16 100 90 80 E - 70 I 60 ai 50 c j2 40 o 30 20 -Channel Width -Landslides 4000 3000 1950 1960 1980 Year 1990 2000 2010 Norrish Reach 17 60 50 ? I 40 I o = 30 20 -Channel Width "Landslides 1950 1960 . 1970 1980 1990 2000 2010 Year Norrish Reach 18 40 — 30 10 - Channel Width -Landslkles 5000 1950 1960 1970 1980 1990 2000 2010 Year Norrish Reach 19 20 g 1 1 0 o -Channel Wkllh -Landslkles 5000 4000 3000 1950 1960 1970 1980 1990 2000 2010 Year Norrish Reach 20 10 1960 1970 1980 Year 2000 2010 146 £ •a 5 10-Chapman Reach 2 - Channel Width "Landslides 1970 Year 20 Chapman Reach 5 - Channel Width "Landslides 1940 1950 1960 1970 Year 1980 1990 2000 147 Chapman Reach 6 5 10 - Channel Wkdth "Landslktes 5000 4000 3000 1000 1940 1950 1960 1970 1980 1990 2000 Year Chapman Reach 7 _ 15 £ i io-| - Channel Width • Landslides 2000 1940 1950 1960 1970 • 1980 1990 2000 Year Chapman Reach 8 20 ? | 15 | CD I 10 -Channel Wkdth "Landslkle3 4000 1950 1960 1970 1980 1990 2000 Year Chapman Reach 9 S 10 -Channel Width "Landslkies 1950 1960 1970 1980 19 Year • 0 2000 Chapman Reach 10 - Channel Wkfth "Landslides Chapman Reach 11 - Channel Width "Landslides 2000 1950 1960 1970 1980 1990 2000 Year 148 Chapman Reach 12 20 ? £ 15 •D 1 10 Chapman Reach 13 30 £ . 20 - Channel Width "Landslkdes 1950 1960 1970 1980 Year Chapman Reach 14 30 i . 20 JZ TJ 3 10 1000 750 ID CO 500 2 1950 1960 1970 1980 1990 2000 Year 1990 2000 7500 5000 . 2500 1950 1960 1970 1980 1990 2000 Year Chapman Reach 15 20 15 TJ 5 10 • 7500 - Channel Width -LandslkJes 0 -1950 1960 1970 1980 Year Chapman Reach 16 40 35 30 \ ? s  25 jo 5 20 0) I 15 10 H 5 1970 1980 Year Chapman Reach 17 20 ? I 15 g 10 5000 - Channel Width •Landslides 149 30 25 20 1 15 c I 10 20 ? £ 15 5 <1) 5 10 20 S 10 Chapman Reach 18 1970 1980 Year Chapman Reach 19 1950 1960 1970 1980 19 Year Chapman Reach 20 15000 1950 1960 1970 1980 1990 2000 Year 40 35 _ 30 E £ 2 5 T3 3 20 at I 15 .c o 10 5 0 -1950 20 ? £ 15 •o S QJ c 10 0 -1950 £. 10 . C « C c x: 5 O Chapman Reach 21 Chapman Reach 22 1970 1980 Year Chapman Reach 23 - Channel Width "Landslides o -1950 9000 8000 7000 6000 "jE 5000 (A to 4000 « c 3000 2000 .1000 1970 1980 Year 150 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0091815/manifest

Comment

Related Items