Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Spatial impact trends on debris flow fans in southwestern British Columbia Zubrycky, Sophia 2020

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

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2020_november_zubrycky_sophia.pdf [ 59.8MB ]
Metadata
JSON: 24-1.0391879.json
JSON-LD: 24-1.0391879-ld.json
RDF/XML (Pretty): 24-1.0391879-rdf.xml
RDF/JSON: 24-1.0391879-rdf.json
Turtle: 24-1.0391879-turtle.txt
N-Triples: 24-1.0391879-rdf-ntriples.txt
Original Record: 24-1.0391879-source.json
Full Text
24-1.0391879-fulltext.txt
Citation
24-1.0391879.ris

Full Text

Spatial Impact Trends on Debris Flow Fans inSouthwestern British ColumbiabySophia ZubryckyB.A.Sc. Geological Engineering, Queen’s University, 2013A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFMaster of Applied ScienceinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Geological Engineering)The University of British Columbia(Vancouver)June 2020© Sophia Zubrycky, 2020The following individuals certify that they have read, and recommend to the Faculty of Graduate andPostdoctoral Studies for acceptance, the thesis entitled:Spatial Impact Trends on Debris Flow Fans in Southwestern British Columbiasubmitted by Sophia Zubrycky in partial fulfillment of the requirements for the degree of Masterof Applied Science in Geological Engineering.Examining Committee:Dr. Scott McDougall, Assistant Professor, Geological Engineering, UBCSupervisorDr. Erik Eberhardt, Professor, Geological Engineering, UBCSupervisory Committee MemberDr. Brett Eaton, Professor, Geography, UBCSupervisory Committee MemberDr. Matthias Jakob, Geoscientist, BGC Engineering Inc.Supervisory Committee MemberDr. Mitch D’Arcy, Assistant Professor, Department of Earth, Ocean and Atmospheric Sciences, UBCAdditional ExamineriiAbstractForecasting the spatial impact of debris flows is challenging due to complex runout behaviour,such as variable mobility and channel avulsions. Practitioners often base the probability of runoutexceedance on a fan, or define avulsion scenarios, on judgement. To support decision making,spatial impact trends were studied at thirty active debris flow fans in southwestern British Columbia(SWBC), Canada. 176 debris flow impact areas covering an average observation period of 74 yearswere mapped using orthorectified historical airphotos, satellite imagery, topographic basemaps, lidar,and field observations. A graphical plotting method was developed that converts geospatial mappingto spatial impact heatmaps normalized by the fan boundary, allowing for comparison of runouttrends across fans in the dataset. Probability of spatial impact was analyzed in two components:runout down-fan (i.e., how far debris flows tend to travel past the apex toward the fan toe) and runoutcross-fan (i.e., how far debris flows tend to deviate from the previous flow path). For fans in SWBC,there is a characteristic decay in spatial impact probability from the fan apex and the previous flowpath, represented by a normal and log-normal distribution for normalized runout in the down-fan andcross-fan components, respectively. Differences in spatial impact trends can be explained, in part,by event volume, Melton ratio, fan truncation, and fan activity, however not by fan morphometrics,such as the slope or the point at which channelization is lost. A tool was created that transposes theempirical runout distributions onto a fan to assist in risk-based decision making. Future work mayinvolve fitting functions to the spatial impact data for a more robust and adaptable forecasting tool.iiiLay SummaryDebris flows are extremely rapid landslides comprised of debris and water that travel downsteep mountain creeks. Estimating the chance of being impacted by a debris flow is important tounderstanding the risk to the public and infrastructure. This work is challenging because debris flowscan travel long distances or suddenly change directions. To help with our understanding of likelyfuture debris flow impacts, a historical record of debris flow impacts dating back to 1922 was mappedat 30 sites in southwestern British Columbia. By looking at these data in new ways, we can identifyareas most susceptible to impacts, and what factors allow prediction of debris flow travel distanceand flow paths. Debris flow volume and sediment mixture are key variables in explaining those twocharacteristics.ivPrefaceSome of the data presented in Figure 2.4 and Table 3.6 in Chapter 2 were published in conferencepaper Zubrycky, S., Mitchell, A., Aaron, J., and McDougall, S. (2019) Preliminary calibration of anumerical runout model for debris flows in southwestern British Columbia, in Debris-Flow HazardsMitigation: Mechanics, Prediction, and Assessment (pp. 911–918), Golden, CO. Numerical modelcalibration was performed by A. Mitchell and I using methods and code developed by J. Aaron.The database in Chapter 3 Section 3.1 was presented as part of a conference poster/extendedabstract Zubrycky, S., Bonneau, D., McDougall, S., Jakob, M., and Hutchinson, J. (2018) Empiricalprediction of debris flow avulsion and runout exceedance probability using a debris flow inventoryfrom Southwestern British Columbia, poster session presented at the meeting of the 7th CanadianGeohazards Conference, Canmore, Alberta. The preliminary debris flow event database was compiledby D. Bonneau, which I subsequently edited and expanded.A version of the material in Chapter 4 has been accepted to the XIII International Symposiumon Landslides, with co-authors A. Mitchell and S. McDougall. I am the lead author of this work,responsible for all areas of major concept formation, data collection, analysis, and manuscriptcomposition. A. Mitchell operated the lidar drone, processed the lidar data, helped with field work,and provided review. S. McDougall was the supervisory author, involved throughout the project inconcept formation and manuscript edits.Many of the airphotos in Table 3.2 were scanned with help from S. Ghadirianniari and K. Matson.Figure 2.2 appears in this thesis with permission from John Wiley and Sons, and Figure 4.17with permission from Elsevier. Debris flow lobe mapping presented in Figure 4.13 was adapted fromde Haas et al. (2018a) with permission from the author.vTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiLay Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiList of Supplementary Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviiList of Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviiiAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.3 Research Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.4 Research Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.5 Study Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.5.1 Geologic Setting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.5.2 Climate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.5.3 Previous Research in the Study Area . . . . . . . . . . . . . . . . . . . . . 92 A Comprehensive Review of Debris Flow Runout . . . . . . . . . . . . . . . . . . . . 112.1 Debris Flow Processes and Forms . . . . . . . . . . . . . . . . . . . . . . . . . . 112.2 Debris Flow Fan Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.3 Debris Flow Hazard and Risk Assessment . . . . . . . . . . . . . . . . . . . . . . 162.4 Factors Affecting Debris Flow Mobility . . . . . . . . . . . . . . . . . . . . . . . 202.4.1 Event Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202.4.2 Composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212.4.3 Path Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22vi2.5 Understanding Debris Flow Avulsion . . . . . . . . . . . . . . . . . . . . . . . . . 232.5.1 Avulsion Triggers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.5.2 Estimating Avulsion Probability . . . . . . . . . . . . . . . . . . . . . . . 262.5.3 Locating Avulsion Points . . . . . . . . . . . . . . . . . . . . . . . . . . . 272.6 Empirical-Statistical Methods for Forecasting Debris Flow Runout . . . . . . . . . 292.6.1 Travel Path . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302.6.2 Deposit Dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322.6.3 Volume-Area Relationships . . . . . . . . . . . . . . . . . . . . . . . . . 332.6.4 Volume Balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352.6.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352.7 Challenges and Knowledge Gaps . . . . . . . . . . . . . . . . . . . . . . . . . . . 382.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393 Creating a Geospatial Dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413.1 Fan Site Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413.2 Geomorphic Fan Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463.2.1 Fan Boundary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473.2.2 Fan Apex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483.2.3 Impact Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483.2.4 Flow Path . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483.3 Data Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493.3.1 Airphotos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493.3.2 Satellite Imagery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 543.3.3 Lidar and Orthophotos . . . . . . . . . . . . . . . . . . . . . . . . . . . . 573.3.4 TRIM DEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 593.3.5 Field Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 603.4 Estimating Event Volumes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 643.4.1 Lidar Change Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . 643.4.2 Features in Post-Event Lidar . . . . . . . . . . . . . . . . . . . . . . . . . 663.4.3 Field Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 673.4.4 Volume-Area Relationship . . . . . . . . . . . . . . . . . . . . . . . . . . 673.5 Avulsion Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 723.6 Fan Site Descriptors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 743.6.1 Melton Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 753.6.2 Watershed Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 753.6.3 Fan Slope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 753.6.4 Average Fan Channel Slope . . . . . . . . . . . . . . . . . . . . . . . . . 773.6.5 Fan Elevation Relief Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . 773.6.6 Fan Intersection Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . 783.6.7 Hydrogeomorphic Process Recognition . . . . . . . . . . . . . . . . . . . 803.6.8 Source Geology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 813.6.9 Fan Truncation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 813.6.10 Fan Activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 823.6.11 Summary of Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . 823.7 Mapping Certainty Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . 83vii3.8 Dataset Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 853.9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 874 Extraction and Analysis of Spatial Impact Trends on Debris Flow Fans . . . . . . . 894.1 Creating Fan-Normalized Spatial Impact Heatmaps . . . . . . . . . . . . . . . . . 894.1.1 Relative to the Fan Axis . . . . . . . . . . . . . . . . . . . . . . . . . . . 904.1.2 Relative to the Previous Flow Path . . . . . . . . . . . . . . . . . . . . . . 904.1.3 Code Workflow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 914.1.4 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 944.2 Regional Spatial Impact Trends on Fans in Southwestern British Columbia . . . . . 954.2.1 Spatial Impact Heatmaps . . . . . . . . . . . . . . . . . . . . . . . . . . . 954.2.2 Maximum Runout Distributions . . . . . . . . . . . . . . . . . . . . . . . 984.2.3 Comparison to an External Case Study: Kamikamihori Fan, Japan . . . . . 1034.2.4 Comparison to Conceptual Avulsion Scenarios . . . . . . . . . . . . . . . 1074.3 Local Spatial Impact Trends at Two Locations in Southwestern British Columbia . 1104.3.1 Mount Currie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1114.3.2 Fountain Ridge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1164.3.3 Comparison and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 1214.4 Factors Affecting Spatial Impact Trends on Fans in SWBC . . . . . . . . . . . . . 1254.4.1 Statistical Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1254.4.2 Event Volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1304.4.3 Melton Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1304.4.4 Watershed Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1314.4.5 Fan and Channel Slope . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1324.4.6 Fan Elevation Relief Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . 1324.4.7 Normalized Fan Intersection Point . . . . . . . . . . . . . . . . . . . . . . 1334.4.8 Source Geology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1344.4.9 Fan Truncation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1354.4.10 Fan Activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1364.4.11 Summary and Implications . . . . . . . . . . . . . . . . . . . . . . . . . . 1364.5 Fan-Normalized Empirical Runout Estimator Tool . . . . . . . . . . . . . . . . . . 1374.5.1 Code Workflow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1384.5.2 Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1394.5.3 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1414.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1415 Conclusions and Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . 1445.1 Summary of Main Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1445.2 Implications for Hazard and Risk Assessments . . . . . . . . . . . . . . . . . . . . 1475.3 Recommendations for Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . 1485.4 Closure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153Appendix A Fan Site Summaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167viiiAppendix B Supplementary Material: GIS Data . . . . . . . . . . . . . . . . . . . . . 198Appendix C Supplementary Material: MATLAB code . . . . . . . . . . . . . . . . . . 199ixList of TablesTable 3.1 List of fan sites in this study. . . . . . . . . . . . . . . . . . . . . . . . . . . . 45Table 3.2 Summary of airphoto and Metashape airphoto orthomosaic coverage at each fansite. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53Table 3.3 Summary of external ALS lidar data sources. . . . . . . . . . . . . . . . . . . . 58Table 3.4 Summary of RPAS lidar data collected. . . . . . . . . . . . . . . . . . . . . . . 58Table 3.5 Summary of field work completed. . . . . . . . . . . . . . . . . . . . . . . . . 62Table 3.6 Event volumes and areas. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69Table 3.7 Description of impact area mapping certainty classes. . . . . . . . . . . . . . . 84Table 4.1 Summary of morphometric variables at Mount Currie and Fountain Ridge fans. . 122xList of FiguresFigure 1.1 Regional geology and physiographic regions of the study area. Provincial digitalbedrock geology accessed from Cui et al. (2017), physiographic regions forCanada accessed from Bostock (2014), and Quaternary volcanic fields from theGaribaldi volcanic belt assembled from Wilson (2019). . . . . . . . . . . . . . 6Figure 1.2 Monthly climate normals (1971-2000) for climate stations across the study area.Locations shown on map in Figure 1.3. Station data accessed from PCIC dataportal. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8Figure 1.3 Gridded precipitation data for November (left) and July (right) climate normals(1971-2000) from PCIC (2014). High-resolution climatology grid with a pixelsize of about 800 m is derived from station data interpolated using the Parameter-elevation Regressions on Independent Slopes Model (PRISM). Monthly climatestation data shown in Figure 1.2. Ecoregions defined by Demarchi (2011). . . . 9Figure 2.1 Schematic diagram of typical debris flow processes, forms, and impacts (not toscale). Adapted from Pierson (1986) and Lau (2017). . . . . . . . . . . . . . . 14Figure 2.2 Schematic of autogenic cycles on experimental debris flow fans and fluvial fansand deltas from de Haas et al. (2016). . . . . . . . . . . . . . . . . . . . . . . 16Figure 2.3 Conceptual debris flow avulsion scenarios and mobility that may be consideredin a QRA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19Figure 2.4 Preliminary correlations by Zubrycky et al. (2019) to calibrated Voellmy frictioncoefficient using data from debris flow events presented in this thesis. Numericalmodelling was completed in Dan3D (McDougall & Hungr, 2004). . . . . . . . 20Figure 2.5 Conceptual avulsion triggers. . . . . . . . . . . . . . . . . . . . . . . . . . . . 25Figure 2.6 Schematic (not to scale) depicting empirical approaches for forecasting debrisflow runout. a) Travel path, such as total travel path length (LT), path length rel-ative to the fan apex (Lf) (e.g., Rickenmann, 1999), or angle of reach (α) definedby the ratio of fall height (H) to horizontal travel distance (L) (e.g., Corominas,1996); b) deposit dimensions, such as deposition length (Ld) and maximumlateral deposit width (Wd) (e.g., Tang et al., 2012); c) planimetric inundationarea (A) correlated to debris flow volume (V) (e.g., Griswold & Iverson, 2008);and d) volume balance rules quantifying entrainment and deposition along aflow path, by which the total travel distance is defined where the cumulative flowvolume is zero, i.e., the sum of the volume deposited (Vd) is equal to the sum ofthe volume entrained (Ve) (e.g., Fannin & Wise, 2001). . . . . . . . . . . . . . 30xiFigure 2.7 Power law scaling relationships with a 2/3 slope between volume and area fornon-volcanic debris flows, over the domain of the respective volumes for eachdataset. Trendline for lahars provided for reference. . . . . . . . . . . . . . . . 34Figure 2.8 Predictor variables used in empirical debris flow runout relationships from areview of 44 published sources, categorized by methodology. . . . . . . . . . . 36Figure 3.1 Location map of preliminary debris flow event inventory for SWBC. BC regionaldistricts labelled for reference. . . . . . . . . . . . . . . . . . . . . . . . . . . 43Figure 3.2 Location map of fan sites in this study. . . . . . . . . . . . . . . . . . . . . . . 44Figure 3.3 Example of geomorphic mapping at Currie C using a) lidar hillshade to definethe apex and fan boundary; and b) 1996 airphoto orthomosaic to delineate animpact area and flow path (20 m contours derived from lidar). 2017 ALS bareearth hillshade courtesy of SLRD. . . . . . . . . . . . . . . . . . . . . . . . . 47Figure 3.4 1965 orthomosaic covering fan and watershed for Fergusson (red arrow) createdwith Metashape using three airphotos at a scale of 1:31,860 and TRIM DEMwith a 25 m pixel size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52Figure 3.5 Debris flows at Currie C and D fans sometime between July and August 2014.a) RapidEye satellite imagery captured on August 7th, 2014 (Planet, 2019);b) dNDVI calculated between July and August 2014 RapidEye bands; and c)impact areas mapped using dNDVI results, field data, and features in post-eventlidar. 2017 ALS bare earth hillshade courtesy of SLRD. . . . . . . . . . . . . 56Figure 3.6 Debris flow at Currie D sometime between July 3 and 12, 2019. a) Planetscopesatellite imagery captured on July 20th, 2019 (Planet, 2019); b) dNDVI cal-culated between May and July 2019 Planetscope bands; and c) results of lidarchange detection between 2017 and 2019 surfaces (0.3 m limit of detection), andimpact area mapped using field data, orthophotos, and lidar change detectionresults. 2017 ALS bare earth hillshade courtesy of SLRD. . . . . . . . . . . . 56Figure 3.7 November 23, 2017 debris flows at Cheam E and W fans. a) Planetscope satelliteimagery captured on July 5th, 2018 (Planet, 2019); b) dNDVI calculated between2018 and 2017 Planetscope bands; and c) impact areas mapped using field dataand high resolution Google Earth imagery. 2017 ALS bare earth hillshadecourtesy of BC MOTI. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57Figure 3.8 Photos of RPAS data collection at a) Currie D and b) Fountain S. . . . . . . . 58Figure 3.9 Interpretation of deposit sequencing at Fountain S lower fan based on Planet(2019) satellite imagery and topographic features in 2019 RPAS lidar DEM. . . 59Figure 3.10 An example of GPS data from field traverses at three fan sites on the LillooetRiver. 2015 ALS bare earth lidar hillshade courtesy of Brian Menounos fromthe University of Northern British Columbia, and John Clague and GioachinoRoberti from Simon Fraser University. . . . . . . . . . . . . . . . . . . . . . . 62Figure 3.11 Examples of field evidence used to identify debris flow processes and delineateimpact areas on fans. a) Lateral boulder levee; b) incised U-shaped channel; c)overgrown paleochannel; d) terminal lobe; e) log jam; f) bouldery channel plug;g) mudline; h) inverse grading; i) matrix supported deposits; j) boulder studdeddebris lobe; k) megaclast; and l) logs with frayed ends. . . . . . . . . . . . . . 63xiiFigure 3.12 Examples of impact area boundaries such as a) edge of a channel levee; b) recentdeposit abutting an older mossy lobe; and c) distal extent of mudwave impactsdownslope of main deposit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64Figure 3.13 2019 debris flow at Currie D. 1) 2019 RPAS orthophoto and 2017 ALS bareearth hillshade; and 2) M3C2 change detection analysis between 2019 RPASpoint cloud and 2017 ALS point clouds, showing areas of a) scour, b) lobedeposition, but not c) muddy afterflow. Change detection clipped to 2019 impactarea mapped with aerial imagery and field GPS. Representative photos of eacharea are shown in Figure 3.14. 2019 RPAS data processed by Andrew Mitchelland 2017 ALS lidar provided by SLRD. . . . . . . . . . . . . . . . . . . . . . 65Figure 3.14 Post-event photos of the 2019 debris flow at Currie D for different locationsalong the flow path (locations in Figure 3.13). a) Deeply incised channel atthe upper fan, showing recent scour and bank erosion; b) thick, coarse, lobedeposit from main avulsion; and c) muddy afterflow deposits on the floodplaindownstream of the fan toe. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66Figure 3.15 Workflow to approximate lobe thickness and volume with post-event lidar topog-raphy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67Figure 3.16 Volume-area relationships for debris flows in SWBC for deposit area (A) andimpact area (Ai). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70Figure 3.17 Comparison of SWBC volume-area data to empirical relationships for non-volcanic debris flows, over the domain of the respective volumes for each dataset.Trendline for lahars provided for reference. . . . . . . . . . . . . . . . . . . . 71Figure 3.18 Distribution of event volumes for the SWBC dataset (110). . . . . . . . . . . . 72Figure 3.19 Debris flow avulsion classification scheme. . . . . . . . . . . . . . . . . . . . 73Figure 3.20 Distribution of avulsion classes (Figure 3.19) for impact areas in the SWBCdataset. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74Figure 3.21 Box and whisker plot comparing different methods for calculating average fanslope across the fan sites. Elevations derived from TRIM DEM with a 25 m pixelsize. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76Figure 3.22 Comparison of overall fan slope to average slope along the fan channel. . . . . 77Figure 3.23 Intersection determined with lidar DEM at Currie D. 2017 ALS bare earth DEMcourtesy of SLRD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79Figure 3.24 Intersection determined with field observations at Ross. 20 m contours derivedfrom TRIM DEM. Sketches are based on field cross-sections and are not to scale. 80Figure 3.25 Fan sites plotted on typical hydrogeomorphic process recognition charts withboundaries by (left) Wilford et al. (2004) and (right) Bardou (2002) and Bertrandet al. (2013). Fan site labels correspond to Table 3.1. . . . . . . . . . . . . . . 81Figure 3.26 Distributions of morphometric and qualitative data describing the SWBC fan sites. 83Figure 3.27 Distribution of data certainty classes for SWBC impact areas. . . . . . . . . . 85Figure 4.1 Example of the fan-normalized plotting method for one impact area relative tothe fan axis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90Figure 4.2 Example of the fan-normalized plotting method for one impact area relative tothe previous flow path. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91Figure 4.3 Example shapefile inputs for plotting. . . . . . . . . . . . . . . . . . . . . . . 92xiiiFigure 4.4 Initializing measurement grid. a) Normalizing fan dimensions; b) circularmeasurement grid with n radial increments in the x dimension, and m angularincrements in the y dimension; and c) grid nodes stored in an m×n array. . . . 92Figure 4.5 Reshaping and summing impact area arrays. a) Intersection of an impact areawith the measurement grid; b) re-ordering impact area array relative to fan axisand summing; and c) re-ordering impact area array relative to previous flow path(i−1) and summing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93Figure 4.6 Examples of summed and normalized impact area plots for one fan site, relativeto the (left) fan axis and (right) previous flow path. . . . . . . . . . . . . . . . 94Figure 4.7 Regional debris flow spatial impact heatmaps for SWBC based on 176 mappedimpact areas across 30 fans. a) Fan-normalized, arc lengths measured relative tothe fan axis; b) unnormalized, arc lengths measured relative to the fan axis; c)fan-normalized, arc lengths measured relative to the previous flow path; and d)unnormalized, arc lengths measured relative to the previous flow path. . . . . . 96Figure 4.8 Smoothing of the fan-normalized spatial impact surface (heatmap relative to theprevious flow path, non-directional) using filters in Surfer ® (Golden Software,LLC, 2018). a) Raw data; and b) 3 passes of a 5×5 maximum value filter and10 passes of a Gaussian low-pass filter. . . . . . . . . . . . . . . . . . . . . . . 97Figure 4.9 Isolines extracted from regional fan-normalized spatial impact heatmap. a,b)Raw data; and c,d) data smoothed in MATLAB using LOWESS (locally weightedscatterplot smoothing). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98Figure 4.10 Fan-normalized maximum runout distributions for the down-fan (left) and cross-fan (right) components based on 30 fans in SWBC. . . . . . . . . . . . . . . . 101Figure 4.11 Maximum runout distributions for the down-fan (left) and cross-fan (right)components based on 30 fans in SWBC. . . . . . . . . . . . . . . . . . . . . . 102Figure 4.12 Relationship between maximum runout in the down-fan (x) and cross-fan (y)components, with normalized (left) and unnormalized (right) scales. Eigenvec-tors of the covariance matrix (Σ) scaled by the respective eigenvalue are plottedin red. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103Figure 4.13 Depositional history at the Kamikamihori fan, Japan. Impact areas adapted andmapped using figures compiled by de Haas et al. (2018a). . . . . . . . . . . . . 104Figure 4.14 Fan-normalized spatial impact heatmaps a) relative to the fan axis; and b) relativeto the previous flow path, for the Kamikamihori fan based on 17 impact areas. . 104Figure 4.15 Comparison of maximum runout distributions in the down-fan and cross-fancomponents for the Kamikamihori fan (17 impact areas) to the regional SWBCdataset (176 impact areas, 30 fans). . . . . . . . . . . . . . . . . . . . . . . . 106Figure 4.16 Steep, bouldery, clast-supported, deposit front plugging the channel on theproximal fan at Currie D (evidence of localized debris flow impacts not visiblein aerial imagery). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107Figure 4.17 Figure from de Haas et al. (2018a) illustrating conceptual avulsion patterns basedon varying flow volume sequences. . . . . . . . . . . . . . . . . . . . . . . . 109Figure 4.18 Fan-normalized cumulative runout exceedance distributions for conceptual avul-sion scenarios proposed by de Haas et al. (2018a) (Figure 4.17) compared toempirical data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110Figure 4.19 Location map of Mount Currie and Fountain Ridge fans. . . . . . . . . . . . . 111xivFigure 4.20 Overview of Mount Currie with main geomorphic features mapped. 2017 ALSbare earth hillshade courtesy of SLRD. . . . . . . . . . . . . . . . . . . . . . 112Figure 4.21 Mount Currie field photographs. a) Incised channel at the upper fan of Currie C;b) bouldery lobe, mid to lower fan at Currie D; and c) boulder-studded sandydeposit at the lower fan of Currie B, Green River floodplain in the distance. . . 113Figure 4.22 Debris flow impact area mapping at Mount Currie fans. 2017 ALS bare earthhillshade courtesy of SLRD. . . . . . . . . . . . . . . . . . . . . . . . . . . . 114Figure 4.23 Fan-normalized spatial impact area heatmaps relative to the fan axis (left) andthe previous flow path (right) for the Mount Currie fans. . . . . . . . . . . . . 115Figure 4.24 Oblique view of Fountain Ridge in Google Earth. Fans and watersheds areoutlined in white, and talus slopes in blue. Fountain N is supplied by a 1.2 kmlong talus chute. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117Figure 4.25 Fountain Ridge field photographs. a) Channelized reach on the upper FountainN fan; b) near-vertical cemented channel side-wall exposing flow sequences andinverse grading, mid-fan Fountain N; and c) thin terminal lobe, distal FountainN fan. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118Figure 4.26 Comparison of debris flow deposit morphology and textures at the distal FountainRidge fans. Sheet-like deposits more typical at Fountain N, compared to lobate,coarser-grained deposits at Fountain S. 2019 bare earth lidar collected by RPAS(drone). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119Figure 4.27 Debris flow impact area mapping at Fountain Ridge fans. 2019 bare earth lidarcollected by RPAS (drone) overlain on 1997 orthorectified airphoto scene. . . . 120Figure 4.28 Fan-normalized spatial impact area heatmaps relative to the fan axis (left) andthe previous flow path (right) for the Fountain Ridge fans. . . . . . . . . . . . 121Figure 4.29 Time series showing the evolution of maximum debris flow runout in both down-fan and cross-fan components, along with volume estimates where available (seeSection 3.4), at Mount Currie and Fountain Ridge fans. . . . . . . . . . . . . . 124Figure 4.30 Comparison of cumulative runout exceedance distributions at Mount Currie andFountain Ridge fans. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125Figure 4.31 Comparison of cumulative runout exceedance distributions using sub-samples ofimpact areas from the SWBC dataset. Each column of plots corresponds to avariable by which the samples are separated, and each row is a different runoutmetric (down-fan or cross-fan; normalized or unnormalized). Sample groups arepartitioned by variable quartiles (Q1, Q3). KS test p-value between upper andlower quarters are bold if the samples are from different distributions (p-value <0.05). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127Figure 4.32 Comparison of cumulative runout exceedance distributions using sub-samples ofimpact areas from the SWBC dataset. Each column of plots corresponds to avariable by which the samples are separated, and each row is a different runoutmetric (down-fan or cross-fan; normalized or unnormalized). Sample groups arepartitioned by variable quartiles (Q1, Q3). KS test p-value between upper andlower quarters are bold if the samples are from different distributions (p-value <0.05). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128xvFigure 4.33 Comparison of cumulative runout exceedance distributions using sub-samples ofimpact areas from the SWBC dataset. Each column of plots corresponds to avariable by which the samples are partitioned, and each row is a different runoutmetric (down-fan or cross-fan; normalized or unnormalized). KS test p-valueare bold if the samples are from different distributions (p-value < 0.05). . . . . 129Figure 4.34 (Left) Distribution of avulsion nodes along the longitudinal position on the fanand (right) relationship between longitudinal position of the avulsion node onthe fan relative to the fan intersection point. . . . . . . . . . . . . . . . . . . . 134Figure 4.35 Probability density of fan-normalized maximum runouts grouped by fan truncation.135Figure 4.36 Components of the empirical runout estimator tool. a) Input shapefiles; b)measurement grid centered on fan apex, including nodes along flow path; c)empirical data source: filtered and smoothed fan-normalized spatial impactheatmap; and d) empirical data sampled at each measurement grid node andexported as a georeferenced raster, where it can be contoured, clipped, or usedfor calculations in GIS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139Figure 4.37 Probability of runout exceedance heatmaps derived from subsets of the SWBCdataset based on volume class. The Catiline fan is used as an example predictionspace. 2014 ALS bare earth lidar hillshade courtesy of SLRD. . . . . . . . . . 140xviList of Supplementary MaterialGIS data (Appendix B)MATLAB code (Appendix C)Supplementary material can be found on cIRcle (UBC digital repository).Refer to appendices for description of data.xviiList of AcronymsALS Airborne laser scanningBC British ColumbiaBGC BGC Engineering Inc.DEM Digital elevation modelecdf Empirical cumulative distribution functionF-M Frequency-magnitudeGCP Ground control pointGIC Geographic Information CentreGIS Geographic information systemGPS Global positioning systemKS Kolmogorov-SmirnovLidar Light detection and rangingNDVI Normalized difference vegetation indexpdf Probability density functionPCIC Pacific Climate Impacts ConsortiumQRA Qualitative risk assessmentRPAS Remotely piloted aerial systemSfM Structure from MotionSLRD Squamish Lillooet Regional DistrictSWBC Southwestern British ColumbiaTRIM Terrain Resource Information ManagementUBC University of British ColumbiaUTM Universal Transverse MercatorxviiiAcknowledgmentsI am very grateful and humbled to have been mentored by my supervisor, Scott McDougall.Without his wisdom, practical guidance, and unwavering support, this thesis would not have beenpossible. Scott follows in the footsteps (and fills the shoes of) the late Oldrich Hungr, who I wouldalso like to acknowledge as a guiding force throughout my research. Thank you to my thesisand examining committee, Brett Eaton, Erik Eberhardt, and Matthias Jakob, for your stimulatingdiscussions and thoughtful review.To my office mates and colleagues, Jordan Aaron, Afshin Amini, Vincent An, Ally Brown, VivianCristina Dias, Sahar Ghadirianniari, Negar Ghahramani, Anya Leenman, Andrew Mitchell, MasoudRahjoo, Brie Shaw, Lauren Vincent, and Siobhan (Whadcoat) Velazquez; thank you for your input,friendship, and keeping me sane. Special shout out to Andrew, who has been a brilliant collaborator,excellent field partner, and like a second supervisor to me.I am also indebted to my field partners, and grateful for their patience and resilience during fieldwork: David Bonneau, Bea Collier-Pandya, Vivian Cristina Dias, Carie-Ann Lau, Marin McDougall,Scott McDougall, Andrew Mitchell, Joshua Nicholas, Natalie Pedersen, Jose´ Pullarello, StephanieTarnai, and Bill and Siobhan Velazquez. I would also like to acknowledge Sahar Ghadirianniari andKelsey Matson for their help scanning airphotos.I am very fortunate to have had support from BGC, and mentorship from the people that makeup BGC, including Bea Collier-Pandya, Kris Holm, Matthias Jakob, Carie-Ann Lau, Phil LeSueur,Emily Moase, and Alex Strouth. Special thank you to Alex for having my back and helping me seethe bigger picture.A lot of the data used in this thesis came from people and organizations who went out of theirxixway to help me. I would like to express my gratitude to David Bonneau and Jean Hutchinson fromQueen’s University, John Clague and Gioachino Roberti from Simon Fraser University, TrevorEvans from Canadian National Railway, Gord Hunter from the BC Ministry of Transportation andInfrastructure, Brian Menounos from the University of Northern British Columbia, Mike Tinholtfrom BC Hydro, Ryan Wainwright from the Squamish Lillooet Regional District, and John Whittallfrom BGC, for their generosity in sharing data. Lastly, I would like to acknowledge Erin and RobElliott, Sharon Kamenka, and the Ts’kw’aylaxw First Nation for land access permissions.Finally, thank you to my family, friends, partners, and teammates; you have supported me inmore ways than I can express.Funding for this research was provided by the Natural Sciences and Engineering ResearchCouncil of Canada (NSERC) and scholarships given by The Department of Earth, Ocean andAtmospheric Sciences.xxChapter 1IntroductionIn this chapter, the overall context for the thesis is established by introducing the researchproblem, stating research objectives and hypotheses, and describing the overall research approach.These sections are followed by a general description of the study area.1.1 Problem StatementIn mountainous regions, many communities and infrastructure projects are built on fans at themouths of steep creeks, which may be subject to episodic debris flows. Forecasting the spatialimpact of debris flows is an important part of hazard mapping, risk assessment, and mitigationdesign, but is challenging due to complex physical processes. Debris flow mobility, defined asthe ability to travel long distances and/or inundate large areas, depends largely on volume (e.g.,Corominas, 1996; Griswold & Iverson, 2008), but also flow composition and topographic controls toa certain extent. Rheological parameters used in semi-empirical numerical models such as Dan3D(McDougall & Hungr, 2004) or RAMMS (Christen et al., 2010) can be adjusted to simulate variousflow mobility for a given volume, however, there is presently little guidance available to practitionersto do so (McDougall, 2017). These models are typically calibrated through back-analysis, requiringpre- and post-event data and objective calibration methods (e.g., Aaron et al., 2019). Empirical-statistical methods (e.g., Corominas, 1996; Griswold & Iverson, 2008; Rickenmann, 1999) provide asimple yet practical alternative, but must also be calibrated to local datasets. Additionally, empirical1relationships are prone to considerable scatter, although this variability can be used to establish limitsof confidence in runout estimates for probabilistic assessments (McDougall, 2017).Debris flows are prone to avulsion, defined as a sporadic deviation of flow from an establishedflow path. Avulsions are formative processes on debris flow fans, shifting the active channel andlocus of deposition (and hazard) through space and time. Until recently, much of our understandingof avulsion processes is from observations of fluvial systems (de Haas et al., 2018a; Densmoreet al., 2019). There is currently little guidance for predicting where and when the next avulsionmight occur. Although three-dimensional numerical models can help indicate potential avulsionassociated with superelevation and runup around channel bends, they currently lack the capability tosimulate avulsions caused by sporadic channel blocking by coarse lobes or woody debris (McDougall,2017). Estimating the probability of avulsion from direct observation may not be feasible due to longreturn periods for debris flows and even longer return periods for avulsion (de Haas et al., 2018a).Reconstructing fan history requires a significant amount of effort, and there are a limited numberof well-studied fans in the literature to infer typical avulsion rates, which can be highly variablebetween fan settings (de Haas et al., 2018a).In the study area of southwestern British Columbia (SWBC), debris flow hazard and risk assess-ments are becoming common to support planning and decision making. These include local fanstudies (e.g., BGC, 2015; 2018b) and regional prioritization works (e.g., two case studies summa-rized by Sturzenegger et al. (2019) for central BC). In both these cases, expert judgement was anintegral part of debris flow runout forecasting, such as selecting various mobility conditions, avulsionscenarios, and interpreting numerical modelling outputs. There must be a continued effort to bolsterexpert judgement, and in some cases, challenge practitioner bias, with empirical observations andstatistical analyses. To date, there have been no systematic studies of runout evolution trends on fansin SWBC, and very few runout prediction methods in general that are either probabilistic, or considerflow deviation due to avulsion. This thesis aims to address these key knowledge gaps.1.2 Research ObjectivesTo address the challenges described in Section 1.1, the main research objectives are:21. Create a rich geospatial dataset documenting debris flow impacts with high spatial accuracyacross numerous fans in SWBC.2. Develop a systematic method to extract, visualize, and compare spatial impact trends acrossnumerous fans.3. Test statistical differences in spatial impact trends for groups of fans or events using easilymeasurable variables.4. Provide data-driven guidance to practitioners for estimating probability of runout exceedanceon a fan area using case studies in SWBC.1.3 Research HypothesisAlong with addressing the research objectives, the following hypotheses were tested:1. Spatial impact trends relative to the previous flow path, both down and cross-fan, exist, andcan be generalized for a group of fans.2. Differences in spatial impact trends for groups of fans or events can be explained, in part, withmorphometric or geotechnical characteristics.1.4 Research ApproachFirst, a comprehensive literature review was completed to generate a conceptual model of whatfactors affect debris flow mobility and avulsion based on our current state of knowledge. Existingempirical runout methods were also summarized, and key challenges and knowledge gaps identified,to be addressed in this thesis and future work. The literature review is provided in Chapter 2.The data compilation phase consisted of collecting high quality field and remote sensing dataacross 30 fans in SWBC. Impact areas and flow paths dating back to the beginning of the airphotorecord were mapped with an ensemble of data sources, including airphotos, satellite imagery, lidar,and field observations. Event volumes were reconstructed where possible, and morphometricvariables were calculated for each fan site. The dataset, along with the data compilation process, isdescribed in Chapter 3.Once the dataset was complete, spatial impact trends were aggregated across multiple fans using3the fan area as a normalizer, with the fraction of impacted areas a proxy for probability of impact. Tosimplify the data analysis, runout was considered in two components: runout down-fan (i.e., howfar debris flows tend to travel past the apex toward the fan toe) and runout cross-fan (i.e., how fardebris flows deviate from the previous flow path). Distributions of maximum down-fan and cross-fanrunout are a proxy for mobility and avulsion behaviour, respectively. Regional and local trends arediscussed, and differences in these distributions were tested using characteristics of the fan site orevent as discriminators. Data analysis and interpretation of results is presented in Chapter 4.Lastly, the main findings and implications for hazard and risk assessments were described, alongwith recommendations for future research. These conclusions are found in Chapter 5.1.5 Study Area1.5.1 Geologic SettingThe study area is located in southwestern British Columbia (SWBC), characterized by ruggedmountains, deep valleys, and plateaus sculpted by Pleistocene glaciation. Most of the study sitesare located in the southern Coast Mountains (Pacific Ranges) physiographic region, with a fewsites in the Fraser Lowlands and bordering the northern Cascade Mountains and Thompson Plateau(Figure 1.1).The study area is mostly underlain by Middle Jurassic to Eocene granitic rocks of the CoastPlutonic Complex, largely granodiorites with some quartz diorite and diorite, which overprintaccretionary terranes of sedimentary and volcanic rocks of Middle Jurassic age and older (Bustinet al., 2013). High-grade regional metamorphism is closely associated with plutonism, as well aswith major structures, consisting of northwest and north-trending contractional and strike-slip faultsystems (Monger & Journeay, 1994). Quaternary volcanic rocks in the study area are part of theGaribaldi volcanic belt, the northern segment of the Cascade volcanic arc, characterized by a chainof intermediate composition volcanoes with evidence of extensive glaciovolcanism (Kelman et al.,2002).During the most recent Pleistocene glacial episode, almost all of BC was covered by the4Cordilleran ice sheet, reaching its maximum extent about 17,000 years ago (Clague & Ward, 2011).Glaciation sculpted the Coast Mountains, shaping fjords and u-shaped valleys that dissect the terrain,depositing sequences of glacial till, glaciofluvial, glaciolacustrine, and minor glaciomarine sediments(Church & Ryder, 2010). Contemporaneous and post-glacial volcanism have formed prominentedifices, including Mount Garibaldi, Mount Cayley, and Mount Meager. Deglaciation was largelycompleted 11,500 years ago, accompanied by isostatic uplift and post-glacial dissection of valley-fill(Ryder, 1971; Ryder et al., 1991). Alpine glaciers are still present in the study area at high elevations,although these are vanishing with recent climate change (Walker & Pellatt, 2003). Following glacialdebuttressing, mass wasting and fluvial reworking resulted in a pulse of sedimentation (paraglacialprocesses), followed by more stabilized slopes and a relaxation in sediment supply (Ballantyne, 2002;Church & Ryder, 2010). These processes formed talus slopes, colluvial cones, alluvial fans, dissectedterraces, floodplains, and deltas that dominate the contemporary landscape. Debris flows are adominant hillslope process by which sediments are delivered to valley floors, along with rockfallsand rock avalanches (Church & Ryder, 2010).5######!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(121° W121° W122° W122° W123° W123° W124° W124° W125° W125° W51° N51° N50° N50° N49° N49° N0 40 8020 Kilometers±British Columbia AlbertaUSA!( Study sites# Quaternary volcanic fieldsPhysiographic regionBedrock typeVolcanicIntrusiveSedimentaryMetamorphicUltramaficThompson PlateauPacific RangesCascadeMountainsFraser LowlandsFigure 1.1. Regional geology and physiographic regions of the study area. Provincial digitalbedrock geology accessed from Cui et al. (2017), physiographic regions for Canadaaccessed from Bostock (2014), and Quaternary volcanic fields from the Garibaldi volcanicbelt assembled from Wilson (2019).1.5.2 ClimateThe climate of SWBC is generally moist and mild, but varies considerably across the studyarea. Pacific oceanic storms bring heavy rain and snow from late fall to winter along the coastof the Pacific Ranges. In the Lower Mainland, Pacific storms stalling against the mountains canalso bring about intense, orographically enhanced, precipitation (Demarchi, 2011). Here, summersare typically dry and warm with occasional rainy periods. Further inland, the Interior TransitionRanges ecoregion lies in the rainshadow of the Coast Mountains, consisting of hot summers andcool, dry winters (Demarchi, 2011). Debris flows on the western flank of the Pacific Ranges can be6triggered by heavy rainfall or rain-on-snow events from October to December, while debris flows inthe northeastern quadrant of the study area may be associated with spring rain-on-snow events orsummer thunderstorms (Bovis & Jakob, 1999). Microscale weather processes in the catchments, suchas cells of high intensity rainfall from orographic uplift and local wind pattern, are also importantdebris flow triggering mechanisms (Jakob & Lambert, 2009).Figure 1.2 shows monthly climate normals from 1971-2000 at four stations across the studyarea accessed from the Pacific Climate Impacts Consortium (PCIC) data portal. Figure 1.3 showsgridded monthly precipitation data (PCIC, 2014) for the wettest (November) and driest (July) months,highlighting rain-shadow and topographic effects across the study area.According to the PCIC Plan2Adapt Climate Information Tool (PCIC, 2012), projected climatechanges for the study area over the next century include an increase in temperature and overallannual precipitation, with drier summers and wetter winters. These projections are based on themean temperature and precipitation change from the 1961-1990 baseline using 30 climate changeprojections (15 Global Climate Models, two global greenhouse gas emission scenarios) for theSquamish Lillooet, Fraser Valley, and Metro Vancouver Regional Districts (PCIC, 2012). Theseclimate projections could mean more wildfires, glacial and permafrost changes, beetle infestations,and more landslide triggering storms in the winters, all of which are known to influence debrisflow activity (Jakob, 2019). A study by Jakob & Lambert (2009) also supports increased landslidefrequency along the southwest coast of BC based on climate change models that predict increasedantecedent and short-term precipitation in the twenty-first century.7Figure 1.2. Monthly climate normals (1971-2000) for climate stations across the study area.Locations shown on map in Figure 1.3. Station data accessed from PCIC data portal.8Interior TransitionRangesPacific RangesGeorgia-Puget BasinLower Mainland4321122° W122° W123° W123° W124° W124° W51° N51° N50° N50° N49° N49° N0 40 8020 Kilometers±British Columbia AlbertaNovember climatenormals (1971-2000)monthly precip.USAPacific RangesGeorgia-Puget BasinInterior TransitionRangesLower Mainland4321122° W122° W123° W123° W124° W124° W51° N51° N50° N50° N49° N49° N±July climate normals(1971-2000) monthlyprecip.!( Climate stations! Study sitesEcoregions906 mm23 mm209 mm20 mmFigure 1.3. Gridded precipitation data for November (left) and July (right) climate normals(1971-2000) from PCIC (2014). High-resolution climatology grid with a pixel size ofabout 800 m is derived from station data interpolated using the Parameter-elevationRegressions on Independent Slopes Model (PRISM). Monthly climate station data shownin Figure 1.2. Ecoregions defined by Demarchi (2011).1.5.3 Previous Research in the Study AreaTo date, there have been no comprehensive studies of avulsion trends on debris flow fans in BC.Papers by Hungr et al. (1984) and VanDine (1985) laid the groundwork for debris flow hazard analysisand remedial measures in western Canada. The most relevant work related to debris flow runout inthe study area is by Jordan (1994), who studied dynamic behaviour and physical properties of debrisflows in the Squamish and upper Lillooet River drainages. Jakob et al. (1997) studied morphometricand geotechnical controls on debris flow frequency and magnitude, including many of the study sitesin this thesis. Lau (2017) examined morphometric controls on scour depth on temperate alluvial fansin southern BC, including a detailed case study on a debris flow with extreme channel scour includedin this inventory. Relevant research for coastal BC using a debris flow inventory from the QueenCharlotte Islands (Haida Gwaii) northwest of the study area include: empirical-statistical runout9models by Fannin & Rollerson (1993) and Fannin & Wise (2001); a study relating volume to slopelength by Hungr et al. (2008); and an examination of controls on debris flow mobility by Guthrieet al. (2010).10Chapter 2A Comprehensive Review of Debris FlowRunoutThis literature review provides an overview of debris flow runout and avulsion processes onfans, focusing on implications for hazard and risk assessments. Here, debris flow and fan evolutionprocesses are described, establishing a conceptual model for factors that affect debris flow mobilityand avulsion. Components of a hazard and risk assessment are defined, and existing empiricalmethods for forecasting debris flow runout are discussed. Finally, the main challenges and knowledgegaps to be addressed in this thesis and future work are outlined.2.1 Debris Flow Processes and FormsA debris flow is defined by Hungr et al. (2014) as: “Very rapid to extremely rapid surging flowof saturated debris in a steep channel. Strong entrainment of material and water from the flowpath”. Debris is loose, unsorted material of low plasticity produced by mass wasting processes,weathering, glacier transport, explosive volcanism, or human activity (Hungr et al., 2001). Debrisflows can be triggered by heavy precipitation, rapid snowmelt, mass movements in the catchment,and outburst flooding. Typically, most of the debris flow volume is derived from entrainment, withthe initiating volume small in comparison (Hungr et al., 2014). A debris flow may be comprisedof a single surge or many surges, which consist of a steep, coarse-grained front followed by a tail11of dilute sediment-charged (hyperconcentrated) afterflow (schematic in Figure 2.1) (e.g., Hungr,2005; Iverson, 1997; Pierson, 1986). Thick surge fronts magnify the peak discharge, which can be anorder of magnitude greater than the most extreme hydrological flood (Hungr et al., 2014). Velocitiescan range from 0.5 to 20 m/s (Lorenzini & Mazza, 2004) with measured and back-calculated peakvelocities of 3 to 12 m/s recorded at some Canadian creeks (VanDine, 1985). Velocities are typicallyhighest along steep and confined channel reaches of the catchment, decreasing at the fan apex wherea loss in confinement and decrease in slope initiate deposition processes.The schematic in Figure 2.1 helps depict the debris flow processes and forms described here. Atthe fan, narrow boulder levees typically form on either side of the flow path as coarse materials areadvected to the flow edges (Blair & McPherson, 1998; Costa, 1984; Johnson et al., 2012). Debriscan also be deposited in the channel as channel plugs (de Haas et al., 2018a; Whipple & Dunne,1992). As the debris flow surges down-fan, accompanied by a reduction in slope, loss of channelconfinement, and selective boulder depletion, the flow spreads and loses momentum, depositing aslobes (Blair & McPherson, 1998; de Haas et al., 2019). Debris flows are also prone to avulsion,where debris abandons the main channel, diverting flow to elsewhere on the fan (discussed furtherin Section 2.2). A watery afterflow may continue beyond the terminal lobes as the falling limbof the debris surge continues down slope, or as the catchment continues to drain rainfall (Blair &McPherson, 1998; Hungr, 2005). The upper fan is typically dominated by levee deposition whilelobate forms are more common on the lower fan (Blair & McPherson, 1998). Fans are generallyconsidered a depositional landform, but debris flows also erode and entrain fan sediments, governedby basal shear stresses, grain collisional stresses, or destabilization and collapse of channel banks(Schu¨rch et al., 2011b, and references therein).Debris flows are part of a wide and continuous spectrum of hydrogeomorphic processes withvarying sediment sizes and particle-size distributions affecting flow properties (Pierson, 2004). Inorder of increasing sediment concentration, these processes include clear water floods, debris floods,hyperconcentrated flows, and mudflows/debris flows. Debris flows can be differentiated from theother processes by this definition by Hungr et al. (2014): “(1) the peak discharge is more than threetimes greater than that of a major flood flow, or (2) mean solids volume concentration at the surge12peak greater than about 60% and the water and solid phases thoroughly mixed”. Mudflows are similarto debris flows but are distinguished texturally as containing a significant content of saturated plasticfines and lacking coarse fragments, typical in regions of deep weathering (Costa, 1984; Hungr et al.,2014). Hyperconcentrated flows are two-phase non-Newtonian fluids with sediment concentrationsabout 20-60% by volume (Pierson, 2005). Hungr et al. (2014) defines a debris flood as: “Very rapidflow of water, heavily charged with debris, in a steep channel. Peak discharge comparable to thatof a water flood”. The main difference between debris floods and hyperconcentrated flows/debrisflows/mudflows is that flow properties are governed by fluid flow rather than the interaction of solidand fluid forces (Hungr et al., 2001; Iverson, 1997). Clear water/stream flow floods have less than20% solids by volume with distinct sedimentary structures from fluid flow (e.g., stratified, well sorted,upward fining deposits, clast imbrication) (Costa, 1984; Wilford et al., 2004).The distinction of hydrogeomorphic processes is important for hazard management becausedifferent hazard characteristics are associated with typical flow types (Wilford et al., 2004). Debrisflows can be extremely destructive due to high velocities, flow depths, and the movement of largeboulders. Except for catastrophic dam-breach outburst floods, debris floods usually do not developimpact forces comparable to debris flows (Hungr et al., 2001). Field evidence (described in Sec-tion 3.3.5) and morphometrics (described in Section 3.6.7) can help distinguish the dominant processtype. In general, debris flow processes are more typical of steep drainages of less than a few squarekilometers, while debris floods are associated with larger watersheds with greater hydrologic flooddischarges (Hungr et al., 2014; Wilford et al., 2004). However, all processes may occur within asingle drainage or may be present at different times during an event or a single surge (Costa, 1984).13Figure 2.1. Schematic diagram of typical debris flow processes, forms, and impacts (not toscale). Adapted from Pierson (1986) and Lau (2017).2.2 Debris Flow Fan EvolutionA fan’s semi-conical morphology is the product of avulsion sequences that shift the activechannel and locus of deposition through space and time (de Haas et al., 2018a; 2019). Until recently,there have been few systematic studies investigating controls on debris flow avulsion, and much ofour understanding is from observations of fluvial systems (de Haas et al., 2018a; Densmore et al.,2019). Research on autogenic controls (internal thresholds and feedback response) on fluvial fandynamics show that fan morphology evolves through repeated cycles of incision, backfilling, and14spreading/avulsion (Figure 2.2) (e.g., Reitz & Jerolmack, 2012; Schumm et al., 1987). Evidencefrom the fluvial literature suggests that avulsion rates increase with sediment supply (e.g., Ashworthet al., 2004; Bryant et al., 1995), with small aggrading floods critical to preconditioning avulsions byreducing channel capacity to some threshold (e.g., Field, 2001; Jones & Schumm, 1999).de Haas et al. (2016) observed similar autogenic cycles on experimental debris flow fans, butwith different processes (Figure 2.2). After upstream migration of the depocenter toward the fanapex (i.e., backstepping), an avulsion is triggered toward a topographically favourable path. Afteravulsing, backstepping recommences once debris flows reached a maximum runout, and the cyclerepeats. One of the main differences between fluvial and debris flow end-members is that cycles onfluvial fans are controlled by progressive aggradation operating continuously in time, while debrisflow deposition is more localized in space and time (de Haas et al., 2016). A single debris flow lobecan trigger an avulsion, leading to more chaotic avulsion patterns on debris flow fans compared tofluvial fan systems (de Haas et al., 2018a). Runoff erosion and secondary fluvial processes betweendebris flows were not replicated in the experiments by de Haas et al. (2016) since the watery tail-endof the experimental debris flow was prevented to bury or rework the initial deposit; these resultsrepresent an idealized end-member of the hydrogeomorphic process spectrum.de Haas et al. (2018a) validated the experimental work with an analysis of spatio-temporalpatterns on 16 well-studied fans from around the world. Patterns on natural fans were significantlymore chaotic compared to the experimental findings due to variations in magnitude, composition,and rheology of the flows (de Haas et al., 2018a). Overall however, de Haas et al. (2018a) observedthat debris flows on natural fans also follow cycles of channel plugging, backstepping of depositiontowards the fan apex, avulsion towards a topographic low, and establishment of a new active channel.For the fans in that study, avulsions appeared to occur approximately every 3 to 8 flows, but thisis very dependent on the fan environment; major channel shifts may require more flows betweenavulsion events, or a complete avulsion cycle could be initiated within a single event with multiplesurge sequences (de Haas et al., 2018a).15Figure 2.2. Schematic of autogenic cycles on experimental debris flow fans and fluvial fansand deltas from de Haas et al. (2016).2.3 Debris Flow Hazard and Risk AssessmentIn mountainous regions, communities and infrastructure are often built upon fans requiringdetailed hazard and risk assessments to understand potential risks posed by hydrogeomorphichazards. Risk is the probability of a hazardous event and its potential consequences. A quantitativerisk assessment (QRA) is a systematic and quantitative framework for evaluating risk. Risk can beexpressed analytically using Equation 2.1 or something similar (e.g., Corominas et al., 2014; Daiet al., 2002; Fell, 1994):Risk =n∑i=1P(H)i×P(S|H)i×P(T |S)i×Vi×E (2.1)16where, for i of n hazard scenarios, P(H) is the annual probability of the hazard occurring, P(S|H) isthe spatial probability that the hazard will reach the element at risk, P(T|S) is the temporal probabilitythat the element at risk will be present if the hazard reaches its location, V is the vulnerability orthe probability of loss of life if the element is impacted, and E is the value of the element at risk ornumber of people at risk in the case of life loss risk.Hazard assessments are generally limited to the P(H) and P(S|H) terms by considering theprobability of a hazard occurring and its intensity without specifying the exposure or consequencesto the elements at risk (Corominas et al., 2014). For debris flow hazard assessments, this usuallyinvolves identifying a potential debris flow hazard, determining event magnitudes for a range ofreturn periods (i.e., developing frequency-magnitude (F-M) relationships), and conducting a runoutassessment. A combination of P(H), P(S|H), and P(T|S) is sometimes referred to as encounterprobability. The P(T|S) represent exposure to the hazard, while the V and E terms represent theconsequences.Defining a volume for a given return period is a sensitive parameter in the QRA as volumecan propagate into other terms in Equation 2.1, such as P(S|H), where volume is typically aninput for runout models. Establishing reliable F-M curves can be a costly and time-consumingendeavor, requiring detailed fan reconstructions using a variety of absolute and relative datingmethods, supplemented by statistical analyses to extrapolate probabilities outside of the observedrecord (Jakob, 2019). When detailed studies are not practical, debris production has been correlatedwith catchment morphometrics, such as debris contributing areas, relief, drainage density, andruggedness (e.g., Bovis & Jakob, 1999; D’Agostino & Marchi, 2001; de Haas & Densmore, 2019).Jakob et al. (2016) showed how regional F-M relationships might be developed by normalizing anensemble of established F-M curves by fan volume or area.Debris flow runout assessments involve forecasting debris flow motion, such as how far andhow fast a debris flow will travel. Forecasting runout behaviour can be challenging due to thecomplexity of physical processes, as described in Section 2.1. There are many evolving tools andtechniques, ranging from simple empirical methods (discussed further in Section 2.6) to analyticalmethods. Empirical methods are typically applied at the regional scale or for screening level studies,17while detailed QRAs or engineering design would more likely involve process-based modelling withnumerical models, such as Dan3D (McDougall & Hungr, 2004), RAMMS (Christen et al., 2010), andFLOW-2D (O’Brien et al., 1993). Numerical models and GIS technologies are increasingly commonfor developing hazard and risk maps (Quan Luna et al., 2014). Numerical model outputs can beused to estimate vulnerability by associating building damage to debris flow intensity, calculatedfrom numerical outputs such as flow depths, velocities, and impact pressures (e.g., Jakob et al., 2012;Kang & Kim, 2016; Quan Luna et al., 2011).McDougall (2017) identified some challenges with modelling flow-like landslides, such assensitivity to topographic resolution, selection of model input parameters, and simulating suddenchannel obstructions causing avulsion. Semi-empirical models such as Dan3D (McDougall &Hungr, 2004) simulate bulk flow behaviour with a fixed rheology and a few calibrated parameters.Calibration through back-analysis requires detailed event documentation and pre-event topography.Even if calibrated, it is uncertain whether future events can be simulated with similar rheologicalmodels or calibrated parameters. Furthermore, although a fixed rheology may adequately simulatebulk flow, it does not capture flow heterogeneity inherent to debris flow processes (Iverson, 2003).Mathematical models such as D-Claw (George & Iverson, 2014) are able to simulate the effects ofevolving dilatancy and flow phases from initiation to deposition, but require more model parameters.Numerical models may simulate avulsions associated with superelevation and runup, but avulsionscaused by channel blockages from coarse deposits or woody debris require ad hoc topographicadjustments to simulate with a semi-empirical numerical model.Many runout prediction methods are deterministic. Practitioners must use judgement to convertrunout calculations or modelling outputs to a probability of runout exceedance, depending onanticipated mobility (irrespective of volume) or avulsion scenarios, depicted in Figure 2.3. Asystematic way to account for different mobility or avulsion behaviours in a QRA is to identifycredible sub-scenarios, assign a conditional probability, and model each sub-scenario separately.There is currently little guidance available to practitioners for identifying credible sub-scenariosand for assigning probabilities. A wide spectrum of mobility and avulsion behaviours make for acomplicated event tree. Alternatively, varying mobility could be modelled with a Monte-Carlo style18analysis by sampling model parameters from probability density functions (PDF) (e.g., Aaron, 2017;Quan Luna, 2012; Scheidl & Rickenmann, 2010). Monte-Carlo style analyses require a rich databaseof systematically calibrated parameters to develop probability density functions, and would not beefficient for complex models that require hours to complete one model run.Figure 2.3. Conceptual debris flow avulsion scenarios and mobility that may be considered in aQRA.Not all hazard and risk-based decision making follow the procedures outlined here. For instance,regional or preliminary studies often employ landslide susceptibility mapping. Landslide susceptibil-ity, as defined by Fell et al. (2008), is “a quantitative or qualitative assessment of the classification,volume (or area), and spatial distribution of landslides which exist or potentially may occur in anarea”. In other words, a susceptibility analysis involves identifying the hazard and potential runoutwithout explicitly considering a temporal probability. The Flow-R model (Horton et al., 2013) is aGIS-based regional susceptibility model with automatic source area delineation and flow propagationusing a spreading algorithm and simple frictional laws. Since flow propagation calculations arebased on a unit energy balance, there is no specification of event volumes or scenarios. Flow-Rmodel outputs can be interpreted as areas that could potentially be reached by debris flows, with anassociated relative susceptibility, but are not equivalent to a spatial probability of impact map.In the context of hazard and risk, this thesis focuses on questions related to the P(S|H) term inEquation 2.1, and questions related to different mobility and avulsion scenarios shown in Figure 2.3.192.4 Factors Affecting Debris Flow MobilityMobility is a measure of a landslide’s ability to either runout long distances and/or inundatelarge areas. Controls on debris flow mobility have been studied extensively through experiments andfield observations. Mechanistic interpretations from the literature are summarized here, forming aconceptual model for what factors affect debris flow mobility.2.4.1 Event ConditionsIt has been well established that debris flow volume has a large effect on runout, with greatervolumes having more momentum and the tendency to spread further (e.g., Corominas, 1996; Griswold& Iverson, 2008; Legros, 2002). There is considerable scatter in these trends which may be attributedto other factors, discussed in the following Sections 2.4.2 and 2.4.3. A preliminary study by Zubryckyet al. (2019) using data from this thesis found a potential inverse correlation between a calibratedVoellmy friction coefficient and volume (Figure 2.4). A similar volume dependency was also observedby Schraml et al. (2015).Figure 2.4. Preliminary correlations by Zubrycky et al. (2019) to calibrated Voellmy frictioncoefficient using data from debris flow events presented in this thesis. Numerical modellingwas completed in Dan3D (McDougall & Hungr, 2004).Other factors specific to an event scenario might also influence mobility, such as triggeringconditions, discharge hydrograph (Whipple, 1992), number of surges (Chen et al., 2017), and degreevolume generated from progressive entrainment (Frank et al., 2015). Flow height is influenced in part20by the flow hydrograph, which affects the degree of erosion and deposition (Schu¨rch et al., 2011b).Chen et al. (2017) showed numerically that the inundated areas and runout distances of successivedebris flows are smaller (i.e., a multi-surge event) than those of concurrent debris flows due to lowermobility of smaller individual events and blockage by the earlier debris flows. Although the mobilityof an individual surge may be controlled by other factors, the total impact area of an entire eventwould be influenced by the number of surges, hypothetically.2.4.2 CompositionDebris flow composition can be described by water content, grain-size distribution, lithology,and the amount of woody debris or organics. It has been well established in soil mechanics thatpore-fluid pressures influence the strength of a sediment-water mixture. The effect of grain-sizedistribution on debris flow rheology has been studied experimentally (e.g., Major & Iverson, 1999;Parsons et al., 2001; Phillips & Davies, 1991) and observed geomorphically (e.g., Whipple & Dunne,1992). Large-scale flume experiments by Major & Iverson (1999) show deposition at coarse grainedflow margins is governed by frictional resistance in the absence of high pore-fluid pressure. Kaitnaet al. (2016) found that the primary manner in which grain-size distribution controls excess porepressure is in limiting pore pressure dissipation. Similarly, experimental findings by de Haas et al.(2015) show that increasing the coarse fraction enhances mobility to an extent, but an excess ofcoarse material enhances pore pressure diffusivity, increasing frontal friction and stalling the flow.Due to the grain-size heterogeneity within a single surge, debris flows should not be represented witha fixed rheology (Iverson, 2003). However, the bulk mobility of an event may be informed by thecatchment lithology since weathering of the source rock would influence the grain-size distribution ofa typical flow. Field evidence from the study area collected by Jordan (1994) found that fine-textureddebris flows from weak, clay-rich Quaternary volcanic rocks exhibit long runout on gentle gradientscompared to coarse-textured debris flows from granitic sources. A study by Tiranti & Deangeli(2015) showed how catchment lithology may help inform the selection of a rheological model.Slurries with high proportions of silt and clay usually have lower yield strengths and viscositiesmaking them generally more mobile (Whipple & Dunne, 1992). Lahars with high clay contents from21hydrothermally altered volcanic rocks have higher mobility than granular debris flows (Griswold& Iverson, 2008). Even within lahar populations, cohesive flows with more than 3 to 5 percentclay-sized sediment are more mobile compared to non-cohesive flows (Scott et al., 1995). Zhanget al. (2013) found that muddy debris flows in the Wenchuan Earthquake Zone with a higher finescontent were more mobile compared to flows with larger clasts and less than 2 percent silt and clay,deriving different empirical runout equations based on the flow type. The experiments by de Haaset al. (2015) found that an increase in clay content enhances mobility due to retained excess porepressures, however too much clay creates a viscous flow with reduced runout.The effect of organics and woody debris on mobility are poorly understood. May (2002) foundthe runout length had a strong influence on the accumulation of wood as the flow traveled. Lancasteret al. (2003) hypothesized that one way in which large woody debris may reduce runout is byentanglement at the flow front causing wood jamming in channels.2.4.3 Path CharacteristicsTopography can also exert significant control on debris flow mobility (e.g., Corominas, 1996).Path characteristics that influence mobility include elevation loss, channel gradients, path curvature,channel confinement, obstacles, and interaction with forest stands. Many of these factors influencemobility by dissipating energy and promoting deposition. Debris flows that initiate higher in thewatershed have higher potential energy, while those that interact with obstructions such as naturaltopographic features or human-made structures would dissipate energy. Dense forests have beenfound to suppress debris flow runout (e.g., Booth et al., 2020; Ishikawa et al., 2003), however thismay only apply for debris flows up to a certain magnitude. Sharp channel bends also reduce flowvelocity, thus limiting mobility (e.g., Benda & Cundy, 1990; Fannin & Wise, 2001).Assuming steady, uniform, gravity-driven flow, yield stress can be calculated assuming a flowdensity, depth, and surface slope (e.g., Johnson & Rodine, 1984; Whipple & Dunne, 1992). It hasbeen well observed that a decrease in slope initiates deposition and limits runout (e.g., Benda &Cundy, 1990; de Haas et al., 2015; Fannin & Wise, 2001; Miller & Burnett, 2008). However, factorsthat set the fan slope, such as lithology, grain-size, or process type (e.g., Blair & McPherson, 1998;22Hooke, 1968) may exert more of a control on mobility than the flow interacting with the slope itself.Based on a preliminary numerical model calibration exercise using data from this thesis, there isan apparent correlation between the calibrated Voellmy friction coefficient and the lower channelgradients (Figure 2.4). Scheidl & Rickenmann (2010) correlated empirical mobility coefficients tothe average fan slope and the average channel slope. Experimental findings by de Haas et al. (2015)show that topographic controls, such as increased runouts with increased outflow plain slopes, werenegligible compared to the effect of debris flow composition.Channelized or confined flows typically exhibit longer runouts (e.g., Cannon, 1989; Garcı´a-Ruizet al., 1999; Zhang et al., 2013). Flow depths are thicker when concentrated in a channel, promotingentrainment and thus propagation down-fan; once unconfined, debris tends to spreads to some criticalthickness, and movement is halted once all debris is deposited (Cannon, 1989; Fannin & Rollerson,1993; Miller & Burnett, 2008; Schu¨rch et al., 2011b). Along channelized reaches, water may beincorporated into the flow, further enhancing mobility. With a reduction in the channel capacity(cross-sectional area), frictional interactions with the channel sidewalls, such as boulder or logjamming, may lead to the formation of channel plugs (e.g., de Haas et al., 2018a).2.5 Understanding Debris Flow AvulsionControls on debris flow avulsion are much less well studied than those affecting mobility. Asdescribed in Section 2.2, debris flow fan evolution processes can be chaotic and occur over varioustime scales. In this section, factors that may influence avulsion likelihood and location on a debrisflow fan are described based on the current state of knowledge. Autogenic (intrinsic) controls onavulsion (e.g., de Haas et al., 2016) are the primary focus of this literature review, although allogenic(extrinsic) factors such as climate, base-level, and tectonics also play a role (Stouthamer & Berendsen,2007).2.5.1 Avulsion TriggersThere are various triggers, or physical drivers, that may cause a debris flow to avulse. Thesetriggers may be considered in a hazard or risk assessment, or when describing an event forensically.23Avulsion triggers have been grouped into three general scenarios, as shown in Figure 2.5: 1)overtopping (e.g., de Haas et al., 2018b); 2) superelevation (e.g., Field, 2001); and 3) channelblockages (e.g., de Haas et al., 2018a, de Haas et al., 2019, Whipple & Dunne, 1992). Many ofthese triggers, or hybrids, may be at play during a single event, possibly interacting with one-another.Progressive aggradation is more typical of fluvial processes (e.g., Bryant et al., 1995; de Haas et al.,2016; Field, 2001), but might be an important trigger for mixed-process fans or debris flows with anabundance of inter-surge or precursory flooding. de Haas et al. (2019) found channel-plug formationto be the dominant mechanism for triggering avulsions in Saline Valley, California, while Millard et al.(2006) found channel crossing structures from forestry operations were associated with avulsionsfor debris flows in coastal BC. Bank failure has not been identified in the literature as a potentialmechanism but is considered here conceptually.24Figure 2.5. Conceptual avulsion triggers.252.5.2 Estimating Avulsion ProbabilityEstimating the probability of avulsion is challenging because avulsion rates vary betweenfan environments and through time. There are few fans that have been spatially and temporallyreconstructed to help constrain typical avulsion recurrence intervals. Based on a study of four veryactive fans with frequent avulsions, de Haas et al. (2018a) found avulsions occurred every 3 to 8flows. Studies at alluvial fans in Owens Valley, California (e.g., D’Arcy et al., 2015; Du¨hnforth et al.,2007) and at the Illgraben fan, Switzerland (Schu¨rch et al., 2016) dated fan surfaces with a variety oftechniques including cosmogenic radionucliides of boulders. These studies are useful for looking atthe long-term evolution of fan sectors over thousands of years, but provide limited help to resolve theprobability of an impending debris flow avulsing (de Haas et al., 2018a).Certain fan environments may be more prone to avulsions than others. Fuller (2012) summarizedphysical variables that affect alluvial fan avulsions, including those related to flow discharge, sedimenttransport, fan physiography, channel condition, and allogenic factors. A recent study by Pedersonet al. (2015) related compensational stacking, or the tendency to fill topographic lows throughavulsion, to measurable fan characteristics. Based on the internal stratigraphy of three fans inColorado, Pederson et al. (2015) found that areas with typical debris-flow characteristics (abundantcoarse clasts, thick units, large lobes, high clay content) tend to stack more compensationally thanareas with typical stream-flow characteristics (thinner deposits, less clay and coarse clasts) (Santiet al., 2017). Avulsions may be more likely at fans with characteristically thick lobes compared totheir channel depths. As a proxy for probability of avulsion, de Haas et al. (2019) estimated theprobability for a channel plug to have sufficient thickness to induce avulsion by comparing typicalchannel depths to lobe thicknesses across nine debris flow fans in Saline Valley, California.Based on fan evolution studies by de Haas et al. (2016, 2018a) described in Section 2.2, fanswith recent channel plugging in the active channel or a backward propagation of the depocentermay be a strong indicator of impending avulsion. An experimental study by de Haas et al. (2018b)showed that fans experiencing abundant small flows followed by a large flow were more likely toavulse; sequences of smaller flows cause channel plugging, while larger flows have sufficient volumeto leave the active channel. However, flows with large flow depths may also erode the existing26channel, enhancing channelization and reducing the probability of avulsion (Densmore et al., 2019;Schu¨rch et al., 2011b). There may be an optimal frequency-magnitude distribution for which avulsionfrequency is maximized (de Haas et al., 2018a), but the possibility of erosion and self-channelizationmust somehow be accounted for. An experimental study by Deijns (2018) found that althoughthe debris flow frequency-magnitude distribution seems to be the major controlling factor in fandevelopment, changes in sediment composition also influence the avulsion behaviour on a timescaleof a few events; gravel-rich flows (higher coarse-fraction) increased erosion and inhibited avulsion inthe experiments, while gravel-poor experiments were more likely to plug the fan channel. Deijns(2018) describes that avulsion behaviour on experimental fans is controlled by an interplay betweenvolume sequences, fan topography, and debris flow composition.Considering the avulsion mechanism of boulder or wood jamming, fans with channel blockingstructures (e.g., roads, bridges, culverts, check dams) would be more susceptible to avulsion.2.5.3 Locating Avulsion PointsCurrently, selecting avulsion points for a hazard or risk assessment is mostly based on expertjudgement. In alluvial architecture modelling, avulsion points are simulated when some threshold isexceeded along a channel. Thresholds used in these models to induce river avulsions include theratio between cross-valley slope and down-valley slope (e.g., Mackey & Bridge, 1995; Slingerland &Smith, 1998; To¨rnqvist & S. Bridge, 2002), and the ratio between superelevation (height differencebetween the levee crest and the average floodplain elevation) and the bankfull channel depth (e.g.,Mohrig et al., 2000). Mackey & Bridge (1995) included the ratio of flood discharge to a flooddischarge threshold to model avulsion probability along a channel length. Using this analog fordebris flows, one could simply determine where along the channel the peak discharge is great enoughto overwhelm the channel conveyance capacity. This may work for a single surge with a high peakdischarge, but it assumes the channel capacity does not change during the course of an event. Dueto complex flow sequencing, channel bed entrainment, and sporadic channel plugging, locatingpotential avulsions is not always straightforward.A report by Millard et al. (2006) for fans in coastal BC found avulsions were most frequent27immediately downstream of the location at which the channel merges with the contemporary fansurface (i.e., the intersection point, as defined by Hooke (1967)), with frequency declining from thispoint toward the fan toe. Pederson et al. (2015) found avulsion tendency increased with distancefrom the fan apex, likely attributed to longitudinal variation in fan morphology, such as channelconfinement decreasing down-fan.The balance between typical lobe thickness and channel depth down-fan may be useful foridentifying avulsion hotspots (de Haas et al., 2018a). Therefore, a likely avulsion point is related tothe combined probability of a debris flow lobe stopping at a given location and having a thicknessthat approximately equals or exceeds the local channel depth (de Haas et al., 2019). From a study ofnine debris flow fans in Saline Valley, California, de Haas et al. (2019) found that channel plugginghas a similar likelihood at all radial distances from the fan apex, apart from areas of fan-head incision,since both lobe thickness and channel depth decrease with distance down-fan. For fans in coastal BC,Millard et al. (2006) found the channel depth did not appear to have a strong effect on the location ofavulsions.Given the avulsion triggers depicted in Figure 2.5, avulsion locations may be associated withchannel bends (superelevation) or channel blocking structures. The possibility of these locationsbeing credible avulsion points compared to any other location along the channel associated withother triggers is not well understood.Rather than trying to predict avulsion locations, an understanding of avulsion dynamics couldbe used instead to delineate areas susceptible to impact from avulsion. Long-term avulsion pronesectors may be identified by analyzing radial variations in fan topography since deposition tends togradually shift toward topographically lower parts on a fan (de Haas et al., 2018a). Paleochannelsmay also indicate preferential avulsion flow paths. Typical avulsion opening angles between old andnew pathways have not been documented, so it is not clear whether an avulsion is more likely togradually shift laterally or occupy a completely different fan sector (de Haas et al., 2018a; Densmoreet al., 2019).282.6 Empirical-Statistical Methods for Forecasting Debris FlowRunoutEmpirical methods are a simple, practical, and widely used approach for estimating debris flowrunout. They are derived from observation and guided by some knowledge of physical processes,where our understanding may be limited. Many of these methods are based on simple geometriccorrelations with easily obtainable predictor variables that can be calibrated to local datasets ortransposed to similar environments. Empirical tools can be useful for preliminary local assessmentsand regional studies in the absence of site-specific data, detailed topographic data, and resources toundertake numerical modelling.Empirical methods have been well summarized by Rickenmann (1999), Rickenmann (2005),Hu¨rlimann et al. (2008), Scheidl & Rickenmann (2010), and others. Here, the most prominentrelationships from the literature are highlighted, updated with some new approaches, followed bya discussion addressing challenges and limitations of these methods. Empirical relationships havebeen grouped into four general approaches, as shown in Figure 2.6:29Figure 2.6. Schematic (not to scale) depicting empirical approaches for forecasting debris flowrunout. a) Travel path, such as total travel path length (LT), path length relative to the fanapex (Lf) (e.g., Rickenmann, 1999), or angle of reach (α) defined by the ratio of fall height(H) to horizontal travel distance (L) (e.g., Corominas, 1996); b) deposit dimensions, suchas deposition length (Ld) and maximum lateral deposit width (Wd) (e.g., Tang et al., 2012);c) planimetric inundation area (A) correlated to debris flow volume (V) (e.g., Griswold& Iverson, 2008); and d) volume balance rules quantifying entrainment and depositionalong a flow path, by which the total travel distance is defined where the cumulative flowvolume is zero, i.e., the sum of the volume deposited (Vd) is equal to the sum of thevolume entrained (Ve) (e.g., Fannin & Wise, 2001).2.6.1 Travel PathRunout is commonly represented by the angle of reach (α), or fahrbo¨schung, which is thearctangent of the ratio of the fall height (H) to the horizontal travel distance (L), measured fromthe head of the source landslide to the furthest runout extent (Figure 2.6a). In other words, it is theinclination of the line projected from the landslide crest to the deposit toe. This parameter was firstdefined by Heim (1932), who designated α as a relative index of the mobility of rock avalanches,describing energy loss due to friction. Heim (1932), Scheidegger (1973), and many others havecorrelated α to landslide volume, showing an inverse relationship to event volume (V). Corominas30(1996) presented the first and most comprehensive review examining the effect of V on the H/Lratio for various landslide types including debris flows, finding a continuous reduction of H/L withan increase in V (m3) (Equation 2.2). Equation 2.2 is from a dataset of 71 debris flows, debrisslides, and debris avalanches (excluding mudflows and mudslides) triggered by storms in northernSpain (Corominas, 1996). Corominas (1996) found scattering in the relationship was mostly due tomechanisms of motion, obstacles or topographic constraints on the path.log(H/L) =−0.105log(V )−0.012 (2.2)Rickenmann (1999) derived relationships for L as a function of V and H, where the product of Vand H can be considered as potential energy of the mass movement.Regional angle of reach distributions can be used to define preliminary hazard zones withoutspecifying hazard scenarios and associated volumes. Rickenmann & Zimmermann (1993) defined aminimum H/L ratio of 0.19 using a dataset of 600 debris flows in the Swiss Alps, which is comparableto a rule of thumb once used in Japan of 0.2 (Takahashi, personal communication, 1994, as citedin Bathurst et al., 1997). Minimum H/L ratios for matrix supported flows (0.07) were found tobe lower than coarser-grained, clast supported flows (0.19) (Rickenmann & Zimmermann, 1993).Zimmermann et al. (1997) defined a lower envelope for H/L using the catchment area, where largercatchment areas are correlated to smaller H/L ratios (i.e., longer runout). A recent application ofthe angle α in hazard management tools is the GIS-based regional susceptibility model Flow-R(Horton et al., 2013). Flow propagation in Flow-R uses a spreading algorithm and energy balancelaws, where energy loss due to friction can be parameterized by the angle of reach. Travel pathlength measured horizontally, such as the total travel path (LT) or path relative to the fan apex (Lf)shown in Figure 2.6a, neglects the vertical component of energy loss. From a comprehensive studyof various long runout landslides, Legros (2002) found that travel distance depends primarily onvolume, while height just adds scatter to the correlation. A recent empirical study of rock avalanchesby Mitchell et al. (2020) found contrary evidence, with predicted runout distances highly sensitive tothe fall height, although initiating conditions for rock avalanches are unique to many debris flows.31Rickenmann (1999) provides a geometric scaling equation relating V (m3) to Lf (m) based on adataset of 140 debris flows from various sources (Equation 2.3). Rickenmann (1999) describes thatrunout is better represented by Lf since local changes in the channel geometry on the fan and differentmaterial properties are relatively more important compared to the entire travel path in the basin.Equation 2.3 is not recommended by Rickenmann (1999) for practical application as the scatterbetween predicted and observed values is too large.L f = 15V13 (2.3)A recent analysis of runout distances on depositional fans in the Wenchuan earthquake zone byZhou et al. (2019) derived empirical multivariate equations suitable for prediction of Lf using twovariables: volume and internal catchment relief.Models that use stopping criteria based on path geometry to predict total runout (e.g., Benda &Cundy, 1990; Burton & Bathurst, 1998; Miller & Burnett, 2008) are generally used for sedimenttransport modelling in the catchment. Stopping criteria are based on the assumption that debris flowstend to deposit sediment where the channel gradient declines and/or at high tributary junction angles.2.6.2 Deposit DimensionsPredicting deposit dimensions, such as maximum deposit length (Ld) and maximum lateraldeposit width (Wd) shown in Figure 2.6b, are important for hazard assessments on fans where debrisflow transport in the catchment is of less importance compared to deposition patterns on the fan.Most of these relationships use volume as a predictor variable (e.g., Chen et al., 2007; Ikeya, 1981;Yu et al., 2006) as there is a natural scaling relationship (discussed further in Section 2.6.3). Chenet al. (2007) considered the effect of deposit shape on these relationships. Tang et al. (2012) usedstepwise multiple regression analysis without volume as a candidate variable due to uncertainty in itsestimation. Instead, the model by Tang et al. (2012) uses catchment area, catchment relief, and anestimate of the volume of removable sediment in the catchment to estimate maximum depositionwidth and length. Some of these models assume the onset of deposition starts at the fan apex, makingLd equivalent to Lf described previously.322.6.3 Volume-Area RelationshipsAssuming landslide deposits are geometrically similar, there is a ratio between volume (V, m3) andthe planimetric area (A, m2) based on physical scaling laws that follows the power law relationshipin Equation 2.4, where c is a calibrated coefficient (Hungr & Evans, 1993; Iverson et al., 1998).Griswold & Iverson (2008) calibrated this coefficient for rock avalanches and debris flows (c = 20),finding that power law equations with a specified slope of 2/3 are statistically indistinguishable fromthe best-fit power law regressions. Many other authors have calibrated the volume-area power lawrelationships to local and global debris flow datasets (e.g., Berti & Simoni, 2007; Booth et al., 2020;Crosta et al., 2003; D’Agostino et al., 2010; Scheidl & Rickenmann, 2008; Webb et al., 2008; Yuet al., 2006), as summarized in Figure 2.7.A = cV 2/3 (2.4)33Figure 2.7. Power law scaling relationships with a 2/3 slope between volume and area fornon-volcanic debris flows, over the domain of the respective volumes for each dataset.Trendline for lahars provided for reference.The c coefficient can be interpreted as a mobility coefficient, irresepective of volume. Asdescribed in Section 2.4, mobility can be related to properties of the flow or topographic constraints.A c coefficient of 200 for lahars (debris flows from volcanoes) mean lahars typically inundatea planimetric area roughly ten times greater than a debris flow of the same volume. Scheidl &Rickenmann (2010) correlated c to the average fan slope and the average channel slope. The c34coefficients from the non-volcanic data sources shown in Figure 2.7 vary from 6.2 to 40. The variancecould be related to many of the factors listed in Section 2.4, including different geology, climate, andgeomorphic setting, or in part measurement inconsistency (Simoni et al., 2011). Landslide volumesare often estimated by multiplying an area covered by an estimated average thickness (Legros, 2002),so there is an inherent relationship between the two variables due to measurement.Volume-area relationships are implemented in inundation mapping programs such as LAHARZ(Schilling, 1998), DFLOWZ (Berti & Simoni, 2014), and TopRunDF (Scheidl & Rickenmann,2010). They are also used to reconstruct frequency-magnitude curves by associating a volume with ahistorical inundation area (Jakob, 2005).2.6.4 Volume BalanceVolume balance methods determine the total runout based on the point at which volume entrainedequals volume deposited (Figure 2.6d). Rulesets for erosion and deposition are based on theobservation that debris flows tend to entrain material through steep, confined sections of the travelpath, and deposit material at unconfined sections with lower gradients (Benda & Cundy, 1990; Fannin& Wise, 2001). Based on field survey data of 449 debris flows from glaciated hillslopes in coastalBC that were clear-cut logged, Fannin & Rollerson (1993) found the deposition of channelizedevents to be influenced by the ratio of channel width to channel gradient, with the onset of depositionexpected when the ratio exceeds unity. Extending this work, Fannin & Wise (2001) developed anempirical-statistical model that calculates volume change based on reach geometry and slope anglefor different flow types (unconfined, confined, and transition). Miller & Burnett (2008) calibratedentrained/deposited volumes and probability of debris flow termination to empirical data using thefollowing attributes along the flow path: forest-cover class, gradient, flow path confinement, andjunction angle.2.6.5 DiscussionEmpirical-statistical methods require a robust database of field observations for validation. Thenumber of events in a dataset may be limited and biased since large magnitude debris flows areinfrequent and smaller events often go undetected. Applying an empirical relationship outside of35the dataset area must be done with caution, and if so, the relationship should be from a comparablegeographic area with similar geologic conditions (Rickenmann, 2005). Using observations of pastevents to forecast the runout of future events must also be done with caution as physical conditions(e.g., triggers, process type, topography, climate) will likely change with time.From a review of 44 journal articles, conference papers, and technical reports, predictor variablesused in various empirical runout models for debris flows (not including volcanic debris flows, such aslahars) are summarized in Figure 2.8. Although the count is biased by the empirical methods includedhere and the variables measured for each study, Figure 2.8 shows the variety of variables fit to explainrunout for different datasets, with volume being the most common by far. For a discussion of themechanisms that affect debris flow mobility, including some of these variables, refer to Section 2.4.Figure 2.8. Predictor variables used in empirical debris flow runout relationships from a reviewof 44 published sources, categorized by methodology.Debris flow volume is difficult to ascertain, both forensically and for forecasting purposes,resulting in high uncertainty with predictions. Forensically, in the absence of pre- and post-eventlidar, volumes are typically estimated by multiplying an area covered by an estimated average36thickness (Legros, 2002). Deposits can be eroded or obscured by younger deposits, making eventreconstruction prone to error (Jakob, 1996). Event reconstructions in the literature do not consistentlyinclude estimates of accuracy, precision, and error (Santi, 2014). Simoni et al. (2011) describes thescatter in the volume-area relationships for a homogeneous dataset is likely attributed to volumemeasurement errors rather than differences in flow mobility. There is considerable uncertainty whenforecasting potential release volumes, and volume-based relationships are usually not practical forregional analyses.The second most common predictor was elevation loss from a landslide source due to the numberof relationships that use angle of reach. Estimating the location of the source zone can be challenging,especially if there are multiple source zones or if the debris flow volume is derived from progressiveentrainment (e.g., Hungr et al., 2008). A few studies have used different datums, such as Prochaskaet al. (2008) who defined the H/L ratio from the mid-point elevation of the drainage channel as itis more straightforward to identify. Similarly, Rickenmann (1999), Zhou et al. (2019), and othersmeasure runout from the fan apex.Due to the complexity and variability of debris flow processes, or perhaps measurement error,many of the empirical relationships show considerable scatter providing only order of magnitudeestimates (Rickenmann, 1999). Variation within a dataset can be used to establish uncertainty forprediction, using prediction intervals as a proxy for probability of runout exceedance (McDougall,2017; Mitchell et al., 2020).Since uncertainty is inherent to natural processes like debris flows, deterministic estimates ofrunout cannot be reliable. As described in Section 2.3, risk assessments usually require runoutestimates expressed as conditional probabilities. Different empirical approaches have been used toexpress debris flow runout probabilistically. Chen et al. (2007) derived different equations basedon a non-exceedance probability using the cumulative distribution of the data. Simoni et al. (2011)assigned uncertainty factors to the volume-area relationship to associate probabilistic meaning tomodel results. The model by Fannin & Wise (2001) samples from a user-defined probability densityfunction, with the number of simulations that surpass a given travel distance linked to the probabilityof exceedance. Bathurst et al. (1997) developed a logistic regression model for percentage of sediment37delivery to streams for shallow landslides that evolve into debris flows. Miller & Burnett (2008)assumed an exponential decrease in runout probability with distance travelled based on channelattributes.Many of the flow path or volume balance models were developed for sediment transport in thewatershed and may not be applicable for hazard management on developed fans. There were norelationships that directly addressed avulsion mechanisms on the fan since most of the methods areconcerned with mobility away from a source, neglecting the possibility of major lateral diversions.Schu¨rch et al. (2011a) developed a novel stochastic fan evolution model based on empiricalequations that attempts to model short and long-term fan behaviour driven by incision, aggradation,and avulsion over a sequence of events. In this model, flow volumes and sediment concentrationsare sampled from probability density functions, a flow path is routed with a flow routing algorithm,empirical rules are applied for deposition, erosion, and stopping, and the process is iterated with anupdated digital elevation model. Avulsions would be simulated when the topography is altered insuch a way that a channel is blocked and/or there is a new topographically favourable flow path to adifferent part of the fan. Preliminary results show this model can be used to highlight locations in thechannel network where avulsions are most likely to occur and what conditions are most likely to leadto avulsions (Schu¨rch et al., 2011a). One foreseen limitation is having enough data to generate inputprobability density functions and to calibrate empirical coefficients, as well as observations on realfans to validate the overall results.2.7 Challenges and Knowledge GapsDebris flows are complex, multi-phase processes with runout that varies in space and time. Ithas been well established that event volume increases mobility, but its effect on avulsion is moreequivocal; a large peak discharge is more likely to overwhelm the channel capacity but it may alsobe erosive and self-channelizing. Irrespective of volume, flow composition and path characteristicsaffect debris flow mobility and avulsions, but it is unclear which variables are the most important toconsider when forecasting runout. Although we are starting to understand avulsion mechanisms andindicators of impending avulsion, there is currently no broadly applicable, rigorous, and repeatable38way to define avulsion locations and associated probabilities.Considering variability in debris flow mobility and avulsion scenarios, runout assessments shouldbe probabilistic. Currently, most numerical models are unequipped to complete probabilistic analysesor simulate avulsions. Complex, multi-phase mathematical models may eventually simulate channelblockages, but these types of models are still in development. Empirical methods are a practicalalternative based on observations of real events. There are currently no empirical methods for debrisflows that consider flow deviation due to avulsion, and few that are probabilistic. Empirical modelvariables should be easy to obtain and measure with consistency and accuracy, however many of theexisting relationships require a total volume and initiating point to be estimated a-priori, which aregenerally uncertain.Although experimental and numerical studies can provide great insight into runout behaviourunder controlled conditions, it is important to continue to collect data on natural debris flow fans,both for model validation, but also for developing empirical relationships and expert judgement.There remains a need to collect high-quality and consistent field measurements (with estimates oferror) immediately following debris flow events to generate more reliable databases for statisticalanalysis.2.8 Summary• Debris flows are extremely rapid, surging, multi-phase mass movements down steep creeksthat exhibit dynamic feedbacks between entrainment and deposition. They exist on a wideand continuous spectrum of hydrogeomorphic processes with varying water content and flowbehaviour.• Hazard and risk assessments on fans require forecasting debris flow motion, which shouldconsider variability in mobility and the possibility of channel avulsions.• Debris flow mobility is highly correlated to volume, but flow composition and interaction withthe path have some effect too. Most generally, high volume events with high sustained porefluid pressures travelling down steep, channelized, and unobstructed paths are highly mobile.• Avulsions move deposition across a fan surface. They can be triggered by volume overwhelm-39ing the channel, superelevation of flow, and channel blockages. Despite recent advances in ourunderstanding of these processes, there is still little evidence to help forecast avulsion locationsand frequency on a fan.• Empirical methods are a practical and widely used approach to forecast runout based on directobservation and guided by knowledge of physical processes, where our understanding may belimited. There are a wide variety of empirical tools, including angle of reach and volume-areascaling relationships, that provide insight into controls on mobility. Currently, there are noempirical methods that consider both down-fan and cross-fan variability in runout trends,which is important for risk-based decision making considering variable mobility and avulsionbehaviours inherent to debris flow processes. This thesis addresses this key knowledge gap.40Chapter 3Creating a Geospatial DatasetThe geospatial dataset consists of a historical archive of spatial impact areas and flow pathsfor 30 fans in SWBC. This chapter details the process of creating this dataset, including study siteselection, geomorphic fan mapping methodologies, acquisition and processing of remote sensing data,and field methods. Methodology for calculating debris flow volumes and morphometric variablesare described, along with data summaries. Data quality classes are presented to give the readeran intuition for mapping quality, followed by a discussion of the limitations of this dataset. Thereliability of the analyses in Chapter 4 hinges on the quality and thoroughness of the data presentedhere. There remains a need to collect high-quality field and remote sensing data, both immediatelyfollowing debris flows and back through time, to continue growing rich geospatial datasets; thischapter provides some guidance on how to do so. Mapping and field photographs for each fan sitecan be found in Appendix A, and geomorphic mapping shapefiles with metadata are provided inAppendix B.3.1 Fan Site SelectionThe 30 fans in this study were selected from a larger event database compiled as part of this work,consisting of 98 fans and hundreds of events (Figure 3.1). The event database was initiated in 2017 byD. Bonneau from Queen’s University, who compiled data on events in Squamish Lillooet RegionalDistrict (SLRD) from news articles, DriveBC records, theses, publications, engineering reports,41satellite imagery, and personal observations (not published). Bonneau’s preliminary inventory wasmanually checked and filtered to exclude rock avalanches, debris avalanches, and suspected floodingprocesses, and was expanded to include recorded events from the Metro Vancouver and Fraser ValleyRegional Districts using similar sources listed above. The main references used to compile theinventory include: Hungr et al. (1984), Hungr et al. (1987), Evans & Lister (1984), VanDine (1985),Jordan (1994), Jakob (1996), Jakob et al. (1997), Jakob et al. (2016), Friele & Clague (2004), BGC(2004), Blais-Stevens & Septer (2008), and Sutton (2011). Well-documented and field mapped debrisflows at Haida Gwaii (Queen Charlotte Islands) (Fannin & Rollerson, 1993; Fannin & Wise, 2001),and debris flows recorded elsewhere in western BC (e.g., Geertsema et al., 2009) were not includedin this inventory due to their poor proximity for field work.The fan sites selected for this study are shown in Figure 3.2 and listed in Table 3.1 (coordinatespresented in NAD83 UTM Zone 10 N). The location of the anonymous case is undisclosed. Thesefan sites were manually chosen from the preliminary inventory using the following criteria:• Presence of an active debris flow fan landform with minimal fan modification and a visiblechannel.• A legacy of debris flows with impact areas on the fan visible in airphotos or satellite imagery,and/or well-preserved deposits, and/or well-documented debris flows that have been fieldmapped.42!( !(!(!(!(!(!(!( !(!(!(!( !(!(!(!( !(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!( !(!(!(!(!(!(!(!(!(!(!( !(!(!( !(!(!(!(!(!(!(!( !(!(!(!(!(!(!(!(!(!(!(!(!(!(Fraser ValleyMetro VancouverSquamish-LillooetCopyright:© 2014 Esri121° W122° W122° W123° W123° W124° W124° W125° W125° W51° N51° N50° N50° N49° N49° N0 40 8020 Kilometers±British Columbia AlbertaUSAFigure 3.1. Location map of preliminary debris flow event inventory for SWBC. BC regionaldistricts labelled for reference.43!!!!!!!!!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!((!(!(!(!(12 3784 6515282320192226272110131411122524181716299HopeSurreyLyttonSquamishWhistlerLillooetPembertonChilliwackCopyright:© 2014 Esri121° W122° W122° W123° W123° W124° W124° W50° N 50° N49° N 49° N0 40 8020 Kilometers±British Columbia AlbertaUSAFigure 3.2. Location map of fan sites in this study.Many of the fans in the study area have some form of fan modification (e.g., logging, development,roads, bridges, berms, channelization works), so this potential influence could not be fully avoided(e.g., Hope with an engineered deflection berm). Fans with large engineered barriers or fillingbasins were not included in this study. Debris flows that transport directly into a waterbody withoutmuch visible deposition on the fan, or where evidence of deposition is not preserved, were not goodcandidates for spatial impact mapping. This eliminated creeks along the Sea to Sky corridor (Blais-Stevens & Septer, 2008), with many events flowing directly into Howe Sound or filling engineered44Table 3.1. List of fan sites in this study.Label Name Easting (m) Northing (m) Key Reference1 Abandoned 461334 5558054 This thesis2 Endurance 473304 5548766 Jakob (1996)3 Terminal 475132 5549082 Jakob (1996)4 Middle Lillooet W 474226 5606122 Jakob (1996), also known asClearwater Creek5 Middle Lillooet C 475684 5605380 Jordan (1994)6 Middle Lillooet E 477147 5604529 Jordan (1994)7 Petersen 488876 5589224 Jakob (1996)8 Upper Rutherford 498423 5571312 This thesis9 No Law 500317 5570349 Jakob (1996)10 Sootip 500332 5569470 This thesis11 Lower Rutherford W 503650 5570723 This thesis12 Lower Rutherford E 503916 5570658 This thesis13 Ross 503460 5587279 This thesis14 Nightmare 503560 5587870 Jordan (1994), Jakob (1996), alsoknown as Lower Ryan15 Fergusson 515823 5625672 Jakob (1996)16 Currie B 515965 5570375 BGC (2018a)17 Currie C 516839 5570140 BGC (2018a)18 Currie D 517411 5570114 BGC (2018a)19 Deepa 523562 5588576 Jakob (1996)20 Neff 529458 5593780 Lau (2017)21 Catiline 535567 5568500 BGC (2015)22 Fern 537775 5462900 This thesis23 Bear 550167 5616375 BGC (2018b)24 Fountain S 578798 5616465 Jordan (1994), Jakob (1996), alsoknown as Gunbarrel II25 Fountain N 579160 5615932 Jordan (1994), Jakob (1996), alsoknown as Gunbarrel I26 Cheam W 594952 5451397 DriveBC27 Cheam E 595063 5451381 DriveBC28 Hope 613674 5469474 Jakob et al. (1997)29 Allard 616587 5489888 This thesis- Anonymous - - This thesis45debris basins. Creeks at the Mount Meager volcanic complex, a Quaternary volcano in the studyarea with a history of large volcanic debris flows (e.g., Friele et al., 2008), were either well-incisedinto their fan surfaces, and/or consist of large multi-process fan complexes, including fans formedby rock avalanches. Moreover, it was difficult to distinguish individual events on very active fansdue continuous overprinting of previous deposits. A debris flow creek dissecting a presumed rockavalanche deposit was also excluded since the debris flow fan was not well distinguished from therock avalanche. Fans formed under paraglacial conditions were not candidates for this study sincethe fan area is not representative of recent debris flow activity. Paraglacial fans have been exposed torapid shifts in base level and reduction of debris supply post-glaciation, causing erosion by trunkstreams, fanhead trenching, and fan dissection, leaving large portions of the fan inactive (Ryder,1971).3.2 Geomorphic Fan MappingThe following section describes the geomorphic features mapped to create this geospatial dataset.All mapping was completed in ArcGIS using an ensemble of remote sensing and field data describedin Section 3.3, including airphotos, satellite imagery, lidar, contour basemaps, and field data. Anexample is shown in Figure 3.3. Mapping at each fan site can be found in Appendix A along withshapefiles in Appendix B.46!( !(0 200100 Meters!( ApexFanWatershedImpact areaFlow path±a bFigure 3.3. Example of geomorphic mapping at Currie C using a) lidar hillshade to define theapex and fan boundary; and b) 1996 airphoto orthomosaic to delineate an impact area andflow path (20 m contours derived from lidar). 2017 ALS bare earth hillshade courtesy ofSLRD.3.2.1 Fan BoundaryFans are sloping, semi-conical landforms at the mouth of a mountain basin formed by thedeposition of sediment discharged by hydrogeomorphic processes such as floods, debris floods, anddebris flows. They are planimetrically fan-shaped with contours that bow in the down-slope direction(Bull, 1977). For geohazard risk assessments, the fan landform is often used as a zoning tool andrepresents an area susceptible to hydrogeomorphic hazard impact. The fan boundary can also beinterpreted as a statistical upper-bound of its formative processes, except in cases where the fanhas been eroded or buried. Given these descriptions, the fan boundary was mapped using the slopeand shape of contours, and to include evidence of debris flow processes such as lobes, levees, andchannels from the same sediment source. For coalescing fans, the orientations of channels and leveeswere used to differentiate sediment sources. For fans that coalesce with other landforms such asfloodplains or talus slopes, the boundary was set to differentiate the dominant geomorphic process.473.2.2 Fan ApexThe fan apex is the highest point of the fan landform. It represents a transition between mostlyconveyance in the steep and confined drainages of the watershed to the onset of deposition at thefan. In the context of hazard and risk, elements at risk are typically located below this point sincepermanent development in the watershed is less common, especially in Canada. The fan apexwas visually determined using contours and aerial imagery as the point of loss of lateral channelconfinement from the basin valley slopes.3.2.3 Impact AreaAn impact area is defined here as any area below the fan apex that has been impacted by adebris flow, or multiple debris flows, over a certain time period. The impact area is distinguishedfrom the deposit area as it represents any areas of debris flow erosion, transport, and deposition (seeschematics in Figure 3.16). For the earliest observation record, the impact area is referred to as the“baseline” and represents the area most recently impacted by debris flows. Although it is possiblethat a spectrum of hydrogeomorphic process types are present within these defined areas, they havebeen referred to as debris flow impact areas for brevity (discussed further in Section 3.8). Mappingcertainty was qualified for each impact area using classes defined in Section 3.7. In total, 176 impactareas (146 not including baseline impact areas) were mapped across 30 fans in this dataset. 110 ofthese impact areas were flagged as most likely an impact area associated with a single debris flowevent (discussed further in Section 3.7).3.2.4 Flow PathThe flow path is defined here as a line from the fan apex to the toe along either the active channel,the center of deposition if there is no defined channel, or the path of steepest descent past the toeof the deposit. If there were multiple active channels, the most active (i.e., most incised or mostnotable deposition) was selected (e.g., Figure 3.3). In the context of geohazard assessments, this linerepresents the most likely flow path of future hydrogeomorphic processes at a given time. A flowpath was defined for each of the 176 impact areas in this dataset.483.3 Data Sources3.3.1 AirphotosHundreds of historical airphotos of the fan, watershed, and surrounding areas for each study sitewere scanned from the UBC Geographic Information Centre (GIC) Airphoto Library, as summarizedin Table 3.2. The GIC airphoto archive includes over 2.5 million airphotos at various scales acrossBC dating back to 1922. An average of 15 observation points over an average span of 58 years wereavailable from airphotos. Those lacking adequate scale to discern geomorphic features, or thoseobscured by clouds or snow, were not included. Some of the airphotos previously scanned at Catiline,Bear, Cheam E, and Cheam W were shared by BGC Engineering Inc. (BGC).Airphotos were inspected for changes in erosion and deposition at the fan sites. Indicators ofdebris flow activity include the formation of levees and lobes, removal of vegetation, and changesin channelization including deepening/widening of channels, channel infilling, and migration byavulsion. Airphotos were selected for georeferencing if there was notable change since the previousphoto, or as a baseline. An example of an impact area delineated using airphotos is shown inFigure 3.3.Where possible, airphotos were orthorectified using Agisoft Metashape Professional (Metashape)(Agisoft LLC., 2019). Metashape is a commercial software that creates three-dimensional modelsfrom overlapping photographs using Structure from Motion (SfM). For this work, Metashape wasused solely to orthorectify airphotos and not for reconstructing topography. Airphoto flight lines thatwere good candidates for orthorectifying in Metashape were those with three or more overlappingairphotos over the fan area, a small enough scale for accurate tie point selection (generally lessthan 1:30,000), and a large enough scale to cover variable terrain elevations (generally greater than1:5,000). An example of an orthomosaic made at Fergusson with only 3 airphotos at a scale of1:31,860 and a Digital Elevation Model (DEM) with a 25 m pixel size is shown in Figure 3.4. Otherorthomosaics can be seen as basemap imagery in Appendix A for some of the fan sites. In total, about60 orthomosaics were created in Metashape using the following workflow, adapted from Roberti(2018):491. Scan airphotos. All airphotos were scanned at 800 dots per inch, which ensures about a 1m pixel size given the upper bound of the expected model photo scale (1:30,000), followingguidelines by Linder (2016). At least three overlapping airphotos over the fan site were used,but ideally more to include side overlap.2. Mask photos. Photo frames and inscriptions were manually removed with “Intelligent Scis-sors”.3. Align photos. Metashape generates a sparse point cloud by matching features between photos.Photos were initially aligned with the “high” resolution setting, and re-aligned with a lowersetting (less points) if the model was over-fitting and generated an unrealistic point cloud.Most photos were successfully aligned with the “medium” setting.4. Georeference. Ground Control Points (GCP) were placed using lidar and imagery instead ofbeing field-collected. In Metashape, the coordinate system was set to NAD83 UTM Zone 10N. In ArcGIS, a point shapefile was created with the same coordinate system to store GCPs.One by one, GCPs were selected in ArcGIS and corresponding markers added to the airphotosin Metashape. If available, lidar and orthophotos were used to pick GCPs, otherwise, pointswere selected with DigitalGlobe satellite imagery (tile layer by Esri, 0.5 m resolution). Findingreliable GCPs was the most challenging and time-consuming step, especially with lowerquality imagery or where the landscape has changed (e.g., forestry, construction, flooding,variable snow cover, etc.). Ideally GCPs are in flat, stable, and easily recognizable locationsthat constrain the coordinate in the x, y, and z directions (Roberti, 2018). For this work, GCPswere placed on the following features: corners of buildings or bridge abutments, road orpath intersections, powerline footings, distinct bedrock features or lineations, bedrock creekjunctions, large boulders, small ponds, or sometimes the end of a large fallen tree. Changinggeomorphic features such as channels or river banks were not used. About 2-5 GCPs werechosen per airphoto, but this varied depending on the airphoto scale and availability of reliableGCPs. GCPs were also selected to be evenly spaced across the airphotos and to cover a rangeof elevations. After GCP selection, elevations were extracted to the point file in ArcGIS froma DEM, either lidar or TRIM. The point shapefile was exported from ArcGIS to a .csv file and50imported into Metashape to associate the x,y,z coordinates to each marker ID.5. Optimize. Metashape adjusts camera alignment and updates the sparse cloud based onthe GCPs to improve accuracy. Once optimized, the GCPs with the lowest planimetricaccuracy were unselected, and the model was re-optimized to improve accuracy. This was aniterative process and sometimes additional GCPs were added, removed, or adjusted to attain aplanimetric accuracy less than about 10 m for each GCP.6. Build dense point cloud. Based on the camera positions, Metashape calculates a depth mapto generate a dense point cloud. Aggressive depth filtering was used to omit outliers, which isrecommended for aerial photography.7. Build mesh. Metashape interpolates polygons between points to create a surface. The “Heightfield” setting that interpolates along the z axis (ideal for aerial photography) and a “high”polygon count was used, per recommendations by Roberti (2018).8. Build orthomosaic. Default settings were used. Orthomosaic was exported as a .tif file foruse in ArcGIS.For scenes without sufficient overlap, airphotos were georeferenced in ArcGIS by aligning theairphoto image raster with control points. About 10 evenly distributed control points were selectedfor each airphoto using the methodology described above for Metashape (Step 4). Different transfor-mations were tested to warp the image to match the control points. Although the resolutions of theArcGIS georeferenced airphotos are higher than those of the Metashape orthomosaics, georeferencingin ArcGIS is not as reliable due to aerial photogrammetry distortions, such as varied terrain or cameratilt. It was found that the airphotos georeferenced in ArcGIS were subject to minor distortions or didnot align with the base imagery along steep slopes. Since Metashape uses photogrammetry principlesto preserve 3-dimensional scenes, the Metashape orthomosaics were more reliable as georeferencedbase imagery. However, the airphotos georeferenced in ArcGIS were still useful for observing changeand guiding mapping, in concert with other remote sensing data.51Figure 3.4. 1965 orthomosaic covering fan and watershed for Fergusson (red arrow) createdwith Metashape using three airphotos at a scale of 1:31,860 and TRIM DEM with a 25 mpixel size.52Table 3.2. Summary of airphoto and Metashape airphoto orthomosaic coverage at each fan site.Fan site(s) (Count) Airphoto years (Count) Methashape orthomosaic yearsAbandoned (11) 1947, 1964, 1967, 1976, 1980, 1981, 1986, 1990, 1994, 1999, 2005 (5) 1947, 1967, 1976, 1986, 1994Endurance andTerminal(10) 1947, 1948, 1964, 1974, 1976, 1980, 1982, 1986, 1990, 1994 (5) 1947, 1964, 1976, 1986, 1994Middle Lillooet W, C (14) 1947, 1948, 1965, 1973, 1976, 1979, 1980, 1981, 1982, 1986, 1990, 1994, 1997, 2006 (7) 1965, 1973, 1976, 1979, 1986, 1990, 1994Middle Lillooet E (15) 1947, 1948, 1962, 1965, 1973, 1976, 1979, 1980, 1981, 1982, 1986, 1990, 1994, 1997, 2006 (7) 1965, 1973, 1976, 1979, 1986, 1990, 1994Petersen (15) 1947, 1948, 1962, 1964, 1969, 1973, 1978, 1981, 1982, 1986, 1987, 1994, 2005 (4) 1948, 1964, 1981, 1994Upper Rutherford (14) 1948, 1949, 1962, 1964, 1969, 1973, 1977, 1980, 1981, 1982, 1990, 1994, 2003, 2004 (4) 1948, 1973, 1981, 1994No Law, Sootip (13) 1948, 1949, 1962, 1964, 1969, 1973, 1980, 1981, 1982, 1990, 1994, 2003, 2004 (4) 1948, 1973, 1981, 1994Lower Rutherford E,W(12) 1947, 1948, 1962, 1964, 1969, 1973, 1977, 1980, 1981, 1990, 1994, 2003 (4) 1948, 1973, 1981, 1994Ross, Nightmare (14) 1946, 1965, 1969, 1973, 1976, 1977, 1978, 1980, 1981, 1982, 1986, 1990, 1994, 2005 (4) 1969, 1973, 1980, 1986Fergusson (10) 1947, 1964, 1965, 1975, 1978, 1979, 1987, 1993, 1997, 2005 (3) 1947, 1965, 1987Currie B (18) 1946, 1947, 1948, 1949, 1958, 1962, 1964, 1969, 1971, 1973, 1974, 1977, 1980, 1981, 1986, 1990,1994, 2004(7) 1946, 1958, 1964, 1969, 1980, 1990, 1994Currie C, D (19) 1946, 1947, 1948, 1949, 1958, 1962, 1964, 1966, 1969, 1971, 1973, 1974, 1977, 1980, 1981, 1986,1990, 1994, 2004(7) 1946, 1958, 1964, 1969, 1980, 1990, 1994Deepa (14) 1946, 1947, 1958, 1965, 1969, 1973, 1974, 1977, 1981, 1988, 1990, 1994, 1997, 2005 (3) 1946, 1958, 1969Neff (17) 1946, 1947, 1962, 1965, 1967, 1969, 1974, 1980, 1981, 1987, 1988, 1990, 1993, 1994, 1997, 2005,2006(4) 1946, 1969, 1988, 1993Catiline (13) 1948, 1962, 1967, 1969, 1979, 1980, 1981, 1987, 1990, 1993, 1994, 1997, 2005 (-)Fern (13) 1940, 1953, 1963, 1967, 1968, 1979, 1980, 1982, 1991, 1993, 1995, 1996, 2009 (1) 1982Bear (12) 1947, 1948, 1951, 1964, 1965, 1969, 1975, 1987, 1992, 1997, 2004, 2005 (1) 1948Fountain N, S (13) 1948, 1959, 1964, 1965, 1966, 1967, 1975, 1987, 1992, 1993, 1995, 1997, 2004 (7) 1948, 1959, 1964, 1975, 1993, 1997, 2004Cheam W, E (29) 1928, 1947, 1953, 1954, 1959, 1961, 1963, 1966, 1968, 1974, 1975, 1977, 1978, 1979, 1980, 1981,1982, 1983, 1986, 1987, 1991, 1992, 1993, 1995, 1996, 1999, 2002, 2003, 2009(5) 1928, 1947, 1953, 1961, 1968Hope (14) 1947, 1966, 1968, 1974, 1978, 1979, 1981, 1983, 1986, 1988, 1990, 1992, 1996, 2002 (1) 1996Allard (13) 1947, 1954, 1961, 1965, 1969, 1978, 1979, 1983, 1989, 1991, 1992, 1996, 2002 (5) 1947, 1961, 1969, 1983, 1996Anonymous (10) 1957, 1959, 1963, 1966, 1976, 1982, 1989, 1992, 1996, 2009 (-)533.3.2 Satellite ImageryMultispectral satellite imagery was accessed from Planet (2019) as part of Planet’s Educationand Research Program. Weekly to monthly 5-band RapidEye Ortho Tile imagery with a 5 m pixelsize was available since 2009, and daily to weekly 4-band PlanetScope Scene imagery with a 3 mpixel size was available since 2016. Temporal resolutions varied depending on cloud cover or forestfire smoke, but monthly imagery since 2009 and weekly imagery since 2017 was reliably accessedfor the entire study area from Planet. DigitalGlobe (tile layer by Esri) imagery within the last 3-5years at a 0.5 m resolution was available for the entire study area. Satellite imagery hosted by GoogleEarth at various temporal and spatial resolutions was also used.Planet (2019) satellite images were manually inspected in Planet Explorer to identify changes indeposition. For areas of visible change, pre- and post- event multispectral tiles were downloadedfrom Planet Explorer for mapping in ArcGIS. Change in brightness due to the removal of vegetationwas often sufficient to delineate impact areas by visual inspection, but multispectral data was alsoused to supplement mapping. The normalized difference vegetation index (NDVI) is a spectral indexcommonly used for landslide detection and susceptibility mapping (e.g., Chen et al., 2018; Fiorucciet al., 2019; Martha et al., 2010; Miura, 2019; van Westen et al., 2008). NDVI is an indicator ofgreen biomass and is calculated from near infrared (NIR) and red (R) spectral reflectances (unitless)(Equation 3.1, Rouse et al., 1974; Tucker, 1979). NDVI is an index from 0 to 1 with high valuescorresponding to dense vegetation and low values to bare rock or soil. In the event that a debris flowdisturbs vegetation, a change in NDVI between n satellite images (dNDVI, Equation 3.2) may helpwith identifying an event and delineating impact extents.NDV I =NIR−RNIR+R(3.1)dNDV I = NDV In−NDV In−1 (3.2)dNDVI calculated for three events is shown in Figures 3.5-3.7 with varying levels of successfor impact area delineation. dNDVI was calculated with Planet satellite imagery in ArcGIS using54Image Analysis and Raster Math (i.e., pixel differencing). Low values (warm colours) indicate adecrease in vegetation, possibly due to the removal of vegetation or sedimentation, and high values(cool colours) indicate an increase in vegetation. Impact areas mapped using an ensemble of lidarchange detection (where available), features in post-event lidar, orthophotos, satellite imagery, and/orfield observations, are shown as a visual comparison to the dNDVI results.The 2014 debris flows at Currie C and D fans shown in Figure 3.5 are visible in the dNDVIraster since these events impacted previously forested areas and sufficiently disturbed the canopy.Figure 3.5 is an example of ideal conditions for delineating impact areas with dNDVI. Figure 3.6shows dNDVI compared to the lidar change detection results (described in Section 3.4.1) for a debrisflow at Currie D in 2019. The dNDVI raster showed impacts to areas previously vegetated, such asthe mudwave downstream of the main avulsion lobe, which was outside the limit of detection fromthe lidar change detection, but verified in the field. As expected, this method did not detect changewhere the lobe overprinted previous deposits. dNDVI greater than 0 in Figure 3.6 shows areas wherelobes are starting to revegetate on adjacent fans. Lastly, the 2017 debris flows at Cheam E and Wfans shown in Figure 3.7 were not well-delineated with dNDVI due to dense vegetation, except nearthe fan apex, which has been cleared by snow avalanches and debris, or along cutlines at the lowerfan. dNDVI may be used as a way to automatically detect events and generate event inventories(discussed further in Section 5.3), however this was not attempted because of the limited availabilityand accessibility of pre- and post-event satellite imagery from Planet.55a b cFigure 3.5. Debris flows at Currie C and D fans sometime between July and August 2014.a) RapidEye satellite imagery captured on August 7th, 2014 (Planet, 2019); b) dNDVIcalculated between July and August 2014 RapidEye bands; and c) impact areas mappedusing dNDVI results, field data, and features in post-event lidar. 2017 ALS bare earthhillshade courtesy of SLRD.a b cFigure 3.6. Debris flow at Currie D sometime between July 3 and 12, 2019. a) Planetscopesatellite imagery captured on July 20th, 2019 (Planet, 2019); b) dNDVI calculated betweenMay and July 2019 Planetscope bands; and c) results of lidar change detection between2017 and 2019 surfaces (0.3 m limit of detection), and impact area mapped using fielddata, orthophotos, and lidar change detection results. 2017 ALS bare earth hillshadecourtesy of SLRD.56a b cFigure 3.7. November 23, 2017 debris flows at Cheam E and W fans. a) Planetscope satelliteimagery captured on July 5th, 2018 (Planet, 2019); b) dNDVI calculated between 2018 and2017 Planetscope bands; and c) impact areas mapped using field data and high resolutionGoogle Earth imagery. 2017 ALS bare earth hillshade courtesy of BC MOTI.3.3.3 Lidar and OrthophotosLidar bare earth DEMs and orthophotos were available for 17 fan sites. Airborne laser scanning(ALS) lidar and orthophotos were acquired piecemeal from various agencies and institutions, assummarized in Table 3.3. Remotely piloted aerial system (RPAS, or drone) lidar and orthophotosat 3 fans were collected in September and October of 2019 using the UBC Geohazards ResearchTeam drone. RPAS lidar details are summarized in Table 3.4. The platform is a Phoenix LidarSystems MiniRanger ULS, which consists of a DJI Matrice M600 Pro drone equipped with a RieglminiVUX laser scanner, Northrop Grumman uIMU, single-antenna dual frequency GNSS receiver,and a Sony A6000 camera with 16 mm prime lens. Data processing was completed using NovAtelInertial Explorer v.8.80, Phoenix SpatialExplorer v.4.0.3, and TerraSolid v.019. A photo of RPASdata collection is shown in Figure 3.8, and an example of the final RPAS lidar product can be seen inFigure 3.9 as the bare earth hillshade.57Table 3.3. Summary of external ALS lidar data sources.Fan site(s) Year Source CoveragePavilion1 2011 BC Ministry of Transportation andInfrastructureFanCatiline 2014 Squamish Lillooet Regional District Fan and watershedAllard 2015 Canadian National Railway Fan and watershedNeff 2015 BC Hydro FanMiddle Lillooet E, W, and C 2015 University of Northern BritishColumbia, Simon Fraser UniversityFanCheam E and W, Hope 2017 BC Ministry of Transportation andInfrastructureFan and watershedCurrie B, C, and D, Bear 2017 Squamish Lillooet Regional District Fan and watershedAnonymous 2013, 2019 Metro Vancouver Fan and watershed1Fan site not part of this dataset except for volume-area relationship (Section 3.4.4).Table 3.4. Summary of RPAS lidar data collected.Fan site Dates flown Average pointdensity (pts/m2)Average pointspacing (m)Fan area covered(km2)Currie D October 1-2, 2019 6.7 0.38 0.50Fountain N October 23, 2019 8.1 0.35 0.95Fountain S September 18-19, 2019 8.1 0.35 0.44Figure 3.8. Photos of RPAS data collection at a) Currie D and b) Fountain S.58Lidar contours and hillshades were used to map impacts from debris flows, including levees,lobes, and channels, and to delineate the fan and watershed boundary. For three events in the dataset,pre- and post-event lidar was available for change detection analysis between the DEMs to estimateevent volume and impact area, described further in Section 3.4.1.In some cases, post-event lidar was used to confirm relative sequencing of recent lobe depositionbased on cross-cutting relationships and superposition of lobes. An example of lobe superposition isshown in Figure 3.9 with well-preserved deposits at the lower Fountain S fan. Approximate datesand extents were first determined with satellite imagery (described in Section 3.3.2) and topographicfeatures in the lidar DEM were used to refine mapping. The lower fan at Fountain S represents anidealized case. Further up-slope, the Fountain S fan is more channelized with many overlappingimpact areas, so sequencing is not well-preserved in the topography. Deposition at the Fountain Nlower fan is sheet-like, consisting of channelized surfaces rather than distinct lobes (Figure 3.9), sorelative sequencing using topography was not possible here.4104124144164180 20 40 60 80 100Elevation (m)Distance (m)?3943963984004020 20 40 60 80 100Elevation (m)Distance (m)???200920152018201820152009200920142015Section ASection BChannel plugFountain S lower fanlobate depositsFountain N lower fan sheet-like depositsFigure 3.9. Interpretation of deposit sequencing at Fountain S lower fan based on Planet (2019)satellite imagery and topographic features in 2019 RPAS lidar DEM.3.3.4 TRIM DEMThe Terrain Resource Information Management (TRIM) 1:20,000 DEM was used where lidarwas not available. Elevation contours with a 20 m resolution and gridded map tiles with a 25 m pixel59size were accessed from the BC Geographic Data Catalogue.3.3.5 Field DataGeomorphic field mapping was completed at 18 of the 30 fans in the dataset over the period ofJuly to November of 2018. Additional field observations at Currie D, Fountain N, and Fountain Swere made during collection of RPAS lidar in September and October of 2019. Six additional fanswere also visited in the field as part of this work, but were excluded from the final analysis, asidefrom the volume-area data at Pavilion (Section 3.4.4). Fans selected for field work were based on siteaccessibility, suspected recent debris flow activity, and priority to supplement or verify observationsmade using remote sensing data. A list of fan sites visited in the field is provided in Table 3.5. Fielddata from Catiline (BGC, 2015), Bear (BGC, 2018b), and the Anonymous site were collected andshared by BGC.The main objective of the field work was to field-truth debris flow impact areas identified inairphotos and satellite imagery. Approximately 2-6 hours were spent at each fan with priority givento traversing recent flow paths and deposit boundaries. Field work consisted of hiking the fan tothe fan apex and mapping debris flow lobes, levees, and channels. Fan boundaries, paleochannels,bedrock outcrops, and other landforms were recorded where possible. Any fan modifications weredocumented, including roads, bridges, berms, and culverts. Occasionally, the channel was hiked pastthe fan apex, but investigation of the watershed was beyond the scope of this work.Topographic basemaps, georeferenced historical airphotos, satellite imagery, and lidar hillshadeswere accessed in the field with a GPS-enabled iPad tablet using the Avenza Maps app (AvenzaSystems Inc.). Observation waypoints and tracks (e.g., Figure 3.10) were recorded with a GarminGPSMAP 64s GPS unit with a position accuracy within 5 to 10 m under normal conditions. Hundredsof photos were taken during the field work and georeferenced to the GPS track timestamp to assist inmapping. Select field photos for each fan are shown in Appendix A, as well as throughout this thesis.Where available, the following measurements and observations were made in the field:• Deposition angle, superelevation angle, channel gradient, fan slope, and bank slope with aSuunto clinometer.60• Deposit thickness, levee dimensions, high water marks or mudlines, impact scar height, runupheight, flow width, channel dimensions, and maximum cross-sections at bedrock controlledreaches with a measuring tape or Bushnell rangefinder.• Description of debris, including grain-size (D50, D90, Dmax), sorting, structure, presence oforganics, and clast lithology.The following geomorphic (landform) and sedimentologic (deposit) evidence was used to identifydebris flow processes on a fan (Costa, 1984; Giraud, 2005; Jakob, 2005; Lau, 2017; Pierson, 2004):• Lateral levees along channel margins (Figure 3.11a).• Trapezoidal to U-shaped channels with evidence of scour (Figure 3.11b).• Paleochannels indicating previous flow paths (Figure 3.11c).• Debris lobes, typically with positive relief (convex) and lobate margins (Figure 3.11d).• Local damming by log jams or boulder clusters (channel plugs) (Figure 3.11e,f).• Evidence of upstream runup, such as mud coatings and embedded gravel, on trees or otherobstacles (Figure 3.11g).• Inversely graded deposits lacking sorting or imbrication (Figure 3.11h).• Matrix-supported deposits (Figure 3.11i), open-work structure also possible.• Accumulation of coarse clasts at deposit margins or boulder studded surfaces (Figure 3.11j).• Presence of megaclasts (Figure 3.11k).• Buried logs with frayed ends (Figure 3.11l).Impact area boundaries were delineated in the field where possible, which involved placing GPSwaypoints or tracks along the edges of levees, lobes, and distal extents of muddy deposits (for recentevents) (Figure 3.12). Observations of relative age helped with mapping, although these can be quitevariable depending on the fan setting, fan activity, and climate. Given a distinguishable differencein relative age, and only where deposits have not been altered by subsequent flows, the followingobservations helped distinguish relative debris flow vintage in the field:• Relative age of vegetation established on deposit surfaces.• Degree of oxidation, moss cover, and lichen growth on deposit boulders (Figure 3.12).• Channel activity, including scour, erosion, and deposition.61Table 3.5. Summary of field work completed.Fan site Easting (m) Northing (m) Date(s) of field workMiddle Lillooet W 474226 5606122 September 8, 2018Middle Lillooet C 475684 5605380 September 8, 2018Middle Lillooet E 477147 5604529 September 7, 2018Upper Rutherford 498423 5571312 September 1, 2018No Law 500317 5570349 August 31-September 1,2018Lower Rutherford W 503650 5570723 August 29,2018Lower Rutherford E 503916 5570658 September 1, 2018Ross 503460 5587279 August 28-29, 2018Currie B 515965 5570375 August 15, 2018Currie C 516839 5570140 August 16, 2018Currie D 517411 5570114 August 16, 2018; October 1-2, 2019Neff 529458 5593780 August 7-8, 2018Fern 537775 5462900 July 29, 2018Fountain S 578798 5616465 September 23, 2018; September 18-19,2019; October 23-24, 2019Fountain N 579160 5615932 September 23, 2018; September 18-19,2019; October 23-24, 2019Cheam W 594952 5451397 September 13, 2018Cheam E 595063 5451381 September 13, 2018Anonymous - - November 4, 2018Pavilion1 589951 5634992 September 22, 20181Fan site not part of this dataset except for volume-area relationship (Section 3.4.4).!(!(!(!(!(!(!(!( !(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!( !(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!( !( !(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(!(0 500 1,000250 Meters±!( GPS waypointsGPS tracksMiddle Lillooet WMiddle Lillooet C Middle Lillooet EFigure 3.10. An example of GPS data from field traverses at three fan sites on the Lillooet River.2015 ALS bare earth lidar hillshade courtesy of Brian Menounos from the University ofNorthern British Columbia, and John Clague and Gioachino Roberti from Simon FraserUniversity.62a b cd e fg h ij k lFigure 3.11. Examples of field evidence used to identify debris flow processes and delineateimpact areas on fans. a) Lateral boulder levee; b) incised U-shaped channel; c) overgrownpaleochannel; d) terminal lobe; e) log jam; f) bouldery channel plug; g) mudline;h) inverse grading; i) matrix supported deposits; j) boulder studded debris lobe; k)megaclast; and l) logs with frayed ends.63Figure 3.12. Examples of impact area boundaries such as a) edge of a channel levee; b) recentdeposit abutting an older mossy lobe; and c) distal extent of mudwave impacts downslopeof main deposit.3.4 Estimating Event VolumesEvent volumes were estimated for 110 impact areas that were most likely from a single debrisflow event (discussed further in Section 3.7). Volumes were estimated directly for 16 events with pre-and post-event lidar (Section 3.4.1), post-event lidar topography (Section 3.4.2), and field estimatesof deposit thickness (Section 3.4.3). The direct estimates were used to develop local volume-arearelationships to estimate volumes for the remaining 96 events in the dataset (Section 3.4.4).3.4.1 Lidar Change DetectionThe highest confidence volume estimates come from sites with pre- and post-event lidar. Achange detection analysis was completed at Currie D between 2017 ALS and 2019 RPAS lidardatasets. A debris flow occurring sometime between July 3 and 12 of 2019 was identified usingsatellite imagery. Lidar and orthophotos were collected by RPAS on October 1 and 2, 2019 to capturepost-event topography for part of the fan area (Figure 3.13). Change detection between the 2017ALS and 2019 RPAS bare earth point clouds was completed in CloudCompare (2019) using theMultiscale Model to Model Cloud Comparison (M3C2) (Lague et al., 2013), with a limit of detectionof ±0.3 m based on the standard deviation of the differences between unchanged areas (Abella´net al., 2009). Negative change of 35,000 m3 and positive change of 95,000 m3 were calculated inCloudCompare between the 2.5D lidar surfaces. As shown in Figure 3.13, the muddy afterflowinundating the floodplain downstream of the main avulsion lobe was not detected in the lidar change64detection. An additional 5,000 m3 of sediment was estimated based on field observations assumingan average mud thickness of 0.3 m across the inundation area mapped with aerial imagery and fieldGPS.Volume estimates by change detection analysis were available for two other events in the dataset:BGC provided a volume estimate for the Anonymous event (pers. comm., J. Whittall, 2019), andthe M3C2 analysis by Lau (2017) was used for the 2015 event at Neff. Field estimates of depositthickness were used to verify the volume estimate by BGC, and field observations and post-eventlidar features were used to adjust the volume estimate outside of the lidar overlap presented by Lau(2017).Elevation Change (m)AvulsionMuddy afterflowabc1 2Figure 3.13. 2019 debris flow at Currie D. 1) 2019 RPAS orthophoto and 2017 ALS bare earthhillshade; and 2) M3C2 change detection analysis between 2019 RPAS point cloud and2017 ALS point clouds, showing areas of a) scour, b) lobe deposition, but not c) muddyafterflow. Change detection clipped to 2019 impact area mapped with aerial imagery andfield GPS. Representative photos of each area are shown in Figure 3.14. 2019 RPAS dataprocessed by Andrew Mitchell and 2017 ALS lidar provided by SLRD.65Figure 3.14. Post-event photos of the 2019 debris flow at Currie D for different locations alongthe flow path (locations in Figure 3.13). a) Deeply incised channel at the upper fan,showing recent scour and bank erosion; b) thick, coarse, lobe deposit from main avulsion;and c) muddy afterflow deposits on the floodplain downstream of the fan toe.3.4.2 Features in Post-Event LidarDeposit features preserved in post-event lidar topography were used to estimate volumes forthree events in the dataset. Lobe volumes were estimated with ArcGIS assuming a planar depositbase (example shown in Figure 3.15). A polygon outside the lobe boundary was draped to the DEMand interpolated with a triangulated irregular network (TIN) to approximate the pre-event topography.The TIN was converted to a DEM and subtracted from the lidar DEM to create a coarse DEMof difference (difference raster), with the sum of the difference raster pixels used to approximatethe lobe volume. The difference raster is also useful for approximating lobe or levee thicknesseswithout manually extracting cross sections. A similar approach was used by de Haas et al. (2019)to estimate lobe thickness from lidar elevation profiles assuming a planar base. There is significanterror associated with this method since a deposit base cannot be approximated with a planar surface.As a test, this method was compared to the M3C2 analysis at Currie D (Figure 3.13) using a portionof the main avulsion lobe. In this case, the interpolation method using post-event lidar topographyunderestimated the total deposit volume by 30% compared to the change detection results. Depositsat Currie D and Fountain S were good candidates for this method because of thick, well-definedlobes with minimal post-event modification. Wide deposits that require long interpolation distances,lobes deposited in a channel, and deposits that have been subsequently modified or overprinted, arenot good candidates for this method.662. Triangulate lobe boundary to approximate pre-event surface and rasterize1. Delineate lobe boundary and drape to lidar DEM3. Subtract lidar DEM from triangulated pre-event raster Figure 3.15. Workflow to approximate lobe thickness and volume with post-event lidar topog-raphy.3.4.3 Field MeasurementsField measurements were used to constrain volumes for 10 events in the dataset. Five of thesevolumes were provided by others from post-event field inspections. For the other five events presentedhere, the sum of the mean deposit thickness multiplied by a representative deposit area were used toapproximate the total volume. Field estimates of deposit thickness include measurements such asthe height of debris piled against an obstacle, the thickness of a deposit layer exposed in channelwalls, or the thickness of a lobe measured from a datum surface. Error of the volume estimate wascalculated using the lower bound and upper bound deposit thickness estimates. Since this volumecalculation is dependent on the sum of areas, there is an inherent relationship between the twovariables, introducing some circularity to the volume-area relationship.3.4.4 Volume-Area RelationshipA summary of the areas and volumes used to develop the volume-area relationship is providedin Table 3.6. The impact area is distinguished from the deposit area as it includes the entire areaimpacted on the fan below the fan apex, such as areas of erosion, transportation, and deposition(described in Section 3.2.3). Area error was estimated by multiplying the maximum cell-size of thedata or imagery by the polygon perimeter, which accounts for measurement errors due to resolution,but not epistemic errors (e.g., cases where debris flow impacts are not visible through dense canopy,67evidence of deposition is removed or modified, etc.).Volume-area relationships are shown in Figure 3.16 for debris flows in SWBC. Regressions fordeposit area (A, m2) and impact area (Ai, m2) as a function of total volume (V, m3) are provided inEquations 3.3 and 3.4, respectively. In Figure 3.17, volume-deposit area data for SWBC is comparedto other local and global published power-law relationships (previously discussed in Section 2.6.3).The 2/3 exponent is preferable for the volume-deposit area relationship (Equation 3.3) because itfollows a physical scaling relationship, and the quality of its fitting is comparable to the best-fitmodel, consistent with findings by Berti & Simoni (2007) and Griswold & Iverson (2008). Forthe volume-impact area relationship (Equation 3.4), the best-fit model is used because the physicalscaling relationship is less relevant. Equation 3.4 was used to approximate volumes for the remainingevents in the dataset since it is difficult to distinguish areas of deposition from erosion using remotesensing data. A histogram of event volumes for the resulting 110 event SWBC dataset is provided inFigure 3.18.A = 33V 2/3 (3.3)Ai = 12V 0.78 (3.4)68Table 3.6. Event volumes and areas.Fan site Event date Deposit area (m2) Impact area (m2) Volume (m3) Area mapping method(s) Volume estimation methodMiddle Lillooet C Sept or Oct 2015 31,500 ± 4,000 31,500 ± 4,000 45,000 ± 10,000 Field GPS, satellite imagery Field estimates of deposit thicknessMiddle Lillooet E Oct 2015 16,500 ± 3,400 19,700 ± 4,100 20,000 ± 5,000 Field GPS, satellite imagery Field estimates of deposit thicknessCurrie B Late fall 2016 orearly 2017304,000 ± 8,000 310,000 ± 7,900 500,000 ± 150,000 Lidar, orthophoto, field GPS Features in post-event lidar, fieldestimates of deposit thicknessCurrie D July or Aug 2014 34,200 ± 1,800 103,900 ± 6,600 70,000 ± 20,000 Lidar, satellite imagery(dNDVI), field GPSFeatures in post-event lidar, fieldestimates of deposit thicknessCurrie D Between July 3and 12, 201976,500 ± 4,000 103,900 ± 6,600 100,000 ± 5,000 Lidar, orthophoto, field GPS,satellite imagery (dNDVI),Google EarthRPAS lidar change detection, fieldestimates of deposit thickness, featuresin post-event lidarNeff Sept 20, 2015 140,500 ± 2,700 151,800 ± 4,100 220,000 ± 30,000 Lidar, orthophoto, field GPS Lidar change detection (Lau, 2017),field estimates of deposit thickness,features in post-event lidarCatiline Sept 28, 2010 26,500 ± 3,700 35,300 ± 5,100 17,500 ± 2,500 Field GPS (Cordilleran, 2010) Field estimates of deposit thickness(Cordilleran, 2010)Catiline Aug 30, 2013 24,000 ± 3,800 24,000 ± 3,800 17,500 ± 7,500 Field GPS (Cordilleran, 2013) Field estimates of deposit thickness(Cordilleran, 2013)Bear July 30, 2016 98,100 ± 5,000 100,700 ± 5,700 67,000 ± 20,000 Lidar, orthophoto, GPS (BGC,2018b), satellite imagery(dNDVI)Field estimates of deposit thickness(BGC, 2018b)Fountain S Between Aug 2and Aug 6, 201816,700 ± 3,600 23,700 ± 2,900 15,000 ± 5,000 Lidar, satellite imagery(dNDVI)Features in post-event lidarFountain S1 Between Sept 11and Sept 23, 201514,400 ± 2,000 15,000 ± 5,000 Lidar, satellite imagery(dNDVI), field GPSFeatures in post-event lidar, fieldestimates of deposit thicknessCheam W November 23,201786,000 ± 13,600 86,000 ± 13,600 65,000 ± 15,000 Field GPS, satellite imagery(dNDVI), Google EarthField estimates of deposit thicknessCheam E Nov 23, 2017 60,700 ± 10,400 60,700 ± 10,400 45,000 ± 15,000 GPS, satellite imagery(dNDVI), Google EarthField estimates of deposit thicknessHope Nov 8, 1995 29,900 ± 6,100 33,100 ± 7,200 50,000 ± 12,500 Orthorectified airphoto Field estimates of deposit thickness(Jakob et al., 1997)Anonymous - 21,800 ± 1,800 23,000 ± 2,000 15,000 ± 5,000 GPS Lidar change detection (courtesy ofBGC), field estimates of depositthicknessPavilion2 Aug 20, 2014 9,400 ± 2,200 11,000 ± 2,900 7,500 ± 1,500 Field GPS, satellite imagery Field estimates of deposit thicknessPierce Creek2 Nov 28, 1995 52,500 ± 6,800 52,500 ± 6,800 63,000 ± 15,000 Orthorectified airphoto Field estimates of deposit thickness(Jakob et al., 1997)1Upper portions of the deposit overprinted by subsequent flows, so volume and deposit area associated with lower deposition lobes. Not included in impact volume-area relationship.2Fan site not part of this dataset except for volume-area relationship.69Figure 3.16. Volume-area relationships for debris flows in SWBC for deposit area (A) andimpact area (Ai).70Figure 3.17. Comparison of SWBC volume-area data to empirical relationships for non-volcanic debris flows, over the domain of the respective volumes for each dataset.Trendline for lahars provided for reference.71Figure 3.18. Distribution of event volumes for the SWBC dataset (110).3.5 Avulsion ClassificationIn this work, an avulsion is defined as any deviation of flow from an established flow path duringa debris flow. The classification scheme in Figure 3.19 was made to describe the types of avulsions(or lack thereof) observed in the dataset, based on the location, magnitude, and surface expression ofdebris flow impacts. During the mapping process, impact areas were assigned an avulsion class (ormultiple classes), summarized in Figure 3.20. Out of the 146 impact areas (not including baselineimpact areas), 86% were classified as having some form of avulsion or spreading across the fan,while 35% corresponded to a positional shift of the active channel on the fan (classes 4 and 5). Localchannelized avulsions (class 3) were the most common, and often coincident with the other avulsiontypes. Major avulsions were more commonly initiated on the upper fan (class 5B), while lateral shiftswere relatively more common on the lower fan (class 4A). The avulsion classification serves as aqualitative way to describe fan impacts from an empirical dataset; a quantitative analysis of runoutand avulsion trends is presented in Chapter 4.72Figure 3.19. Debris flow avulsion classification scheme.73Figure 3.20. Distribution of avulsion classes (Figure 3.19) for impact areas in the SWBCdataset.3.6 Fan Site DescriptorsThis section presents the quantitative and qualitative variables used to describe the fan sites inthe dataset, including definitions, calculations, and data sources. A data summary of all the variablesis provided in Section 3.6.11.Geomorphometry is the quantitative study of topography (Pike, 2000). Geomorphometric(morphometric) variables have been used to discriminate hydrogeomorphic process types (e.g.,Wilford et al., 2004), predict debris flow activity (e.g., Bovis & Jakob, 1999), and as variables inempirical runout relationships (described in Section 2.6). Fan and watershed morphometric variableswere calculated for each fan site for two purposes: 1) to describe the morphology of the fan sitesand characterize dominant hydrogeomorphic process types; and 2) to be tested as discriminators toexplain differences in mobility and avulsion trends (Section 4.4). Only a few variables were includedin this work, in part due to the quality and variability of topographic data across the study area, butalso to focus on variables that may be related to debris flow runout trends. Morphometric variableswere selected based on results from previous studies that used morphometric variables to predictrunout or scour (e.g., Lau, 2017; Scheidl & Rickenmann, 2010), and include Melton ratio, watershed74area, fan slope, average fan channel slope, fan elevation relief ratio, and fan intersection point. Othervariables, such as measures of fan roughness, might be included in future work, as discussed inSection 5.3.In addition to the morphometric variables, three qualitative variables (i.e., classifiers) wereincluded as descriptors: source geology, fan truncation, and a relative fan activity class.3.6.1 Melton RatioThe Melton ratio (R), as defined by Melton (1965), is determined with Equation 3.5:R =Hw√Aw(3.5)where Hw is the watershed relief (km), and Aw is the planimetric watershed area (km2). TheMelton ratio is used to describe watershed ruggedness, with larger values corresponding to a morerugged (steep) watershed. The Melton ratio can be used to differentiate hydrogeomorphic processtypes, as shown in Section 3.6.7.3.6.2 Watershed AreaThe watershed area is the planimetric area of the watershed boundary upstream of the fan apex.Watershed boundaries were obtained from the Freshwater Atlas of BC (GeoBC, 2009), or calculatedusing Global MapperTM (v18) “Create Watershed” tool with lidar or TRIM DEMs, and modifiedmanually using elevation contours where necessary. Watershed shapefiles are provided in Appendix BGIS files.3.6.3 Fan SlopeThe fan slope was calculated as the slope of a straight line from the fan apex to the fan boundarythrough the fan centroid, with elevations derived from either lidar or TRIM DEMs. There aremany ways to calculate fan slope, some of which have been explored here. Slopes calculated usingfour different methods across the 30 fans are compared in Figure 3.21. The first method uses themean slope from a slope raster across the entire fan area. This method is sensitive to the DEM75resolution, with higher resolution DEMs capturing human-made cut slopes, channel sidewalls, orother geomorphic features included in the fan area. As shown in Figure 3.21, slopes calculated withthe slope raster are lower compared to the other methods, likely because of the comparatively largearea of the lower fan with gentler gradients. The second method is the slope of a straight line fromthe maximum to the minimum elevation along the fan boundary. Although this is straightforwardto compute, it may result in steep slopes along a maximum relief across the fan. To consistentlymeasure slope in the down-fan direction (i.e., radially away from the fan apex), the third methodtried was the slope of a straight line from the apex to a point along the fan toe. The downside of thismethod is it requires differentiating the toe from the rest of the fan boundary. The fourth method,using a line from the apex through the fan centroid, was used for this study because the centroid bedetermined objectively, calculations are not very sensitive to DEM resolution, and the distribution ofslopes was comparable to the third method.Figure 3.21. Box and whisker plot comparing different methods for calculating average fanslope across the fan sites. Elevations derived from TRIM DEM with a 25 m pixel size.763.6.4 Average Fan Channel SlopeThe average channel slope was calculated as the average slope between points along the lengthof the main channel from the fan apex to toe. A 25 m point sampling interval was used to matchthe coarsest DEM resolution (TRIM). Where available, the main channel was mapped with lidartopography following the guidelines for mapping flow paths described in Section 3.2.4. Otherwise,flow lines from the Freshwater Atlas of BC (GeoBC, 2009) were clipped to the fan area. As shownin Figure 3.22, the average fan channel slope is closely related to the overall fan slope, althoughtypically lower since the channel slope is the average of slope segments along a curved path.Figure 3.22. Comparison of overall fan slope to average slope along the fan channel.3.6.5 Fan Elevation Relief RatioThe elevation relief ratio (ERR) is defined by Equation 3.6 (Wood & Snell, 1960):ERR =z¯− zminzmax− zmin (3.6)where z, zmin, and zmax, are the mean, minimum, and maximum elevation (m), respectively. TheERR is mathematically equivalent to the hypsometric integral (Pike & Wilson, 1971), both of whichare used to approximate basin profile curvature. In this study, ERR is used as a very simple way toquantify the profile curvature of the fan, using elevations from the fan area instead of the watershed.77An ERR close to 0.5 corresponds to a more planar fan surface, while an ERR less than 0.5 is concave.There are other ways to quantify fan concavity, such as fitting curves to longitudinal elevation orslope profiles (e.g., Williams et al., 2006); these methods may be explored in future work with accessto higher quality topographic data across all fans in the dataset.3.6.6 Fan Intersection PointThe fan intersection point is where the main channel intersects the fan surface, usually somewheremid-fan, and represents the transition from fan incision to deposition (Hooke, 1967). The intersectionpoint is not a typical measurement used in morphometric analyses, and there is no objective wayto calculate it with topographic data alone, but it is included here to quantify channel morphologyand loss of confinement. The intersection point was defined manually using lidar DEMs and fieldobservations. In this work, the intersection point is reported as the distance from the fan apex to theaverage intersection point (usually a range) normalized by the maximum fan length (i.e., averagenormalized position down-fan). Due to the availability of lidar data and field observations, theintersection point was determined for 23 of the 30 fans.As an example, Figure 3.23 depicts how the intersection determined with a lidar DEM for CurrieD. The active channel thalweg was offset to either side of the channel outside of the levees to representthe proximal average fan surface elevation, and the elevation difference between the thalweg and theaverage fan surface (i.e., relative channel fan incision) was calculated at each point along the channel.This is similar to the methodology applied by Lau (2017) for quantifying relative fan scour. Theintersection is where the relative incision approaches zero. Representative cross-sections and photosin Figure 3.23 show fan incision for the channel up-slope of the intersection (A-A’), and depositionon top of the fan surface down-slope of the intersection (B-B’).Figure 3.24 depicts how the intersection was determined for fans without lidar using fieldobservations. In Figure 3.24, the intersection is somewhere between points b and c, although thelocation may have been artificially modified since portions of the channel were excavated upslope ofthe logging road. In all cases, judgement was used to select the representative intersection location.78100200300400500600700-10-5051015200 200 400 600 800 1000 1200 1400 1600Elevation (m asl)Relative channel fan incision (m)Distance along channel (m)Relative channel fan incisionThalweg elevationAverage fan surface (offset) elevationErosion / transport dominant2362382402420 20 40 60Elevation (m asl)Distance (m)Active channel thalwegIntersection3923943963984004024044060 20 40 60Elevation (m asl)Distance (m)A-A’B-B’Section B-B’Section A-A’OffsetsRelative channel fan incisionThalweg elev.Offset elev.Photo at A-A’Photo at B-B’IntersectionB-B’A-A’Deposition dominant(m asl)(m)(m asl)Figure 3.23. Intersection determined with lidar DEM at Currie D. 2017 ALS bare earth DEMcourtesy of SLRD.79aabbccIntersectionExcavated channel portionFan surfaceFan surfaceLobe surfaceChannel topographically confinedChannel moderately incised into fan surfaceDepositional lobes,no defined active channelBedrock outcroppingFigure 3.24. Intersection determined with field observations at Ross. 20 m contours derivedfrom TRIM DEM. Sketches are based on field cross-sections and are not to scale.3.6.7 Hydrogeomorphic Process RecognitionMorphometric variables were used to examine the dominant hydrogeomorphic processes for thefan sites using process boundaries by Wilford et al. (2004), Bardou (2002), and Bertrand et al. (2013)(Figure 3.25). The boundary by Bertrand et al. (2013) in Figure 3.25 discriminates debris flow andfluvial fans based on a statistical analysis of 620 catchments from a global dataset. In general, small,steep basins with higher Melton ratios are more typical of debris flow processes. Debris flows are thedominant classified flow type for most of the fan sites, with lesser debris flood/mixed-process fans.As described in Section 2.1, hydrogeomorphic processes exist on a spectrum, and all fans may beprone to all process types.80Figure 3.25. Fan sites plotted on typical hydrogeomorphic process recognition charts withboundaries by (left) Wilford et al. (2004) and (right) Bardou (2002) and Bertrand et al.(2013). Fan site labels correspond to Table 3.1.3.6.8 Source GeologySource (basin) geology was subdivided into four groups based on rock classes from provincialdigital bedrock geology mapping at a scale of 1:50,000 to 1:250,000 (Cui et al., 2017): intrusive,sedimentary, volcanic, or metamorphic. For basins with multiple rock classes, classification wasbased on the geologic unit with the largest area. Basins were also classified as either supply-unlimited(i.e., transport-limited, almost an unlimited amount of sediment available), or supply-limited (i.e.,weathering-limited, time is required to recharge channels with debris before the next event), asdefined by Jakob (1996) and Bovis & Jakob (1999). Since the intrusive basins were classified assupply-limited due competent bedrock limiting sediment delivery, the source geology classifier isconsidered an adequate proxy for supply conditions in this study.3.6.9 Fan TruncationFans in the SWBC dataset are situated in various geomorphic settings that may impact debrisflow runout. As a preliminary way to describe the interactions of debris flows with geomorphic81processes beyond the fan extents, fans were classified as truncated if a body of water, such as a riveror lake, abuts the fan toe. Erosion at the fan boundary may undersize the fan area, and evidence ofimpact is lost as debris flows enter the water body. Fans with mostly unconstrained deposition onto afloodplain or terrace, or where there is minimal geomorphic interaction with the body of water, areconsidered not truncated.3.6.10 Fan ActivityRelative fan activity varies across the fan sites. The number of impact areas (not including thebaseline) recorded at each fan ranges from 1 to 14, with an average of about 5 impact areas per fanover an average observation length of 74 years. The most active fans in the dataset are the ones atMount Currie (Currie B, C and D) and Fountain Ridge (Fountain N and S), described further inSection 4.3. The impact recurrence period for each fan was calculated by dividing the length of theobservation record by the number of mapped impact areas. This number is related to the numberof available airphotos, the frequency of visible change in the airphoto and satellite record, and thenumber of debris flow events mapped in the field, and is not necessarily equivalent to the return periodof debris flow activity at each fan. The average recurrence interval of mapped hydrogeomorphicchange across the fan sites is about 24 years.3.6.11 Summary of VariablesHistograms summarizing the distribution of quantitative and qualitative variables describing thefan sites are shown in Figure 3.26. Data for each fan site is summarized in Appendix A, and can beaccessed digitally as shapefile metadata provided in Appendix B.82Figure 3.26. Distributions of morphometric and qualitative data describing the SWBC fan sites.3.7 Mapping Certainty ClassificationThe certainty (i.e., quality) of impact area mapping was given a qualitative rating using the classesdescribed in Table 3.7. The temporal certainty class reflects the observation frequency interval, and issomewhat related to the likelihood that the mapped impact area is from a single event. In other words,impact areas with dates that are approximately constrained are more likely (but not always) to be frommultiple events, especially at very active fans. However, it is possible that an event constrained to a83Table 3.7. Description of impact area mapping certainty classes.Class Description Temporal Certainty Spatial Certainty1 WellconstrainedEye witness account or eventrecorded; date constrained withimagery to within 1 year; dateverified by other dating means.Recent event with well-preserved deposits or flowmarkers verified in the field; pre- and post-event lidaror deposits visible in lidar; high quality pre- andpost-event satellite imagery or orthophotos; sparsecanopy or canopy removed by event.2 ModeratelyconstrainedDate constrained with imageryto within 10 years.Less recent event with moderately preserved fieldevidence; some deposits or channels visible in lidar;moderate to high quality pre- and post- event satelliteimagery or orthophotos; parts of the impact areaobscured by canopy or fan disturbance.3 ApproximatelyconstrainedDate constrained with imageryto within 10-20 years.Impact area mapped using only airphotos or satelliteimagery; moderate quality pre- and post- eventsatellite imagery or orthophotos; lidar unavailable ordeposits obscured by subsequent events; impact areaobscured by canopy or fan disturbance.year with satellite imagery is the sum of a series of debris flows, or that an impact area constrainedto a 20 year interval is from a single event. 110 of the 176 impact areas were flagged to be mostlikely associated with a single debris flow event. This distinction was made on a case-by-case basisconsidering the temporal certainty class and the relative activity of the fan.The spatial certainty represents confidence in the impact area mapping, dictated by the qualityof remote sensing data and field data. High volume events that disturbed large portions of the fansurface and were easy to observe in imagery, or those field mapped shortly after occurring, wereconsidered well constrained. Figure 3.27 shows the number of impact areas for each class. Thereis an inverse distribution in the data certainty, with a lesser proportion of well spatially constrainedimpact areas compared to their temporal constraint. This is due, in part, to the nature of the imagerysources. Satellite data (Section 3.3.2) provides high temporal resolutions (weekly to monthly scenes)but at lower spatial resolutions (3-5 m pixel sizes), compared to the lesser amount of high resolutionairphotos (Section 3.3.1) with inconsistent observation periods (about 1-20 years between photos).Out of the entire dataset, only 20 impact areas had both a temporal and spatial certainty class of1. Data certainty classes for each impact area can be found as metadata with the GIS shapefilesprovided in Appendix B.84Figure 3.27. Distribution of data certainty classes for SWBC impact areas.3.8 Dataset LimitationsThe compiled dataset has limitations that should be considered throughout this thesis. Due to theensemble of data sources used, mapping was completed at varying levels of confidence, as addressedby the data certainty classification defined in Section 3.7. The monitoring interval was not constantthrough time or consistent between fans, and it remains uncertain whether many of the impact areasare from multiple flows or a single event. There is data censoring since small channel-pluggingevents not visible in imagery or mappable in the field are underrepresented, although repeated RPASlidar campaigns may fill this gap in the future. Since the airphoto record spans less than a century,the chances of capturing an extreme event on each fan are very low. Events that overprint previousdeposits (e.g., Figure 3.6) or flow under dense canopy (e.g., Figure 3.7) are generally undetected inairphotos and satellite imagery. Distal debris flow impacts, such as the full extents of sheet floodingor mudwave deposits beyond the main deposit lobes (e.g., Figure 3.12c) may only be detected withhigh quality post-event satellite imagery or timely post-event field investigations. Post-depositionalprocesses including damage cleanup or subsequent storm events may obfuscate evidence. Forestryroad building, clear cutting, and snow avalanches may also obscure mapping. The spatial accuracyof geomorphic mapping is controlled by many factors, including the availability of lidar, accuracy of85the georeferencing process, pixel size of the imagery, and the accuracy of a handheld GPS. Althoughgeomorphic mapping is subject to interpretation, effort was made to employ a consistent methodology,as described in Section 3.2.Sometimes impacts from debris flows are difficult to distinguish from other hydrogeomorphicprocess types using imagery alone, in the absence of obvious lobes and levees. As shown inSection 3.6.7, morphometric variables were used to help distinguish the dominant hydrogeomorphicprocess types for the fans in this dataset, with most fans clustered close to debris flow processes.Trimlines or landslide scars were sometimes, but very rarely, identified in the watershed as an indicatorof a possible debris flow initiating processes, and only some process types were field-truthed usingthe criteria listed in Section 3.3.5. Moreover, a spectrum of sediment concentrations and processtypes may be present within a single event or surge, as discussed in Section 2.1. Considering processtype uncertainties, and whether a mapped change is from a single event or many, the impact areasrepresent the migration of debris flow-dominant hydrogeomorphic processes on debris flow fans.There are other methods to reconstruct fan history for frequency-magnitude analysis, as sum-marized by Jakob (2005, 2013). Test trenching and borehole drilling, along with tephrachronologyand radiocarbon dating methods, can determine a chronology of events and associated thicknessdating back to the early fan record. Unfortunately, reconstructing 3-dimensional fan architecture isexpensive, time consuming, and invasive. Dendrochronology can be used to reconstruct areas par-tially affected by debris flow inundation by dating tree ring growth reactions to external disturbances(Schneuwly-Bollschweiler & Stoffel, 2013). It is a less intensive method and provides a somewhatcontinuous record of debris flow activity over a few hundred years, given adequate conditions (Jakob,2013). Lichenometry (e.g., Jomelli, 2013), cosmogenic radionuclides (e.g., Ivy-Ochs et al., 2013),and weathering fractures in boulders (D’Arcy et al., 2015), have also been used to date landforms.Although these methods provide a more continuous and complete data record for each fan, they lackthe spatial detail awarded by airphoto analysis, lidar interpretation, and post-event field investigationsfor more recent flows, such as the delineation of flow paths and distal debris flow impacts.As discussed in Section 3.4, there is a lot of uncertainty associated with estimating event volumes.Only three events in the dataset have pre- and post-event lidar available for a change detection analysis.86The volume-impact area relationship was based on 16 direct measurements, many of which involveusing an area to estimate the volume assuming a representative thickness, introducing circularity inthe relationship. To address these uncertainties, methodologies used to estimate volume have beendescribed and estimates of error have been provided. In this study, the volumes are only used toplace events in order-of-magnitude volume classes, and therefore the use of the volume-impact arearelationship for the remaining 96 events is reasonable since the 95% prediction interval spans abouthalf an order of magnitude (Figure 3.16).Lastly, high quality topographic data was not available across the entire study area, and manymorphometric variable calculations were completed with a 25 m resolution DEM. Therefore, onlya few simple morphometric variables that were relatively insensitive to topographic resolution andeasy to calculate have been included.3.9 Summary• 176 debris flow impact areas and flow paths were mapped across 30 fan sites dating back to1928, with 110 impact areas likely associated with a single debris flow event. The averagerecurrence interval of mapped hydrogeomorphic change across the fan sites is about 24 years.• 30 fan sites were selected from a larger debris flow event inventory compiled for SWBC basedon the presence of a well-defined fan landform and a legacy of mappable debris flow events.• Geomorphic mapping was completed in ArcGIS using an ensemble of remote sensing dataand field data to define the fan boundary, fan apex, impact areas, and flow paths. The impactarea is defined as any area below the fan apex that has been impacted by a debris flow, ormultiple debris flows, over a certain time period. These areas represent the migration ofhydrogeomorphic processes on debris flow fans.• Remote sensing data consisted of hundreds of historical airphotos, satellite imagery, topo-graphic basemaps, lidar, and orthophotos. About 60 airphoto scenes were orthorectified usingAgisoft Metashape Professional (Agisoft LLC., 2019). Change detection with the spectralindex NDVI (Normalized Difference Vegetation Index) was used for impact area mapping withsatellite data, which consisted of daily to monthly images dating back to 2009 available through87Planet (2019). Lidar was available for 16 of the fan sites, including lidar and orthophotoscollected during field work using a remotely piloted aerial system (RPAS, or drone) at three ofthe fan sites.• Geomorphic field mapping was completed at 18 fan sites, in which the fan was hiked tothe apex to delineate debris flow features, including lobes, levees, deposit boundaries, andchannels.• Event volumes were calculated for 16 events using lidar change detection, features in post-eventlidar, and field data. This data was used to develop a local volume-impact area relationship toestimate volumes for the remaining 96 events in the dataset.• A classification scheme was developed to describe the different types of avulsions (or lackthereof) based on the location, magnitude, and surface expression of debris flow impacts. 86%of the impact areas mapped involved some form of avulsion or spreading across the fan, while35% corresponded to a positional shift of the active channel.• The fan sites in the dataset were described by morphometric variables (Melton ratio, watershedarea, fan slope, average fan channel slope, fan elevation relief ratio, and fan intersection point),source geology, truncation by a body of water, and fan activity. These variables will be used inSection 4.4 to stratify runout trends.• Despite a dataset with varying levels of spatial and temporal accuracy, the compiled data is aunique and thorough record of spatial impacts on debris flow fans in SWBC.88Chapter 4Extraction and Analysis of SpatialImpact Trends on Debris Flow FansThis chapter is an analysis of the geospatial dataset described in Chapter 3. A novel methodto extract and aggregate runout trends across fans is presented, creating fan-normalized spatialimpact heatmaps. Using this method, regional spatial impact trends for the SWBC dataset aredeveloped and compared to other data. Mobility and avulsion trends at five very active fans in thedataset are described. Using the entire SWBC dataset, differences in runout distributions grouped byevent volume, morphometric variables, and other descriptors, are tested to examine which variablesdiscriminate different mobility and avulsion trends. Lastly, a simple tool that transposes the empiricaldata onto another fan is presented, showing potential applications for data-driven runout assessments.4.1 Creating Fan-Normalized Spatial Impact HeatmapsA new graphical method is presented to extract and visualize spatial impact trends on debris flowfans. This method builds on metrics proposed by Densmore et al. (2019), where avulsion size isquantified by the opening angle of the avulsion and the radial distance of the avulsion site, as wellas techniques used by de Haas et al. (2018a) to summarize deposition patterns over time based onrunout distance and flow angle measured from the fan apex.Impact areas are extracted using a circular grid centered on the fan apex and normalized by89the maximum fan length and fan arc length. The maximum fan dimensions can be interpreted asstatistical upper-bounds of its formative debris flows. In the measurement grid, zones of increasingradii represent mobility down-fan, and arc length offsets represent lateral shifts across the fan relativeto two datums: the fan axis (Section 4.1.1) or the previous flow path (Section 4.1.2). Multiplenormalized impact area plots are summed to create a heatmap for a fan or a group of fans. The dataextraction process is described in more detail in the code workflow in Section 4.1.3.4.1.1 Relative to the Fan AxisThe fan axis is a line bisecting the fan through the apex and the fan centroid. Although thefan axis is determined planimetrically, it can be interpreted as a reasonable proxy for the overalldepocenter of the fan landform (considered here as the location of maximum deposit thickness) sincecross-fan profiles tend to be convex (e.g., Blair & McPherson, 1998; Whipple & Dunne, 1992). Asshown in Figure 4.1, the fan axis is used as a datum for measuring cross-fan offsets for individualimpact area polygons. The overlapping impact area polygons are summed to create a heatmap, with“hotspots” used to identify areas on an individual fan most impacted based on the data record.Figure 4.1. Example of the fan-normalized plotting method for one impact area relative to thefan axis.4.1.2 Relative to the Previous Flow PathFor each impact area, a flow path was defined from apex to toe using imagery and topographyas either the active channel, or the center of deposition if there is no defined channel, or the path of90steepest descent past the toe of the deposit. As shown in Figure 4.2, the impact area plotted relativeto the previous flow path highlights flow path deviations, including locations and extents of avulsions.A spatio-temporal component is captured since the datum is relative to impacts from a previoustime. Using the impact area in Figure 4.2 as an example, the flow avulsed from the channel about athird of the way down-fan, extending across almost half of the maximum fan arc length. Avulsiontrends are thus recorded in these plots, which are typically difficult to ascertain when looking atimpact area mapping in GIS, and tedious to measure manually for many events. When impact areasare summed creating a heatmap, hotspots represent areas most likely impacted relative to a currentchannel configuration based on historical data. Interpretations and applications of these plots arediscussed throughout this chapter.Figure 4.2. Example of the fan-normalized plotting method for one impact area relative to theprevious flow path.4.1.3 Code WorkflowThe following steps briefly describe the code workflow for converting GIS mapping to fan-normalized spatial impact heatmaps for a site. The code was implemented in MATLAB (R2019b).1. Load shapefiles. For a fan site with i impact areas, the fan apex (1 point), fan boundary (1polygon), impact areas (i polygons, ordered sequentially), and flow paths (i polylines, orderedsequentially) are loaded as separate shapefiles in UTM coordinates (Figure 4.3). The fanpolygon, impact area polygons, and flow path polylines should intersect the fan apex.91Figure 4.3. Example shapefile inputs for plotting.2. Initialize a measurement grid centered on the fan apex. The grid resolution is specified bythe number of nodes down the maximum fan length (x′) and across the maximum fan angle (y′).A grid resolution of 500 was used in this study. The grid is sized to span twice the length of thefan in the radial component to measure runout past the fan toe, x ∈ {0,2x′}, and 360 degreesin the angular component to capture all fan orientations, y ∈ {0,360}. The number of radial(n) and angular (m) increments in the grid are calculated using the specified grid resolution,and the grid nodes are stored in an m×n array. Grid dimensions are shown in Figure 4.4. They component can be converted from degrees to arc lengths using the angle and radius at eachnode (equation in Figure 4.4c). Due east was set as an arbitrary datum for 0 degrees.Figure 4.4. Initializing measurement grid. a) Normalizing fan dimensions; b) circular measure-ment grid with n radial increments in the x dimension, and m angular increments in the ydimension; and c) grid nodes stored in an m×n array.3. Intersect impact area shapefiles with grid. Arrays are populated with ones and zeros toindicate if the grid node intersects the shapefile, forming the z dimension. An example for oneimpact area is shown in Figure 4.5a.924. Reorder and sum impact area arrays:• Relative to fan axis. Columns are reordered relative to the bearing of a line connectingthe fan apex to the fan centroid (Figure 4.5b).• Relative to previous flow path. The bearing to each node along a flow path (n) for eachflow path (i) is stored in an i×n array. For each impact area array (i), each column (n) isreordered relative to the bearing along the previous flow path (i−1) (Figure 4.5c).Figure 4.5. Reshaping and summing impact area arrays. a) Intersection of an impact area withthe measurement grid; b) re-ordering impact area array relative to fan axis and summing;and c) re-ordering impact area array relative to previous flow path (i−1) and summing.5. Plot summed arrays on normalized axes. The x component is normalized by the maximumfan length (x′), calculated as the maximum planimetric distance from the apex to a point on thefan boundary. The y component is normalized by the maximum fan arc length (y′), calculatedas the maximum arc length intersecting the fan along the measurement grid. Normalizingfan dimensions are shown in Figure 4.4a. The z summation arrays are re-sampled over the93normalized x ∈ {0,2} and y ∈ {−1,1} vectors at the specified grid resolution to reduce thesize of the array (500 used in this study). Fan-normalized spatial impact heatmaps (summationplots) for one fan are shown in Figure 4.6.Figure 4.6. Examples of summed and normalized impact area plots for one fan site, relative tothe (left) fan axis and (right) previous flow path.4.1.4 LimitationsAside from mapping and data record uncertainties (described in Section 3.8), the main limitationsassociated with the fan-normalized spatial impact heatmaps are related to normalization and referencedatums. The fan boundary is an imperfect normalizer because fans truncated by valleys, rivers, glacialfeatures, and/or coalescing fans would be undersized, whereas fans formed largely under paraglacialconditions may be oversized, potentially skewing trends (although paraglacial fans were not includedin this dataset). Not all fans are idealistic semi-conical shapes, and in some cases, normalizingdimensions reflect external geomorphic and topographic constraints rather than debris flow runouttrends. Topography is not considered since all measurements are planimetric. This method is notideal for fans with bifurcating channels or flows with multiple flow paths since measurements fromthe previous flow path are relative to a single line. The heatmaps treat all impacts as equal; there isno differentiation between impact energy based on flow thickness, composition, or speed. Futurework incorporating flow intensity is discussed in Section 5.3. Lastly, different fan environments(i.e., geology, climate, topography, fan morphology) may preclude aggregating impact areas acrossmultiple fans, which is made possible with normalized axes. As with any empirical method, discretionis required when interpreting regional heatmap aggregates.944.2 Regional Spatial Impact Trends on Fans in Southwestern BritishColumbia4.2.1 Spatial Impact HeatmapsSpatial impact heatmaps comprised of 176 impact areas across 30 fans in SWBC are shown inFigure 4.7, relative to the fan axis (a,b) and relative to the previous flow path (c,d). The heatmaprelative to the fan axis shows the variety (and chaos) of impacts and flow paths across the fan space.The heatmap relative to the previous flow path shows how most impact areas follow the previousflow path, with some deviations from avulsion and/or lateral spreading. Plots b and d in Figure 4.7are not normalized by the fan dimensions, and preserve scaling. From the fan apex, almost all debrisflows impact within ±60◦ relative to the fan axis or previous flow path.Overall, the heatmaps in Figure 4.7 show a decay in the fraction of impacted areas away from thefan apex, and for the plots relative to the previous flow path, a decay away from the active channel.Figure 4.8 shows smoothing of the fan-normalized heatmap (surface) using filters in Surfer ® (GoldenSoftware, LLC, 2018). The plots in Figure 4.8 are an oblique view of a symmetrical version ofFigure 4.7c, where directionality was removed by transposing impacted grid cells along the y = 0axis. The general shape of the surfaces in Figure 4.8 can be interpreted as a bivariate empiricalcumulative runout exceedance distribution function, representing the fraction of events exceeding acertain distance relative to an active channel. Isolines extracted from the fan-normalized heatmaps(i.e., draped incrementally along both axes of the fan-normalized spatial impact surface) are shownin Figure 4.9.95Figure 4.7. Regional debris flow spatial impact heatmaps for SWBC based on 176 mappedimpact areas across 30 fans. a) Fan-normalized, arc lengths measured relative to the fanaxis; b) unnormalized, arc lengths measured relative to the fan axis; c) fan-normalized,arc lengths measured relative to the previous flow path; and d) unnormalized, arc lengthsmeasured relative to the previous flow path.96Figure 4.8. Smoothing of the fan-normalized spatial impact surface (heatmap relative to theprevious flow path, non-directional) using filters in Surfer ® (Golden Software, LLC,2018). a) Raw data; and b) 3 passes of a 5×5 maximum value filter and 10 passes of aGaussian low-pass filter.97Figure 4.9. Isolines extracted from regional fan-normalized spatial impact heatmap. a,b) Rawdata; and c,d) data smoothed in MATLAB using LOWESS (locally weighted scatterplotsmoothing).4.2.2 Maximum Runout DistributionsDistributions of maximum runout in the down-fan (x) and cross-fan (y) components were extractedfrom the spatial impact heatmaps relative to the previous flow path. Probability density functions(pdf) and empirical cumulative runout exceedance distribution functions (ecdf’, or cumulative runoutexceedance for short) are shown in Figure 4.10 (fan-normalized) and Figure 4.11 (unnormalized)98for the SWBC dataset. Runout exceedances are presented instead of a non-exceedance probabilitytypical of an empirical cumulative distribution function (ecdf) due to the applicability in hazardand risk calculations, specifically, the probability that a hazard will reach the element at risk (seeSection 2.3). The ecdf’ (also known as the complementary cumulative distribution function) is equalto 1−ecdf.Statistical distributions were fit to the data using the maximum likelihood estimation method andselected based on goodness of fit metrics (e.g., Akaike information criterion, Bayesian informationcriterion). Fan-normalized distributions (Figure 4.10) follow normal and log-normal distributions inthe down-fan and cross-fan components, respectively. Although the logistic distribution marginallyout-performed the normal distribution for maximum runout in the down-fan component, the normaldistribution is presented here due to similarities with the log-normal distribution, such as calculatingthe location (mean, µ) or scale (standard deviation, σ) parameters. The unnormalized distributionsshown in Figure 4.11 generally follow a similar distribution to the fan-normalized data, althoughthe maximum runout in the down-fan component is better represented by a Gamma distributionwith heavier tails and a positive skew. Based on the SWBC dataset, about 90% of the debris flowsimpacted past 50% of the maximum length down the fan, while less than 10% avulsed beyond 50%of the maximum arc length across the fan.Since the cumulative runout exceedance curves are projections of the maximum runout inthe down-fan and cross-fan dimensions, the probabilities cannot be directly combined for a spatialprobability of impact at a location on a fan. Instead, a two-dimensional cumulative runout exceedancedistribution can be represented by either the heatmap in Figure 4.8, the isolines in Figure 4.9, ora bivariate distribution fit to the data in Figure 4.12. An example of how the fan-normalizedspatial impact heatmaps can be transformed for use in hazard and risk calculations is described inSection 4.5.1.There is a steep reduction in impact areas with runout recorded past the maximum fan length(x = 1), as shown by the normalized pdf for the down-fan component in Figure 4.10. The paucity ofdebris flow impacts here might be a result of fan truncation by a water body, such as a lake or river,or another topographic or physical obstacle. Debris flows may have runout further if unconstrained99by these features, with a hypothetical distribution resembling the bell-shaped tail of the normaldistribution (implications of fan truncation is discussed further in Section 4.4.9). Data censoring mayalso be present at the upper tail of the distribution (closer to the fan apex, 0 < x < 0.4). Smaller, lessmobile events that mostly deposit in the upper-fan channel are not discernible in aerial imagery, andthus would not be captured in the data record. Similarly, debris flows that remained in the activechannel may be underrepresented in the cross-fan distribution. Without data censoring, the peak ofthe cross-fan distribution in Figure 4.10 might be closer to the previous flow path (y = 0), with ahigher density of impact areas between 0 < y < 0.1.There is a slight positive covariance between the two orthogonal runout dimensions (Figure 4.12),indicating that debris flows that deviate from the previous flow path also typically have longer runoutsin the down-fan direction. The covariance is partly explained because arc lengths are calculated witha radius (i.e., the position down-fan), but there are many impact areas where the points of maximumrunout for each dimension are at different locations, such as those with bifurcating flow paths or avariable spreading width. Figure 4.12 provides evidence that debris flow mobility and avulsion mightbe related. A possible physical interpretation for this relationship is related to event volume, in whichlarge magnitude events with sufficient energy travel long distances down-fan (e.g., Corominas, 1996)are also likely to avulse the channel (e.g., de Haas et al., 2018b). The relationship between volumeand runout is discussed further in Section 4.4.100Figure 4.10. Fan-normalized maximum runout distributions for the down-fan (left) and cross-fan (right) components based on 30 fans in SWBC.101Figure 4.11. Maximum runout distributions for the down-fan (left) and cross-fan (right) com-ponents based on 30 fans in SWBC.102Figure 4.12. Relationship between maximum runout in the down-fan (x) and cross-fan (y)components, with normalized (left) and unnormalized (right) scales. Eigenvectors of thecovariance matrix (Σ) scaled by the respective eigenvalue are plotted in red.4.2.3 Comparison to an External Case Study: Kamikamihori Fan, JapanThe regional SWBC distributions were compared to runout distributions at the Kamikamihori fan,a well-studied and monitored debris flow creek on the slopes of the active volcano Mount Yakedakein the northern Japanese Alps (e.g., Suwa et al. 2009; Suwa et al. 2011; de Haas et al. 2018a). Overten debris flows per year occurred in the decade following the last phreatic eruption in 1962, and anobservation station installed in 1970 has recorded data from upwards of 91 debris flow events (Suwaet al., 2011). Although the geologic conditions differ from the SWBC dataset, the Kamikamihori fanwas chosen because deposit extents have been mapped following each event since 1978.The Kamikamihori fan dataset consists of 17 events over a span of 27 years. Depositional historyfrom 1978 to 2005 compiled by de Haas et al. (2018a) was manually georeferenced (Figure 4.13) andthe polygons converted into fan-normalized spatial impact heatmaps (Figure 4.14). Lobe sequencescorresponding to a single event were combined into one impact area, and boundaries were extendedto the fan apex following contours or channels, consistent with impact area mapping for the SWBCdataset. Since aerial photos, satellite images, or field data were not available for these events, flowpaths were approximated as the center of the impact area path in the down-fan direction. Theapproximate fan area and apex were estimated using contour maps and satellite imagery.103Sources: Esri, HERE, Garmin, Intermap, increment P Corp., GEBCO, USGS,FAO, NPS, NRCAN, GeoBase, IGN, Kadaster NL, Ordnance Survey, Esri Japan,METI, Esri China (Hong Kong), swisstopo, © OpenStreetMap contributors, andthe GIS User Community0 200 400100 Meters±impact_area20052004200219991998199719961995198819851983198019791978I t rFigure 4.13. Depositional history at the Kamikamihori fan, Japan. Impact areas adapted andmapped using figures compiled by de Haas et al. (2018a).Figure 4.14. Fan-normalized spatial impact heatmaps a) relative to the fan axis; and b) relativeto the previous flow path, for the Kamikamihori fan based on 17 impact areas.104Figure 4.15 shows the maximum runout distributions at the Kamikamihori fan (blue) comparedto the entire SWBC dataset (grey). The distribution of normalized runout in the cross-fan componentis similar to the SWBC dataset, however, there is a clear difference in the down-fan component; amuch higher proportion of debris flows recorded at Kamikamihori terminate closer to the fan apex.As summarized by de Haas et al. (2018a), the deposition on the Kamikamihori fan follows patternsof channel plugging, backstepping, and avulsion, with successive deposits migrating up-fan untila flow of sufficient magnitude initiates a large avulsion. An example of this pattern is the four lowmobility channel-blocking events from 1985 to 1996, followed by a relatively large debris flow in1997 diverting flow from the south to the north side of the fan. Debris flow lobes at the Kamikamihorifan have been distinguished into two groups: “swollen” (steep bouldery fronts, clast-supported,fan-proximal) and “flat” (thin deposits lacking steep fronts, matrix-supported) (Suwa et al. 2009;Suwa et al. 2011). The relative proportion of these two groups is interpreted to be reflected in thetwo peaks of the down-fan pdf near 0.2 and 0.8 of the normalized fan length (Figure 4.15). Channel-blocking lobes with steep fronts and open-work structure have also been observed at the SWBC fans(e.g., Figure 4.16), however, these localized channel-blocking events are difficult to distinguish inaerial imagery compared to large magnitude events and/or major avulsions. Futhermore, some ofthe impact areas from the SWBC dataset might be the sum of a few smaller events, the runouts ofwhich would not be documented. The characteristic down-fan runout exceedance curve with a higherproportion of fan-proximal deposits afforded by a continuously monitored debris flow channel mayprovide justification for adjusting the upper tail of the regional SWBC distribution.105Figure 4.15. Comparison of maximum runout distributions in the down-fan and cross-fancomponents for the Kamikamihori fan (17 impact areas) to the regional SWBC dataset(176 impact areas, 30 fans).106Figure 4.16. Steep, bouldery, clast-supported, deposit front plugging the channel on the prox-imal fan at Currie D (evidence of localized debris flow impacts not visible in aerialimagery).4.2.4 Comparison to Conceptual Avulsion ScenariosIn a recent study of the spatio-temporal evolution of debris flow fans, de Haas et al. (2018a)postulated conceptual avulsion patterns, as observed on fans from around the world. Figure 4.17by de Haas et al. (2018a) illustrates conceptually the influence of flow volume sequencing and fantopography on runout and avulsion patterns. de Haas et al. (2018a) presented three scenarios: a)backstepping from a sequence of smaller flows followed by an avulsion during a large flow; b)avulsion through multiple channels, with the most topographically favourable flow path forming themain channel, followed by progressive backfilling of side channels; and c) gradual lateral shiftingtowards a topographic low during a sequence of similar-sized flows.For illustrative purposes, fan-normalized cumulative runout exceedance distributions were ex-tracted for the three conceptual scenarios in Figure 4.17 using the methods described in Section 4.1.Conceptual cumulative runout exceedance distributions in both down-fan and cross-fan componentsare shown in Figure 4.18 compared to empirical data. The conceptual distributions appear as stepfunctions since only one cycle with three events per scenario is plotted, but a characteristic curve107shape can be idealized from them. As identified by de Haas et al. (2018a), scenario (a) cycles areobserved on the Kamikamihori fan (Section 4.2.3), with a higher proportion of short runout eventsdue to backstepping processes toward the fan apex. Conceptual scenario (b) is distinguished fromthe other distributions as having the largest area under the runout exceedance curve in the cross-fancomponent; this curve shape is due to multiple channels becoming activated, and therefore debris flowimpacts directed further away from the previous flow path. Fan analogues for scenario (b) from theSWBC dataset are Abandoned or Cheam E, but are not shown in Figure 4.18 (refer to Appendix A forindividual plots). The Currie D runout exceedance distribution closely resembles the runout patternsof scenario (c), particularly in the cross-fan component. Debris flows at Currie D mostly follow theprevious flow path, with gradual lateral shifting and overprinting of previous deposits. Spatial impacttrends at Currie D are described in more detail in Section 4.3.1. The SWBC aggregate distributionlies somewhere between the conceptual curves, showing an ensemble of avulsion patterns. Basedon this preliminary work, the conceptual patterns described by de Haas et al. (2018a) are realisticanalogues of spatial runout patterns observed on real debris flow fans.It is likely that these conceptual avulsion patterns evolve through time with changing climate,supply conditions, and fan topography. Furthermore, many cycles may occur during a single event, ora cycle may be disrupted by extreme system perturbations or stochastic processes. The average runoutdistribution can be represented by an ensemble, as shown by the SWBC aggregate in Figure 4.18, buttheoretical distributions based on topography (relative location of steepest descent) and flow volumesequencing may be derived to refine runout forecasting. This approach is shown conceptually inFigure 4.18 based on the work by de Haas et al. (2018a), but future work may involve developingthese conceptual runout exceedance curves.108Figure 4.17. Figure from de Haas et al. (2018a) illustrating conceptual avulsion patterns basedon varying flow volume sequences.109Figure 4.18. Fan-normalized cumulative runout exceedance distributions for conceptual avul-sion scenarios proposed by de Haas et al. (2018a) (Figure 4.17) compared to empiricaldata.4.3 Local Spatial Impact Trends at Two Locations in SouthwesternBritish ColumbiaIn this section, local spatial impact trends are examined more closely at two locations from theSWBC dataset: three fans at Mount Currie near Pemberton, and two fans at Fountain Ridge nearLillooet (Figure 4.19). These sites were selected because of high rates of debris flow activity (10-15impact areas at each fan), as well as the presence of lidar data and field observations to support amore detailed analysis.110!!!###WhistlerLillooetPembertonCopyright:© 2014 Esri121°40'0"W121°40'0"W122°0'0"W122°0'0"W122°20'0"W122°20'0"W122°40'0"W122°40'0"W123°0'0"W123°0'0"W51°0'0"N51°0'0"N50°40'0"N50°40'0"N50°20'0"N50°20'0"N0 20 4010 Kilometers±British Columbia AlbertaUSA Fountain RidgeMount CurrieFigure 4.19. Location map of Mount Currie and Fountain Ridge fans.4.3.1 Mount CurrieThree conjoined fans emanate from the steep, precipitous, north facing slopes of Mount Currie(Figure 4.20). Mount Currie is a northeast trending glacial areˆte ridge consisting of foliated quartzdiorites with a strong joint control of relief (Bovis & Evans, 1995). The mountain ridge is dissectedby linear tension cracks, scarps, and trenches, with a prominent 1.7 km long linear scarp obliqueto the ridge axis (shown on the lidar in Figure 4.20) likely associated with gravitational movement(Bovis & Evans, 1995; Thompson et al., 1997). Frequent rockfalls, rock slides, and debris slides fillthe watershed gullies with colluvium (BGC, 2018a). In addition to debris flows, snow avalanchescommonly reach the fan. Despite Currie B having the largest watershed area, Currie D has thelargest fan area. Although differences in fan areas might be attributed to different weathering rates orkinematic mechanisms in the respective watersheds, it is more likely a portion of the Currie B fan111was buried by floodplain sediments (anabranches of the Green River are seen abutting the fan toe ofCurrie B in the early airphoto record).The fans are highly channelized, with large bouldery levees present at the upper fans, cobblydeposits with a sandy matrix typical on the lower fans, and evidence of sediment plumes inundatingthe Green River floodplain. The active channels are incised 10 to 15 m into the fans near their apexes.Typical debris flow features and deposits at the Mount Currie fans are shown in Figure 4.21.±Currie B0 1,000 2,000500 MetersCurrie CCurrie DGreen RiverMain ridge scarp Main talus slopesFansWatershedsFigure 4.20. Overview of Mount Currie with main geomorphic features mapped. 2017 ALSbare earth hillshade courtesy of SLRD.112abcFigure 4.21. Mount Currie field photographs. a) Incised channel at the upper fan of Currie C;b) bouldery lobe, mid to lower fan at Currie D; and c) boulder-studded sandy deposit atthe lower fan of Currie B, Green River floodplain in the distance.113Impact area mapping at the Mount Currie fans is shown in Figure 4.22, and spatial impactheatmaps in Figure 4.23. The largest debris flow in the dataset occurred at Currie B, with a volumeof approximately 500,000 m3. Part of the flow avulsed from the main channel at the channel benddownslope of the apex, while the bulk of the deposit followed the previous flow path, inundatingthe floodplain (sandy debris field pictured in Figure 4.21c). Debris flows recorded at Currie Care markedly of smaller magnitudes and more channelized compared to the other fans. The mostimpact areas mapped at any of the fans in the dataset is at Currie D (15), the most recent of whichoccurred during the summer of 2019 with a volume of about 100,000 m3 (see Section 3.4.1 for changedetection results). Debris flows recorded at Currie D remain channelized at the upper fan, and oftendeposit thick, coarse-grained terminal lobes at the distal fan, overlapping and side-stepping previousdeposits here. No major avulsions shifting deposition to the east side of the fan were observed atCurrie D in the airphoto record.±Legend              fan_areasyear_img2019201720162015201420132011201020092004199419901986198019771969196219581946Currie BCurrie C Currie DFan boundaryImpact areaGreen River0 500 1,000250 MetersFigure 4.22. Debris flow impact area mapping at Mount Currie fans. 2017 ALS bare earthhillshade courtesy of SLRD.114Figure 4.23. Fan-normalized spatial impact area heatmaps relative to the fan axis (left) and theprevious flow path (right) for the Mount Currie fans.1154.3.2 Fountain RidgeFountain Ridge is a north-northwest trending ridge of folded, deformed, and highly weatheredsedimentary rocks of the Jackass Mountain Group, including greywackes, argillites, and conglomer-ates (Duffell & McTaggart, 1952). Two very active conjoined debris flow fans have formed on top ofa river terrace east of the Fraser River, and are truncated by a kame terrace to the south (Figure 4.24)(Ryder, 1969). Debris flow channels are fed by constant raveling of extensive talus slopes from steep,small basins (Jordan, 1994).Jordan (1994) found that debris flows at Fountain Ridge appear to have relatively low velocities,high viscosities, and deposit most of their sediment load in well-developed levees. This morphologyand flow behaviour is common of arid environments with lower water contents (Jordan, 1994;Whipple & Dunne, 1992). Typical debris flow features and deposits at Fountain Ridge are shownin Figure 4.25. Debris flows form narrow levee-confined channels with small lobes that breakthrough the levees, or thin (less than 2 m thick) terminal lobes of uniform thickness on the distal fans,consisting of mostly gravels and cobbles. Based on one grain-size sample taken by Jordan (1994),32% of the debris was matrix (smaller than 4 mm), and 21% was cobbles and boulders. Cemented,matrix-supported flow sequences were preserved in near-vertical channel banks at Fountain N(Figure 4.25b), showing inverse grading.Differences in deposit textures and morphology at Fountain N and S are shown in Figure 4.26.Deposition at the lower Fountain N fan is more sheet-like, consisting of thin and wide deposits withchannelized surfaces, while deposits at Fountain S tend to be self-channelizing and lobate. Based onfield observations, the grain-size of the terminal lobes at Fountain S appear to be coarser and moreuniformly graded compared to the deposits at Fountain N. Geomorphic evidence corroborating ahigher coarse fraction for the Fountain S flows is the formation of prominent levees from coarsematerials advected to the flow edges.116Figure 4.24. Oblique view of Fountain Ridge in Google Earth. Fans and watersheds are outlinedin white, and talus slopes in blue. Fountain N is supplied by a 1.2 km long talus chute.117abcFigure 4.25. Fountain Ridge field photographs. a) Channelized reach on the upper Fountain Nfan; b) near-vertical cemented channel side-wall exposing flow sequences and inversegrading, mid-fan Fountain N; and c) thin terminal lobe, distal Fountain N fan.118AB: Fountain S depositsBA: Fountain N depositsFigure 4.26. Comparison of debris flow deposit morphology and textures at the distal FountainRidge fans. Sheet-like deposits more typical at Fountain N, compared to lobate, coarser-grained deposits at Fountain S. 2019 bare earth lidar collected by RPAS (drone).Impact area mapping at the Fountain Ridge fans is shown in Figure 4.27, and spatial impactheatmaps are shown in Figure 4.28. The largest event is estimated to be approximately 170,000 m3 atFountain N, while flows at Fountain S are less than 40,000 m3. Avulsions are common at both fans.At Fountain N, major avulsions occur near the fan apex, causing back-and-forth switches betweenthe north and south sectors, while avulsions at Fountain S occur at various locations down-fan aslobes break through levees. The two main avulsion paths at Fountain N are aligned roughly withthe orientations of the main basin drainage and the talus chute. Debris flow runout distances aremarkedly longer at Fountain N (up to 1.7 km long, the furthest in the SWBC dataset), resulting in alarge fan area.Differences in deposit morphology, mobility, and avulsion patterns at Fountain N and S cannotbe attributed to different source geologies or climates. A possible reason for these differences is thepresence of a 1.2 km long talus slope in the comparatively larger watershed at Fountain N, providinga constant and unlimited supply of fine-grained material directly to the fan (Figure 4.24). The channelis incised 10-15 m into a Holocene fan near the apex, providing additional sediment from entrainment.119At Fountain N, it is likely that progressive aggradation or plugging near the fan apex primes thechannel for avulsions.Spatial impact trends at Fountain S are affected by both the texture of the sediment supply andtopography. Fountain S has a much less extensive talus source and a smaller watershed comparedto Fountain N. Debris with a slightly lower fines content is more diffusive of excess pore pressures,resulting in less mobile flows (e.g., de Haas et al., 2015; Whipple & Dunne, 1992). Although theFountain S fan abuts a kame terrace, contemporary debris flow lobes typically terminate short of it,whereas debris flows at Fountain N runout further at comparable fan gradients. Fountain S flow pathsare topographically forced to the south away from the depositionally-dominant northern fan system.±Fountain SFountain NLegend                 fan_areasyear_img2018201520142010200920041992198719751966196419591948Fan boundaryImpact areaKame Terrace0 500 1,000250 MetersFigure 4.27. Debris flow impact area mapping at Fountain Ridge fans. 2019 bare earth lidarcollected by RPAS (drone) overlain on 1997 orthorectified airphoto scene.120Figure 4.28. Fan-normalized spatial impact area heatmaps relative to the fan axis (left) and theprevious flow path (right) for the Fountain Ridge fans.4.3.3 Comparison and DiscussionMount Currie and Fountain Ridge are very active fan complexes with distinctive geologic settingsand climates. Morphometric variables at each fan site are summarized in Table 4.1. As describedin the previous sections, debris at Mount Currie is comprised of cobbly, bouldery crystalline rock,while Fountain Ridge debris is derived from weathered sedimentary rocks, and is comparatively121Table 4.1. Summary of morphometric variables at Mount Currie and Fountain Ridge fans.Fan site Fan area(km2)Overall fanslope (◦)Average fanchannel slope (◦)Watershedarea (km2)Watershedrelief (km)Melton ratioCurrie B 0.4 10.7 8.2 2.7 2.1 1.3Currie C 0.4 17.5 15.8 1.2 1.7 1.6Currie D 1.3 14.1 14.7 1.7 1.5 1.1Fountain N 1.2 10.6 11.4 0.9 1.1 1.2Fountain S 0.4 14.3 12.4 0.4 1.0 1.7fine-textured. Mount Currie receives more precipitation, while Fountain Ridge is located in a morearid region of SWBC (see Section 1.5.2 for climate data).Comparing the spatial impact heatmaps for Mount Currie in Figure 4.23, hotspots are moreconcentrated along the previous flow path and near the fan apex. At Mount Currie, channels aresteep and deeply incised, with major avulsions more typical past 20% of the maximum fan length(except for large volume events at Currie B). In contrast, spatial impact heatmaps for Fountain Ridgein Figure 4.28 show less concentrated impact area hotspots, indicating frequent shifts in flow paths;channels on the upper fan at Fountain Ridge are less stable, and avulsions are common near the fanapex.Figure 4.29 shows a time series of maximum fan-normalized runout extents. The apparentincrease in frequency after 2009 is due to access to Planet satellite imagery with high temporalresolution. Based on the impact area mapping, there are no obvious patterns of backstepping in thedown-fan component followed by an avulsion, although it is likely smaller channel plugging eventshave been censored. There appears to be a correlation between event volume and runout distance,although the largest avulsion magnitudes at Fountain N and S were associated with relatively smallvolumes. Figure 4.29 shows cycles of high magnitude avulsions at Fountain N (frequency andamplitude of the cross-fan time series), whereas the other fans exhibit more gradual cycles of fanmigration.Fan-normalized and unnormalized cumulative runout exceedance distributions between the fansites are compared in Figure 4.30. Fountain N has the longest runout distances in both down-fan andcross-fan dimensions, which might be attributed to larger event volumes, enhanced mobility from122high fines content, and a constant sediment supply from watershed talus slopes priming channels foravulsion. Currie D, with similar down-fan runout distances and event volumes to Fountain N, has amarkedly lower cross-fan mobility, with most events remaining channelized until the distal fan. Thecomparatively different runout trends at Currie D might be related to many factors, including failuremechanisms and source geology. For instance, rock toppling and sliding in the upper watershed couldtrigger debris flows with large peak discharges and high velocities; evidence of high impact energieswas observed in the field, including the destruction and burial of large trees, and splintered logs withfrayed ends buried in debris. These types of debris flows are more likely to erode the channel intoits steep fan (itself a product of coarse-grained, granitic debris with a high friction angle), thereforechannelizing the flow, and only deviating from the flow path at the distal fan where confinement islost, or if there is an event with sufficient peak discharge to overwhelm the channel capacity.At a high level, a comparison of spatial impact trends at two very active debris flow fanswith differing geologic settings help add to the conceptual model of factors that influence debrisflow mobility (Section 2.4) and avulsion (Section 2.5). Based on observations at Mount Currie andFountain Ridge, debris flow volumes and peak discharge, source geology, sediment supply conditions,grain-size distribution, fan topography, and fan incision, influence the migration of impact areasthrough time. Climate differences between the two sites may also affect runout trends, but were notexplored as part of this work. In Section 4.4, a statistical approach is taken to test what variables areassociated with different runout trends using the entire SWBC dataset.123Figure 4.29. Time series showing the evolution of maximum debris flow runout in both down-fan and cross-fan components, along with volume estimates where available (see Sec-tion 3.4), at Mount Currie and Fountain Ridge fans.124Figure 4.30. Comparison of cumulative runout exceedance distributions at Mount Currie andFountain Ridge fans.4.4 Factors Affecting Spatial Impact Trends on Fans in SWBC4.4.1 Statistical ApproachTo test the research hypothesis that differences in debris flow spatial impact trends can beexplained, in part, with morphometric or geotechnical characteristics, the following statisticalapproach was taken:1251. Impact areas from the SWBC dataset were separated into groups based on event volume(Section 3.4), and site characteristics described by morphometric variables, source geology,fan truncation, and rates of debris flow activity (Section 3.6). Variables were limited to thosethat were relatively simple to obtain given the availability of data across all fan sites, andthat might be related to mobility and avulsion based on the literature review (Sections 2.4and 2.5). Partitioning impact areas by event volume is not fan-specific, whereas the othervariables stratify trends by fan site. For continuous variables, the population was split into 3groups using the lower (Q1) and upper (Q3) quartiles. In this preliminary analysis, quartileswere selected as a simple way to compare upper and lower sample groups without sacrificingthe sample size. For categorical variables, the dataset was split into two groups to maximizesample size.2. Cumulative runout exceedance distributions relative to the previous flow path were generatedfor each group, for both the down-fan and cross-fan components, and with fan-normalized andunnormalized runout distances.3. A two-sample Kolmogorov-Smirnov (KS) test was used to test if there is a statisticallysignificant difference in the distributions of the sample groups for each variable. The KS testis non-parametric (i.e., does not assume a distribution) and the test statistic is the maximumabsolute distance between the empirical cumulative distribution functions. The null hypothesisis that both samples come from the same distribution; if the null hypothesis is rejected (p-value< 0.05), the samples are from different distributions, and the variable used to separate thesample might explain variation in runout trends. For continuous variables, the differencebetween the lower (<Q1) and upper quarters (>Q3) are considered. The two-sample KS testwas completed in MATLAB.Results of the statistical analysis are displayed in Figures 4.31, 4.32 and 4.33. P-values are boldwhere the null hypothesis is rejected (p-value < 0.05), indicating the variable might be a discriminatorfor the spatial impact metric. Conversely, a case where the null hypothesis is accepted might indicatethat the ensemble distribution adequately represents the probability of runout exceedance, regardlessof the discriminating variable. A discussion for each variable is provided in the following sections.126Figure 4.31. Comparison of cumulative runout exceedance distributions using sub-samples ofimpact areas from the SWBC dataset. Each column of plots corresponds to a variable bywhich the samples are separated, and each row is a different runout metric (down-fanor cross-fan; normalized or unnormalized). Sample groups are partitioned by variablequartiles (Q1, Q3). KS test p-value between upper and lower quarters are bold if thesamples are from different distributions (p-value < 0.05).127Figure 4.32. Comparison of cumulative runout exceedance distributions using sub-samples ofimpact areas from the SWBC dataset. Each column of plots corresponds to a variable bywhich the samples are separated, and each row is a different runout metric (down-fanor cross-fan; normalized or unnormalized). Sample groups are partitioned by variablequartiles (Q1, Q3). KS test p-value between upper and lower quarters are bold if thesamples are from different distributions (p-value < 0.05).128Figure 4.33. Comparison of cumulative runout exceedance distributions using sub-samples ofimpact areas from the SWBC dataset. Each column of plots corresponds to a variable bywhich the samples are partitioned, and each row is a different runout metric (down-fan orcross-fan; normalized or unnormalized). KS test p-value are bold if the samples are fromdifferent distributions (p-value < 0.05).1294.4.2 Event VolumeThere is a positive correlation between volume and maximum runout distributions, with largermagnitude events typically travelling farther in the down-fan and cross-fan dimensions (Figure 4.31a).This result is consistent with empirical findings from the literature (e.g., Corominas, 1996; Griswold& Iverson, 2008). Since volume was estimated using the planimetric impact area for a majority ofthe events in this study, there is an inherent relationship between the maximum runout extents andvolume. For fan-normalized runout, the results show that larger magnitude events are more mobilein the down-fan direction relative to other fan-formative events. Although there is a weak positivecorrelation for normalized runout in the cross-fan component, the differences are not statisticallysignificant. The implications of these findings are that the volume of a debris flow may not beas significant when forecasting the probability of avulsion compared to mobility in the down-fandirection. Examples of this finding can be seen at the Fountain N and Fountain S fans, where thelargest magnitude avulsions were not associated with the largest event volumes in their respective datarecord (refer to Figure 4.29). Competing mechanisms may explain the weaker correlation betweenvolume and avulsions; although large magnitude events would have a sufficient peak discharge toovertop the active channel, they may also erode the channel bed, enhancing channelization (e.g.,Schu¨rch et al., 2011b).4.4.3 Melton RatioThere is a statistically significant difference in the normalized runout distributions for fansgrouped by Melton ratio (Figure 4.31b). In the down-fan component, lower Melton Ratios (lessrugged watersheds) are associated with events that terminate closer to the fan toe, whereas higherMelton ratios (more rugged watersheds) have a higher proportion of short-runout events. A possibleinterpretation for this trend is through the association of Melton ratio with hydrogeomorphic processtype (e.g., Jackson et al., 1984; Wilford et al., 2004) (see Section 3.6.7). Events on mixed-processfans (lower Melton ratios), possibly debris flow-flood hybrids, would have higher water contentsand thus higher mobility, whereas debris flows with high sediment concentrations are more likely toform channel plugs, terminating mid-fan. Secondary processes, such as stream-flow and flooding130between debris flow events, may also erode debris flow deposits and redistribute sediment down-fan,enhancing connectivity to the distal fan. The normalized cross-fan runout trends corroborate thisinterpretation; there is a higher proportion of events that follow the previous flow path for the lowestMelton ratios compared to the highest Melton ratios (although the inter-quartile range has very fewhigh magnitude avulsions). Debris flows have characteristically higher peak discharges comparedto debris floods (Hungr et al., 2014), which may increase the probability of the channel capacitybeing overwhelmed. Similar distinction between process type was found by Pederson et al. (2015)through stratigraphic analysis; deposits with typical debris flow characteristics tended to stack morecompensationally (i.e., avulse) compared to areas with typical stream-flow characteristics (Santiet al., 2017).4.4.4 Watershed AreaRunout distributions stratified by watershed area mirror trends found by the Melton ratio (sincewatershed area is used to calculate the Melton ratio), but were not statistically significant from oneanother, except for the unnormalized data in the down-fan component (Figure 4.31c). A likely reasonfor the unnormalized down-fan component outlier (<Q1) is because the Fountain N and S fans areincluded in this sample, with characteristically long runout lengths and small watershed areas. Thewatershed area is hypothesized to influence spatial impact trends in different, potentially competing,ways. Large watersheds generate higher water discharges, and therefore more fluidized, mobile flows(Tang et al., 2012); for this reason, watershed area has been used as a variable to predict runoutdistances in some empirical runout relationships (e.g., Tang et al., 2012; Zimmermann et al., 1997).Larger watersheds might (but not always) contain more contributing debris flow source areas, andhigher sediment inputs would aggrade fan channels through time, triggering subsequent avulsions.Conversely, larger watersheds are associated with debris flood and stream-flow processes (e.g.,Millard et al., 2006; Wilford et al., 2004), and therefore deposits that stack less compensationally (i.e,less likely to avulse) (Pederson et al., 2015). Lau (2017) found watershed area to be an importantvariable contributing to channel scour for alluvial fans in SWBC, which could hypothetically reducecross-fan impacts due to increased confinement.1314.4.5 Fan and Channel SlopeNeither the fan slope nor channel slope discriminate differences in runout exceedance distributions(Figure 4.32a,b). Although the fan and channel slope are closely related (see Figure 3.22), theywere calculated differently (refer to Sections 3.6.3 and 3.6.4), and have slightly different physicalinterpretations. The fan slope represents the overall slope of the fan landform, whereas the averagechannel slope is more representative of contemporary processes along the active channel. Thechannel slope, and to some extent the overall fan slope, fluctuate in time, and future studies withpre-event topographic measurements may yield different results.It is hypothesized that fans with gentler slopes are correlated to higher mobility because they areassociated with both finer-grained flows (Blair & McPherson, 1998) and mixed-process fans (Bardou,2002; Bertrand et al., 2013; Scheidl & Rickenmann, 2010). A preliminary numerical modellingstudy by Zubrycky et al. (2019) using debris flow events from this thesis found a potential positivecorrelation between calibrated Voellmy friction coefficients and channel gradients (see Figure 2.4).However, this trend may be moderated by the interaction of topography with the flowing mass, inwhich a decrease in slope initiates deposition. Hypothesized causative links to runout in the cross-fancomponent are also enigmatic; either steep fans are reflective of debris flow-dominant processesthat generate coarse-grained, channel plugging events that trigger avulsions (de Haas et al., 2018a;Pederson et al., 2015), or steep fans are more likely to be incised, in which flow is concentrated alongthe active channel (Lau, 2017).4.4.6 Fan Elevation Relief RatioFor groups separated by the fan ERR, statistically significant differences were found betweenthe runout exceedance distributions in the down-fan component (Figure 4.32c). For the normalizeddata, events reaching the distal fan were more common for the (relatively) more planar fan surfaces(i.e., fans with higher ERRs). A theoretical interpretation explaining the high proportion of debrisflows terminating mid-fan with higher concavity is the reduction in fan slope exerts a centripetalacceleration, stalling the flowing mass. Williams et al. (2006) found planar fan slopes to be associatedwith debris flow processes, while concave-upward shapes are more typical of fluvially-fed fans.132Given this association, the runout trends contradict trends found using the Melton Ratio, wherehigh mobility events reaching the distal fan were more common for mixed-process watersheds.It should be acknowledged that the range of fan ERRs is narrow (0.21-0.42), with all fans beingconcave-upward. Further research with a wider range of fan profiles and different measures ofconcavity are recommended to validate these results.4.4.7 Normalized Fan Intersection PointBased on the position at which the main channel intersects the fan surface, there is no statisticallysignificant difference in the runout exceedance distributions for the SWBC dataset (Figure 4.32d).Since the normalized fan intersection point is a reasonable proxy for the extent of fan incision alongthe fan length, it was hypothesized that fans with intersection points on the distal fan would havelower cross-fan runout magnitudes compared to fans where flow confinement is lost on the upperfan. By accepting the null hypothesis however, the degree of channelization on a fan may not be astrong indicator for forecasting spatial impact trends. The intersection point fluctuates in time, andfuture work using pre-event topography for measures of fan incision and channelization should beconsidered to further these results.The relationship between the fan intersection point and the avulsion node (position on the fanwhere the avulsion path deviates from the previous flow path) is shown in Figure 4.34 for the SWBCdataset. Although there is no trend between the two locations, most of the avulsion nodes in theSWBC dataset occur upstream of the intersection point, contrary to findings by Millard et al. (2006),who found avulsions were most frequent immediately downstream of the intersection point for fansin coastal BC. Based on the distribution of avulsion node positions down the fan, avulsions weremost frequent immediately downstream of the fan apex and at around 30% of the maximum fanlength, and declining toward the fan toe.133Figure 4.34. (Left) Distribution of avulsion nodes along the longitudinal position on the fanand (right) relationship between longitudinal position of the avulsion node on the fanrelative to the fan intersection point.4.4.8 Source GeologyImpact areas were separated into two groups based on source geology: granitic rocks, and non-granitic rocks (sedimentary, metamorphic, and volcanic rock types). The potential effects of sourcegeology on spatial impact trends is previously discussed in Section 4.3, comparing Mount Currie fans(granodiorites) to Fountain Ridge fans (weathered sedimentary rocks). Based on data from the entirestudy area, there is a statistically significant difference in the absolute down-fan runout distributionbetween source geology groups (Figure 4.33a). This finding is consistent with the understanding thatcoarse-grained, frictional, granitic debris flows have typically shorter runouts compared to debriscomprised of sedimentary or volcanic rocks with a larger proportion of fine-grained material andhigher clay contents. However, fan-normalized runout exceedance curves in the down-fan directionare very similar for both groups of rock types, which supports the use of fan-normalized runoutexceedance distributions, irrespective of source geology.It is hypothesized that supply conditions influence cross-fan runout distributions, in that supply-unlimited basins with a more constant sediment feed would plug or aggrade the channel, providingoptimal conditions for avulsion. Since granitic basins in the dataset were classified as supply limited134(Jakob, 1996), the source geology groups are a reasonable proxy for supply conditions, in additionto grain-size distributions and rheological properties. For the SWBC dataset, granitic debris flowsappear to have shorter cross-fan mobility compared to the other rock types, however the difference isnot statistically significant.4.4.9 Fan TruncationImpact areas separated by fan toe truncation have different distributions in the down-fan com-ponent (Figure 4.33b). Longer absolute runouts are observed on the non-truncated fans, but whenrunout is normalized by the fan boundary, the truncated fans have a higher proportion of debrisflows that terminate closer to the mapped fan boundary. In comparison, maximum runouts fornon-truncated fans have a more normal distribution, with a mean closer to 75% of the maximum fanlength (Figure 4.35). These differences may be attributed to normalizing with a shorter fan length,where the true fan boundary reflective of a fan’s formative debris flows has been inundated or erodedby a water body. Furthermore, debris flows that runout past the fan toe would not be captured in thedata record, resulting in a high proportion of mapped impact areas terminating at the fan toe. Spatialimpact trends grouped by fan truncation may also reflect interaction with downstream conditions,such as backwater effects caused by an impounded water body, or fan entrenchment from riverincision at the toe, requiring further study.Figure 4.35. Probability density of fan-normalized maximum runouts grouped by fan trunca-tion.1354.4.10 Fan Activityde Haas et al. (2018a) and de Haas et al. (2018b) show avulsion trends might be influenced byfrequency-magnitude distributions; fans with abundant, small, channel-plugging events followedby a large magnitude event with sufficient volume to overwhelm the channel capacity are idealfor high rates of avulsion. Due to the limited data record, frequency-magnitude relationships werenot derived for the SWBC dataset. Instead, the relative fan activity (Section 3.6.10) was testedas a discriminator using two groups: observed change from debris flow processes every 1 to 10years (very active fans), or greater than 10 years (relatively less active fans). The only distributionsthat were statistically different were runout in the down-fan direction, where the more active fanshad characteristically longer absolute runout distances, but with a comparatively lower proportionreaching the distal fan (Figure 4.33c). In other words, for the SWBC dataset, fans with higher ratesof activity experience more debris flows that terminate short of the fan boundary. Based on the workby de Haas et al. (2018a) and de Haas et al. (2018b), these relatively low mobility events may serveto backfill channels, causing subsequent debris flows to avulse; although we see a slight increase innormalized cross-fan runouts for the more active fans in the dataset, the difference is not statisticallysignificant.4.4.11 Summary and ImplicationsDifferences in runout distributions in the down-fan and cross-fan components for the SWBCdataset were tested using simple variables that describe the event and fan characteristics. The purposeof the statistical analysis was to test the hypothesis that certain morphometric and geotechnicalcharacteristics influence mobility and avulsion trends on fans. The results have implications forrunout analyses; variables that discriminate spatial impact trends can be used to stratify the empiricaldataset for forward prediction, whereas variables with no discernable effect may not be as relevant.From this preliminary study, event volume had the most significant impact on debris flowmobility, consistent with findings from the literature. However, volume may not be as significantwhen forecasting the probability of avulsion compared to mobility in the down-fan direction.Some differences were found for spatial impact trends stratified by variables related to hydrogeo-136morphic process type (Melton ratio, watershed area), in which runouts on mixed-process fans weremore mobile in the down-fan component, but not in the cross-fan component (i.e., less avulsions).Spatial impact trends from a wider range of alluvial fan types should be studied to test this hypothesissince most of the fans in the SWBC dataset are classified as debris flow-dominant (Section 3.6.7).Overall, most of the morphometric variables related to the fan (fan slope, channel slope, and in-tersection point) did not separate the runout distributions into statistically significant groups. Oneinterpretation of this result is the fan morphometric variables used are incongruent with the timescale of this dataset (e.g., if the average channel slope fluctuates on the decadal scale, or conversely,the fan slope is more representative of centuries of debris flow activity).Source geology, fan truncation, and fan activity had no impacts on the cross-fan runout distri-butions. In the down-fan component, normalized runout trends for granitic rocks were statisticallysimilar to those of the other geology types in the study area, supporting the use of fan-normalizeddistributions irrespective of source geology. Normalized trends in the down-fan component differedfor fans truncated by a waterbody, which might be related to an undersized fan length normalizer anddebris flow impacts not being recorded past the fan toe. Lastly, more active fans appeared to have ahigher proportion of events terminating on the upper fan.Overall, the statistical assessment presented here is a preliminary effort to test differences inmobility and avulsion trends using a rich geospatial dataset. The results are meant to enhancepractitioner judgement when using empirical data in forward analyses (Section 4.5.1), and to formhypotheses that should continue to be tested using different datasets or methods (Section 5.3).4.5 Fan-Normalized Empirical Runout Estimator ToolThis section describes a tool that transposes the fan-normalized spatial impact heatmaps derivedin this chapter onto another fan relative to an active channel. The tool is useful for visualizingempirical runout trends across a fan, and may be applicable for preliminary hazard assessments,regional prioritization studies, or supporting expert judgement. The fan used as an example (Catiline)is part of the empirical dataset; it is used strictly to illustrate the methodology, and not as a validationexercise, nor are the results to be interpreted for any risk assessment.1374.5.1 Code WorkflowThe workflow for converting fan-normalized spatial impact heatmaps to probability of runoutexceedance contours for a fan relative to an active channel is described below. The code wasimplemented in MATLAB (R2019b), and is provided in Appendix C, along with sample shapefilesand empirical grids (i.e., fan-normalized spatial impact heatmaps relative to the previous flow path).1. Load shapefiles. For a fan site, the fan apex (1 point), fan boundary (1 polygon), and flowpath (1 polyline), are loaded as separate shapefiles in UTM coordinates (Figure 4.36a). Theflow path must intersect the apex and extend past the fan toe.2. Initialize a measurement grid centered on the fan apex. For a specified grid resolution andradius size, grid nodes are created along a series of circles centered on the fan apex. Grid xand y coordinates are stored in an array, including nodes along the flow path. A grid with 50nodes down and across the maximum fan extents, and sized 1.2 × the maximum fan length, isshown in Figure 4.36b.3. Load empirical data. A fan-normalized spatial impact heatmap grid relative to the previousflow path is selected as the empirical data (Figure 4.36c), with dimensions x∈ {0,2}, y∈ {0,1},and z ∈ {0,1}. The empirical data can be a subset of the full dataset depending on theapplication (e.g., volume class, geology), as described in Section 4.4. Grids were filtered inanother MATLAB script and smoothed in Surfer® (Golden Software, LLC, 2018). Empiricalgrids for the entire SWBC dataset are provided in Appendix C: one of the raw data, and onesmoothed (5 passes of a 5×5 maximum value filter and 10 passes of a Gaussian low-passfilter).4. Sample empirical data. For each grid node, the normalized distance from fan apex (x) andnormalized arc length offset from the flow path (y) are calculated, and a z value is extractedfrom the empirical grid.5. Export raster. Measurement grid nodes are interpolated over a raster with a specified pixelsize (m) and exported as a GeoTIFF or Surfer® grid file. Figure 4.36d shows the output rasterclipped to the fan area and contoured in ArcGIS.138ApexFanActive channel or flow path(a) (b)(c) (d)01.0Empirical probabilityof runout exceedanceEmpirical probabilityof runout exceedance (z)Figure 4.36. Components of the empirical runout estimator tool. a) Input shapefiles; b) mea-surement grid centered on fan apex, including nodes along flow path; c) empirical datasource: filtered and smoothed fan-normalized spatial impact heatmap; and d) empiricaldata sampled at each measurement grid node and exported as a georeferenced raster,where it can be contoured, clipped, or used for calculations in GIS.4.5.2 ApplicationThe main application of the empirical runout exceedance tool is to visualize the fan-normalizedspatial impact heatmaps onto another fan space for forecasting purposes. The output grid andcontours represent an empirical probability of runout exceedance; although a statistical model hasnot been fit to the data at this stage, the raw empirical data is useful for risk-based decision making.139In a regional study, empirical runout distributions can be used to prioritize fan sites for further study.For instance, the probability of runout exceedance derived from a regional dataset multiplied by anaverage annual probability of occurrence (for a debris flow of any size) approximates the overallencounter probability for any location on the fan without explicitly modelling different volume oravulsion scenarios. This encounter probability applies for an impending debris flow event, whilelong term fan evolution studies would require a different approach to what is described here.Figure 4.37 shows an example of the empirical data partitioned into different volume classes,showing varying spatial impact distributions for each. For local studies, the shape and extents of theempirical contours can provide guidance to practitioners for converting numerical modelling resultsinto probability of spatial impact for risk calculations. In a Bayesian statistical framework, an expertopinion (prior), based on a local observations, site-specific interpretations, or modelling results, canbe updated with the empirically derived probability of runout exceedance contours (likelihood), toform a posterior distribution of spatial impact probabilities. A similar Bayesian approach was usedby Nolde & Joe (2013) to incorporate expert judgement for more precise estimates of debris flowreturn periods.Empirical probabilityof runout exceedance0     1.0<10,000 m3 (n=18) 10,000-100,000 m3 (n=84) >100,000 m3 (n=8)Figure 4.37. Probability of runout exceedance heatmaps derived from subsets of the SWBCdataset based on volume class. The Catiline fan is used as an example prediction space.2014 ALS bare earth lidar hillshade courtesy of SLRD.1404.5.3 LimitationsThe empirical runout exceedance tool shares the limitations with creating the fan-normalizedspatial impact heatmaps, such as normalizing assumptions, discussed previously in Section 4.1.4.This tool is not meant for predicting fan-specific avulsion paths, rather, a possible distribution ofrunout extents from an empirical sample. Results depend on the selection of a flow path a-priori,which may not always be clear. The tool does not work for multiple or bifurcating flow paths, norwould it be appropriate for channels with mitigation structures. Runout exceedance for flow pathswith sharp channel bends may not be realistic because x and y measurements would be oblique tothe measurement grid. Since arc length offsets are measured relative to a line, the channel width isnot taken into account; there may be some cases where the probability of runout exceedance decaysprematurely in the cross-fan component for wide channel sections, and vice versa. Probability ofrunout exceedance heatmaps and contours are sensitive to the number of samples in the empiricaldataset, smoothing of the empirical grid, number of nodes in the measurement grid, and the pixelsize of the output raster. If the empirical grid is not smoothed enough, or a small sub-sample isused, artefacts of avulsion pathways from one fan in the dataset will be spuriously projected onto theprediction fan space; an example of this effect can be seen in Figure 4.37 for the >100,000 m3 volumeclass since it only has 8 events in the subset. To avoid overfitting, future work may involve fitting threedimensional functions or statistical models, rather than projecting the data itself, discussed furtherin Section 5.3. As is the case with any empirical tool, judgement must be used when interpretingoutputs and presenting results.4.6 Summary1. A new graphical method is developed to extract and summarize debris flow runout trends,creating spatial impact heatmaps. The main application is to aggregate trends across differentfans creating an empirical runout distribution normalized by the fan boundary. The heatmapsare also useful in highlighting avulsion “hotspots” and measuring avulsion magnitudes.2. The maximum fan-normalized runout distributions relative to the previous flow path follow anormal distribution in the down-fan component, and a lognormal distribution in the cross-fan141component. These distributions provide an understanding of relative down-fan and cross-fanmobility calibrated to an empirical dataset. Based on the SWBC dataset, about 90% of thedebris flows impacted past 50% of the maximum length down the fan, while less than 10%avulsed beyond 50% of the maximum arc length across the fan.3. In comparing the SWBC fan-normalized maximum runout distributions to a monitored fan inJapan, cross-fan impact trends were found to be similar, but the distributions of runout in thedown-fan component were not; a higher proportion of debris flows terminating on the upperfan were recorded at the monitored fan, and may justify adjusting the upper tail of the SWBCdataset distribution to account for missing data (such as smaller debris flows not detected inaerial imagery).4. Runout distributions based on theoretical avulsion cycles were compared to case studies,showing conceptually that these patterns are observed on real debris flow fans.5. Runout and avulsion trends were analyzed and compared for groups of fans at two locationswith very high rates of debris flow activity (Mount Currie and Fountain Ridge). Debrisflow volume, peak discharge, source geology, sediment supply, grain-size distribution, fantopography, and fan incision, are hypothesized to influence the migration of impact areasthrough time for these case studies.6. The following variables were tested as discriminators for differences in down-fan and cross-fanrunout distributions: event volume, Melton ratio, watershed area, fan slope, channel slope, fanelevation relief ratio, fan intersection point, source geology, fan truncation, and fan activity.Event volume had the most significant impact stratifying spatial impact trends, with largermagnitudes corresponding to more mobile runout in the down-fan component. Volume wasnot a statistically significant indicator for cross-fan runout offsets. Fans with lower Meltonratios tend to have impacts more concentrated along the previous flow path and reaching thedistal fan extents. Most fan morphometrics and source geology had no significant impact onnormalized runout patterns. Down-fan runout distributions were unique for fans truncatedby a water body, although the differences might be attributed to a truncated fan length andrunout not recorded past the fan toe. Fans with higher event frequencies also had more events142terminating short of the fan boundary.7. Avulsion nodes (locations) were most common immediately downstream of the fan apex, andabout 30% of the maximum fan length, with frequencies declining toward the fan toe.8. A tool was developed to transpose fan-normalized spatial impact heatmaps from the SWBCempirical dataset onto another fan for guidance in risk-based decision making. The code anddata are provided in Appendix C.143Chapter 5Conclusions and RecommendationsThis chapter summarizes the work completed and highlights the main findings, addressing theresearch objectives and hypotheses. More detailed summaries are provided at the end of each chapterin Sections 2.8, 3.9, and 4.6. Implications for practitioners are discussed, followed by a list of ideasfor future work.5.1 Summary of Main FindingsThis work was undertaken to better understand spatial impact trends on debris flow fans. Cur-rently, there is little research guiding practitioners in estimating probability of spatial impacts ona fan considering various mobility and avulsion scenarios. A comprehensive literature review wascompleted to form a conceptual model of factors that affect debris flow runout. According to theliterature, high mobility events are generally associated with large volumes and fall heights, highsustained pore pressures, and steep, channelized travel paths. Avulsions are comparatively lessunderstood, but there is some evidence that debris composition, lobe thickness, preceding events, andthe frequency-magnitude distribution also influence the probability of avulsion. Based on descriptionsfrom the literature, avulsion triggers were grouped into three scenarios: overtopping, superelevation,and various channel blocking mechanisms, including channel plugs and progressive aggradation.From a literature review of 44 empirical runout relationships with various runout assessment method-ologies, volume was by far the most common predictor, followed by elevation loss from the source144zone (i.e., fall height), both of which can be difficult to ascertain. Very few of the methods reviewedwere probabilistic, and none consider cross-fan impacts via avulsion mechanisms explicitly.One of the major contributions from this work is the creation and documentation of a uniquespatial record of debris flow impacts in SWBC, which can be continuously added to and used forfuture analyses. This geospatial dataset consists of 176 debris flow impact areas and flow paths across30 fan sites. In this work, an impact area is defined as any area below the fan apex that has beenimpacted by a debris flow, or multiple debris flows, over a certain time period. Geomorphic mappingwas completed using an ensemble of remote sensing and field data, including hundreds of historicalairphotos dating back to 1928, satellite imagery with high temporal resolution dating back to 2009,topographic basemaps, lidar, and orthophotos. Change detection with the spectral index NDVI wasuseful for delineating impact areas with satellite imagery in some cases, working best on sparselyforested or clearcut fans, and for debris flows that disturb the canopy and do not overprint recentdeposits. Lidar was available for 16 of the fan sites, including lidar and orthophotos collected duringfield work with a RPAS (drone) at three fans. Geomorphic field mapping was completed at 18 fan sitesto delineate lobes, levees, and channels. Field observations were documented, including flow depths,superelevation angles, deposit thickness, deposition angles, and debris composition. The geospatialdataset also consists of fan and watershed boundaries, fan apex locations, and morphometric variablescalculated with lidar or freely available DEMs. A classification scheme was developed to describe thedifferent types of avulsions (or lack thereof) based on the location, magnitude, and surface expressionof debris flow impacts. Of all the impact areas mapped, 86% had some form of avulsion or spreadingacross the fan, with local channelized avulsions the most common type. 35% of the impact areascorresponded to a shift in the position of the channel on the fan. Data certainty classes were alsodefined to give the reader a sense of spatial and temporal accuracy of the impact area mapping. Thedataset, including GIS shapefiles with associated metadata, are provided in Appendix B.As part of the data compilation process, local volume-area relationships were derived basedon lidar change detection, features in post-event lidar, and field data. Impact areas, deposit areas,and approximate event volumes, along with estimates of error, have been documented for 16 eventsin the dataset. The volume-area relationship for SWBC was compared to nine other relationships145for non-volcanic debris flows from the literature and had the same regression coefficient as therelationship for granular debris flows from the Italian Alps. The volume-impact area relationship(differentiated from the volume-deposit area relationship) was used to approximate volumes for theremaining events in the dataset using the mapped impact areas.A novel plotting method was devised to extract runout trends from the geospatial data in boththe down-fan and cross-fan components using a circular grid centered on the fan apex. Zones ofincreasing radii on the grid represent runout down-fan, and arc length offsets represent lateral shiftsacross the fan relative to the fan axis or previous flow path. Spatial impact heatmaps were created bysumming the plotted impact areas. Heatmaps across different fans were combined by normalizingrunout to the maximum fan length (down-fan) and arc length (cross-fan). The ensemble heatmapfor all the SWBC impact areas shows that most debris flows impact along the previous flow path,with the probability of impact decaying from the apex and away from the active channel. Almost alldebris flows impact within ±60◦ relative to the fan axis or previous flow path.Maximum runouts in the down-fan and cross-fan components were extracted from the dataset.Based on the SWBC dataset, about 90% of the debris flows impacted past 50% of the maximumlength down the fan, while less than 10% avulsed beyond 50% of the maximum arc length across thefan. Avulsions were most common immediately downstream of the fan apex and at about 30% of themaximum fan length, with instances decreasing toward the fan toe. Maximum normalized runout inthe down-fan component can be represented by a normal distribution, while the cross-fan follows alog-normal distribution.From a more thorough comparison of spatial impact at five very active fans, it appears that debrisflow volume, peak discharge, source geology, sediment supply conditions, grain-size distribution, fantopography, and fan incision, play a role in mobility and avulsion patterns. For two coalescing fanswith the same climate and source geology, the effect of sediment supply and grain-size distribution onspatial-temporal impacts is made apparent. As is the case for alluvial fans, a more constant sedimentsupply was the likely cause of a very laterally unstable fan system, shown by one of the case studieswith an extensive talus slope. The other case study had coarser-grained and more uniformly gradeddebris flows, forming lobate deposits with characteristically shorter runouts.146Using the entire SWBC dataset, a preliminary statistical analysis was completed to test thehypothesis that differences in spatial impact trends for groups of fans or events can be explained,in part, with morphometric or geotechnical characteristics. Event volume, unsurprisingly, had asignificant influence stratifying spatial impact trends, with larger magnitude volumes correspondingto more mobile runout in the down-fan component. Volume was not a statistically significant discrim-inator for normalized runout in the cross-fan component, as some of the largest magnitude avulsionsin the dataset were not associated with the largest magnitude volumes. There were statisticallysignificant differences based on the Melton ratio, which is compelling for the interpretation thatimpacts on mixed-process fans tend to reach further down-fan but remain closer to the active channel.Differences in mobility and avulsion, however, were not explained by fan morphometrics, such asthe slope or the point at which channelization is lost. Granitic debris flows tended to travel shorterdistances, although when runout is normalized by the fan, the distributions were statistically similarto the other source geology types. Separating down-fan runout distributions based on fan truncationor fan activity is warranted given statistically significant differences between the two populations.Overall, there was no clear morphometric discriminator for spatial impact trends on debris flow fans,warranting further study.Lastly, a tool was developed that transposes the empirical runout distributions from the SWBCdataset onto a fan to assist in risk-based decision making. The code and empirical forecasting dataare provided in Appendix C.5.2 Implications for Hazard and Risk AssessmentsEstimating the probability of debris flow impact is an important part of calculating risk for landzoning or hazard mitigation efforts on fans. This work provides a new perspective on debris flow fansusceptibility to impact based on real debris flow events. The methods used here are different fromother empirical methods in that runout trends are represented on the fan space and in two dimensions.By measuring cross-fan runouts from the previous flow path, typical avulsion locations and anglesare uncovered.The maximum fan boundary can be interpreted as a statistical upper-bound of runout from its147formative debris flows. Normalizing by the fan boundary allows for runout trends on groups of fans tobe compared and combined. Although the fan boundary is an imperfect normalizer (e.g., in the caseof truncated fans), the fan landform can be identified somewhat consistently for forward prediction.As shown in Section 4.5.1, normalized regional aggregates can be transferred to other fans to estimateencounter probability on a fan using empirical data. With enough reconstructed fans in the dataset,the ensemble heatmap captures regional frequency-magnitude distributions, mobility behaviours, andavulsion scenarios, without having to specify these a priori. The regional spatial impact heatmapscould be useful for validating hazard maps made with other methods, regional fan susceptibilitymapping, or prioritization studies. In a preliminary statistical assessment, discriminators includingthe Melton ratio, fan truncation, and fan activity might be site-specific variables to consider whencustomizing fan-normalized trends. Future work should involve fitting functions to the data for amore robust and adaptable forecasting tool, as described in Section 5.3. It should be made explicitthat use of this data for forecasting purposes applies to the next debris flow event, and not long termfan evolution directly.Considering numerical modelling for a specific volume and flow path, the cumulative runoutexceedance curves in the down-fan component for a certain volume class could be used to transformmodelling outputs from deterministic to probabilistic. When modelling an avulsion, the conditionalprobability could be approximated using the cumulative runout exceedance curves in the cross-fancomponent, considering an anticipated avulsion path. Although there was a clear relationship betweenflow volume and down-fan mobility, the probability of avulsion may not be as sensitive to flowvolume.The potential applications proposed here are not meant to replace expert judgement or the needfor numerical modelling in certain cases, rather, the empirical spatial impact trends are anotherresource available to support risk-based decision making.5.3 Recommendations for Future WorkOpportunities to improve and extend the work presented in this thesis are as follows:1. Semi-automated inventory generation. Rather than manually inspecting satellite images, a148semi-automated workflow could be developed to detect debris flow events on fans. A recentstudy by Deijns et al. (2020) used NDVI calculated from Lansat imagery to determine thetiming of landslides over a 33-year period. For a debris flow fan area, a time series of high-resolution satellite data (maximum 5 m pixel size) could be extracted, and differences inspectral indices calculated as an indicator of change (refer to Section 3.3.2 for examples). Thismethod might perform poorly for densely vegetated fans or fans with logging and frequentsnow avalanches, and the model would have to be trained to filter for cloud cover, geometricdistortions, and seasonality. Feature detection using high resolution DEMs is also an emergingfield, and could be used to associate lobe boundaries with dates detected from the satellite data.Similar to concepts used by Eisank et al. (2014) for delineating drumlins, debris flow lobeshapes could be detected in lidar, and methods used in Section 3.4.2 could be used to constrainthe event volume from lobes. Fountain S would be an ideal test fan for generating debris flowinventories with satellite and lidar data.2. Incorporating other dating methods. Methods such as dendrochronology and surface expo-sure dating methods could be used to extend the data record for long-term debris flow evolutionstudies, at the cost of lower spatial accuracy. Dendrochronology might be used to constrainsome of the events in this dataset to a year.3. Continued monitoring campaigns. There remains a need to collect high quality and consis-tent field and remote sensing data immediately following debris flow events. It is recommendedthat annual RPAS (drone) lidar scans are continued at Mount Currie at Fountain Ridge formore accurate estimates of debris flow volume and erosion, and to capture events undetectedwith satellite imagery. Baselines should be established at other fans in this dataset. Ideally,ALS scans should be taken following major events to measure volume change in the watershed.Monitoring equipment, such as cameras, geophones, and rain gauges, would be useful at themost active fans (e.g., Currie D and Fountain N) to alert a research team when an event occursfor timely field investigations. This equipment would also generate useful data for futureresearch, such as information about surge sequencing, velocity measurements, and storm data.4. Expanding the dataset. More fans and impact areas should be added to the dataset to capture149variability in runouts and a larger spectrum of event magnitudes. Repeating the statisticalanalysis for fans outside the study area with different climates and geologic settings, or toinclude a wider range of hydrogeomorphic processes, would provide more opportunity to testthe hypotheses.5. Expanding the predictor variables. Due to the limited topographic data available across thefan sites, only a limited selection of morphometric variables was practical for this work. Uponaccess to high quality lidar data across more of the fan sites, future work may incorporate moreaccurate measures of fan incision and curvature. Other variables that would be compelling totest, as available, include fan roughness, number of channels on a fan, channel curvature, areaof contributing source zones in the watershed, grain-size distributions, climate variables suchas annual precipitation, intensity and duration of rainfall during an event, number of surgesduring an event, peak discharge, location and volume of the initiating mass, and the runout ofthe previous event(s).6. Deriving statistical distributions or functions for probability of runout exceedance. Fu-ture work may involve fitting three dimensional functions or statistical models to the impactarea heatmaps (i.e., bivariate empirical cumulative runout exceedance distribution functions),resulting in robust and adaptable models for runout prediction, without overfitting to theempirical data. For instance, statistical model parameters (e.g., location, scale, shape) could bea function of site-specific attributes or theoretical avulsion sequence patterns.7. Testing and validation with laboratory flume table experiments. Spatial impact heatmapscould be compared to flume table debris flow experiments, applying the same measurementapproach with a grid centered on the fan apex and avulsions measured relative to the previousflow path. Extending the work by de Haas et al. (2015) and de Haas et al. (2018b), variablessuch as the peak discharge, number of surges, grain-size distribution, and water content, couldbe systematically changed to test the impacts on debris flow avulsion and mobility trends.Furthermore, experimental work allows the opportunity to measure the fan topography andchannel geometry before and after every event, which can be used to examine the interactionof debris flows with their path.1508. Methods to estimate probability of avulsion based on longitudinal channel thresholdexceedance. Although this thesis begins to elucidate fan-scale avulsion “hotspots”, methodsto estimate the location and conditional probability of an avulsion along a channel are stillundeveloped. Future research might involve determining discharge thresholds at which anavulsion is imminent for a given channel configuration. Another approach might involvedetermining the probability of avulsion versus conveyance with a logistic regression functioncalibrated to variables related to the flow (depth, velocity, grain-size) and channel (slope,cross-sectional area, planimetric curvature). A similar philosophy was applied with volume-balance runout models by Miller & Burnett (2008) and Fannin & Wise (2001), in which theprobability of erosion versus deposition was estimated using path characteristics. Numerouscase studies with pre- and post-event lidar topography (e.g., 2019 event at Currie D), or flumetable experiments, would be required to calibrate these types of models.9. Extending the statistical analysis. The preliminary statistical analysis presented in this thesisis meant to unearth general runout trends with the available data, and to examine what variablesshould be considered for future work. With more data, the statistical analysis might be extendedto include other classification and regression techniques. Rather than splitting the predictorvariables into groups based on quartiles, future work may involve optimizing the value atwhich a variable maximizes the differences between groups. Discriminant analyses, decisiontree algorithms, and various clustering approaches might be considered for predicting spatialimpact distributions.10. Incorporating intensity mapping. In this work, spatial impact was treated as a binary variable(impacted/not impacted) without consideration of flow intensity used to estimate vulnerabilityin a QRA. Effort should be undertaken to document the impact energy of debris flows, such asmeasures of structural damage, extent of vegetation removal, velocity estimates (e.g., runup,superelevation), grain-size, and flow depths, as an additional dimension to scale the spatialimpact heatmaps. This documentation was completed for a few events in the dataset, but notextensively enough for a statistical analysis.11. Hosting the dataset and code online. The geospatial dataset and code should be hosted151online through an open-source web application. A workflow could be created that allows usersto add their own data, filter the empirical dataset or use pre-calibrated functions, and exportcustom probability of runout exceedance contours for a fan or group of fans.5.4 ClosureThis work was undertaken to better understand debris flow fan hazard susceptibility throughempirical observation. The research objectives (Section 1.2) have been accomplished: a geospatialdataset documenting debris flow impacts was created, and spatial impact trends extracted for usein forecasting. The research hypotheses (Section 1.3) have been tested: spatial impact distributionswere fit for runout in both down-fan and cross-fan components, and variables that discriminate runouttrends were tested. Much research remains to be done before debris flow avulsion and mobility canbe routinely predicted, however, the work presented here provides a unique empirical approach forsuch analysis.152ReferencesAaron, J. (2017). Advancement and Calibration of a 3D Numerical Model for Landslide RunoutAnalysis. Ph.D Thesis, University of British Columbia, Vancouver, BC. → page 19Aaron, J., McDougall, S., & Nolde, N. (2019). Two methodologies to calibrate landslide runoutmodels. Landslides, 16(5), 907–920, https://doi.org/10.1007/s10346-018-1116-8. → page 1Abella´n, A., Jaboyedoff, M., Oppikofer, T., & Vilaplana, J. M. (2009). Detection of millimetricdeformation using a terrestrial laser scanner: experiment and application to a rockfall event.Natural Hazards and Earth System Sciences, 9(2), 365–372,https://doi.org/10.5194/nhess-9-365-2009. → page 64Agisoft LLC. (2019). Agisoft Metashape Professional (Version 1.5.2) (Software)https://www.agisoft.com/downloads/installer/. → pages 49, 87Ashworth, P. J., Best, J. L., & Jones, M. (2004). Relationship between sediment supply and avulsionfrequency in braided rivers. Geology, 32(1), 21–24, https://doi.org/10.1130/G19919.1. →page 15Ballantyne, C. K. (2002). Paraglacial geomorphology. Quaternary Science Reviews, 21(18-19),1935–2017, https://doi.org/10.1016/S0277-3791(02)00005-7. → page 5Bardou, E. (2002). Methodologie de diagnostic des laves torrentielles sur un bassin versant alpin.Ph.D Thesis, E´cole Polytechnique Fe´de´rale de Lausanne, Lausanne, Switzerland. → pagesxiii, 80, 81, 132Bathurst, J. C., Burton, A., & Ward, T. J. (1997). Debris Flow Run-Out and Landslide SedimentDelivery Model Tests. Journal of Hydraulic Engineering, 123(5), 410–419,https://doi.org/10.1061/(ASCE)0733-9429(1997)123:5(410). → pages 31, 37Benda, L. E. & Cundy, T. W. (1990). Predicting deposition of debris flows in mountain channels.Canadian Geotechnical Journal, 27(4), 409–417, https://doi.org/10.1139/t90-057. → pages22, 32, 35Berti, M. & Simoni, A. (2007). Prediction of debris flow inundation areas using empirical mobilityrelationships. Geomorphology, 90(1-2), 144–161,https://doi.org/10.1016/j.geomorph.2007.01.014. → pages 33, 68Berti, M. & Simoni, A. (2014). DFLOWZ: A free program to evaluate the area potentially inundatedby a debris flow. Computers & Geosciences, 67, 14–23,https://doi.org/10.1016/j.cageo.2014.02.002. → page 35Bertrand, M., Lie´bault, F., & Pie´gay, H. (2013). Debris-flow susceptibility of upland catchments.Natural Hazards, 67(2), 497–511, https://doi.org/10.1007/s11069-013-0575-4. → pagesxiii, 80, 81, 132BGC (2004). Preliminary Debris Flow Hazard Assessment of Field, Carratt, Eng, McNab, and DaleCreeks, Hatzic Valley. Technical Report Project No. 0378-001-03. → page 42153BGC (2015). Catiline Creek Debris-Flow Hazard and Risk Assessment. Technical Report ProjectNo. 1358001. → pages 2, 45, 60BGC (2018a). Mount Currie Landslide Risk Assessment. Technical Report Project No. 1358004. →pages 45, 111BGC (2018b). Seton Portage Area Integrated Hydrogeomorphic Risk Assessment. Technical ReportProject No. 1358005. → pages 2, 45, 60, 69Blair, T. C. & McPherson, J. G. (1998). Recent debris-flow processes and resultant form and faciesof the Dolomite alluvial fan, Owens Valley, California. Journal of Sedimentary Research,68(5), 800–818, https://doi.org/10.2110/jsr.68.800. → pages 12, 22, 90, 132Blais-Stevens, A. & Septer, D. (2008). Historical accounts of landslides and flooding events alongthe Sea to Sky Corridor, British Columbia, from 1855-2007. Open-File 5741, GeologicalSurvey of Canada. → pages 42, 44Booth, A. M., Sifford, C., Vascik, B., Siebert, C., & Buma, B. (2020). Large wood inhibits debrisflow runout in forested southeast Alaska. Earth Surface Processes and Landforms,https://doi.org/10.1002/esp.4830. → pages 22, 33Bostock, H. (2014). Physiographic regions of Canada. Open-File Geological Survey of Canada “A”Series Map 1254A. Scale 1:5,000,000. → pages xi, 6Bovis, M. J. & Evans, S. G. (1995). Rock slope movements along the Mount Currie “fault scarp”,southern Coast Mountains, British Columbia. Canadian Journal of Earth Sciences, 32(12),2015–2020, https://doi.org/10.1139/e95-154. → page 111Bovis, M. J. & Jakob, M. (1999). The role of debris supply conditions in predicting debris flowactivity. Earth Surface Processes and Landforms, 24(11), 1039–1054,https://doi.org/10.1002/(SICI)1096-9837(199910)24:11〈1039::AID-ESP29〉3.0.CO;2-U. →pages 7, 17, 74, 81Bryant, M., Falk, P., & Paola, C. (1995). Experimental study of avulsion frequency and rate ofdeposition. Geology, 23(4), 365–368,https://doi.org/10.1130/0091-7613(1995)023〈0365:ESOAFA〉2.3.CO;2. → pages 15, 24Bull, W. B. (1977). The alluvial-fan environment. Progress in Physical Geography: Earth andEnvironment, 1(2), 222–270, https://doi.org/10.1177/030913337700100202. → page 47Burton, A. & Bathurst, J. C. (1998). Physically based modelling of shallow landslide sediment yieldat a catchment scale. Environmental Geology, 35(2-3), 89–99,https://doi.org/10.1007/s002540050296. → page 32Bustin, A. M., Clowes, R. M., Monger, J. W., & Journeay, J. M. (2013). The southern CoastMountains, British Columbia: New interpretations from geological, seismic reflection, andgravity data. Canadian Journal of Earth Sciences, 50(10), 1033–1050,https://doi.org/10.1139/cjes-2012-0122. → page 4Cannon, S. (1989). An evaluation of the travel-distance potential of debris flows. MiscellaneousPublication 89-2, Utah Geological and Mineral Survey, Salt Lake City, Utah. → page 23Chen, H. X., Zhang, L. M., Gao, L., Yuan, Q., Lu, T., Xiang, B., & Zhuang, W. L. (2017).Simulation of interactions among multiple debris flows. Landslides, 14(2), 595–615,https://doi.org/10.1007/s10346-016-0710-x. → pages 20, 21Chen, J., Chang, S., Tsang, Y., & Shieh, C. (2007). Empirical relationships for deposited length ofdebris flows - A case study in Taiwan. In C. Chen & J. Major (Eds.), Debris-Flow HazardsMitigation: Mechanics, Prediction, and Assessment (pp. 525–530). Chengdu, China. → pages32, 37154Chen, Z., Zhang, Y., Ouyang, C., Zhang, F., & Ma, J. (2018). Automated Landslides Detection forMountain Cities Using Multi-Temporal Remote Sensing Imagery. Sensors, 18(3),https://doi.org/10.3390/s18030821. → page 54Christen, M., Kowalski, J., & Bartelt, P. (2010). RAMMS: Numerical simulation of dense snowavalanches in three-dimensional terrain. Cold Regions Science and Technology, 63(1-2), 1–14,https://doi.org/10.1016/j.coldregions.2010.04.005. → pages 1, 18Church, M. & Ryder, J. M. (2010). Physiography of British Columbia. In R. G. Pike, T. E. Redding,R. D. Moore, R. D. Winkler, & K. D. Bladon (Eds.), Compendium of Forest Hydrology andGeomorphology in British Columbia, volume 1 of 2 of Land Management Handbook 66 (pp.17–45). Victoria, BC: Ministry of Forests and Range Forest Science Program. → page 5Clague, J. J. & Ward, B. (2011). Pleistocene glaciation of British Columbia. In J. Ehlers, P. Gibbard,& P. Hughes (Eds.), Quaternary Glaciations - Extent and Chronology, volume 15 ofDevelopments in Quaternary Sciences (pp. 563–573). Elsevier. → page 5CloudCompare (2019). CloudCompare (version 2.10) [GPL Software]. Retrieved fromhttp://www.cloudcompare.org/. → page 64Cordilleran (2010). Catalina Creek emergency debris flow assessment. Memo to Ministry of Forestsand Range, Squamish, BC. → page 69Cordilleran (2013). Catalina Creek debris flow, August 30, 2013. Memo to Ministry of Forests,Lands and Natural Resource Operations, Squamish, BC. → page 69Corominas, J. (1996). The angle of reach as a mobility index for small and large landslides.Canadian Geotechnical Journal, 33, https://doi.org/10.1139/t96-005. → pagesxi, 1, 20, 22, 30, 31, 100, 130Corominas, J., et al. (2014). Recommendations for the quantitative analysis of landslide risk.Bulletin of Engineering Geology and the Environment, 73(2), 209–263,https://doi.org/10.1007/s10064-013-0538-8. → pages 16, 17Costa, J. (1984). Physical geomorphology of debris flows. In J. Costa & P. Fleisher (Eds.),Developments and Applications of Geomorphology (pp. 268–317). Springer. → pages12, 13, 61Crosta, G. B., Cucchiaro, S., & Frattini, P. (2003). Validation of semi-empirical relationships for thedefinition of debris-flow behaviour in granular materials. In D. Rickenmann & C. Chen (Eds.),Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment (pp. 821–831).:Millpress, Rotterdam. → page 33Cui, Y., Schiarizza, P., & Diakow, L. (2017). British Columbia digital geology. Open-File 2017-8,British Columbia Ministry of Energy, Mines and Petroleum Resources. Data version2019-12-19. → pages xi, 6, 81D’Agostino, V., Cesca, M., & Marchi, L. (2010). Field and laboratory investigations of runoutdistances of debris flows in the Dolomites (Eastern Italian Alps). Geomorphology, 115(3-4),294–304, https://doi.org/10.1016/j.geomorph.2009.06.032. → page 33D’Agostino, V. & Marchi, L. (2001). Debris flow magnitude in the Eastern Italian Alps: Datacollection and analysis. Physics and Chemistry of the Earth, Part C: Solar, Terrestrial &Planetary Science, 26, 657–663, https://doi.org/10.1016/S1464-1917(01)00064-2. → page 17Dai, F. C., Lee, C. F., & Ngai, Y. Y. (2002). Landslide risk assessment and management: anoverview. Engineering Geology, 64(1), 65–87,https://doi.org/10.1016/S0013-7952(01)00093-X. → page 16D’Arcy, M., Roda Boluda, D. C., Whittaker, A. C., & Carpineti, A. (2015). Dating alluvial fan155surfaces in Owens Valley, California, using weathering fractures in boulders. Earth SurfaceProcesses and Landforms, 40(4), 487–501, https://doi.org/10.1002/esp.3649. → pages 26, 86de Haas, T., Braat, L., Leuven, J. R., Lokhorst, I. R., & Kleinhans, M. G. (2015). Effects of debrisflow composition on runout, depositional mechanisms, and deposit morphology in laboratoryexperiments. Journal of Geophysical Research: Earth Surface, 120(9), 1949–1972,https://doi.org/10.1002/2015JF003525. → pages 21, 22, 23, 120, 150de Haas, T. & Densmore, A. L. (2019). Debris-flow volume quantile prediction from catchmentmorphometry. Geology, 47(8), 791–794, https://doi.org/10.1130/G45950.1. → page 17de Haas, T., Densmore, A. L., den Hond, T., & Cox, N. J. (2019). Fan-Surface Evidence forDebris-Flow Avulsion Controls and Probabilities, Saline Valley, California. Journal ofGeophysical Research: Earth Surface, 124(5), 1118–1138,https://doi.org/10.1029/2018JF004815. → pages 12, 14, 24, 26, 28, 66de Haas, T., Densmore, A. L., Stoffel, M., Suwa, H., Imaizumi, F., Ballesteros-Ca´novas, J., &Wasklewicz, T. (2018a). Avulsions and the spatio-temporal evolution of debris-flow fans.Earth-Science Reviews, 177, 53–75, https://doi.org/10.1016/j.earscirev.2017.11.007. → pagesv, xiv, 2, 12, 14, 15, 23, 24, 26, 27, 28, 89, 103, 104, 105, 107, 108, 109, 110, 132, 136de Haas, T., Kruijt, A., & Densmore, A. L. (2018b). Effects of debris-flow magnitude–frequencydistribution on avulsions and fan development. Earth Surface Processes and Landforms,43(13), 2779–2793, https://doi.org/10.1002/esp.4432. → pages 24, 26, 100, 136, 150de Haas, T., van den Berg, W., Braat, L., & Kleinhans, M. G. (2016). Autogenic avulsion,channelization and backfilling dynamics of debris-flow fans. Sedimentology, 63(6),1596–1619, https://doi.org/10.1111/sed.12275. → pages xi, 15, 16, 23, 24, 26Deijns, A. A. (2018). The effect of debris-flow sediment composition change on avulsion behaviorand debris-flow fan development. Master’s thesis, Universiteit Utrecht, Utrecht, Netherlands.→ page 27Deijns, A. A., Bevington, A. R., van Zadelhoff, F., de Jong, S. M., Geertsema, M., & McDougall, S.(2020). Semi-automated detection of landslide timing using harmonic modelling of satelliteimagery, Buckinghorse River, Canada. International Journal of Applied Earth Observationand Geoinformation, 84, 101943, https://doi.org/10.1016/j.jag.2019.101943. → page 149Demarchi, D. A. (2011). An Introduction to the Ecoregions of British Columbia. Victoria, BC:Ecosystem Information Section, Ministry of Environment, 3rd edition. → pages xi, 6, 9Densmore, A. L., de Haas, T., McArdell, B., & Schu¨rch, P. (2019). Making sense of avulsions ondebris-flow fans. In J. Kean, J. Coe, P. Santi, & B. Guillen (Eds.), Debris-Flow HazardsMitigation: Mechanics, Prediction, and Assessment (pp. 637–644). Golden, CO. → pages2, 14, 27, 28, 89Duffell, S. & McTaggart, K. (1952). Ashcroft map-area, British Columbia. Memoir 262, GeologicalSurvey of Canada. → page 116Du¨hnforth, M., Densmore, A. L., Ivy-Ochs, S., Allen, P. A., & Kubik, P. W. (2007). Timing andpatterns of debris flow deposition on Shepherd and Symmes creek fans, Owens Valley,California, deduced from cosmogenic 10Be. Journal of Geophysical Research: Earth Surface,112(F3), https://doi.org/10.1029/2006JF000562. → page 26Eisank, C., Smith, M., & Hillier, J. (2014). Assessment of multiresolution segmentation fordelimiting drumlins in digital elevation models. Geomorphology, 214(100), 452–464,https://doi.org/10.1016/j.geomorph.2014.02.028. → page 149Evans, S. G. & Lister, D. R. (1984). The geomorphic effects of July 1983 rainstorms in the southern156Cordillera and their impact on transportation facilities. In Current Research, Part B (pp.223–235). Geological Survey of Canada. Paper 84-1B. → page 42Fannin, R. J. & Rollerson, T. P. (1993). Debris flows: some physical characteristics and behaviour.Canadian Geotechnical Journal, 30(1), 71–81, https://doi.org/10.1139/t93-007. → pages10, 23, 35, 42Fannin, R. J. & Wise, M. P. (2001). An empirical-statistical model for debris flow travel distance.Canadian Geotechnical Journal, 38(5), 982–994, https://doi.org/10.1139/cgj-38-5-982. →pages xi, 10, 22, 30, 35, 37, 42, 151Fell, R. (1994). Landslide risk assessment and acceptable risk. Canadian Geotechnical Journal,31(2), 261–272, https://doi.org/10.1139/t94-031. → page 16Fell, R., Corominas, J., Bonnard, C., Cascini, L., Leroi, E., & Savage, W. Z. (2008). Guidelines forlandslide susceptibility, hazard and risk zoning for land use planning. Engineering Geology,102(3), 85–98, https://doi.org/10.1016/j.enggeo.2008.03.022. → page 19Field, J. (2001). Channel Avulsion on Alluvial Fans in Southern Arizona. Geomorphology, 37,93–104, https://doi.org/10.1016/S0169-555X(00)00064-7. → pages 15, 24Fiorucci, F., Ardizzone, F., Mondini, A. C., Viero, A., & Guzzetti, F. (2019). Visual interpretation ofstereoscopic NDVI satellite images to map rainfall-induced landslides. Landslides, 16(1),165–174, https://doi.org/10.1007/s10346-018-1069-y. → page 54Frank, F., McArdell, B. W., Huggel, C., & Vieli, A. (2015). The importance of entrainment andbulking on debris flow runout modeling: examples from the Swiss Alps. Natural Hazards andEarth System Sciences, 15(11), 2569–2583, https://doi.org/10.5194/nhess-15-2569-2015. →page 20Friele, P. & Clague, J. J. (2004). Large Holocene landslides from Pylon Peak, southwestern BritishColumbia. Canadian Journal of Earth Sciences, 41(2), 165–182,https://doi.org/10.1139/e03-089. → page 42Friele, P., Jakob, M., & Clague, J. J. (2008). Hazard and risk from large landslides from MountMeager volcano, British Columbia, Canada. Georisk: Assessment and Management of Riskfor Engineered Systems and Geohazards, 2(1), 48–64,https://doi.org/10.1080/17499510801958711. → page 46Fuller, J. E. (2012). Evaluation of avulsion potential on active alluvial fans in central and westernArizona. Arizona Geological Survey Contributed Report CR-12-D. → page 26Garcı´a-Ruiz, J. M., Beguerı´a, S., Lorente, A., & Martı´, C. (1999). Comparing Debris FlowRelationships in the Alps and in the Pyrenees. Technical Report EVG1-CT-1999-00007,Instituto Pirenaico de Ecologı´a, Zaragoza, Spain. → page 23Geertsema, M., Schwab, J., Blais-Stevens, A., & Sakals, M. (2009). Landslides impacting linearinfrastructure in west central British Columbia. Natural Hazards: Journal of the InternationalSociety for the Prevention and Mitigation of Natural Hazards, 48(1), 59–72,https://doi.org/10.1007/s11069-008-9248-0. → page 42GeoBC (2009). Freshwater Atlas of British Columbia https://www2.gov.bc.ca/gov/content/data/geographic-data-services/topographic-data/freshwater. → pages 75, 77George, D. L. & Iverson, R. M. (2014). A depth-averaged debris-flow model that includes the effectsof evolving dilatancy. II. Numerical predictions and experimental tests. Proceedings of theRoyal Society A: Mathematical, Physical and Engineering Sciences, 470(2170),https://doi.org/10.1098/rspa.2013.0819. → page 18Giraud, R. (2005). Guidelines for the geologic evaluation of debris-flow hazards on alluvial fans in157Utah. Miscellaneous Publication 06-6, Utah Geological Survey. → page 61Golden Software, LLC (2018). Surfer v15.5 www.goldensoftware.com. → pages xiv, 95, 97, 138Griswold, J. P. & Iverson, R. M. (2008). Mobility statistics and automated hazard mapping fordebris flows and rock avalanches. Open-File Report 2007–5276, U.S. Geological Survey,Reston, Virginia. → pages xi, 1, 20, 22, 30, 33, 68, 130Guthrie, R. H., Hockin, A., Colquhoun, L., Nagy, T., Evans, S. G., & Ayles, C. (2010). Anexamination of controls on debris flow mobility: Evidence from coastal British Columbia.Geomorphology, 114(4), 601–613, https://doi.org/10.1016/j.geomorph.2009.09.021. → page10Heim, A. (1932). Bergsturz und Menschenleben (Landslides and Human Lives). Vancouver: BitechPress. Translated by N. Smermer, 1989. → page 30Hooke, R. L. (1967). Processes on Arid-Region Alluvial Fans. The Journal of Geology, 75(4),438–460, https://doi.org/10.1086/627271. → pages 28, 78Hooke, R. L. (1968). Steady-state relationships on arid-region alluvial fans in closed basins.American Journal of Science, 266(8), 609–629, https://doi.org/10.2475/ajs.266.8.609. → page23Horton, P., Jaboyedoff, M., Rudaz, B., & Zimmermann, M. (2013). Flow-R, a model forsusceptibility mapping of debris flows and other gravitational hazards at a regional scale.Natural Hazards and Earth System Science, 13(4), 869–885,https://doi.org/10.5194/nhess-13-869-2013. → pages 19, 31Hungr, O. (2005). Classification and terminology. In M. Jakob & O. Hungr (Eds.), Debris-flowHazards and Related Phenomena (pp. 9–23). Berlin, Heidelberg: Springer Praxis Books.10.1007/3-540-27129-5 2. → page 12Hungr, O. & Evans, S. G. (1993). The Failure Behaviour of Large Rockslides in MountainousRegions. Open-File 2598, Geological Survey of Canada. → page 33Hungr, O., Evans, S. G., & Hutchinson, I. N. (2001). A Review of the Classification of Landslides ofthe Flow Type. Environmental & Engineering Geoscience, 7(3), 221–238. → pages 11, 13Hungr, O., Leroueil, S., & Picarelli, L. (2014). The Varnes classification of landslide types, anupdate. Landslides, 11(2), 167–194, https://doi.org/10.1007/s10346-013-0436-y. → pages11, 12, 13, 131Hungr, O., McDougall, S., Wise, M. P., & Cullen, M. (2008). Magnitude–frequency relationships ofdebris flows and debris avalanches in relation to slope relief. Geomorphology, 96(3), 355–365,https://doi.org/10.1016/j.geomorph.2007.03.020. → pages 10, 37Hungr, O., Morgan, G. C., & Kellerhals, R. (1984). Quantitative analysis of debris torrent hazardsfor design of remedial measures. Canadian Geotechnical Journal, 21(4), 663–677,https://doi.org/10.1139/t84-073. → pages 9, 42Hungr, O., Morgan, G. C., VanDine, D. F., & Lister, D. R. (1987). Debris flow defenses in BritishColumbia. In J. E. Costa & G. F. Wieczorek (Eds.), Debris Flows/Avalanches: Process,Recognition, and Mitigation, volume 7 of Reviews in Engineering Geology (pp. 201–222).Geological Society of America. https://doi.org/10.1130/REG7. → page 42Hu¨rlimann, M., Rickenmann, D., Medina, V., & Bateman, A. (2008). Evaluation of approaches tocalculate debris-flow parameters for hazard assessment. Engineering Geology, 102(3-4),152–163, https://doi.org/10.1016/j.enggeo.2008.03.012. → page 29Ikeya, H. (1981). A method of designation for area in danger of debris flow. In Erosion andSediment Transport in Pacific Rim Steeplands, volume 132 (pp. 578–588). Christchurch, New158Zealand. → page 32Ishikawa, Y., Kawakami, S., Morimoto, C., & Mizuhara, K. (2003). Suppression of debrismovement by forests and damage to forests by debris deposition. Journal of Forest Research,8(1), 37–47, https://doi.org/10.1007/s103100300004. → page 22Iverson, R. M. (1997). The physics of debris flows. Reviews of Geophysics, 35(3), 245–296,https://doi.org/10.1029/97RG00426. → pages 12, 13Iverson, R. M. (2003). The debris-flow rheology myth. In D. Rickenmann & C. Chen (Eds.),Debris-flow Hazards Mitigation: Mechanics, Prediction, and Assessment (pp. 303–314). →pages 18, 21Iverson, R. M., Schilling, S. P., & Vallance, J. W. (1998). Objective delineation of lahar-inundationhazard zones. Geological Society of America Bulletin, 110(8), 972–984,https://doi.org/10.1130/0016-7606(1998)110〈0972:ODOLIH〉2.3.CO;2. → page 33Ivy-Ochs, S., Du¨hnforth, M., Densmore, A. L., & Alfimov, V. (2013). Dating Fan Deposits withCosmogenic Nuclides. In M. Schneuwly-Bollschweiler, M. Stoffel, & F. Rudolf-Miklau(Eds.), Dating Torrential Processes on Fans and Cones: Methods and Their Application forHazard and Risk Assessment (pp. 243–263). Dordrecht: Springer Netherlands. → page 86Jackson, L., Kostaschuk, R., & MacDonald, G. (1984). Identification of debris flow hazard onalluvial fans in the Canadian Rocky Mountains. Review in Engineering Geology, 7, 115–124,https://doi.org/10.13140/2.1.2321.1206. → page 130Jakob, M. (1996). Morphometric and Geotechnical Controls of Debris Flow Frequency andMagnitude in Southwestern British Columbia. Ph.D Thesis, University of British Columbia,Vancouver, BC. → pages 37, 42, 45, 81, 135Jakob, M. (2005). Debris-flow hazard analysis. In M. Jakob & O. Hungr (Eds.), Debris-flowHazards and Related Phenomena (pp. 411–443). Berlin, Heidelberg: Springer Praxis Books.10.1007/3-540-27129-5 17. → pages 35, 61, 86Jakob, M. (2013). Events on Fans and Cones: Recurrence Interval and Magnitude. In M.Schneuwly-Bollschweiler, M. Stoffel, & F. Rudolf-Miklau (Eds.), Dating Torrential Processeson Fans and Cones: Methods and Their Application for Hazard and Risk Assessment (pp.95–108). Dordrecht: Springer Netherlands. → page 86Jakob, M. (2019). Debris-flow hazard assessments - a practitioner’s view. In J. Kean, J. Coe, P. Santi,& B. Guillen (Eds.), Debris-Flow Hazards Mitigation: Mechanics, Prediction, andAssessment (pp. 716–723). Golden, CO. → pages 7, 17Jakob, M., Bale, S., McDougall, S., & Friele, P. (2016). Regional debris-flow and debris-floodfrequency-magnitude curves. In Proceedings of GeoVancouver 2016 Conference Vancouver,BC. → pages 17, 42Jakob, M., Hungr, O., & Thompson, B. (1997). Two debris flows with anomalously high magnitude.In C. Chen (Ed.), Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment(pp. 382–394). San Francisco, CA. → pages 9, 42, 45, 69Jakob, M. & Lambert, S. (2009). Climate change effects on landslides along the southwest coast ofBritish Columbia. Geomorphology, 107(3-4), 275–284,https://doi.org/10.1016/j.geomorph.2008.12.009. → page 7Jakob, M., Stein, D., & Ulmi, M. (2012). Vulnerability of buildings to debris flow impact. NaturalHazards, 60(2), 241–261, https://doi.org/10.1007/s11069-011-0007-2. → page 18Johnson, A. & Rodine, J. (1984). Debris flow. In D. Brunsden & D. Prior (Eds.), Slope Instability(pp. 257–361). New York: John Wiley & Sons Inc. → page 22159Johnson, C. G., Kokelaar, B. P., Iverson, R. M., Logan, M., LaHusen, R. G., & Gray, J. (2012).Grain-size segregation and levee formation in geophysical mass flows. Journal of GeophysicalResearch: Earth Surface, 117(F1), https://doi.org/10.1029/2011JF002185. → page 12Jomelli, V. (2013). Lichenometric Dating of Debris Avalanche Deposits with an Example from theFrench Alps. In M. Schneuwly-Bollschweiler, M. Stoffel, & F. Rudolf-Miklau (Eds.), DatingTorrential Processes on Fans and Cones: Methods and Their Application for Hazard and RiskAssessment (pp. 211–224). Dordrecht: Springer Netherlands. → page 86Jones, L. S. & Schumm, S. A. (1999). Causes of avulsion: an overview. In N. Smith & J. Rogers(Eds.), Fluvial Sedimentology VI (pp. 171–178). John Wiley & Sons Inc. → page 15Jordan, R. P. (1994). Debris flows in the southern Coast Mountains, British Columbia: dynamicbehaviour and physical properties. Ph.D Thesis, University of British Columbia. → pages9, 21, 42, 45, 116Kaitna, R., Palucis, M. C., Yohannes, B., Hill, K. M., & Dietrich, W. E. (2016). Effects of coarsegrain size distribution and fine particle content on pore fluid pressure and shear behavior inexperimental debris flows. Journal of Geophysical Research: Earth Surface, 121(2), 415–441,https://doi.org/10.1002/2015JF003725. → page 21Kang, H. & Kim, Y. (2016). The physical vulnerability of different types of building structure todebris flow events. Natural Hazards, 80(3), 1475–1493,https://doi.org/10.1007/s11069-015-2032-z. → page 18Kelman, M. C., Russell, J. K., & Hickson, C. J. (2002). Effusive intermediate glaciovolcanism in theGaribaldi volcanic belt, southwestern British Columbia, Canada. Geological Society, London,Special Publications, 202(1), 195–211, https://doi.org/10.1144/GSL.SP.2002.202.01.10. →page 4Lague, D., Brodu, N., & Leroux, J. (2013). Accurate 3D comparison of complex topography withterrestrial laser scanner: Application to the Rangitikei canyon (N-Z). ISPRS Journal ofPhotogrammetry and Remote Sensing, 82, 10–26,https://doi.org/10.1016/j.isprsjprs.2013.04.009. → page 64Lancaster, S., Hayes, S., & Grant, G. (2003). Effects of wood on debris flow runout in smallmountain watersheds. Water Resources Research, 39, 1168,https://doi.org/10.1029/2001wr001227. → page 22Lau, C.-A. (2017). Channel scour on temperate alluvial fans in British Columbia. Master’s thesis,Simon Fraser University. → pages xi, 9, 14, 45, 61, 65, 69, 74, 78, 131, 132Legros, F. (2002). The mobility of long-runout landslides. Engineering Geology, 63,https://doi.org/10.1016/S0013-7952(01)00090-4. → pages 20, 35, 37Linder, W. (2016). Digital Photogrammetry: A Practical Course. Berlin Heidelberg:Springer-Verlag, 4th edition. https://doi.org/10.1007/978-3-662-50463-5. → page 50Lorenzini, G. & Mazza, N. (2004). Debris flow: Phenomenology and rheological modelling. WitPress. → page 12Mackey, S. D. & Bridge, J. S. (1995). Three-dimensional model of alluvial stratigraphy; theory andapplications. Journal of Sedimentary Research, 65(1b), 7–31,https://doi.org/10.1306/D42681D5-2B26-11D7-8648000102C1865D. → page 27Major, J. J. & Iverson, R. M. (1999). Debris-flow deposition: Effects of pore-fluid pressure andfriction concentrated at flow margins. Geological Society of America Bulletin, 111,1424–1434, https://doi.org/10.1130/0016-7606(1999)111〈1424:DFDEOP〉2.3.CO;2. → page21160Martha, T. R., Kerle, N., Jetten, V., van Westen, C. J., & Kumar, K. V. (2010). Characterisingspectral, spatial and morphometric properties of landslides for semi-automatic detection usingobject-oriented methods. Geomorphology, 116(1), 24–36,https://doi.org/10.1016/j.geomorph.2009.10.004. → page 54May, C. L. (2002). Debris flows through different forest age classes in the central Oregon CoastRange. Journal of the American Water Resources Association, 38(4), 1097–1113,https://doi.org/10.1111/j.1752-1688.2002.tb05549.x. → page 22McDougall, S. (2017). 2014 Canadian Geotechnical Colloquium: Landslide runout analysis —current practice and challenges. Canadian Geotechnical Journal, 54(5), 605–620,https://doi.org/10.1139/cgj-2016-0104. → pages 1, 2, 18, 37McDougall, S. & Hungr, O. (2004). A model for the analysis of rapid landslide motion acrossthree-dimensional terrain. Canadian Geotechnical Journal, 41(6), 1084–1097,https://doi.org/10.1139/t04-052. → pages xi, 1, 18, 20Melton, M. A. (1965). The Geomorphic and Paleoclimatic Significance of Alluvial Deposits inSouthern Arizona. The Journal of Geology, 73(1), 1–38, https://doi.org/10.1086/627044. →page 75Millard, T., Wilford, D., & Oden, M. (2006). Coastal fan destabilization and forest management.Technical Report TR-034, Research Section, Coast Forest Region, BC Ministry of Forests,Nanaimo, BC. → pages 24, 27, 28, 131, 133Miller, D. J. & Burnett, K. M. (2008). A probabilistic model of debris-flow delivery to streamchannels, demonstrated for the Coast Range of Oregon, USA. Geomorphology, 94(1-2),184–205, https://doi.org/10.1016/j.geomorph.2007.05.009. → pages 22, 23, 32, 35, 38, 151Mitchell, A., McDougall, S., Nolde, N., Brideau, M.-A., Whittall, J., & Aaron, J. B. (2020). Rockavalanche runout prediction using stochastic analysis of a regional dataset. Landslides, 17(4),777–792, https://doi.org/10.1007/s10346-019-01331-3. → pages 31, 37Miura, H. (2019). Fusion Analysis of Optical Satellite Images and Digital Elevation Model forQuantifying Volume in Debris Flow Disaster. Remote Sensing, 11(9),https://doi.org/10.3390/rs11091096. → page 54Mohrig, D., Heller, P. L., Paola, C., & Lyons, W. J. (2000). Interpreting avulsion process fromancient alluvial sequences: Guadalope-Matarranya system (Northern Spain) and Wasatchformation (Western Colorado). Bulletin of the Geological Society of America, 112(12),1787–1803, https://doi.org/10.1130/0016-7606(2000)112〈1787:IAPFAA〉2.0.CO;2. → page27Monger, J. W. & Journeay, J. M. (1994). Guide to the Geology and Tectonic Evolution of theSouthern Coast Mountains. Open-File 2490, Geological Survey of Canada. → page 4Nolde, N. & Joe, H. (2013). A Bayesian extreme value analysis of debris flows: Extreme ValueAnalysis of Debris Flows. Water Resources Research, 49(10), 7009–7022,https://doi.org/10.1002/wrcr.20494. → page 140O’Brien, J. S., Julien, P. Y., & Fullerton, W. T. (1993). Two-dimensional water flood and mudflowsimulation. Journal of Hydraulic Engineering, 119(2), 244–261,https://doi.org/10.1061/(ASCE)0733-9429(1993)119:2(244). → page 18Parsons, J. D., Whipple, K. X., & Simoni, A. (2001). Experimental study of the grain-flow,fluid-mud transition in debris flows. The Journal of Geology, 109(4), 427–447,https://doi.org/10.1086/320798. → page 21PCIC (2012). Plan2Adapt Climate Information Tool161https://pacificclimate.org/analysis-tools/plan2adapt/. → page 7PCIC (2014). High Resolution Climatology https://data.pacificclimate.org/portal/bc prism/map/. →pages xi, 7, 9Pederson, C. A., Santi, P. M., & Pyles, D. R. (2015). Relating the compensational stacking ofdebris-flow fans to characteristics of their underlying stratigraphy: Implications for geologichazard assessment and mitigation. Geomorphology, 248, 47–56,https://doi.org/10.1016/j.geomorph.2015.06.030. → pages 26, 28, 131, 132Phillips, C. J. & Davies, T. R. H. (1991). Determining rheological parameters of debris flow material.Geomorphology, 4(2), 101–110, https://doi.org/10.1016/0169-555X(91)90022-3. → page 21Pierson, T. C. (1986). Flow behavior of chanellized debris flows, Mount St. Helens, Washington. InA. Abrahams (Ed.), Hillslope Processes (pp. 269–296). Boston: Allen & Unwin. → pagesxi, 12, 14Pierson, T. C. (2004). Distinguishing between Debris Flows and Floods from Field Evidence inSmall Watersheds. Fact Sheet 2004-3142, U.S. Geological Survey. → pages 12, 61Pierson, T. C. (2005). Hyperconcentrated flow — transitional process between water flow and debrisflow. In M. Jakob & O. Hungr (Eds.), Debris-flow Hazards and Related Phenomena (pp.159–202). Berlin, Heidelberg: Springer Praxis Books. 10.1007/3-540-27129-5 8. → page 13Pike, R. J. (2000). Geomorphometry - diversity in quantitative surface analysis. Progress in PhysicalGeography: Earth and Environment, 24(1), 1–20,https://doi.org/10.1177/030913330002400101. → page 74Pike, R. J. & Wilson, S. E. (1971). Elevation-relief ratio, hypsometric integral, and geomorphicarea-altitude analysis. Bulletin of the Geological Society of America, 82(4), 1079–1084,https://doi.org/10.1130/0016-7606(1971)82[1079:ERHIAG]2.0.CO;2. → page 77Planet (2019). Planet Application Program Interface: In Space for Life on Earth. San Fransisco, CA.https://api.planet.com. → pages xii, 54, 56, 57, 59, 88Prochaska, A. B., Santi, P. M., Higgins, J. D., & Cannon, S. H. (2008). Debris-flow runoutpredictions based on the average channel slope (ACS). Engineering Geology, 98(1-2), 29–40,https://doi.org/10.1016/j.enggeo.2008.01.011. → page 37Quan Luna, B. (2012). Chapter 6: Application of a Monte Carlo method to debris flow run-outmodeling. Ph.D Thesis, University of Twente, Enschede, Netherlands. → page 19Quan Luna, B., Blahut, J., van Westen, C. J., Sterlacchini, C. J. S., van Asch, T. W., & Akbas, S. O.(2011). The application of numerical debris flow of modelling for the generation physicalvulnerability curves. Natural Hazards and Earth System Sciences, 11, 2047–2060,https://doi.org/10.5194/nhess-11-2047-2011. → page 18Quan Luna, B., et al. (2014). Methods for debris flow hazard and risk assessment. In T. Asch, J.Corominas, S. Greiving, J.-P. Malet, & S. Sterlacchini (Eds.), Mountain Risks: FromPrediction to Management and Governance (pp. 133–177). Springer. → page 18Reitz, M. D. & Jerolmack, D. J. (2012). Experimental alluvial fan evolution: Channel dynamics,slope controls, and shoreline growth. Journal of Geophysical Research: Earth Surface,117(F2), https://doi.org/10.1029/2011JF002261. → page 15Rickenmann, D. (1999). Empirical Relationships for Debris Flows. Natural Hazards, 19, 47–77,https://doi.org/10.1023/A:1008064220727. → pages xi, 1, 29, 30, 31, 32, 37Rickenmann, D. (2005). Runout prediction methods. In M. Jakob & O. Hungr (Eds.), Debris-flowHazards and Related Phenomena (pp. 305–324). Berlin, Heidelberg: Springer Praxis Books.10.1007/3-540-27129-5 13. → pages 29, 36162Rickenmann, D. & Zimmermann, M. (1993). The 1987 debris flows in Switzerland: documentationand analysis. Geomorphology, 8(2-3), 175–189,https://doi.org/10.1016/0169-555X(93)90036-2. → page 31Roberti, G. (2018). Mount Meager, a glaciated volcano in a changing cryosphere: hazards and riskchallenges. Ph.D Thesis, Simon Fraser University. → pages 49, 50, 51Rouse, J. J., Haas, R., Schell, J., Deering, D., & Harlan, J. (1974). Monitoring the vernaladvancements and retrogradation (green wave effect) of natural vegetation. Type I ProgressReport - Number 7, Texas A&M University, Remote Sensing Center, College Station, TX. →page 54Ryder, J. M. (1969). Alluvial fans of post-glacial environments within British Columbia. Ph.DThesis, University of British Columbia, Vancouver, BC. → page 116Ryder, J. M. (1971). Some Aspects of the Morphometry of Paraglacial Alluvial Fans inSouth-central British Columbia. Canadian Journal of Earth Sciences, 8(10), 1252–1264,https://doi.org/10.1139/e71-114. → pages 5, 46Ryder, J. M., Fulton, R., & Clague, J. J. (1991). The Cordilleran Ice Sheet and the GlacialGeomorphology of Southern and Central British Colombia. Ge´ographie Physique etQuaternaire, 45(3), 365–377, https://doi.org/10.7202/032882ar. → page 5Santi, P. M. (2014). Precision and Accuracy in Debris-Flow Volume Measurement. Environmental &Engineering Geoscience, 20(4), 349–359, https://doi.org/10.2113/gseegeosci.20.4.349. →page 37Santi, P. M., Pyles, D., & Pederson, C. (2017). Debris Flow Avulsion. International Journal ofErosion Control Engineering, 10(1), 67–73, https://doi.org/10.13101/ijece.10.67. → pages26, 131Scheidegger, A. (1973). On the prediction of the reach and velocity of catastrophic landslides. RockMechanics, 5, https://doi.org/10.1007/BF01301796. → page 30Scheidl, C. & Rickenmann, D. (2008). Depositional characteristics and runout of alpine debris flows.In Interpraevent’08, volume 1 (pp. 477–488). Dornbirn, Austria. → page 33Scheidl, C. & Rickenmann, D. (2010). Empirical prediction of debris-flow mobility and depositionon fans. Earth Surface Processes and Landforms, 35(2), 157–173,https://doi.org/10.1002/esp.1897. → pages 19, 23, 29, 34, 35, 74, 132Schilling, S. (1998). LAHARZ: GIS programs for automated mapping of lahar-inundation hazardzones. Open-File Report 98-638, U.S. Geological Survey, Vancouver, Washington. → page 35Schneuwly-Bollschweiler, M. & Stoffel, M. (2013). Dendrogeomorphology – Tracking Past Eventswith Tree Rings. In M. Schneuwly-Bollschweiler, M. Stoffel, & F. Rudolf-Miklau (Eds.),Dating Torrential Processes on Fans and Cones: Methods and Their Application for Hazardand Risk Assessment (pp. 165–178). Dordrecht: Springer Netherlands. → page 86Schraml, K., Thomschitz, B., McArdell, B. W., Graf, C., & Kaitna, R. (2015). Modeling debris-flowrunout patterns on two alpine fans with different dynamic simulation models. Natural Hazardsand Earth System Sciences, 15(7), 1483–1492, https://doi.org/10.5194/nhess-15-1483-2015.→ page 20Schumm, S. A., Mosley, M. P., & Weaver, W. (1987). Experimental Fluvial Geomorphology. NewYork, NY: John Wiley & Sons Inc. → page 15Schu¨rch, P., et al. (2016). Quantitative reconstruction of late Holocene surface evolution on an alpinedebris-flow fan. Geomorphology, 275, 46–57,https://doi.org/10.1016/j.geomorph.2016.09.020. → page 26163Schu¨rch, P., Densmore, A. L., Rosser, N., & McArdell, B. (2011a). A novel debris-flow fanevolution model based on debris flow monitoring and LiDAR topography. In R. Genevois, D.Hamilton, & A. Prestininzi (Eds.), Debris-Flow Hazards Mitigation: Mechanics, Prediction,and Assessment (pp. 263–272). Padua, Italy. https://doi.org/10.4408/IJEGE.2011-03.B-031.→ page 38Schu¨rch, P., Densmore, A. L., Rosser, N. J., & McArdell, B. W. (2011b). Dynamic controls onerosion and deposition on debris-flow fans. Geology, 39(9), 827–830,https://doi.org/10.1130/G32103.1. → pages 12, 21, 23, 27, 130Scott, K. M., Vallance, J. W., & Pringle, P. T. (1995). Sedimentology, behavior, and hazards of debrisflows at Mount Rainier, Washington. Professional Paper 1547, U.S. Geological Survey. →page 22Simoni, A., Mammoliti, M., & Berti, M. (2011). Uncertainty of debris flow mobility relationshipsand its influence on the prediction of inundated areas. Geomorphology, 132(3-4), 249–259,https://doi.org/10.1016/j.geomorph.2011.05.013. → pages 35, 37Slingerland, R. & Smith, N. D. (1998). Necessary conditions for a meandering-river avulsion.Geology, 26(5), 435–438,https://doi.org/10.1130/0091-7613(1998)026〈0435:NCFAMR〉2.3.CO;2. → page 27Stouthamer, E. & Berendsen, H. J. (2007). Avulsion: the relative roles of autogenic and allogenicprocesses. Sedimentary Geology, 198(3-4), 309–325,https://doi.org/10.1016/j.sedgeo.2007.01.017. → page 23Sturzenegger, M., Holm, K., Lau, C.-A., & Jakob, M. (2019). Semi-automated regional scaledebris-flow and debris-flood susceptibility mapping based on digital elevation model metricsand Flow-R software. In J. Kean, J. Coe, P. Santi, & B. Guillen (Eds.), Debris-Flow HazardsMitigation: Mechanics, Prediction, and Assessment (pp. 855–862). Golden, CO. → page 2Sutton, E. (2011). Influence of Hydrometeorological Controls on Debris Flows Near Chilliwack,British Columbia. Master’s thesis, Simon Fraser University, Vancouver, BC. → page 42Suwa, H., Okano, K., & Kanno, T. (2009). Behavior of debris flows monitored on test slopes ofKamikamihorizawa Creek, Mount Yakedake, Japan. International Journal of Erosion ControlEngineering, 2(2), 33–45, https://doi.org/10.13101/ijece.2.33. → pages 103, 105Suwa, H., Okano, K., & Kanno, T. (2011). Forty years of debris flow monitoring atKamikamihorizawa Creek, Mount Yakedake, Japan. In R. Genevois, D. Hamilton, & A.Prestininzi (Eds.), Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment(pp. 605–613). Padua, Italy. → pages 103, 105Tang, C., Zhu, J., Chang, M., Ding, J., & Qi, X. (2012). An empirical–statistical model for predictingdebris-flow runout zones in the Wenchuan earthquake area. Quaternary International, 250,63–73, https://doi.org/10.1016/j.quaint.2010.11.020. → pages xi, 30, 32, 131Thompson, S. C., Clague, J. J., & Evans, S. G. (1997). Holocene Activity of the Mt. Currie Scarp,Coast Mountains, British Columbia, and Implications for its Origin. Environmental andEngineering Geoscience, III(3), 329–348, https://doi.org/10.2113/gseegeosci.III.3.329. →page 111Tiranti, D. & Deangeli, C. (2015). Modeling of debris flow depositional patterns according to thecatchment and sediment source area characteristics. Frontiers in Earth Science, 3, 8,https://doi.org/10.3389/feart.2015.00008. → page 21Tucker, C. J. (1979). Red and photographic infrared linear combinations for monitoring vegetation.Remote Sensing of Environment, 8(2), 127–150,164https://doi.org/10.1016/0034-4257(79)90013-0. → page 54To¨rnqvist, T. & S. Bridge, J. (2002). Spatial variation of overbank aggradation rate and its influenceon avulsion frequency. Sedimentology, 49, 891–905,https://doi.org/10.1046/j.1365-3091.2002.00478.x. → page 27van Westen, C. J., Castellanos, E., & Kuriakose, S. L. (2008). Spatial data for landslidesusceptibility, hazard, and vulnerability assessment: An overview. Engineering Geology,102(3), 112–131, https://doi.org/10.1016/j.enggeo.2008.03.010. → page 54VanDine, D. F. (1985). Debris flows and debris torrents in the Southern Canadian Cordillera.Canadian Geotechnical Journal, 22(1), 44–68, https://doi.org/10.1139/t85-006. → pages9, 12, 42Walker, I. R. & Pellatt, M. G. (2003). Climate Change in Coastal British Columbia — APaleoenvironmental Perspective. Canadian Water Resources Journal, 28(4), 531–566,https://doi.org/10.4296/cwrj2804531. → page 5Webb, R. H., Magirl, C. S., Griffiths, P. G., & Boyer, D. E. (2008). Debris Flows and Floods inSoutheastern Arizona from Extreme Precipitation in July 2006-Magnitude, Frequency, andSediment Delivery. Open-File Report 2008-1274, U.S. Geological Survey, Reston, Virginia.→ page 33Whipple, K. X. (1992). Predicting debris-flow runout and deposition on fans: the importance of theflow hydrograph. In Erosion, Debris-Flows and Environment in Mountain Regions:Proceedings of the Chengdu Symposium. International Association of Hydrological Sciences,volume 209 (pp. 337–345). → page 20Whipple, K. X. & Dunne, T. (1992). The influence of debris-flow rheology on fan morphology,Owens Valley, California. Geological Society of America Bulletin, 104(7), 887–900,https://doi.org/10.1130/0016-7606(1992)104〈0887:TIODFR〉2.3.CO;2. → pages12, 21, 22, 24, 90, 116, 120Wilford, D., Sakals, M., Innes, J., Sidle, R., & Bergerud, W. (2004). Recognition of debris flow,debris flood and flood hazard through watershed morphometrics. Landslides, 1, 61–66,https://doi.org/10.1007/s10346-003-0002-0. → pages xiii, 13, 74, 80, 81, 130, 131Williams, R. M. E., Zimbelman, J. R., & Johnston, A. K. (2006). Aspects of alluvial fan shapeindicative of formation process: A case study in southwestern California with application toMojave Crater fans on Mars. Geophysical Research Letters, 33(10),https://doi.org/10.1029/2005GL025618. → pages 78, 132Wilson, A. (2019). Glaciovolcanism in the Garibaldi volcanic belt: supplementary material. Ph.DThesis, University of British Columbia, Vancouver, BC. https://doi.org/10.14288/1.0388500.→ pages xi, 6Wood, W. F. & Snell, J. (1960). A Quantitative System for Classifying Landforms. U.S. ArmyTechical Report EP-124, Quartermaster Research & Engineering Command. → page 77Yu, F.-C., Chen, C.-Y., Chen, T.-C., Hung, F.-Y., & Lin, S.-C. (2006). A GIS Process forDelimitating Areas Potentially Endangered by Debris Flow. Natural Hazards, 37(1-2),169–189, https://doi.org/10.1007/s11069-005-4666-8. → pages 32, 33Zhang, S., Zhang, L.-M., Chen, H.-X., Yuan, Q., & Pan, H. (2013). Changes in runout distances ofdebris flows over time in the Wenchuan earthquake zone. Journal of Mountain Science, 10(2),281–292, https://doi.org/10.1007/s11629-012-2506-y. → pages 22, 23Zhou, W., Fang, J., Tang, C., & Yang, G. (2019). Empirical relationships for the estimation of debrisflow runout distances on depositional fans in the Wenchuan earthquake zone. Journal of165Hydrology, 577, https://doi.org/10.1016/j.jhydrol.2019.123932. → pages 32, 37Zimmermann, M., Mani, M., Gamma, P., Gsteiger, P., Heiniger, O., & Hunziker, G. (1997).Murganggefahr und Klimaa¨nderung - ein GIS-basierter Ansatz. Zu¨rich: vdf HochschulverlagAG, 1st edition. (in German). → pages 31, 131Zubrycky, S., Mitchell, A., Aaron, J., & McDougall, S. (2019). Preliminary calibration of anumerical runout model for debris flows in southwestern British Columbia. In J. Kean, J. Coe,P. Santi, & B. Guillen (Eds.), Debris-Flow Hazards Mitigation: Mechanics, Prediction, andAssessment (pp. 911–918). Golden, CO. → pages xi, 20, 132166Appendix AFan Site SummariesA one page data summary is provided for each fan in the dataset, including impact area mapping,fan site descriptor variables, spatial impact heatmaps, cumulative runout exceedance distributions,and field photographs (if available).167 Site 1: Abandoned  Easting (m) 461334 Northing (m) 5558054 No. impact areas: 7 No. events: 2 Normalizing fan length (m): 910 Normalizing fan arc-length (m): 943 Normalizing fan angle (°) 82 Watershed area (km2): 2.68 Melton ratio: 0.99 Fan area (km2): 0.41 Fan slope (°): 12.2 Average channel slope (°): 12.8 Fan ERR: 0.23 Normalized intersection point: - Truncated: Yes Bedrock geology: Intrusive        1994 orthophoto 168 Site 2: Endurance  Easting (m) 473304 Northing (m) 5548766 No. impact areas: 5 No. events: 3 Normalizing fan length (m): 637 Normalizing fan arc-length (m): 555 Normalizing fan angle (°) 69 Watershed area (km2): 8.68 Melton ratio: 0.66 Fan area (km2): 0.23 Fan slope (°): 7.4 Average channel slope (°): 7.3 Fan ERR: 0.42 Normalized intersection point: - Truncated: Yes Bedrock geology: Intrusive        1964 orthophoto 169 Site 3: Terminal  Easting (m) 475132 Northing (m) 5549082 No. impact areas: 6 No. events: 3 Normalizing fan length (m): 1292 Normalizing fan arc-length (m): 1199 Normalizing fan angle (°) 75 Watershed area (km2): 9.49 Melton ratio: 0.68 Fan area (km2): 0.78 Fan slope (°): 7.2 Average channel slope (°): 7.2 Fan ERR: 0.38 Normalized intersection point: 0.83 Truncated: Yes Bedrock geology: Volcanic     Northern channel at the fan toe, looking up-fan. Photo provided by Lauren Vincent.  Northern channel at the fan toe, looking down-fan toward Squamish River. Photo provided by Lauren Vincent. 1994 orthophoto 170 Site 4: Middle Lillooet W  Easting (m) 474226 Northing (m) 5606122 No. impact areas: 5 No. events: 2 Normalizing fan length (m): 832 Normalizing fan arc-length (m): 1334 Normalizing fan angle (°) 115 Watershed area (km2): 2.27 Melton ratio: 1.17 Fan area (km2): 0.47 Fan slope (°): 12.7 Average channel slope (°): 9.8 Fan ERR: 0.25 Normalized intersection point: 0.48 Truncated: Yes Bedrock geology: Intrusive     On left bank levee at fan apex looking upslope.   Debris at distal fan, upslope of logging road.  2015 bare earth lidar1  1ALS courtesy of Brian Menounos (University of Northern British Columbia), John Clague, and Gioachino Roberti (Simon Fraser University).  171 Site 5: Middle Lillooet C  Easting (m) 475684 Northing (m) 5605380 No. impact areas: 5 No. events: 4 Normalizing fan length (m): 870 Normalizing fan arc-length (m): 1310 Normalizing fan angle (°) 112 Watershed area (km2): 2.68 Melton ratio: 1.11 Fan area (km2): 0.59 Fan slope (°): 13.7 Average channel slope (°): 11.9 Fan ERR: 0.26 Normalized intersection point: 0.54 Truncated: Yes Bedrock geology: Intrusive     At fan apex, looking upslope.  Debris path on lower fan, overlooking Lillooet river valley.   2015 bare earth lidar1  1ALS courtesy of Brian Menounos (University of Northern British Columbia), John Clague, and Gioachino Roberti (Simon Fraser University).  172 Site 6: Middle Lillooet E  Easting (m) 477147 Northing (m) 5604529 No. impact areas: 7 No. events: 5 Normalizing fan length (m): 810 Normalizing fan arc-length (m): 1287 Normalizing fan angle (°) 115 Watershed area (km2): 2.84 Melton ratio: 1.08 Fan area (km2): 0.52 Fan slope (°): 13.1 Average channel slope (°): 12.5 Fan ERR: 0.27 Normalized intersection point: 0.58 Truncated: Yes Bedrock geology: Intrusive     Incised active channel near fan apex.   Levee and mud line, mid-fan. 2015 bare earth lidar1  1ALS courtesy of Brian Menounos (University of Northern British Columbia), John Clague, and Gioachino Roberti (Simon Fraser University).  173  Site 7: Petersen  Easting (m) 488876 Northing (m) 5589224 No. impact areas: 7 No. events: 3 Normalizing fan length (m): 849 Normalizing fan arc-length (m): 1266 Normalizing fan angle (°) 95 Watershed area (km2): 8.20 Melton ratio: 0.48 Fan area (km2): 0.53 Fan slope (°): 10.7 Average channel slope (°): 10.7 Fan ERR: 0.28 Normalized intersection point: - Truncated: Yes Bedrock geology: Intrusive      1994 orthophoto 174 Site 8: Upper Rutherford  Easting (m) 498423 Northing (m) 5571312 No. impact areas: 2 No. events: 1 Normalizing fan length (m): 1206 Normalizing fan arc-length (m): 1168 Normalizing fan angle (°) 66 Watershed area (km2): 2.72 Melton ratio: 0.88 Fan area (km2): 0.59 Fan slope (°): 12.6 Average channel slope (°): 12.6 Fan ERR: 0.21 Normalized intersection point: - Truncated: Yes Bedrock geology: Intrusive     Lobe near fan apex, west of active channel.  Active channel on eastern side of lower fan. 1981 orthophoto 175 Site 9: No Law  Easting (m) 500317 Northing (m) 5570349 No. impact areas: 4 No. events: 3 Normalizing fan length (m): 587 Normalizing fan arc-length (m): 790 Normalizing fan angle (°) 91 Watershed area (km2): 3.37 Melton ratio: 0.89 Fan area (km2): 0.24 Fan slope (°): 12.9 Average channel slope (°): 11.2 Fan ERR: 0.29 Normalized intersection point: 0.95 Truncated: Yes Bedrock geology: Intrusive     Debris on upper fan, looking upslope toward apex.  Debris field upslope of logging road, mid-fan. 1981 orthophoto 176 Site 10: Sootip  Easting (m) 500332 Northing (m) 5569470 No. impact areas: 3 No. events: 2 Normalizing fan length (m): 387 Normalizing fan arc-length (m): 240 Normalizing fan angle (°) 43 Watershed area (km2): 0.83 Melton ratio: 1.26 Fan area (km2): 0.05 Fan slope (°): 17.2 Average channel slope (°): 16.2 Fan ERR: 0.32 Normalized intersection point: - Truncated: Yes Bedrock geology: Volcanic     Overlooking avulsion lobe and watershed. Photograph taken from opposite side of valley.     1981 orthophoto 177 Site 11: Lower Rutherford W  Easting (m) 503650 Northing (m) 5570723 No. impact areas: 3 No. events: 2 Normalizing fan length (m): 508 Normalizing fan arc-length (m): 375 Normalizing fan angle (°) 57 Watershed area (km2): 2.02 Melton ratio: 1.04 Fan area (km2): 0.14 Fan slope (°): 16.0 Average channel slope (°): 15.9 Fan ERR: 0.36 Normalized intersection point: 0.72 Truncated: Yes Bedrock geology: Intrusive     Boulder levee on upper fan, looking upslope.   Megaclast from 2018 event, upslope of  logging road on lower fan. 1981 orthophoto 178 Site 12: Lower Rutherford E  Easting (m) 503916 Northing (m) 5570658 No. impact areas: 3 No. events: 2 Normalizing fan length (m): 587 Normalizing fan arc-length (m): 268 Normalizing fan angle (°) 50 Watershed area (km2): 1.45 Melton ratio: 1.12 Fan area (km2): 0.12 Fan slope (°): 13.6 Average channel slope (°): 13.2 Fan ERR: 0.40 Normalized intersection point: 0.88 Truncated: Yes Bedrock geology: Intrusive     At fan apex looking downslope.  Bouldery channel plug on upper fan. 1981 orthophoto 179 Site 13: Ross  Easting (m) 503460 Northing (m) 5587279 No. impact areas: 4 No. events: 4 Normalizing fan length (m): 1348 Normalizing fan arc-length (m): 1098 Normalizing fan angle (°) 69 Watershed area (km2): 5.79 Melton ratio: 0.78 Fan area (km2): 0.81 Fan slope (°): 10.4 Average channel slope (°): 11.3 Fan ERR: 0.26 Normalized intersection point: 0.74 Truncated: Yes Bedrock geology: Intrusive     Debris path on upper fan.  Fine-grained deposits on distal fan. 1980 orthophoto 180 Site 14: Nightmare  Easting (m) 503560 Northing (m) 5587870 No. impact areas: 3 No. events: 2 Normalizing fan length (m): 377 Normalizing fan arc-length (m): 501 Normalizing fan angle (°) 111 Watershed area (km2): 1.46 Melton ratio: 1.47 Fan area (km2): 0.09 Fan slope (°): 16.8 Average channel slope (°): 15.6 Fan ERR: 0.40 Normalized intersection point: 0.62 Truncated: Yes Bedrock geology: Intrusive     Overgrown lobe on lower fan.   1980 orthophoto 181 Site 15: Fergusson  Easting (m) 515823 Northing (m) 5625672 No. impact areas: 4 No. events: 2 Normalizing fan length (m): 990 Normalizing fan arc-length (m): 974 Normalizing fan angle (°) 60 Watershed area (km2): 1.36 Melton ratio: 0.87 Fan area (km2): 0.46 Fan slope (°): 10.6 Average channel slope (°): 10.5 Fan ERR: 0.26 Normalized intersection point: - Truncated: No Bedrock geology: Sedimentary                 1965 orthophoto 182 Site 16: Currie B  Easting (m) 515965 Northing (m) 5570375 No. impact areas: 10 No. events: 6 Normalizing fan length (m): 834 Normalizing fan arc-length (m): 865 Normalizing fan angle (°) 76 Watershed area (km2): 2.70 Melton ratio: 1.30 Fan area (km2): 0.36 Fan slope (°): 10.7 Average channel slope (°): 8.2 Fan ERR: 0.25 Normalized intersection point: 0.32 Truncated: No Bedrock geology: Intrusive     Looking upstream at fan apex.   Recent debris contact with mossy lobe, mid fan looking down slope.  2017 bare earth lidar1 1ALS courtesy of Squamish Lillooet Regional District.  183 Site 17: Currie C  Easting (m) 516839 Northing (m) 5570140 No. impact areas: 11 No. events: 9 Normalizing fan length (m): 1045 Normalizing fan arc-length (m): 692 Normalizing fan angle (°) 58 Watershed area (km2): 1.20 Melton ratio: 1.58 Fan area (km2): 0.44 Fan slope (°): 17.5 Average channel slope (°): 15.8 Fan ERR: 0.32 Normalized intersection point: 0.66 Truncated: No Bedrock geology: Intrusive     On the left bank of the incised channel near the fan apex overlooking Currie D deposits.    Paleochannel, mid fan.   2017 bare earth lidar1 1ALS courtesy of Squamish Lillooet Regional District.  184 Site 18: Currie D  Easting (m) 517411 Northing (m) 5570114 No. impact areas: 15 No. events: 11 Normalizing fan length (m): 1655 Normalizing fan arc-length (m): 1612 Normalizing fan angle (°) 70 Watershed area (km2): 1.66 Melton ratio: 1.14 Fan area (km2): 1.28 Fan slope (°): 14.1 Average channel slope (°): 14.7 Fan ERR: 0.21 Normalized intersection point: 0.49 Truncated: No Bedrock geology: Intrusive     Main incised channel, mid fan, looking upstream.  Debris lobes on lower fan, looking west.  2017 bare earth lidar1 1ALS courtesy of Squamish Lillooet Regional District.  185  Site 19: Deepa  Easting (m) 523562 Northing (m) 5588576 No. impact areas: 6 No. events: 4 Normalizing fan length (m): 846 Normalizing fan arc-length (m): 549 Normalizing fan angle (°) 41 Watershed area (km2): 1.12 Melton ratio: 1.22 Fan area (km2): 0.29 Fan slope (°): 15.0 Average channel slope (°): 14.1 Fan ERR: 0.25 Normalized intersection point: - Truncated: No Bedrock geology: Intrusive      1969 orthophoto 186 Site 20: Neff  Easting (m) 529458 Northing (m) 5593780 No. impact areas: 7 No. events: 5 Normalizing fan length (m): 897 Normalizing fan arc-length (m): 824 Normalizing fan angle (°) 85 Watershed area (km2): 3.29 Melton ratio: 1.02 Fan area (km2): 0.42 Fan slope (°): 11.7 Average channel slope (°): 10.9 Fan ERR: 0.24 Normalized intersection point: 0.63 Truncated: No Bedrock geology: Sedimentary     Main channel on upper fan, incised 10-12 m.  Debris on lower fan, looking southeast. 1ALS courtesy of BC Hydro.  2015 bare earth lidar1 187  Site 21: Catiline  Easting (m) 535567 Northing (m) 5568500 No. impact areas: 5 No. events: 4 Normalizing fan length (m): 1278 Normalizing fan arc-length (m): 849 Normalizing fan angle (°) 46 Watershed area (km2): 3.04 Melton ratio: 0.94 Fan area (km2): 0.56 Fan slope (°): 14.4 Average channel slope (°): 13.8 Fan ERR: 0.32 Normalized intersection point: 0.91 Truncated: Yes Bedrock geology: Intrusive     Upper fan channel plug. Photo courtesy of BGC.  Bridge on lower fan. Photo courtesy of BGC. 1ALS courtesy of Squamish Lillooet Regional District.  2014 bare earth lidar1 188 Site 22: Fern  Easting (m) 537775 Northing (m) 5462900 No. impact areas: 4 No. events: 2 Normalizing fan length (m): 868 Normalizing fan arc-length (m): 636 Normalizing fan angle (°) 49 Watershed area (km2): 0.86 Melton ratio: 1.16 Fan area (km2): 0.35 Fan slope (°): 6.3 Average channel slope (°): 6.3 Fan ERR: 0.24 Normalized intersection point: 0.53 Truncated: No Bedrock geology: Intrusive     Debris on upper fan.  Debris on lower fan. 1982 orthophoto 189 Site 23: Bear  Easting (m) 550167 Northing (m) 5616375 No. impact areas: 5 No. events: 4 Normalizing fan length (m): 1030 Normalizing fan arc-length (m): 1319 Normalizing fan angle (°) 85 Watershed area (km2): 2.20 Melton ratio: 1.22 Fan area (km2): 0.67 Fan slope (°): 11.2 Average channel slope (°): 10 Fan ERR: 0.34 Normalized intersection point: 0.18 Truncated: No Bedrock geology: Sedimentary     Incised channel on upper fan. Photo courtesy of BGC.  Channelized deposit, mid fan. Photo courtesy of BGC. 1ALS courtesy of Squamish Lillooet Regional District.  2017 bare earth lidar1 190 Site 24: Fountain N  Easting (m) 578798 Northing (m) 5616465 No. impact areas: 12 No. events: 6 Normalizing fan length (m): 1729 Normalizing fan arc-length (m): 1267 Normalizing fan angle (°) 53 Watershed area (km2): 0.94 Melton ratio: 1.16 Fan area (km2): 1.25 Fan slope (°): 10.6 Average channel slope (°): 11.4 Fan ERR: 0.30 Normalized intersection point: 0.16 Truncated: No Bedrock geology: Sedimentary     Channel incised into paleofan, near fan apex.  Channelized deposit, mid fan.  2019 bare earth lidar, 1997 orthophoto 191 Site 25: Fountain S  Easting (m) 579160 Northing (m) 5615932 No. impact areas: 11 No. events: 7 Normalizing fan length (m): 1347 Normalizing fan arc-length (m): 494 Normalizing fan angle (°) 41 Watershed area (km2): 0.35 Melton ratio: 1.74 Fan area (km2): 0.43 Fan slope (°): 14.3 Average channel slope (°): 12.4 Fan ERR: 0.34 Normalized intersection point: 0.45 Truncated: No Bedrock geology: Sedimentary     Channelized flow path, mid fan.  Terminal lobe. 2019 bare earth lidar, 1997 orthophoto 192 Site 26: Cheam W  Easting (m) 594952 Northing (m) 5451397 No. impact areas: 4 No. events: 3 Normalizing fan length (m): 1088 Normalizing fan arc-length (m): 1094 Normalizing fan angle (°) 73 Watershed area (km2): 1.87 Melton ratio: 1.38 Fan area (km2): 0.63 Fan slope (°): 10.5 Average channel slope (°): 9.4 Fan ERR: 0.28 Normalized intersection point: 0.16 Truncated: No Bedrock geology: Sedimentary     Debris near fan apex, looking upslope.   Recent deposition on lower-mid fan. 2017 bare earth lidar1 1ALS courtesy of BC Ministry of Transportation and Infrastructure. 193 Site 27: Cheam E  Easting (m) 595063 Northing (m) 5451381 No. impact areas: 6 No. events: 2 Normalizing fan length (m): 962 Normalizing fan arc-length (m): 555 Normalizing fan angle (°) 35 Watershed area (km2): 2.25 Melton ratio: 1.26 Fan area (km2): 0.25 Fan slope (°): 11.7 Average channel slope (°): 10.8 Fan ERR: 0.28 Normalized intersection point: 0.97 Truncated: Yes Bedrock geology: Sedimentary     Fan apex.   Active channel, mid fan.  1ALS courtesy of BC Ministry of Transportation and Infrastructure. 2017 bare earth lidar1 194  Site 28: Hope  Easting (m) 613674 Northing (m) 5469474 No. impact areas: 1 No. events: 1 Normalizing fan length (m): 810 Normalizing fan arc-length (m): 595 Normalizing fan angle (°) 53 Watershed area (km2): 0.65 Melton ratio: 1.37 Fan area (km2): 0.28 Fan slope (°): 16.7 Average channel slope (°): 13.4 Fan ERR: 0.39 Normalized intersection point: 0.82 Truncated: No Bedrock geology: Metamorphic      2017 bare earth lidar1 1ALS courtesy of BC Ministry of Transportation and Infrastructure. 195  Site 29: Allard  Easting (m) 616587 Northing (m) 5489888 No. impact areas: 8 No. events: 6 Normalizing fan length (m): 1566 Normalizing fan arc-length (m): 988 Normalizing fan angle (°) 43 Watershed area (km2): 0.65 Melton ratio: 1.12 Fan area (km2): 0.76 Fan slope (°): 15.5 Average channel slope (°): 12.5 Fan ERR: 0.24 Normalized intersection point: 0.70 Truncated: Yes Bedrock geology: Intrusive      2015 bare earth lidar1 1ALS courtesy of Canadian National Railway. 196  Site 30: Anonymous  Easting (m) - Northing (m) - No. impact areas: 1 No. events: 1 Normalizing fan length (m): 445 Normalizing fan arc-length (m): 260 Normalizing fan angle (°) 64 Watershed area (km2): 0.21 Melton ratio: 1.52 Fan area (km2): 0.08 Fan slope (°): 14.7 Average channel slope (°): 14.2 Fan ERR: 0.32 Normalized intersection point: 0.52 Truncated: No Bedrock geology: Intrusive      197Appendix BSupplementary Material: GIS DataGIS data for the SWBC dataset (Anonymous case excluded) have been made available assupplementary material. The files listed below can be found on cIRcle (UBC digital repository)in a .zip file. Shapefiles are projected in the UTM NAD83 Zone 10 coordinate system. Metadataassociated with each shapefile are listed below.1. apex. Apex point locations. Metadata: fan ID, fan name, number of impact areas, number ofevents, number of observation years, observation length (years), normalizing fan radius (m),normalizing fan arc length (m), normalizing fan angle (◦), fan area (km2), overall fan slope(◦), average fan channel slope (◦), fan elevation relief ratio, normalized fan intersection point,watershed area (km2), watershed relief (km), Melton ratio, fan truncation, geology (rock class).2. fan. Fan boundary. Metadata: fan ID, fan name.3. watershed. Watershed boundary. Metadata: fan ID, fan name.4. impact areas. Impact area boundaries. Metadata: fan ID, fan name, impact area ID, year,temporal certainty class, spatial certainty class, avulsion classification, event classification(boolean), impact area (m2), deposit area (m2), volume (m3), volume estimation method, notes.5. flow paths. Flow path lines associated with each impact area. Metadata: fan ID, fan name,impact area ID.6. swbc data.xls. Metadata spreadsheet. First tab defines the metadata fields, and the remainingtabs correspond to data for each shapefile listed above.198Appendix CSupplementary Material: MATLABcodeMATLAB code and example inputs for the workflow described in Section 4.5.1 have been madeavailable as supplementary material. The files listed below can be found on cIRcle (UBC digitalrepository) in a .zip file.1. NERE-DF.m. MATLAB code for the Normalized Empirical Runout Estimator - Debris Flow.2. swbc.grd, swbc smoothed.grd. Empirical fan-normalized impact area heatmaps (ASCII grid)for the SWBC dataset (raw and smoothed).3. apex.shp, fan.shp, channel.shp. Example input geometry shapefiles.199

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}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            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.24.1-0391879/manifest

Comment

Related Items