UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Paleohydrology of the Bella Coola River basin : an assessment of environmental reconstruction Desloges, Joseph R. 1987-12-31

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata


831-UBC_1987_A1 D47.pdf [ 29.07MB ]
JSON: 831-1.0097187.json
JSON-LD: 831-1.0097187-ld.json
RDF/XML (Pretty): 831-1.0097187-rdf.xml
RDF/JSON: 831-1.0097187-rdf.json
Turtle: 831-1.0097187-turtle.txt
N-Triples: 831-1.0097187-rdf-ntriples.txt
Original Record: 831-1.0097187-source.json
Full Text

Full Text

PALEOHYDROLOGY OF THE BELLA COOLA RIVER BASIN AN ASSESSMENT OF ENVIRONMENTAL RECONSTRUCTION By JOSEPH ROBERT DESLOGES B.E.S., University of Waterloo, 1980 M.Sc, University of Wisconsin-Madison, 1982 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY i n THE FACULTY OF GRADUATE STUDIES (Department of Geography) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA OCTOBER 1987 w Joseph Robert Desloges, 1987 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of Geography The University of British Columbia 1956 Main Mall Vancouver, Canada V6T 1Y3 Date October 10, 1987  DE-6f.Vft-n i i ABSTRACT Recent geomorphic and hydrologic environments of a mid-latitude alpine basin are investigated under the integrative theme of paleohydrology. The aims of this research are: 1) to characterize the response of selected biological and geophysical elements to recent climatic change; 2) to determine the resolution and length of paleoenvironmental records in the study area; and 3) to ascertain the significance of observed and inferred environmental change over the Little Ice Age interval. 2 Bella Coola River drains 5050 km of glacierized mountains along the central coast of British Columbia. Biological elements examined on a basin-wide scale included: tree-growth in temperature and moisture-stressed environments, damage to trees in glacial and fluvial settings, pollen variations in a variety of sedimentary deposits and soil development. Geophysical elements include primarily glacio-lacustrine and floodplain sediments, glacier deposits and river channel morphology. A retrospective strategy was adopted by testing initially for the nature of relationships between synoptic climate, basin hydrology and element response during the period of instrument record (1900 AD to present). Inferences about pre-instrument environments were then made using the proxy data. Events of several types are characteristically mixed in a response record. Variations in Douglas and subalpine fir growth, glacio-lacustrine sedimentation rates, glacier fluctuations and shifting of the Bella Coola River reflect a combination of persistent and episodically extreme behavior. Glaciers appear to respond by advancing or retreating after departures in winter precipitation persistent for several years. Extreme events, particularly high-magnitude autumn floods, are not exclusively linked to a particular set of mean climatic departures. This makes inferences from proxy data such as floodplain deposits and flood-damaged i i i vegetation difficult. Periods of increased flood frequency are supposed to relate to an increase in floodplain sedimentation. Except in very favorable circumstances, paleoenvironmental methods do not have the resolution promised. Climatic information recoverable from tree-ring data and glacio-lacustrine sediments is of considerably lower than annual resolution. Statistically based climate models using proxy data as independent variables produce low levels of explained variance. Proxy data sources in the basin were largely restricted to the last 300 to kOO years or Little Ice Age interval. Most glaciers in the basin reached Little Ice Age maxima in the middle of the 19th century in response to below average temperatures and above average precipitation between approximately 1800 and 1855 AD. Tree-ring data and equilibrium line altitudes on glaciers indicate that precipitation was on average 25 to 30% greater than the 1951-1980 mean. Inferred below average temperatures in the early l8th century probably signaled the beginning of the Little Ice Age along the central coast; however, there was not a major response in glaciers until persistent positive departures in precipitation occurred. Recession of glaciers from Little Ice Age maxima was slowed by cooler and wetter conditions between I885 and 1900 AD. The persistence of warmer and drier conditions in the first half of the 20th century was exceptional in comparison with inferred climate of the last 330 years. Major floods in 1805/06, 1826, 1885 and 1896 correspond to intervals of increased precipitation. i v PALEOHYDROLOGY OF THE BELLA COOLA RIVER BASIN: ASSESSMENT OF ENVIRONMENTAL RECONSTRUCTION Chapter I: INTRODUCTION (1.1) Systems Framework 1 (1.2) Operational Approaches 4 Process modelsEmpirical models 5 Inductive models 7 (1.3) Objectives and Research Strategy 9 (1.4) Data Sources 10 (1.5) Organization 3 Chapter I I: STUDY AREA (2.0) Location 15 (2.1) Physiography 8 (2.2) Bedrock Geology 9 (2.3) Late Quaternary Geology 23 (2.it) Climate Setting(2.5) Hydrology 32 Average RegimeMajor Floods During the 20th Century 36 (2.6) Sources and Transfer of Clastic Sediment 8 Upland Sediment Sources itO (2.7) Pedologic Setting 4(2.8) Flora 4(2.9) Settlement and Logging History 50 Chapter III: HYDROPHYSICAL RECORDS OF ENVIRONMENTAL CHANGE: TESTS WITHIN THE INSTRUMENT PERIOD (3.0) Introduction . 52 (3.1) Post-1945 Synoptic Climatology of the N.E. Pacific Sector 5Secular Trends in Temperature and Precipitation 55 Comparison With Other Coastal Regions 57 Atmospheric Circulation 60 Frequency and Seasonal Persistence of Synoptic Types 62 (3.2) Response of Hydrologic Variables to Climate Change 69 Regional Winter Snowpack 6Spr i ng Runoff 73 Spring and Summer Runoff - Bella Coola River 77 Autumn Floods 82 Flood Frequency(3.3) Long-Term, High Resolution Climate Analysis 90 Comparisons With Other Regions 98 (3.4) Conclusions 99 V Chapter IV: GEOMORPHOLOGICAL EVIDENCE OF RESPONSE OF GEOPHYSICAL SYSTEMS TO ENVIRONMENTAL CHANGE: TEST WITHIN THE HISTORICAL RECORD (k.O) Introduction 101 (4.1) Glacier Fluctuations and Recent Climate Change 102 Mass Balance Fluctuations of Glaciers 103 Mass Balance and Climate 107 Glacier Snout Fluctuations 110 (4.2) Recent Fluctuations in Upland Sediment Transfers 115 (4.3) Sediment Sources and Processes of Floodplain Development 121 Processes of Floodplain Development 12Alluvial Sediment Sources 127 (4.4) Equilibrium Conditions and Inferences from Channel Morphology 13Vertical and Lateral Stability of Bella Coola River 2 Inferences From Channel Morphology ' 137 (4.5) Recent Floodplain Sedimentation 140 (4.6) Conclusions 141 Chapter V: SEDIMENT0L0GICAL EVIDENCE OF ENVIRONMENTAL CHANGE: TESTS USING LAKE SEDIMENTATION RATES (5.0) Introduction 148 (5.1) Patterns of Sedimentation in Ape Lake 149 Physical Setting 14Sediment Facies 15^ Recent and Long-Term Sedimentation 156 (5-2) Periodicity of Sedimentation in Ape Lake 164 Results 16(5.3) Chronology and Controls of Sedimentation Rates 169 Recent Sedimentation Chronology 170 Controls of Glacio-Lacustrine Sedimentation Rates 172 (5.4) Inferences from Long-Term Variations in Glacio-Lacustrine Deposition 176 Key Events 8 Hydrocl imati c Inferences l8l (5.5) Summary and Conclusions 182 Chapter VI: BIOLOGICAL EVIDENCE OF ENVIRONMENTAL CHANGE: TESTS USING VARIATIONS IN TREE GROWTH (6.0) Introduction 184 (6.1) Tree-Growth and Climate 18Methods 7 (6.2) Response Characteristics 191 (6.3) Growth Variations and Climate 194 Interior Douglas firs 19** Subalpine fir 19(6.4) Development of Transfer Functions 200 (6.5) Inferences about 20th Century Hydrology From Tree-vi Rings 208 Summer Runoff 20Rainstorm and Snowmelt Flooding 212 (6.6) Summary and Conclusions 215 Chapter VII: INFERENCES ABOUT PRE-INSTRUMENT ENVIRONMENTS IN THE BELLA C00LA RIVER BASIN (7.0) Introduction 217 (7.1) Inferences of Environmental Change From Former Glacial ExtentSnowline Fluctuations and Climate Change 218 Estimates of the Contemporary Climatic Snowline 221 Estimates of Former Snowline and Inferences of Hydrologic Change 223 Little Ice Age Glacial Chronology 227 (7.2) Long-Term, Low and High Resolution Biogeophysical Evidence 230 Fluctuations in Average Temperature and Precipitation 231 Inferences from Lake Sedimentation 235 High-Frequency Variations in Tree Growth, Lake Sedimentation Rates and Related Hydroclimatic I nf erences 237 (7.3) Effects on the Bella Coola Valley 240 Floodplain Age 24Evidence for a Long-Term Flood History 245 Long-Term Floodplain Sedimentation 246 Chapter VIII: SYNTHESIS, CONCLUSIONS AND FUTURE DIRECTIONS (8.1) Methodolgical Implications and Biogeophysical Response Records 252 Recent Hydrocl imat i c Record 253 Tree-Ring Response RecordGlac i o-Lacustr i ne Response Record 256 Glacier Response Record 257 Alluvial Response Record 9 General Conclusion 26l (8.2) Recent Environments of the British Columbia Mid-Coast 262 Little Ice Age Environment 263 (8.3) Future Directions 265 REFERENCES 268 Appendix A: RADIOCARBON DATES OF SELECTED DEPOSITS FROM THE BELLA COOLA BASIN 289 (A.1) Radiocarbon Dating 290 vi i Appendix B: TEXTURAL AND MINERALOGICAL VARIATIONS OF THE BELLA COOLA VALLEY-FILL SEDIMENTS 292 (B.1) Introduction 293 (B.2) Bulk Gravel Samples 29Results ^ (B.3) Mineralogy of the Valley-fill 296 Results 297 Appendix C: RECENT PATTERNS OF SEDIMENTATION IN THE BELLA COOLA VALLEY 304 (Cl) Introduction 305 (C.2) Sampling Design and Methods 30Rambeau Slough 310 East Hagensborg Slough 31Walker Island Slough 2 Snootli Slough 313 Mclellan Road Slough 315 Appendix D: APE LAKE SEDIMENTS: SAMPLING, MEASUREMENT AND SUMMARY OF RESULTS 317 (D.l) Introduction 318 (D.2) Field and Laboratory Procedures 31Compaction and Distortion of Sediments 319 (D.3) Periodicity of Sedimentation at Ape Lake 321 Results 324 Appendix E: TREE-RING SITE CHRONOLOGIES IN AND NEAR BELLA COOLA BASIN 330 (E.1) Introduction 331 (E.2) Tree-Ring Data Acquisition 33Tree-Ring Data 4 Appendix F: APPLICATION OF PALE0HYDR0L0GICAL TECHNIQUES FOR RECONSTRUCTING FORMER CONFIGURATIONS OF GLACIERS IN THE BELLA COOLA RIVER BASIN 3^5 (F.l) Accumulation Area Ratios..... 3^+6 (F.2) Reconstructing Former Climatic Snowlines 3^9 (F.3) Glacier Chronologies for the Bella Coola Basin 350 Methods 351 (F.4) Description of Glacier Chronologies at Selected Sites 5 East Nusatsum Valley 35Tsini Tsini and Borealis Glaciers 358 Fyles Glacier 3&0 Deer Lake and Jacobsen Glaciers 3&1 vi i i LIST OF TABLES (2.1) Summary of formations and constituent mineral assemblages of rock types found in the Bella Coola basin after Baer (1973) and Tipper (1979) 22 (2.2) Hydrometric data for stations in the Bella Coola basin 33 (3-1) Climate Stations in and near the Bella Coola basin 5** (3-2) Classification of synoptic circulation types from Yarnel (1983) and Barry et al. (1982) 65 (3.3) Winter precipitation and synoptic type frequencies 66 (3.4) Summer temperature and synoptic type frequencies 67 (3>5) Winter snow courses in southwestern British Columbia 70 (3-6) Winter snowpack and synoptic type frequencies 74 (3>7) Significant variables and regression coefficients for climate-runoff models 80 (3.8) Recognized trends in 20th century conditions in the N.E. Pacific sector 95 (4.1) Empirical response models for glacier net mass balance and climate in southwestern British Columbia 10(4.2) Winter mass balance on B.C. south coastal glaciers and synoptic type frequency 108 (4.3) Summer mass balance on B.C. coastal glaciers and synoptic type frequency 109 (4.4) Estimates of floodplain sediment volumes for the Bella Coola River between Atnarko/Talchako confluence and mouth 126 (4.5) Morphological data for selected reaches of Bella Coola River 13(5.1) Ape lake morphology 152 (5.2) Pearson's r for varve thickness and hydroclimatic var i ables 173 i x (6.1) Summary of tree ring-width chronologies used in this study 190 (6.2) Correlation matrix between PC scores and hydrocl imati c variables 202 (6.3) Coefficients of transfer functions for winter (Oct.-Apr.) precipitation 203 APPENDIX TABLES (A.l) Summary of radiocarbon derived dates 291 (B.1) Classification of basic lithological and mineral characteristics of bedrock and alluvial sediments of major tributaries to the Bella Coola River 299 (B.2) Constituent minerals of floodplain and channel zone sediments for the Bella Coola valley 300 (D.l) Cesium-137 results for core 84E2 326 (D.2) Ekman core data and standardized chronology for Ape Lake 327 (D.3) Summer and winter lamina thickness for percussion cores from Ape Lake 328 (D.4) Standardized sedimentation chronology for Ape Lake percussion core 329 (F.l) Accumulation area ratios for selected Canadian glaciers 3^7 (F.2) Summary of characteristics of and estimated ages for selected terminal moraines in the Bella Coola basin 356 X LIST OF FIGURES (1.1) A conceptual systems-framework model for environmental reconstruction showing functional relationships and inferential pathways 3 (1.2) Temporal resolution, continuity and sampling interval for biogeophysical and historical/instrumental data sources 11 (2.1) General location of study area and sampling locations 6 (2.2) Bella Coola basin and location of local hydroclimate stations 17 (2.3) Bedrock geology of the Bella Coola basin and mineral sampling locations 20 (2.4) Observed and inferred valley-fill in the lower reaches of Bella Coola River near Snootli Creek 25 (2.5) Hydroclimographs for selected climate stations in the vicinity of the Bella Coola basin 31 (2.6) Bella Coola River annual hydrograph for 1980 35 (2.7) Generalized clastic sediment routing and primary transport processes in Bella Coola basin 39 (2.8) Illustrations of primary clastic sediment sources for the higher-order fluvial network in the Bella Coola basin 44 (2.9) Major sediment transfer routes and storage sites for clastic sediment in the Bella Coola basin 47 (3-1) Seasonal temperature and precipitation departures for coastal and interior regions of west central British Columbia between 1945 and 1983 56 (3.2) Generalized departures of winter precipitation for southern, central and northern coastal areas of British Columbia. General departures in sea surface temperatures for the northeastern Pacific ocean 58 (3-3) 500 mb synoptic scale circulation patterns which produce specified climate departures for coastal British Columbia 61 (3-4) Winter snowpack departures for coastal and interior regions of southwestern British Columbia 72 x i (3-5) Spring runoff departures for three coastal rivers of southwestern British Columbia 76 (3.6) Ten day precipitation time series and associated recurrence intervals for flood generating storms at Bella Coola 83 (3.7) Frequency of autumn storm runoff in the Bella Coola, Homathko and Squamish Rivers between 19^5 and 1985 6 (3-8) Partial duration curves of flow frequency for rivers of the Bella Coola basin 89 (3-9) Contoured surfaces of intra- and interannual climate at Bella Coola: 1904-1983 91 (3-10) Contoured surfaces of intra and interannual climate at Big Creek:1904-1983 2 (4.1) Relative rates of advance and recession fo selected glaciers in the Bella Coola basin 112 (4.2) West Saugstad Glacier in late July of 1985 showing advances position 114 (4.3) Indirect evidence for secular changes in sediment delivery to tributary channels and alluvial fans of Bella Coola River 118 (4.4) Morphology and channel zone sediments in laterally stable and unstable reaches of Bella Coola River 124 (4.5) Alluvial sediment sources for the Bella Coola River floodplain 130 (4.6) Water and bed surface longitudinal profiles of Bella Coola River between Burnt Bridge Creek and river mouth 134 (4.7) Changes in river planform and channel widths for a transitional reach of the Bella Coola River between 1945 and 1984 138 (4.8) Topographic setting of the McCall Flats backwater area of Bella Coola River above Burnt Bridge Creek 142 (4.9) Stratigraphy and pollen concentrations of backwater sediments extracted from the McCall Flats ponds 144 (5-1) Location and bathymetry of Ape Lake, British Columbia 150 xi i (5.2) Photographs of the Ape Lake basin after the October 20, 1984 glacier outburst flood 153 (5-3) Facies distribution, sediment sampling^locations and sources for runoff and sediment at Ape Lake 155 (5.4) Photographs of Ekman cores and cross-correlation of laminae in sediments from the east basin of Ape Lake 157 (5-5) Laminae thickness versus depth for Ekman cores in the east basin of Ape Lake 159 (5.6) Laminae thickness versus depth for percussion cores extracted from the east basin of Ape Lake 160 (5.7) Mean grain size versus standard deviations in grain sizes for samples taken from core 84C2 163 137 (5.8) Concentration of Cs in grouped couplets from core 84E2 compared with atmospheric flux rate to Lake Mi chi gan. 167 (5-9) Composite sedimentation chronology for Ekman sampled sediments of eastern Ape Lake 171 (5*10) Standardized sedimentation chronology for percussion corer sampled sediments of eastern Ape Lake 177 (5-11) Sedimentation anomalies and frequency of micro-laminations in Ape Lake sediments compare to runoff anomalies in the Bella Coola River 180 (6.1) Location of Douglas and subalpine fir sampling sites used for development of tree-ring chronologies 189 (6.2) Response characteristics of growth in Douglas Fir and subalpine fir to variations in selected hydrocl imati c variables... 193 (6.3) High and low frequency growth signals in grouped Douglas fir chronologies from the Bella Coola area 196 (6.4) High and low frequency growth signals in grouped subalpine fir chronologies from the Bella Coola area 198 (6.5) Reconstructed variations in annual precipitation for eastern portions of the Bella Coola River basin 205 (6.6) Reconstructed variations in summer temperature for eastern portions of the Bella Coola River basin 207 xi i i (6.7) Observed, estimated and reconstructed summer runoff indices for Bella Coola River above Burnt Bridge Creek 210 (6.8) A comparison of inter- and intraannual persistence of wet winters ans warms summers with tree growth variability and flooding on the Bella Coola River 213 (7-1) Relationship between relative ELA and climate indices for selected glaciers in southwestern, British Columbia 220 (7.2) Estimates of contemporary and former climatic snowline for eastern and western portions of the Bella Coola basin 22k (7.3) Glacier moraine chronology for the Bella Coola basin and selected areas in the Canadian Cordillera 229 (7.4) Long-term, low frequency variations in reconstructed annual precipitation and the first PC score of subalpine fir 232 (7.5) Long-term, high frequency variations in the first PC scores of Douglas and subalpine fir in the vicinity of the Bella Coola basin 238 (7.6) Spatial distribution of surfaces of different ages on the Bella floodplain 243 (7-7) Stratigraphy, grain size variations, organic matter content and pollen concentrations in vertical accretion sediments of the Brekke site on the Bella Coola floodplain 248 APPENDIX FIGURES (B.1) Downstream variations in grain size for sediments of the Bella Coola River and tributary alluvial fans 295 (B.2) Variations in quartz and lithic fragment content with grain size in alluvial sand samples of Bella Coola River 298 (B.3) Ternary diagram showing discriminant functions developed to distinguish between sediment sources using mineral characteristics of alluvial sediments 301 xiv (C.1) Sampling locations of backchannel and slough deposits of the Bella Coola valley 309 (C.2) Stratigraphy and pollen concentrations for recently active sloughs of the lower Bella Coola River (Rambeau, East Hagensborg and Walker Island si tes) 311 (C.3) Stratigraphy and pollen concentrations for recently active backchannels of the lower Bella Coola River (Snootli and Mclellan Rd. sites) 31^ (D.l) Estimates of differential sediment compaction in Ape Lake sediments using bulk density and laminae thickness 322 (F.l) Illustration of a typical glacier in the Bella Coola basin used to infer paleo equilibrium line altitudes ik8 (F.2) Location of moraine sampling sites in the study basin 352 (F.3) (a) Photograph of recessional moraines in East Nusatsum valley and (b) left lateral moraine of Tsini Tsini Glacier 357 (F.4) Photographs of recessional moraine fields for (a) Borealis and (b) Fyles Glaciers 359 (F.5) Photographs of terminal and recessional moraine fields for (a) Deer Lake and (b) Jacobsen Glaciers 362 XV ACKNOWLEDGEMENTS I would like to thank Dr. Michael Church who as advisor has given me tremendous encouragement throughout the course of this research. As teacher and scientist, his sagacity has both inspired and motivated me to a greater understanding of geomorphology and the earth sciences. This project has benefited substantially from the experience of Dr. Robert Gilbert who was instrumental in all phases of field work at Ape Lake and Dr. June Ryder who provided invaluable guidance in the collection and assessment of evidence for glacier fluctuations in the study area. Dr. Glenn Rouse kindly made available his pollen lab and provided direction and adivce in the preparation and analysis of the pollen data. Drs. John Hay, William Mathews and Olav Slaymaker reviewed early drafts of the thesis and made many valuable suggestions during the execution of this research. Able field assistance was provided by Mark Cantwell, Lome Davies, Dan Hogan, Tom Millard and Jane Weninger. I would also like to thank Sandy Hart for his assistance in preliminary reconnaissance of Bella Coola River. Accommodation and valuable insight to the Bella Coola region was graciously provided by Tony and Lis Karup. Appreciation is extended to my graduate colleagues at UBC who were always willing to discuss the research problem at hand. Drafting of several diagrams was kindly provided by Mary Artmont. My greatest thanks and appreciation goes to Pamela Moss des Loges who not only endured my often cantankerous nature but also assisted in the field and in preparation of this dissertation. Funding, in the form of a post-graduate scholarship from the Natural Sciences and Research Council of Canada and a pre-doctoral fellowship from the Killam Memorial Fund for Advanced Studies is gratefully acknowledged. 1 Chapter I INTRODUCTION It has long been recognized that certain biological and geophysical components of the environment respond in a predictable manner to external forces. Remaining evidence of these components allows inferences to be made about the distribution and magnitude of forces in the past. Confident identification of past environmental settings relies upon some knowledge of the attendant physical processes, the nature of system responses and how the system might be affected when certain boundary conditions are altered. An integrated approach towards understanding environmental change prior to the period of instrument records utilizes evidence from several biogeophysical subsystems in order to assess the consistency of results and to evaluate errors involved in interpretations and related inferences. The purpose of this thesis is to assess the utility of several methodologies in the analysis, interpretation and reconstruction of recent hydrological and geomorphological processes. The test field area is in a mid-latitude, glacierized alpine basin of west-central British Columbia. (1.1) Systems Framework In this work the interactions amongst atmospheric, hydrological, biological and earth surface processes are analyzed around the integrative theme of paleohydrology. Leopold and Miller (195*0 first introduced this term to describe "the interaction of climate, vegetation, stream regimen and runoff obtained under climates different from that of the present". Schumm (1965) suggested that the term be "restricted to that portion of the hydrological cycle that involves the movement of water [and sediment] over the surface of the earth . . .". Brown (1982) has proposed that the term be 2 even more narrowly defined as "the applications of concepts, methods and models derived from hydrogeomorphological studies to fluvial processes, sediments and forms of the past". This latter definition stems from the frequent emphasis on fluvial processes but is considered here as unnecessarily restrictive. The common theme in paleohydrological studies is an attempt to provide a more comprehensive understanding of environmental settings over longer time scales and with increasing temporal resolution. It is convenient to view the environment as a system comprised of several morphogenetic components linked by a series of functional relations and affected by a set of boundary conditions. Figure 1.1 illustrates four morphogenetic subsystems of interest in this study: glacial, glaciolacustrine, fluvial and mass wasting. A set of functional relations between the atmospheric environment and the earth surface environment can be defined. These relations are formulated on the basis of exchanges of energy and mass and are influenced by boundary conditions such as topographic and geologic controls, neotectonics and former glaciation. Energy from the atmospheric and hydrological components of figure 1.1 drives the system resulting in the redistribution of mass and the formation of new sedimentary deposits. Inferences about former hydrological settings, then, are made in the reverse direction from that of the functional relat i ons. The transfer of information between the hydroclimatic environment and any one of the four morphogenetic subsystems identified in figure 1.1 is subject to a variety of temporal and spatial filtering effects. Deposits within any of the subsystems integrate the effect of short-term hydroclimatic processes, or events, hence have a characteristic low resolution of the true amplitude and frequency of climatic and hydrologic variations. For example, maximum advances of ice are frequently marked by 3 : • CLIMATE energy balance water balance circulation regime I GLACIER ACTIVITY mass balance moraines. till, outwash * RUNOFF/ STREAMFLOW SEDIMENT TRANSPORT \ \ GLAC10-LACUSTRINE PROCESSES physical 1imnology bathymetry ALLUVIAL ACTIVITY sediments, channel geometry 1aminated lake sediments floodplain and terrace al1uvi um LAND SURFACE weatheri ng vegetation/ soi 1 s COLLUVIAL ACTIVITY basin morphometry "X" I _L_ 1andslides/ rockslides earthflows LAND USE human decisions SEDIMENTARY DEPOSITS • FUNCTIONAL RELATIONSHIP • INFERENTIAL PATHWAY Figure 1.1 Functional and inferential pathways connecting four morphogenetic subsystems of interest in this project. Not shown are boundary conditions which include geological and neotectonic factors as well as past glacial activity (after Starkel and Thornes, 198l). Other subsystems not considered include eolian, marine and glaciomarine environments. it terminal moraines. However, fluctuations in the direction or rate of ice movement may be a response to a combination of hydroclimatic changes which occur over time scales much shorter than the interval required for formation of the deposit, in contrast, landslides, flood deposits or subaqueous slope failures may represent 'key events' without reference to a particular set of departures in long-term mean conditions of hydroclimate. These two situations, which generally limit the inferences that can be made, are commonly associated with primary morphological sources of information and therefore complicate the exercise. A precursor to successful application of the model in figure 1.1 to past environments is the interpretation of various morphological components such as vegetation, soils, surface forms and sedimentary sequences (Gregory, 1983a). Environmental reconstructions then, can only be made after absolute and relative dating techniques are accurately applied to sequences of sediment and preserved landforms which have clearly defined depositional environments. In this context, a true paleohydrological model is rarely achieved. (1.2) Operational Approaches Brown (1982) suggested that three types of paleoenvironmental modeling can be undertaken: process, empirical and inferential model development. A modified version of this scheme discussed here includes three general categories: (1) process or physical models, (2) empirical models and (3) inductive models. Process Models The most difficult of all three categories is the process model where the functional relationships between various components of the system are 5 quantified and grouped to form a general physical model which operates according to physical and geochemical laws. The model is constructed and calibrated under modern conditions and then applied to past environments. The approach is purely deductive and requires quantitative estimates of input parameters. General circulation models are of this type. Clarke et al. (1984) utilize a process-model, verified using modern outburst floods, to estimate former discharges of ice-dammed glacial Lake Missoula in Washington state. Ice-flow and heat transfer relations are used as governing equations to predict the size of the subglacial discharge tunnel. Various parameters such as lake volume, water temperature and ice-thickness are required input variables. Model similitude and assumptions regarding ice tunnel geometry and flow resistance remain problematic. Power and Young (1979). among others, have used an integrated watershed model to predict future and former runoff from glacier-covered catchments. This type of model is calibrated against observed flow variations and adjusted accord i ngly. The major assumptions underlying process-models are: (1) all dominant processes have been identified and contributing factors parameterized, or, in the case of simulations, unknown but important factors are controlled; (2) governing relationships are properly calibrated for the expected range and variance of input data; and (3) the temporal and spatial resolution of the input data correspond with the calibration data. Violation of the first or second assumption results in an unreliable model whereas violation of the third assumption yields unreliable results. Empirical Models Though the deductive approach used in basis for this second group of models, the process modeling is underlying physical also the processes 6 are not quantified. A particular set of processes (rainfall, stream flow, glacier flow) is related to responses in biological or geophysical attributes of the environment and then relationships are developed, usually using statistical techniques. The relations are then inverted to make predictions of former process based on response chronologies preserved in the environment. These include the quantitative analysis of vegetation growth patterns, variability in glaciolacustrine sedimentation, valley or river channel morphology, paleochannels and fossil fluvial sediments. Nominal, interval or ratio scale data can be used. Mathewes (1973) and Mathewes and Rouse (1975) have attempted qualitative reconstruction of Holocene paleoclimates in the lower Fraser Valley of British Columbia using paleoecological evidence preserved in lake cores. Quantitative estimates of temperature and precipitation for the same area and period were then made using calibration techniques in which the spatial distribution of modern vegetation is calibrated against contemporary climates (Mathewes and Heusser, 1981). These relations are then used to formulate transfer functions in order to estimate former hydroclimatic conditions based on the inferred paleoecology. Two problems complicate the "transfer" of these relations derived from spatial associations. The first is related to the fact that spatial variations in climate can be greater than those experienced at a particular site through time. The second is that factors other than the assumed dependent variable (in this case climate) may explain a significant proportion of the temporal variance and are not included in the development of the transfer function (Imbrie and Webb, 1983) . The major underlying assumption in the retrospective application of empirical models is that an equilibrium has been achieved between a specific attribute and a dominant process for both the calibration and 7 application intervals. Certain biological or geophysical attributes may adjust very slowly to imposed change and thus may not reflect higher-frequency perturbations in the system. In contrast, adjustments may be rapid but sensitive only to forces beyond an exceedence threshold and thus are representative of short-term conditions only. Linear least-squares statistical techniques are commonly used in the formulation of empirical models. The application of these relationships for environmental reconstruction is based on the assumption that the explained portion of the total variance reflects the primary response characteristic of the system and that the residual variance is random "noise". This constraint can seriously affect the significance of interpretations from a linear model if the system is also known to respond significantly to extreme departures. These less frequent, higher magnitude events become 'key' elements which are lost in the assumption of linearity. Inductive Models This category is based primarily on the inductive approach utilizing site-specific evidence for drawing inferences about former environments. Sedimentary sequences (1ithofacies) and the remains or by-products of biological activity are used to reconstruct a series of possible environmental conditions. Ordinal and nominal scale data are used frequently and mostly qualitative results are achieved. Inferences usually relate to a specific set of environmental conditions such as spatial and temporal variations in runoff, or the chronology and distribution of glacier ice movements. Starkel (1984) has used various alluvial sequences to infer the character of climate transitions over the Holocene Epoch in central Europe. Ryder and Thomson (1986) have used several sources of evidence from glacial 8 deposits throughout southwestern British Columbia to infer the timing and magnitude of glacial activity in post-Pleistocene time. The application of several dating techniques is crucial for establishing chronostratigraphic control both locally and regionally. Preservation of datable materials and inaccuracies associated with radiocarbon dating become major limitations of the exercise. Commonly, particularly for alluvial sediments, the stratigraphic record represents extreme or 'key' events only (e.g. flood deposits). Erosional hiatuses or other unconformities may complicate interpretations if a substantial proportion of the intervening sequence is missing or otherwise disturbed. In this way a substantially different record of environmental change may evolve using these data sources. Recognition of these differences is important for successful model development. In summary, empirical and inductive approaches, despite their problems, are commonly employed to draw inferences regarding past hydrological and climatological conditions in the reverse direction from that of the usual functional relations. The process-based approach yields estimates of these functional relations and, then the boundary conditions and initial state are changed to approximate former environments. While initially more attractive, this latter type of model is the most difficult to develop and still frequently suffers from linear parameterizations which do not easily accommodate conditions beyond the calibration interval. The three types of models yield different characteristics of the past environment, and therefore should be viewed as complementary approaches to environmental reconstructions. 9 (1.3) Objectives and Research Strategy The specific study objectives are: (1) to identify and relate the response characteristics of selected geophysical and biological attributes of the study basin to recent fluctuations in hydroclimate; (2) to determine the resolution and length of record available in these attributes and evaluate several methodologies for inferring the character of past hydroclimates; and (3) to assess the within-basin and regional (southwestern British Columbia) significance of measured and inferred environmental change in the study area. The primary emphasis is on problems associated with methods used in the reconstruction of past environments. In particular, consideration is given to sources of error and the strength of inferences drawn from indirect and proxy data which may exist in quantitative, semi-quantitative or qualitative form. Empirical and inductive methods only are considered because of difficulties in applying and linking existing process-based models. The study considers the utility of the model presented in figure 1.1 as a framework for drawing inferences regarding recent paleoenvironments in a mid-latitude alpine basin. In order to assess critically the real resolution of responses in the system and to provide a basis for calibration of paleoenvironmental data, a retrospective strategy has been adopted. This is operationalized by considering first the relationship between hydrologic variability and geophysical responses over the period of recent and detailed instrument records (post-19^5)• The relationships are then tested using less detailed 10 but long-term instrument records as confirmatory data (1900-19^5)• This strategy was adopted as a means of assessing the strength of inferences from biogeophysical data in the pre-instrument period. Inferences about environmental conditions prior to 1900 are then made from the evidence of biogeophysical attributes alone. Initially, it was anticipated that this methodology would allow for a resolution of at least annual, or perhaps even seasonal, biogeophysical responses over the testing period. However, in view of possible persistent trends in biological and geophysical attributes, it was reasonable to expect a resolution only on the order of years to decades. The period after I65O AD defines the study interval. This was found to be consonant with preserved evidence. Details are given in the following sections. (l.k) Data Sources Biological, and geophysical data can be grouped and classified on the basis of temporal continuity, temporal resolution and interval of application (Bradley, 1985)• Some provide only a discontinuous record of changes and may represent only extreme events (e.g. flood deposits), while others may appear to give a continuous record (season-to-season, year-to-year) of changing conditions (e.g. tree-rings). Certain sources are applicable to time scales of millennia or longer and others may yield data concerning only recent (decennial, centennial) processes. The primary data sources for this study can be divided into four groups of attributes: biological, geophysical, historical and archaeological, which are based on the most useful indicators of environmental change in mid-latitude, glacierized alpine basins (figure 1.2). Biological sources include trees, flowering vegetation and lichen. Tree ring-width time series possess a resolution of seasonal (late-11 10" 00 >-x a. o LU U3 oo il ta ^ o —i a ea < < h- o Z K d LU a ° 3 OO S ae i—i — *- = y z < 103 TO2 101 10° 10'' I 1 I 1 I——— I H I— 10 103 102 101 10° 10'1 10 -2 I 1 h 1 Tree-rings »•—^Tree-damage Pollen Lichen Soil Lake Sediment Glacial Deposits Valley-fill Instrumental Written Records Legends Artifacts 10 -2 10" Years Figure 1.2 Biological, geophysical and historical/archaeological sources used for environmental reconstruction in this study. The minimum sampling interval (attribute resolution) in years is given by a solid horizontal line for continuous variables and a dashed line for non-continuous variables. Some attributes have more than one sampling interval. Solid circles indicate the period open for study in years. Resolution of pollen depends on the character of the host sediments. 12 wood/early-wood) to annual (total ring-width) growth variations and are continuous through time. Damage to trees by discrete events, such as floods, or more continuous stress, like insect damage, provide some indication of low-frequency, high-magnitude events but generally these records are discontinuous. The oldest trees available for sampling are usually less than 10^ years. Pollen and spores are produced seasonally, and if preserved in sediments also characterized by a high temporal resolution, may yield a record of changing pollen contributions and thus ecological variability. Mostly, pollen is useful for identifying changes over long-term intervals (10 years) with a resolution of 50 to 500 years (figure 1.2). Lichen and soil development can be used to establish relative ages for different k surfaces as old as 10 years usually with a resolution of 10 to 500 years. In some instances it may be possible to produce continuous form/time curves with an absolute temporal reference but they are difficult to achieve, particularly with soils. Lake deposits form during an influx of sediment from the surrounding terrain into a nearly closed system and may represent several time scales and degrees of continuity depending on sedimentation controls. Low-frequency, high-magnitude deposits may be the result of isolated runoff events which occur over a period of hours or days. These sediments are distributed discontinuously throughout the sequence. Alternatively, autogenous events may occur, such as density underflows, produced by the buildup of deltaic deposits which then exceed stability limits. The event is the consequence of continuous sedimentation, but does not mark an extrinsic event. However, in a long record the changing frequency of failures may be significant. If certain runoff regimes prevail, a continuous record of seasonal or annual sedimentation might be preserved 13 (e.g. annually-laminated sediments or varves). In shallow water away from primary sediment sources, sedimentation rates may be low and represent long 2 3 periods of accumulation (10 - 10 years). Glacial sediments are generally deposited discontinuously over time scales commensurate with the response time of the ice to changes in accumulation and ablation (10 -10 years). A record as old as 10 -10 years may be present if sediments are well-preserved. Valley-fill sediments, downstream from the headwater source areas, also represent discontinuous sedimentation over longer time scales but may incorporate sequences deposited during short-duration floods. In addition to biogeophysical sources, information related to human activity can provide indirect evidence of environmental change (figure 1.2). Data regarding annual or semi-annual changes and extreme short duration events are most common. For western Canada written records prior to the middle and late 1800s are rare. Indian legends and archaeological evidence are also useful, although they are less direct indicators of environmental setting. (1.5) Organization This dissertation is composed of eight chapters. The organization is based on a format for assessing the quality of information transfer from the hydroclimatic environment to the earth surface environment in order to reflect the main objectives. Chapter 2 is a summary of the general physical setting of the Bella Coola River basin and characterization of primary sediment sources in the basin. Chapter 3 is a discussion of variations in several hydrological attributes of the recent and long-term instrument period (1900-1983). Chapter h documents the relationships between these recent forcing factors and the response of selected geophysical components 14 of the landscape. The discussion is restricted to the post-1945 interval for which there are several types of direct records of environmental change. Chapter 5 is devoted to the interpretation and significance of a high-resolution record of lake sedimentation rates as an independent source of information about environmental change. Chapter 6 is an investigation of the response of growth variations in trees, and an assessment of tree-rings as a source of high-resolution data in environmental reconstruction. Chapter 7 is a discussion of low-resolution data sources and interpretation of hydroclimatic variability during the late Neoglacial interval. The regional significance of the reconstructed records is also considered. A summary of results and conclusions is given in Chapter 8. Data not presented in the main text, plus additional topics and analyses, are given in the appendices. 15 Chapter I I STUDY AREA (2.0) Location Starkel and Thornes (198l) suggest that selection of a drainage basin for paleohydrological analysis should be based on the availability of data regarding landscape morphology, process mechanics and earth surface materials. In addition, long-term instrument records of recent hydrologic change are essential for quantitative reconstructions. In British Columbia there is no basin which fits these criteria well. Bella Coola River basin, a partly glacier-covered catchment located on the central coast, approximately 500 km northwest of Vancouver (figure 2.1), was selected because of the availability of recent hydroclimatic data, comparatively early historical records of environmental conditions and the potential for preservation of paleoenvironmental evidence in upland source areas and lowland floodplains. Accessibility, basin size and the existence of several preliminary studies concerning physical processes and recent environmental change in the basin (Tempest, 197^; Church and Russell, 1977; Hart, 198la; Church, 1983) were additional factors. 2 Bella Coola River drains 5050 km of the glacierized Coast Mountains of southwestern British Columbia. Two tributaries join to form Bella Coola 2 River (figure 2.2): the larger Atnarko River (2,430 km ) originates in gently sloping, uplifted terrain to the east, and Talchako River drains 2 1300 km of heavily glacierized mountains to the south. Several smaller tributaries enter the 80 km long Bella Coola River as it flows west to enter North Bentinck Arm (figure 2.2), an inland extension of the Burke Channel and Dean Channel fiord system. Approximately 1% of the catchment is covered by cirque, niche and valley glaciers concentrated in rugged mountainous terrain in the southwest. Two major valley glaciers, Talchako 16 Figure 2.1 General location of study area and hydroclimate stations used in the analysis of climate variations. Axis of the Coast Mountains is an approximate boundary for stations which are referred to in the text as "coastal" and "interior". Also included are locations of glacier and tree-ring data used in the study. Figure 2.2 Bella Coola River basin including major tributaries and the distribution of upland ice. Also included are the locations of climate and river gauging stations. 18 and Jacobsen, along with several smaller north-flowing ice masses, emanate 2 from the Monarch Icefield (total area 350 km ) which straddles the southwest basin perimeter for a distance of kS km. (2.1) Physiography Three different physiographic regions as defined by Tipper (1971a) and more generally by Holland (196*0 can be delineated in Bella Coola basin. To the south of Bella Coola River and west of Talchako River are the rugged, deeply dissected Pacific Ranges of the Coast Mountains, extending from sea level to altitudes greater than 3500 m (figure 2.2). The highest peaks bisect Monarch Icefield, some of which are visible only as nunataks, ranging from 2500 to 3500 m. Local relative relief is similar throughout, averaging 1400 to 1600 m. Drainage directions are dominantly eastward or northward resulting in slope aspects which exhibit a strong bimodal distribution of north/south or east/west and a comparatively high mean slope angle of 27 degrees.* Mountains of the Kitimat Range to the north of Bella Coola River and west of Noosgulch Creek (figure 2.2), rarely exceed maximum elevations of 2500 m. Valley bottoms tend to be wider, average relative relief is lower (1000 m) and the mean slope angle (22^) smaller than those in the Pacific Ranges. Although restricted to headwater regions, landforms typical of the glacial eroded landscape to the south are also found in the Kitimat Range, including U-shaped valleys, cirques, horns, aretes, hanging tributary valleys and several tarns. 1. Mean slope angle was determined as the average of 300 grid-points sampled systematically from 1:50,000 topographic sheets for each physiographic region. A representative slope distance was defined as ±3 100-foot contours measured perpendicular to the slope (except in low-gradient valley bottoms where ±2 was used). Two sub-samples of 50 grid-points each demonstrated that biases due to grid-spacing and orientation were not significant. 19 Contrasting with these two regions is the uplifted and less rugged Fraser Plateau which dominates terrain to the east of Talchako River and Burnt Bridge Creek. Gently undulating and hilly topography (mean slope angle = 7^) characterizes this region in which the average elevation is higher (1300 m) than valley bottom locations to the west. Average relative relief is lower (400 m) than elsewhere in the basin resulting in the formation of hundreds of small lakes along principal drainage lines and on poorly-drained, low-gradient terrain. Charlotte Lake is the largest, 2 covering approximately 65 km (figure 2.2). Drainage from the plateau is concentrated in only a few major streams (Atnarko R., Hotnarko R., Young Cr., Burnt Bridge Cr. and Hunlen Cr. - see figure 2.2) which either have cut deeply into the easily eroded volcanic rocks forming precipitous canyons for tens of kilometers or, where rock types are more resistant, formed waterfalls with vertical drops as large as 250 m. (2.2) Bedrock Geology Physiography and drainage patterns of the Bella Coola basin are influenced by a diversity of rock types and geologic structures characteristic of the Coast Plutonic Complex and Intermontane Volcanic Belt of the western Canadian Cordillera. The geologic setting of the study area described below fits the general pattern of the east slope of the Coast Mountains (cf. Roddick and Hutchinson, 1972). The geology of the Bella Coola area has been mapped in detail by Baer (1973) and more generally by Tipper (1979) (figure 2.3)- The oldest rocks are middle Triassic plutons of quartz-diorite and granodiorite in the southeastern portion of the catchment and on the north flank of Bella Coola River near Stuie. Metamorphism of varying intensities has generated gneissic and schistose foliation in some of these rocks, particularly in Figure 2.3 Distribution of bedrock lithologies in the Bella Coola basin (after Baer (1973) and Tipper (1979)). Triangles are sampling locations used to characterize lithological and mineralogical attributes of sediments derived from tributary basins. O 21 the headwaters of Atnarko River. Volcanic, metavolcanic and sedimentary rocks (Triassic to lower Cretaceous) characterize the central and southern portions of the catchment comprising chlorite-rich greenstone and andesite. The volcanics exhibit widely scattered shear zones occupied by slate and argi11i te. Eocene quartz-monzonites and granodiorites have intruded these older volcanics and form competent, steepened slopes where present. Joint spacing is wide and fracture density is low in these younger plutonic rocks; thus, the quantity of weathered debris is low and the texture is usually large blocks of angular debris. Miocene vesicular basalts and rhyolites are found in the extreme northeast marking the boundary between recent plateau basalts derived from a set of three shield volcanoes of the interior region to the east, and the older volcanic and plutonic rocks which dominate the west (Souther, 1986) . Table 2.1 is a summary of the principal formations, rock types and constituent mineral assemblages found in the Bella Coola bas i n. Topographic expression of structural features and formation boundaries is not well defined in the region. Baer (1973) attributes this to the similarity of mechanical properties (e.g. hardness, jointing) for most rock types. Orientation of individual ranges is not consistent throughout the area except for the regional northwest trend of the Coast Plutonic Complex. Faulting does not appear to be common although relative displacements in volcanic rocks are difficult to observe. Baer (1973) mapped two dominant fault directions, northerly and northeasterly, both of which separate the older chloritized greenstone and granodiorite from the younger andesites. Uplift has occurred during two orogenic events: a less severe Lower Jurassic event and a stronger period of uplift in the late Tertiary. Baer Table 2.1. Summary of formations and constituent mineral assemblages of rock types found in the Bella Coola basin after Baer (1973) and Tipper (1979). Era Period Lithology Minerals Lower/ Pliocene rhyolitic lava very fine-grained mafic minerals occasional obsidian Cenozoic Upper/ Miocene Eocene vesicular basalt quartz monzonite fine-grained mafic minerals small amygdaloids 20-30% quartz and 35-45% pagioclase k-spar as high as 35% biotite, hornblende and muscovite Eocene/ Paleocene granodiorite 25-30% quartz and 40-60% plagioclase minor k-spar, biotite is the only mafic mineral Cretaceous/ Jurassic andesite flows and agglomerates fine-grained plagioclase and minor k-spar plagioclase phenocrysts, 1-5 mm long minor epidote and quartz infilling Mesozoic Middle Jurassic Lower Jurassic black slate/ argillite greenstone fine-grained phyllosilicates with occasional microconglomerate or grit with pebbles of quartz monzonite subhedral to anhedral plagioclase (<0.5 mm) fine grained matrix of chlorite, hornblende, epidote Middle Triassic quartz diorite up to 35% anhedral quartz and small amounts of k-spar plagioclase, biotite, hornblende, epidote and minor chlorite 23 (1973) and Parrish (1982) have attributed most of the present geologic structures to the latter event. Rock shoulders on the eastern flanks of the larger massifs along Talchako River have been interpreted as remnants of a Tertiary erosion surface which extends eastward as the Fraser Plateau. Uplift, estimated to be as much as 1800 m (Souther, 1977. 1986), warped the erosion surface and realigned the drainage such that the Atnarko and Talchako Rivers, which were thought to flow eastward from a divide centered along the axis of the Coast Mountains (Tipper, 1971a. 1971b), are supposed to have been captured by knick point incision of a rapidly growing, orographically-fed, Bella Coola River (Baer, 1973)• (2.3) Late Quaternary Geology The major valley/fiord systems of British Columbia have been deepened by multiple glaciations. The severity of Wisconsinan glacial activity near the end of the Pleistocene has precluded the preservation of ice-contact materials dating from events earlier in the Quaternary. However, the entire B.C. coast probably experienced ice-advances essentially synchronous with the chronology identified for the southern coast (Armstrong, 198l). Deteriorating climate following the Olympia non-glacial period resulted in the advance of ice (28,000 - 15,000 years B.P.; Clague, 198l). Ice flow was from major accumulation areas along the axis of the Coast Mountains. From here ice spread both east and west, initially occupying major valleys and fiords until ice thickness exceeded the local relief, resulting in flow directions less dependent on the underlying topography (Clague, 1981). Erratics found on the shoulders of extinct active shield volcanoes (Rainbow, Ilgachuz and Itcha Mountains) to the east of Bella Coola, along with ice-beveled platforms on some of the highest peaks in the Monarch Icefield, indicate that ice thicknesses exceeded 2000 m (Tipper, 2k 1971a)« Ice flow directions estimated by striae, drumlins, eskers, and fluted terrain show northeasterly flow away from Monarch Icefield towards Charlotte Lake and westerly down the Bella Coola Valley (Baer, 1973; Tipper, 1971b). During the maximum advance, about 15,000 years B.P., ice from the Coast Mountains probably extended as far west as the continental shelf (Clague, 198l) and as far east as Williams Lake, there merging with west-flowing ice from the Cariboo Mountains (Tipper, 1971a). Evidence from the Prince Rupert-Kitimat and Bella Bella areas indicates that deglaciation along the outer coastal lowlands was rapid with glaciers retreating to their respective fiord head positions by 12,500 years B.P. (Clague, 1981; Andrews and Retherford, 1978). Although no dating control is available to the east of the divide, the greater elevation and ice thicknesses there may have resulted in a less rapid retreat. Coastal and valley bottom deposits laid down by the eastward retreating ice were related to the position of sea-level, controlled by isostatic and eustatic responses of the land and ocean to melting ice. Relative sea levels were 150 to 200 m higher until approximately 9000 years B.P. (Andrews and Retherford, 1978; Clague et al., 1982), which resulted in ice fronts terminating in the sea, producing conditions similar to those found today in Glacier Bay, Alaska (Powell, 1983). Glacigenic sediments deposited in proglacial fiords comprise a variety of materials indicative of these changing environmental conditions. The characteristics of the Bella Coola valley-fill have been identified using local borehole and regional data and sedimentary evidence found in outcrops (figure 2.k). Overall the valley-fill represents facies produced by glacial, glaciomarine, marine and alluvial processes. Thick deposits of glaciomarine silty-clay were deposited during the marine inundation phase between 12,500 and 9»000 years B.P. At the same time, tributary streams were depositing 25 c > 0) E +40 gravel-capping floodplain sequence: silty fine sand and fine gravel channel zone gravels: coarse, cobble-boulder gravel to fine sandy gravel cross-valley prograding alluvial fan: coarsening upward sequence from fine gravelly sand to coarse sandy gravel overlain by coarse cobble-boulder gravel downvalley prograding delta of Bella Coola River: coarsening upward from fine sandy gravel to coarse cobble gravel 9580 t 80 yr BP (GSC 3980): wood laminated fine sandy silt and marine mud marine mud: massive silty clay continuation of sequence inferred from regional observations (Howes, 1983; Powell, 1983; Clague, 1985) glaciomarine mud: massive silty clay, somewhat stony (drop stones?) outwash or subaqueous flows at ice margin: gravelly sand basal till bedrock Figure 2.h Observed and inferred valley-fill in the lower reaches of Bella Coola River near Snootli Creek. 26 material in deltas along the margin of Bella Coola valley. Continued ice recession and progradation of the main valley delta produced thick deposits of sand and gravel in a coarsening upwards sequence. Concomitantly, alluvial fans along the valley margin prograded across the valley depositing sandy gravels and coarser cobble material. Till of varying thickness mantles most slopes above the valley floor. Prominent late-glacial features such as terminal or recessional moraines are rare in this catchment as in many other Coast Mountain valleys. Ryder (198l) has attributed this scarcity of deglaciation deposits to one of two possible mechanisms: (1) rapid but orderly retreat of ice fronts allowing for nothing more than brief stationary periods and minimal deposition or (2) in situ ablation of main valley ice which has been cut off from the headwater ice source. A hummocky, kettled ridge extending part way across the narrow Bella Coola valley above Nusatsum River may be a subdued terminal or medial moraine that was formed during a brief pause in late glacial ice retreat. Although morphology indicates that the ridge is ice-contact in origin, exposures of stratigraphy along margins of the deposit are mostly deltaic, and dip directions of foreset beds and constituent lithologies suggest that sediment and water were derived mostly from the upper Bella Coola valley, rather than from the adjacent Nusatsum River. In any event, the bedrock-controlled restriction of the main valley and confluences of the Salloomt and Nusatsum Rivers at this position, appear to have formed a complex depositional environment characterized by glacial, marine and alluvial processes. With the exception of deltaic deposits and a high terrace of marine deposited clayey-silts, late-glacial deposition is found mostly in the upper Bella Coola valley comprising marine deltas, valley bottom kames, 27 till benches, lacustrine silts, eskers and some eolian transported sand deposits. However, the number and size of these deposits are limited. Clague et al. (1982) and Clague (1985) demonstrate that sea levels were within 11 m of present by at least 9720 years B.P. for the Bella Bella area. A date of 3,550 ± 80 years (GSC 396**) was determined for wood sampled near the contact of subareally exposed marine clay and overlying alluvial deltaic sands near the outlet of Salloomt River (elevation is 6l m.s.a.l.). The elevation of this contact indicates that relative sea level in the Bella Coola area was still higher at that time than at present, although marine regression at the site occurred soon after. Evidence from both north and south coast regions indicates that isostatic uplift, sea level adjustments, and ice retreat were all complete by 8,000 years B.P. (Armstrong, 198l; Clague, 1985). Glacier fluctuations during the Holocene Epoch have been documented for a number of different locations along the northeast Pacific rim from Washington to Alaska (Mathews, 1951; Miller, 19&9; Easterbrook and Burke, 1972; Denton and Karlen, 1977; Mann and Ugolini, 1985; Ryder and Thomson, 1986). Early Holocene advances are recognized only in the Aleutians (Black, 1981; 1983), Copper River (Sirkin et al.,1971) and Lituya Bay districts (Mann and Ugolini, I985) of Alaska, and even then appear to be asynchronous suggesting that they were localized events, generally not representative of the warmer and drier conditions which dominated the northeast Pacific between 6,000 and 10,000 years B.P. (Clague, 1981). Miller (1969) and Ryder and Thomson (1986) document evidence for an early Neoglacial advance (Garibaldi Stade) approximately 5.000 years before present. Similar evidence has not yet been found in mountain regions to the north of the Garibaldi Park region in southwestern British Columbia. 28 Several investigators have indicated that an advance of mountain glaciers occurred between 1800 and 3000 years B.P., which varied in intensity, duration and timing between regions, but is generally recognized throughout the coastal region from Alaska to Oregon. Tiedemann Glacier to the south of Bella Coola basin reached its maximum Neoglacial extent during this period, unlike most other glaciers within the region (Ryder and Thomson, 1986). In accord with chronologies developed for Dome Peak and Mt. Rainier in Washington (Crandell and Miller, 1964), Garibaldi Park (Mathews, 1951) and southeast Alaska (Mann and Ugolini, 1985). Ryder and Thomson (1986) found evidence within the Pacific Ranges for ice advances commencing around 1,000 years B.P. with maximum positions achieved by 300 to 400 years B.P. In most regions this was the most pronounced excursion of alpine glaciers throughout the entire Holocene Epoch. A regionally synchronous response of glaciers appears to have occurred during recession from the Little Ice Age maximum, beginning 100 to 350 years B.P. (Field, 1975)- Evidence presented in Chapter 1 indicates that a Little Ice Age chronology, similar to that developed by Ryder et aJ. (unpublished) for regions to the south of the study area, is also applicable to the Bella Coola basin. (2.4) Climate Setting Bella Coola River and the rest of the British Columbia coast fall within the Pacific coast maritime climate regime, characterized by the on shore movement of cyclones and anticyclones generated in the moist, and usually warm, North Pacific region. The intensity and frequency of cyclones vary with latitude and season, and annual climate along the coast is controlled by the seasonal pattern of fronts and air masses. Kendrew and Kerr (1955) recognized the importance of several air masses and their 29 associated fronts. During summer a northward shift of the maritime arctic front towards southeast Alaska is accompanied by the increasing influence of mild Pacific air and the occasional incursion of maritime tropical air along the central and southern B.C. coast. Warm and dry conditions prevail, although the pattern is not invariable, particularly if high pressure ridging is poorly developed, and moisture bearing cyclones migrate southward. The transitional seasons (spring/autumn) are marked by rapid shifts in mean frontal positions and changes in zonal circulation intensity. As the winter season progresses, convergence of major air streams along the north coast bisects the region into a northern half, influenced by maritime arctic air, and a southern half, dominated by warmer and moister maritime Pacific air (Wendland and Bryson, 1981; Yarnel, 1985). Throughout this season the continental arctic front tends to occupy a position along the axis of the Coast Mountains forming a strong ridge of high pressure extending vertically to mid-tropospheric levels and horizontally as far west as the outer coast (Hare and Hay, 1974). Hence, the eastward movement of cyclonic disturbances can be impeded. Fronts nearly always occlude and the warm sector aloft yields high precipitation along the coast. Outbreaks of arctic air are not uncommon early in the winter season (Baudat and Wright, 1969). and it is the timing and duration of these anomalously cold and dry periods which determine snowpack stability and the potential for winter flooding and spring avalanche activity. A succession of cyclones intersects the coast during the winter season and the separation and intensity of these systems are mostly dependent on origin. Kendrew and Kerr (1955) suggest that three source areas for cyclone development are important: (1) a southern group which start as waves on the Polar front in the west Pacific, tracking 30 northeastward towards the B.C. coast; (2) a northern group generated by the effects of Polar air undercutting Maritime Arctic air off Japan and then tracking east into the Gulf of Alaska; and (3) a transitional group forming between fronts in the western and central Pacific. Occlusion and instability through a considerable depth are common to all groups of cyclones as they approach the coast. With decreasing storm separation there is an increase in cumulative rainfall leading to a higher snowpack volume and water content and thus greater potential for high-magnitude winter and spring runoff. Compounding the effects of synoptic-scale circulation features are local and regional topographic controls on temperature and precipitation. Variability of mean temperature, temperature extremes and total precipitation, including the proportions of snowfall and rainfall, all show marked horizontal and vertical gradients. For a series of three stations forming a west-to-east transect from the outer coast to the interior, mean annual temperatures decline from 8.1^C to 3-9^C, and total annual precipitation decreases by an order of magnitude from 4390 mm to 410 mm (figure 2.5)- This strong inland gradient from a mesothermal, perhumid climate to one of continental character occurs over a distance of less than 200 km. Superimposed on the horizontal precipitation gradient are large, orographically induced variations in precipitation with elevation. Using May 1 snowpack water equivalents, recorded 1800 m above the Bella Coola climate station, as a measure of total winter precipitation at this elevation, it appears that, on average, precipitation increases by approximately 100 mm per 100 m of elevation. Approximately 60% of the Bella Coola basin lies in the lee of the Coast Mountains, which is reflected in the distribution and accumulation of snowfall during the winter season. Snow course data, collected for a MAMJJASOND MONTH 600 600 I — 400 c o 5 300 o. u t 200 a. 100 BELLA COOLA 52° 22' N 126° 41" W Elev. 18 m TP 1614 TS 191 TR 1424 mean maximum mean temperature mean minimum ,snowfal1 ralnfalI J F M A M JJ ASO ND MONTH 30 | 20 S -1 10 * c 0 3 -10 o 600 500 i — 400 e o ™ 300 a. £ 200 o. 100 TATLAYOKO LAKE 61° 40' N 124° 24' M Elev. 8S3 m TP 412 urn TS 142 mm TR 270 mn 30 g 20 f 1 o> lOg O 3 J F M A M J JASOND MONTH Figure 2.5 Climographs for a west to east transect through Bella Coola basin. Stations are located in figures 2.1. and 2.2. TP is total precipitation, TR is total rainfall and TS total snowfall (Source: Atmospheric Environment Service, 1981) . 32 limited number of corresponding years from an eastern and a western site at similar elevations within the catchment show that late spring snowpack water equivalents are 8 to 10 times greater in the west (record mean of 1753 mm) than those in the east (record mean of 230 mm). (2.5) Hydrology Average Regime Within the basin, nine river gauges have been operative at various times since 1930, but only five provide data for periods longer than five years after 1948. The combined data series from two stations on the main river cover the period 1948-present, while three of the tributary gauges (Atnarko, Nusatsum, Salloomt) were established during the summer of 19&5 and have been in operation continuously since then. The main river gauge, relocated in 19&5. is only a few kilometers below the confluence of Talchako and Atnarko Rivers, the latter of which is gauged just above the confluence, allowing for estimates of flow in the ungauged Talchako River by simple subtraction. Gauge locations and a summary of hydrometric data from each site are given in figure 2.2 and table 2.2 respectively. Variability in basin-wide runoff is strongly affected by the distribution of snow. It also is affected by the location of short and long-term water storage sites such as lakes and glaciers. Mean annual runoff from the basin above Burnt Bridge Creek is 790 mm with a maximum of 1023 mm recorded in 1976. For partly glacier-covered tributary catchments situated in the southern and western portions of the basin (e.g. Nusatsum River) mean annual runoff is significantly higher, averaging over 2000 mm (table 2.2). Mean annual runoff from the larger Atnarko River draining ice-free terrain is one third of the discharge from Talchako River, which drains a 25% glacier-covered area. Inter-correlation of annual mean and 33 Table 2.2. Hydrometric data for stations in the Bella Coola basin. (Source: Water Survey of Canada, 1982) W.S.C. Station name (number) Interval of Drainage Mean Annual Mean Annual Maximum Daily Discharge X Difference of 3-1 1 Operation Area Discharge Runoff (mm) (m s ) Autumn vs. Spring 2 3-1 (# of years) (km ) (m s ) Spring Autumn (date) (date) Bella Coola R. near Hagensborg (08FB002) 1947-68 4,040 (21) 668 963 (06/11/64) (01/24/68) Bella Coola R. above Burnt Bridge Cr. (08FB007) 1965-present 3,730 (21) 558 703 (06/11/69) (01/23/68) Bella Coola R. at Bella Coola (08FB001) 1929 (1) 405 (07/09/29) 430 (16/10/29) Bella Coola R. at Hagensborg (08FB008) 1930-32 4,800 (2) 561 (10/06/30) 561 (09/17/30) Nusatsum R. (08FB005) 1965-present 269 (21) 83 (16/05/70) 190 (09/27/73) Salloomt R. (08FB004) 1965-present 161 (21) 9.3 1,820 43.3 (16/05/70) 141 (12/16/80) Atnarko R. (08FB006) 1965-present 2,430 (21) 258 (30/05/72) 289 (01/24/68) Tastsquan Cr. (08FB003) 1946-1950 (4) 2.2 2,390 21.6 (06/18/50) 36.5 (24/10/47) Clayton Falls Cr. (08FB009) 1980-present 481 (5) 21.2 (05/21/82) 85.6 (08/09/82) Talchako R. 1965-present 1,300 (21) 62.7 1,521 407 (06/11/67) 552 (01/23/68) 1. Calculated as [Q^/Q^] X 100 where Q,, is the maximum daily discharge due to spring snowmelt and is the maximum daily discharge due to an autumn rainstorm for the period of record. Not computed for stations with a limited length of record. 2. Flows estimated by subtracting Atnarko R. data from Bella Coola R. data above Burnt Bridge Creek. No major tributaries join the system between these two gauging sites. Flows may be underestimated by 2-3Z. 3** annual maximum discharges indicates that Nusatsum/Talchako Rivers are hydrologically distinct from Salloomt/Atnarko Rivers, primarily on the basis of glacier melt contributions to runoff and higher mean elevations. However, the Atnarko River retains a somewhat singular signal due to the much lower annual precipitation and frequency of lakes, which tend to reduce peak discharges. Several seasonal runoff regimes can be identified for Bella Coola River, which are typical of other rivers along the British Columbia coast. Beginning in late April through mid to late June, runoff due to snowmelt is somewhat variable, but steadily increasing discharges are evident. Occasionally, runoff is augmented by rainfall on the ripened snowpack. Following the snowmelt peak in early to middle June (or late May for some rivers), runoff declines until late July, when discharge again increases as glacier melt begins to contribute significant volumes of stored water (figure 2.6). Superimposed on steadily declining river discharges from late August onwards are high-magnitude synoptic runoff events derived from two sources: (1) rainfall derived floods with only minor contributions from melting snow or (2) rain-on-snow events associated with rapidly rising freezing levels following the eastward movement of autumn cyclonic disturbances. Finally, from January to April winter low flows dominate the runoff regime. Maximum daily and instantaneous discharges almost always occur during the short-duration autumn floods (table 2.2), but the highest average runoff is during the summer snow and glacier melt period (figure 2.6). Hart (198lb) has shown that the ratio of maximum instantaneous to mean daily discharge for autumn floods in the Kitimat and Pacific Ranges varies with drainage area but is generally between 1.5 and 1.9. Maximum daily discharges from rainstorm-generated floods in the gauged catchments are on BELLA COOLA RIVER Annual Hydrograph of Mean Daily Flow, 1980 Water Survey of Canada Gauge 08FB007 Ad = 3,430 km2 1 = Rainstorm generated flood 2 = Rain-on-snow flood Winter Low Flow Snowmelt Glacier melt Autumn Rainstorms "jAN I FEB 1 MAR I APR I MA? I JUN 1 JOT 1 AUG ' SEP ' OCT T NOV "DEC ^ Figure 2.6 The 1980 annual hydrograph for Bella Coola River above Burnt Bridge Creek. 36 average 25~40% greater than spring snowmelt floods. However, as glacier cover and local relief increase and basin size decreases, this difference can be as high as 300%. Major Floods During the 20th Century Historical and instrument records show that there were approximately thirteen major flood events between 1896 and I98O. The pre-1948 record of less extreme flooding is probably incomplete, however; all the largest flows were documented. With the exception of a flood in the fall of 1950, all high-magnitude, autumn and spring runoff events on the Bella Coola River were associated with winter seasons (October to April) characterized by positive precipitation anomalies. Only six other seasons in the 88 year record exhibited strongly positive precipitation anomalies and no evidence of responses in the form of significant flooding. Large runoff events were not prevalent in these years because in all six cases spring and summer temperatures were well below normal promoting water storage. The most outstanding flood events occurred in the middle 1930's and 1960's. The four floods (1934, 1936, 1965, 1968) are among the six largest measured or estimated flows on the river. As such they represent 'key events' in the long-term hydrology of the basin because in each instance, particularly the rain-on-snow events, major channel realignment or significant overbank sedimentation occurred. For this reason a brief description of the hydroclimatic circumstances associated with these two sequences is given, and the implications for effects from lower-magnitude events are considered briefly. The flood of October 10, 193^ resulted from the occlusion of fronts associated with an intensified low pressure system (960 mb) to the west of Vancouver Island. A blocking ridge of high pressure over the western United 37 States led to the low stalling as it approached the coast. High-water appears to have been more localized because of significant log jams on the river. Storage of organic debris in the channel was probably significant since the last documented flood was ten years before. Flooding on November 19. 1936 was more severe than the 193** flood for two reasons. First, a series of low pressure systems moving southward from the Gulf of Alaska, preceded the larger system yielding snowfall at middle and at lower levels in the basin. Second, a well developed warm front swept, through the area raising the freezing level substantially. A melting snowpack contributed greatly to the runoff. Synoptic conditions for the October 22, 1965 flood were similar to those in 193**: northward moving disturbances produced seven-day total precipitation of 190 mm prior to the peak discharge; one of the largest flood-related rainfall periods. Ridge development in Alberta was characteristic of the January 23, 1968 flood. Like the 1936 flood, peak discharges were much higher due to substantial snowmelt as the freezing level increased during a four day period prior to the flood. Maximum instantaneous flows from this event were probably the largest to occur this century. With the exception of moderately high spring runoff in late May of 19**8, there were no flood events of notable between 1936 and 1950. The flood of November h, 1950 was the fourth largest estimated flow and appears to have had minimal impact in terms of channel changes. Two factors might explain this. First, both the rising and falling limb of the flood hydrograph were very steep so the peak discharge was of short duration. Second, runoff in two of the preceding three springs was significantly high (although not of major flood proportions) so that any accumulation of organic debris in the channel may have been removed. The implication is 38 that not all large runoff events may have an impact on the river. There is some evidence for incipient instability in certain reaches of the river such that moderate floods may produce substantial localized channel changes. In most cases the instability is related to the occurrence of log jams which, when removed, modify the direction of channel flow. (2.6) Sources and Transfer of Clastic Sediment An important step in determining the trends of basin sediment yield and possible associations with changing hydroclimate is consideration of sediment sources and processes of sediment transfer. The purpose of this section is to document the characteristics of the sediment cascade and assess the distribution of several primary sediment sources and residence times in associated sediment sinks. A conceptual model, shown in figure 2.7, has been adapted from those given in Simons et al. (1982) and Roberts and Church (1986) , and is used here to illustrate the primary sources, transfer and storage of clastic sediment in the Bella Coola River basin. Three primary sediment sources can be identified: (1) in situ weathered bedrock; (2) glacial deposits in proglacial and ice-marginal areas of contemporary glaciers; and (3) clastic materials, including soil, which mantle hillslopes or infilled gullies, the bulk of which was deposited during the late Pleistocene and early Holocene epochs. These materials are mobilized by mass wasting (rockfall/ rockslide, avalanche, debris flow, gully wash, soil creep) and fluvial processes. Storage of recently produced sediment in the upland or headwater areas is in the form of colluvial deposits or partly infilled proglacial lakes which were formed during ice retreat. Residence times will vary as a function of reservoir capacity, slope stability and sediment transfer rates. Fluvial erosion and mass wasting of slope deposits may introduce 39 Weathered Bedrock joint spacing, fracture density Fresh Glacial Deposits terminal, lateral and ground moraines rockfal1/ rockslide mixing Till and In Situ Soil on Hi 11 slopes I debris slide debris flow 4 avalanche i fIuvial transport Colluvium, Talus Composite Debris Slopes I footslope erosion Proglacial Lakes: Lacustrine Deposits f 1 uvial transport Upland Valleys: Channel Storage — . J I 1 channel i erosion i Tributary Alluvial Fans Long-Term Lake Storage T fan undercutting , channel erosion Channel Zone and Floodplain Storage 1 channel migration bank erosion ±_ Marine Delta Figure 2-7 Generalized clastic sediment routing and primary transport processes in Bella Coola basin. ko clastic debris directly to the channel zone of low-order tributaries. Grain sizes are usually very coarse (bouldery to blocky) and thus transport-limited conditions prevail. Fluvial transport moves some of the sediment into downstream storage sites such as tributary alluvial fans, upland valley trains, lakes along principal drainage lines or ultimately to the channel and floodplains of higher order streams. If storage capacities are high and sediment transfer rates low, a transition to supply-limited conditions is likely to occur downstream. Finally, fluvial processes along the main stem of the system will pass sediment through to the basin outlet where sediments are deposited in a growing marine delta. Interpretation from aerial photographs offers the only practical method of evaluating clastic sediment sources and differential movement rates, even though the technique yields limited quantitative information on storage volumes and deposit stability. Two scales of mapping were undertaken to evaluate sediment sources: (1) interpretation of basin-wide features from medium scale (1:21,000 to 1:40,000) black and white air photography and (2) detailed air photograph and field mapping of three 2 2 subcatchments: Salloomt River (169 km ), Burnt Bridge Creek (215 km ), and 2 Nusatsum River (269 km ),*which are representative of the spatial variations in hydrologic, physiographic and geologic conditions within the bas in. Upland Sediment Sources Seven types of sediment sources were identified: composite debris slopes, talus slopes (gravity sorted), alluvial fan or talluvial debris cones, rockfal1/rockslide deposits, in situ mountain top detritus or felsenmeer, lateral and proglacial till deposits, and blankets or veneers of till along forested slopes. kl Composite debris slopes are polygenetic landforms produced by any combination of rockfall, avalanche and debris flow (cf. Rapp, I960; Chandler, 1973; Church et al., 1979)- Highly fractured and jointed greenstone rocks in the basin are an important source for these blocky deposits which form below open rock slopes dissected by numerous avalanche and rockfall chutes. They are also found on confined rock slopes above the flanks of active cirque glaciers, where removal of ice by melting has promoted the failure of rock surfaces. Although widely distributed and of considerable volume throughout the volcanic terrain, these slopes are not overall an important source of onward moving clastic sediment due to the coarse-grained nature of the debris and transport limited conditions which result. Talus slopes are gravity-sorted, rockfall dominated landforms, forming discrete but small sheets or multiple coalescing cones, primarily above treeline, or below steep bedrock cliffs at lower elevations. These deposits occur frequently along the edges of former lava flows and below extension faults in the plateau areas of the northeastern catchment, but are poorly connected to the fluvial network. High magnitude rockfalls/rockslides are defined on the basis of criteria suggested by Mudge (1965) . the most important of which are the slope-base position of coarse blocky debris and a visible upslope failure surface. The largest deposits occur on terrain underlain by plutonic or metamorphic rock types. For example, of the twelve deposits identified in the Nusatsum basin, eight are derived from intrusive rocks and four from volcanic/sedimentary sources. This is significant given that the distribution of rock-type is about the reverse (figure 2.3). Most recognized deposits are at or above treeline and in some cases have impounded small lakes. Estimates of volume are difficult to make because of unknown depths, but appears to range downward from 10^ m^. The largest rockslide mapped occurred in approximately 1951 in the headwaters of the Atnarko River as a result of a foliation plane failure in gneissic rocks g (J.J. Clague, personal communication, 1986). The volume is estimated at 10 3 m . The deposit resulted in minor realignment of the drainage divide between the Atnarko and North Klinaklini Rivers. Alluvial fans and talluvial debris cones are found below low-gradient and steep basin outlets, respectively, and usually contain a perennial surface stream along one margin of the deposit. Fluvial processes, snow avalanches and rockfalls deliver sediment to the surface of these deposits. Perennial streamflow on the surface and active toe erosion by higher order streams indicate that these landforms are an important source of sediment. Block fields are found on elevated rock platforms, ice eroded benches and cols. Sediments are derived by in situ mechanical and chemical weathering of well jointed bedrock, particularly along non-conformable contacts of various volcanic rock types. They are perhaps the least important sediment sources because of limited volumes and stable upland pos i t i ons. Lateral and proglacial till deposits formed by the recession of valley, hanging valley, cirque and niche glaciers are the most important sources of sediment in glacierized portions of the catchment, particularly where the deposits are on steep terrain drained by avalanche and debris torrent channels (figure 2.8a and b). Although the volume of material delivered to the stream is low in relation to the amount available for transport (i.e. low sediment delivery ratio), aggradation and channel braiding are obvious impacts on the fluvial network immediately downstream from these sources. On a basin-wide average, lateral deposits are of less overall importance. Figure 2.8 Illustrations of primary clastic sediment sources for the higher-order fluvial network in Bella Coola basin. (A) Niche glacier on the western flank of the Nusatsum basin supplies a significant amount of sediment to channels below the glacier. The debris torrents are well connected to the high-order fluvial network. (B) Rock glacier/ablation debris in the south fork of Gyllenspetz Creek is dissected by numerous avalanche chutes on easily eroded volcanic rocks. This site is a significant source of sediment for Talchako River. (C) Proglacial lake of Jacobsen Glacier acts as an important storage site for ablation sediments and debris transported by ice marginal streams. (D) Upland plain in the east Nusatsum valley is one of several storage sites for glacial derived sediments. 44 MS **5 The final sources of clastic materials are surface wash, soil creep and bank erosion. Roberts and Church (1986) estimate that surface wash and soil creep, on average, constitute less than 10-20% of the total sediment volume delivered to small, partly logged watersheds in the Queen Charlotte Islands. Relative delivery rates in large, unlogged tributaries, such as most of those in the Bella Coola basin, would be proportionally lower due to the stabilizing effects of the vegetation and a higher frequency of lower-gradient slope segments. Streambanks are an important source for fluvial entrainment of sediment but, estimates of contribution to sediment yield integrated over an entire watershed are difficult to make. Four major types of sediment storage have been identified in the Bella Coola basin. First, relatively deep lakes along principal drainage lines, particularly in the Atnarko watershed, effectively trap all the clastic sediment derived from above each lake site. The largest, Charlotte 2 Lake (65 km ), is fed by creeks which drain an area of approximately 750 2 km , thereby eliminating any significant contribution from this area of the eastern catchment (see figure 2.2 for locations). Residence times in these sites are comparable in length to the interval of deglaciation (10 years). Shallow, proglacial lakes formed during recent glacial retreat (figure 2.8c) are less efficient sinks but equally important over a shorter time 2 scale (10 years) (cf. Smith et al., 1982). Second, upland valley trains form on low-gradient, bedrock-controlled, valley flats at higher elevations in proglacial areas of the southwestern catchment, or on plateau areas of the eastern catchment. 2 Glacial and fluvial sediments derived areas as large as 30 km are stored at these sites (figure 2.8d). Trap efficiency is less than that of lakes 3 4 but residence times are equally long (estimated at 10 - 10 years). Third, rockslide dams of various magnitudes also form effective barriers to the 46 throughput of clastic materials. Their frequency of occurrence is much 3 4 lower, but residence times are also long (10 - 10 years). Finally, colluvium and till, stored on many of the forested hillslopes, are probably the largest sediment sinks in the basin. Ultimate turnover rates are most 4 likely in excess of 10 years. Aerial photographs and field evidence indicate that much of the sediment in the upland tributaries is derived from well-defined, unstable slopes and bank erosion of material already stored within the valley flat. The distribution and volume of the major sediment sources, along with the locations of major storage sites, are plotted in figure 2.9- In several tributaries there are one or two primary sources, such as unstable proglacial deposits (e.g. Nusatsum, Nordschow, Thorsen Creeks) or composite debris slopes (e.g. Burnt Bridge, Nordschow, Noosgulch Creeks). The combined area of the sediment sources in figure 2.9 constitutes less than 16% of the entire watershed. These results suggest that the transfer of 2 3 sediment to the Bella Coola River over a time scale of 10 - 10 years is associated with a relatively small number of upland sources and fluvial activity within several tributary channels. (2.7) Pedologic Setting Soil development in the Bella Coola basin reflects the strong variability in topography, frequent sediment additions from river flooding in the valley and, at higher elevations, slow decomposition of organic material. Within the confines of valley bottoms most soils vary between poorly developed orthic regosols and eutric brunisols (British Columbia Ministry of Agriculture, 1973)* Thin Ah horizons (3 to 5 cm) over fine sand are common on many alluvial surfaces where, additions of mineral matter may exceed organic matter accumulation significantly. Under thick coniferous Figure 2.9 Major clastic sediment sources for the higher-order fluvial network and estimates of residence time in important storage sites. Arrows represent the location of both the sediment sources and primary routes through which sediment is transferred (see text for discussion). ^ 48 forest cover, dystric brunisols may develop into orthic humo-ferric podzols if the site has remained stable for some time. Organic soils are restricted to poorly drained locations surrounded by both coniferous and deciduous vegetation. Typic mesisols and hydric fibrisols occupy many of the water-saturated alpine meadows. (2.8) Flora The floristic component of the landscape provides an essential key to geomorphic activity and hydrologic change within certain growth-time boundaries. The capacity of many species to withstand environmental extremes yet preserve evidence of stressful periods, provides correlations with geomorphic/hydrologic events, as well as a basis for absolute and relative dating. For this purpose some of the dominant vegetation communities are noted. Six biogeoclimatic zones and several subzones have been identified in the Bella Coola basin, each based on the dominance or co-dominance of climax tree species (Robinson and Pojar, 1981; Leaney and Morris, 198l). Along the lower-elevation western periphery where mild, moist, maritime conditions prevail, a thick cover of western red cedar (Thuja plicata Donn) and western hemlock (Tsuga heterophylla (Raf.) Sarg.) is found. The oldest and tallest timber grows on comparatively organic-rich soils of the floodplain and steep colluvial slopes. At elevations below 500 m between Bella Coola and Firvale a mixture of Douglas fir (Pseudotsuga menziesii (Mirb.) Franco) and western hemlock overlie an understory of mosses and devil's club (Oplopanax horridus). This association grades quickly into a Douglas fir/red cedar/moss ecosystem east of Firvale. The Bella Coola River floodplain is dominated by a forest cover indicative of early successional phases on recently flooded surfaces or ^9 areas with high water tables, and late successional species on higher and more stable surfaces. Black cottonwood (Populus trichocarpa Torr. £ Gray), red alder (Alnus Rubra Bong.) and white birch (Betula papyrifera Marsh.) are most evident on recently colonized portions of the floodplain while Western red cedar, Sitka spruce (P/cea sitchensis (Bong.) Carr.), Douglas fir and western hemlock dominate older stable areas. At middle elevations throughout most of the basin (150 to 1000 m) there exists a climax ecosystem comprised of hemlock, amabilis fir (Abies amabilis (Dougl.) Forbes) and occasionally subalpine fir (Abies lasiocarpa (Hook.) Nutt.) . Above 1000 m there is a continuous cover of mountain hemlock, amabilis fir and yellow cedar (Chamaecyparis nootkatensis (D. Don) Spach) although the latter, which occupies a mesic habitat, is present in the east only. A zone of Engelmann spruce (Picea engelmanni i Parry) and subalpine fir occurs between the alpine tundra and mountain hemlock zone. Lodgepole pine (Pinus contorta Dougl.) is a co-dominant species where fires occur frequently, such as the drier Interior Plateau, prohibiting the development of a climax spruce-fir forest . The elevation of the contemporary tree line averages 1600 to 1650 m above sea level, but fluctuates spatially with changes in aspect, slope stability and microclimate. Within the tundra are a variety of alpine meadow species such as mosses, lichen and heather. 50 (2.9) Settlement and Logging History Historical observations and factors influencing settlement patterns provide a potentially valuable source of both direct and indirect evidence of environmental change. Prior to European contact in the late l8th century the Bella Coola valley was populated by Indians of the Salish nation, a splinter group from tribes of the interior and southern coast regions (Kopas, 1970). Estimates by Mcllwraith (19^8) and more recently by Lepofsky (I985) show that at least 2,700 people, concentrated in 27 known village sites, occupied the valley between Stuie and Bella Coola. Since fishing was the primary economic resource, village locations were mostly determined by accessibility to river habitats favorable for salmon. Flooding and channel shifting are known to have influenced settlement locations and the type of structures constructed. First white contact was made in 1793 by a ship from the fleet of Captain Vancouver making explorations of Burke Channel and North and South Bentinck Arms. One month after the first direct contact, Alexander MacKenzie, travelling by land from Fort Chipewyan on Lake Athabasca, descended into the valley along the east flank of Burnt Bridge Creek, the first Caucasian to reach the North American west coast (north of Mexico) by land. No permanent white settlements were established until 1867. four years after Lieutenant Henry S. Palmer made the first survey of the route between Bella Coola and Fort Alexandria on the Fraser River. Several of the village sites were still seasonally occupied when the largest influx of Europeans, mostly Norwegians, settled in the valley in the autumn of 1896. The preliminary land survey by Palmer in 1863* the legal survey of I889-I893 and geological mapping by G.M. Dawson (I878) are the first systematic records of the Bella Coola environment. Exploration and survey notes, journals, letters and diaries from these sources along with limited 51 archaeological evidence from the pre-contact period provide some data regarding environmental conditions over the last several centuries. Interpretation of aerial photographs demonstrates that prior to the early 1950s timber removal was restricted to small plots on the valley bottom and was mostly related to land clearance. Between 1946 and 195^ only 2 about 3 km had been commercially logged, all of it below the Nusatsum River confluence. Up until 1968 an additional 13 kmz had been logged mostly along the steeper slopes above Bella Coola River and in the Salloomt, Nusatsum, Noosgulch and Cacoohtin tributaries. From 1968 to 197** another 8 2 km was logged, much of it on the floodplain above Burnt Bridge Creek and partly on valley slopes of the lower Talchako River. Declining timber prices in the last 5 years have resulted in renewed activity in the more 2 accessible tributaries in the Bella Coola basin. Thus over 18 km has been logged since 1974, much of it concentrated in the Nusatsum and Noomst 2 Creeks and upper Bella Coola River. The total of 43 km represents less than 1% of the basin area but the concentration of commercial timber removal on lower slopes and at tributary junctions in the southwestern catchment would suggest a greater potential for impact on sediment yield. 52 CHAPTER I I I Hydrophysical Records of Environmental Change: Tests Within the Instrument Period (3-0) Introduction Variations in regional climate are ultimately related to fluctuations in atmospheric circulation. These can occur as persistent departures in circulation intensity, shifts in the mean position of atmospheric features, or large-scale changes in circulation pattern. When making inferences about the nature of atmospheric processes and environmental change using predominantly biogeophysical evidence, the functional relationships between various indicators of climate change and climate itself need to be examined for a period in which both the climatic perturbations and environmental responses are known. It is the purpose of this chapter to assess the strength and form of these cause and effect relat i onsh i ps. (3.1) Post-1945 Synoptic Climatology of the N.E. Pacific Sector Precipitation and temperature are commonly available measures of climate, and provide a means for assessing the effects of changing atmospheric circulation. Establishing the nature of secular variations in temperature and precipitation for a given region can be difficult for three reasons: (1) insufficient sampling density for an area which exhibits large spatial variations - this may be due to real differences in local climate, or alternatively, to apparent differences produced by various sampling devices; (2) lack of temporal homogeneity at a station due to relocation or changes in data collection procedures and equipment; and (3) inadequate methods of analyzing and presenting data. Keeping in mind these potential sources of error, data from several climate stations in and around Bella Coola were examined in order to identify the regional response. 53 Eight climate stations, extending as far west as the outer coastal plain and as far east as the central Chilcotin River basin, were used to compile seasonal averages of monthly mean precipitation and temperature (see figure 2.1 for station locations and table 3-l)» Winter is defined as the interval October to April and summer as May to August. September was not included in either seasonal average because of the significant negative correlation between summer temperature and September temperature (i.e. tends to dampen the signal) and relatively low total precipitation during this month. All seasonal data series were entered into a factor analysis which yielded two distinct groups based primarily on variations in precipitation (table 3-1)• They are referred to here as coastal and interior regions and are split between the lower elevation sites to the west and higher elevation stations in the lee of the Coast Mountains. Secular trends in temperatures were similar enough to allow consideration of all stations as one group, but, for consistency, temperature trends are also examined in the two regions. To maximize the regional signal seasonal variables for a given region are standardized, averaged and then plotted as adjusted partial sums using the following scheme: Ptn'Ttn= E n j = l t=l xtj - XJ /n Pt ,T = cumulative sum of standardized seasonal precipitation (P J tn tn 4. IT \ c 4. 4.- . j_ nt or temperature U ) for n stations and t years X . = observed seasonal value for year t and station j X"! = 1951-1980 seasonal mean for station j Adjusted partial sums have advantages over simple moving average techniques. These include giving equal weight to generally wetter and 54 Table 3.1: Climate Stations in and near the Bella Coola River Basin. Latitude Longitude Elevation Years of Record Variables (ra) number of station moves (year) Rotated Components 1 2 Bella Coola 52 20' 125 38' Bella Coola B.C. Hydro Ocean Falls Bella Bella 52 22' 52°21' 52°10' 126 49' 127°41' 128°09' 1895-Present T,P 1961-Prasent 1924-Present T,P 1 C1959) 0.712 0.431 0 0.599 0.395 0 0.812 0.298 P only 1 (1966) 0.839 0.169 INTERIOR Kleena Kleene 51 59' Tatlayoko Lake 51°40' Tatla Lake 51°54' Big Creek 51°10' 124"56' 847 1928-Present 1973-Present 1904-Present T,P T,P T,P T,P 0 0.313 0.746 1 (1979) 0.453 0.865 0 0.239 0.731 1 (1978) 0.121 0.649 OTHER STATIONS percent total variance 54.9Z 25.IX 52 20' 126 15' T,P Anahim Lake 52 21' 52°28' 125 20' 125°18' 1097 1054 1954-1967 1975-1980 T,P T,P 1. See figure 2.1 for locations. 2. T is temperature. P is precipitation. 3. Rotated component loadings for precipitation stations. Stations are grouped according to the strength of their correlation with each factor. See text for discussion. 55 generally drier stations within a region as well as displaying both high frequency (year to year) and low frequency (decennial) trends common to all stations. Secular variations in seasonal precipitation and temperature between 19^5 and 1984 are shown in figure 3»1- Positive slope segments indicate persistently above average conditions whereas negative slopes are indicative of below average conditions. The interval 1945 to 1984 was selected to facilitate comparisons with other hydroclimatic data which were collected for this period only. The 3n year mean (1951-1980) is considered to be a representative of the period average and is also the adopted standard by various atmospheric agencies. Secular Trends in Seasonal Temperature and Precipitation Winter temperatures show the greatest year-to-year variance over the common period, and long-term trends show a high spatial coherence between the coastal and interior regions (figure 3-la). Between 1948 and 1957. consistently below normal temperatures prevailed in both regions, whereas from 1958-1965 this trend reversed itself with above normal winter temperatures dominating. Fluctuations which followed were of lower magnitude, maintaining near average conditions until 1975 when winter temperatures were once again mostly above normal. Summer temperature departures (figure 3-lc) are less pronounced and more persistent within the Interior region. Total October to April precipitation trends after i960 show a strong inter-regional coherence with above normal precipitation from 196l to 1968, average conditions between I969 and 1976, and below normal after 1977 (figure 3-lb). Prior to i960 the trends between regions are inversely related, with the exception of a few years characterized by average precipitation everywhere. Summer precipitation (figure 3-ld) has lower 56 1940 1950 1960 1970 1980 1990 +—*—• coastal 1940 1950 1960 1970 1980 1990 Years (AO) Figure 3-1 Adjusted partial sums of seasonal temperature and precipitation for coastal and interior regions of west-central British Columbia. Departures are from the 1951-1980 station normals and are summed from 19^5 to 1983- Positive slope segments indicate persistent above average departures and negative slopes below average departures. 57 year-to-year variance and the greatest inter-regional differences, which is most likely related to differences in precipitation generating mechanisms during the summer. Convective thunderstorms occur frequently in the Interior region whereas summer rainfall along the coast is mostly derived from cyclonic disturbances. Since 1965» summer precipitation along the coast generally has been above normal and normal to below normal for the Interior. Near average conditions are evident in both regions prior to 1965 except for increases in the late 1950s. Overall, trends in winter climate are towards well below normal coastal precipitation and regional temperatures to 1957. above normal precipitation and temperatures to 1964/65, average conditions everywhere until 1976 and then below normal precipitation and above average temperatures to 1983- Secular variations in summer climate are more variable because of the increasing importance of localized atmospheric i nstabi1i ty. Comparison With Other Coastal Regions of British Columbia Temperature and precipitation trends for most of the province have been examined by Crowe (1963). Powell (1965). Thomas (1975) and Thomson et al. (1984). In each analysis, variably-lengthened, unweighted moving averages were used, a method which is poorly suited for detailed analysis of secular variations in climate when more precise determinations of climate shifts are required (Mitchell, 1966) . More recently, investigations of winter temperature and precipitation fluctuations have been undertaken on the Queen Charlotte Islands (Karanka, 1986) and various stations along the British Columbia and United States west coast (McGuirk, 1982; Yarnel, 1985). A generalized summary of winter precipitation trends for the north, central and southern coasts is presented in figure 3'2. 58 A North Coast ' 1 ' i • i i i i B Central Coast 1 1 1 ' i i i 1 + o South Coast i • j i ^ i i i i - ~\ SST anomalies N.E. Pacific 1 1 i i i i i i i 1 • r r T- 1 r i 1 r— 1 I + o + o Figure 3'2 Generalized departures of winter precipitation from post-1945 normals. Bottom figure is generalized trends in sea surface temperatures. Areas above and below the zero line define periods of above and below normal precipitation, respectively. The magnitudes of departure are not indicated. A) north coast trends from Karanka (1986) (normal period 1900-1983) and Yarnel (I985) (normal period 1948-1980); B) central coast (this study); C) south coast from Powell (1965) (normal period 1900-1960) and Yarnel (1985); D) sea surface temperature anomalies for the northeast Pacific (Chelton, 1984) (normal period 1947-1984) . 59 The most obvious feature of climate change along the northeast Pacific sector is the synchronous shift in 1976 to above average precipitation along the north coast and southeast Alaska, and below average precipitation in the central and south coast regions. As indicated in figure 3-2, these changes were associated with a shift to above normal sea surface temperatures (SST) in the northeast Pacific. Prior to 1976 precipitation trends along the central coast appear to represent a transitional phase between the almost inversely related trends of the northern and southern coasts. For example, above average precipitation along the north coast between 1958 and 1964 persisted until 1969 along the central coast and until 1976 along the south coast, suggesting a southward shift in moisture bearing disturbances through this period (c.f. Namias, 1983; Karanka, 1986) . Between 19^5 and 1957 when winter precipitation was average to below average for much of coastal British Columbia, SSTs were average to above average. Unlike precipitation, winter temperature trends show a much greater spatial and temporal coherence similar to those in figure 3-la. Above average (1958-1964, 1976-1983) and below average temperatures (1948-1957, 1965-1975) for all coastal areas correspond closely to positive and negative SST anomalies, respectively. Secular variations in summer temperature and precipitation have not been examined closely except for the earlier investigations of Crowe (1963) and Powell (1965)• There is some indication of regionally consistent trends with above average temperatures prior to 1952 and average to below average temperatures between 1953 and I960. 60 Atmospheric Circulation Usually, no one feature of the atmospheric circulation can explain adequately variations in surface climates. Sea level pressure patterns, positioning of hemispheric semi-permanent pressure systems, intensity and direction of flow aloft, mean position of seasonal fronts, frequency of storms and orientation of storm tracks are all important components. However, there is increasing evidence of interactive linkages between atmospheric circulation and SST anomalies. It appears from figure 3-2 that much of the recent climatic variability along the B.C. coast is associated with these anomalies. Horel and Wallace (1981) demonstrated that above normal SSTs in the eastern north Pacific result in the development of a high pressure ridge over western North America and a trough in the central Pacific. Ridging along the coast is often associated with a westward shift in the Aleutian Low (Angell and Korshover, 1982). If ridging extends out into the Pacific then there are two possible outcomes: cool/dry conditions as northwesterly airflow aloft is advected southward down the trailing limb of the ridge (figure 3'3a) or warm/dry conditions if the ridge axis is centreed over the coast (figure 3-3b) (Yarnel, 1985)- Wet conditions will prevail if ridging is poorly developed, i.e. zonal flow occurs (figure 3'3c), or if the ridge is displaced eastward by an intensified Aleutian Low, in which case the coast is under the broad, upsweeping, southeastern quadrant of the low (figure 3-3d). Analyses of post-1945 synoptic-scale circulation features by Balling and Lawson (1982) and McGuirk (1982) indicate that flow patterns have been dominantly meridional, characterized by a low Rossby wave number leading to persistent and sometimes extreme climatic departures and abrupt shifts in the character of the climate. Figure 3-3 Synoptic-scale pressure patterns (500 mb) leading to distinct departures of temperature and precipitation in southwestern British Columbia (from Yarnel, 1985)- (A) High pressure ridging extends beyond the coast to produce cool/dry conditions. (B) Ridging is centred over the coast resulting in warm/dry weather. (C) Cooler and wetter conditions occur when ridging is not dominant producing zones of intense 500 mb flow. (D) Ridging occurs to the east of the coast resulting in warm/wet conditions. 62 Based on the trends in figures 3-1 and 3-2 and the results reviewed above, positive SST anomalies between I958 and 1968, associated with ridging over the west coast, promoted above average precipitation along the central and north coasts. Although statistical confirmations cannot be made with the available storm track data, this would correspond to a northward displacement of cyclonic trajectories. Below average precipitation after 1976 for the central and southern coasts occurred when the Aleutian Low shifted westward significantly further enhancing the northward displacement of storm tracks. Frequency and Seasonal Persistence of Synoptic Types Periods of persistent departures in seasonal temperature or precipitation are thought to be characterized by a set of unique synoptic conditions which produce the observed trends (I.e. wet, dry, cool, warm). Each season is comprised of a large number of synoptic types. However, it is likely the frequency or within-season persistence of only certain synoptic types drives the mean departure. This hypothesis is tested here using an objectively defined set of synoptic types which are known to yield certain climatic characteristics. Synoptic circulation types for a particular season and sequence of years were identified using a catalogue of daily 500 mb and daily surface pressure patterns developed by Yarnel (1983) and Barry et al. (1982) for the northeastern Pacific. Both catalogues were constructed using the objective classification scheme of Kirchhofer (1973) in which estimates of daily atmospheric pressure at predefined hemispheric grid points are subjected to a sequence of selection and grouping procedures. Full details of the catalogues and grouping schemes are given in Kirchhofer (1973) and 63 Yarnel (1984). Only potential problems with the technique are considered here. Because of the length of the study period (>10,000 days in both studies) a portion of the pressure/elevation data set is used to define 'key days' representative of specific synoptic groups. The major assumption here is that the smaller data set is representative of the range of synoptic conditions over the longer study period. Problems with the Kirchhofer method include: (1) selection of a cutoff threshold for grouping types - a low threshold produces a larger set of synoptic types and leaves few days unclassified, whereas a high threshold produces fewer types but leaves many unclassified days (Key and Crane, 1986); (2) grid size selection - size will influence the spatial resolution of synoptic patterns with large grids yielding poor definition of mesoscale circulation phenomena; (3) within-group variability - even small changes in the inflection point of a trough or ridge can alter the pattern of temperature or precipitation over an area with no apparent differences in the zonality of circulation; and (4) several key synoptic types may not be classified by the daily analysis during a period when circulation patterns are changing rapidly and when there is a high frequency of missing data. Although a less stringent grouping of synoptic types results from the use of 'key days' and specification of a subjectively defined grouping threshold (Key and Crane, 1986), the same methods were employed in both studies, which enhances the compatibility of interpretations derived from the catalogues. The Yarnel (1983) catalogue, for the period 19^6-1978, is based on eighteen, 500 mb pressure-elevation patterns for a 2.5° X 2.5°» 30-point grid covering the coastal regions of British Columbia and the far eastern Pacific. The Barry et al. (1982) catalogue is based on 31 surface pressure patterns for a 35-point, 5° X 5° diamond-shaped grid over western North America and is for the period 1899 to 1980. Only 6 grid-points in the latter catalogue fall within the coastal B.C. and Gulf of Alaska region. The larger number of synoptic types in the surface pressure catalogue is due to the greater variability in surface pressure patterns and the larger spatial coverage of the grid. Characteristics of synoptic types in each catalogue are summarized in table 3-2. Difference of means tests were carried out to determine if there are significant changes in the frequency of surface and 500 mb synoptic types associated with winter precipitation for: (1) periods of above average and below average precipitation indices along the central coast and (2) groups of years comprised of the most extreme departures in winter precipitation irrespective of ordering in time.* The results presented in table 3-3a show that the frequency of "wet" surface and 500 mb synoptic types are not significantly different between the dry period (1948A9 to 1957/58) and wet period (1959/60 to 1968/69). In contrast, when groups of years characterized by the most extreme departures are tested (table 3-3b), there is a significant difference between wet and dry intervals with an average of ten more days per winter season dominated by "wet" 500 mb types and fifteen more days per season dominated by moisture bearing, surface lows. The same sequence of tests was conducted for cool and warm summer periods, but due to the comparatively weak circulation intensities during the summer, there is greater difficulty in identifying distinct surface pressure patterns associated with extremes in summer temperature. Thus, only the 500 mb patterns are considered (table 3-*0 • Similar conclusions can be made for summer temperature regimes as for winter precipitation in that there is no significant difference in the dominance of a particular 1. Extreme years were defined as those years with climatic departures in excess of one standard deviation of the mean. Sample sizes were determined by the group with the fewest years fitting this criterion. 65 Table 3.2. Synoptic Types from Yarnel (1983), Barry et al. (1982). Group or Type (Yarnel - 500 mb) (Barry - surface) Synoptic Circulation Climate Effect Wet Y-(l,3,4,8,10) B=<3,4,7,11,12, 15,23,31) - southwesterly cyclonic component and advection of moist air due to positive 500mb pressure anomalies over the western USA and central Canada and an intensified Aleutian Low - surface lows are mostly from the west and northwest but also migrate from the south above average precipitation and temperature for the entire B.C. coast (upper air types 1,4) and the central and southern coast (upper air types 3,8,10) Dry Winter Y=(5,9,14,15,17) B=(8,14,18,26,30J north or northwest flow aloft and strong ridging centered on the coast surface lows diverted into Gulf of Alaska or southern US west coast below average temperatures and precipitation for the entire coast (upper air types 5,9) and the southern coast (upper air types 14,15,17) Dry Summer Y-C2.6,11,13,18) B=(5,6,9,13,21) strong zonal flow or ridging centered west of Vancouver Island dry and warm for most of the central and southern coasts (upper air types 2,6) occasional cold lows (11,13,18) Warm Summer Y=(l,4,6,7,15,18) B*»<3,4,6,11,23,25) ridging centered over the coast or inland with advection of warm air aloft from the southwest well above average temperatures for the entire coast (upper air types 1,4) and above average for the central and southern coasts (6,7,15,18) 1. Type numbers refer to the 18 objectively classified daily synoptic-scale patterns (500 mb) of Yarnel (1983) (Y=) and 31 surface pressure types of Barry et al. (1982) (B=). Table 3.3. Winter Precipitation and Synoptic Type Frequencies. A. Significance test between a dry period and a wet period along the central B.C. coast Mean winter precipitation (mean departure) Wet synoptic type frequency^ Wet synoptic 3 type frequency" dry: 1948/49 to 1957/58 wet: 1959/60 to 1968/69 -0.04 ± 0.03 0.07 ± 0.05 82.2 ± 8.3 86.6 ± 8.8 42.5 ± 10.5 44.4 ± 7.8 One-tailed t-test (a = 0.10, df = 18)' t = 2.73 significant t = 1.11 not significant t = 0.46 not significant B. Significance test between wet years and dry years dry years: 1978,71,70, 56,49 -0.19 ± 0.08 78.8 ± 7.0 33.4 ± 7.23 wet years: 1976,74,68, 64,54 0.27 ± 0.08 89.4 ± 5.8 48.0 ± 8.9 One-tailed t-test t - 8.48 t = 3.38 t = 2.80 (<z = 0.10, df = 8) significant significant significant 1. Calculated as the average departure of winter precipitation (Oct-Apr) for the central coast precipitation index from the 1951-1980 normal. 2. Mean number of days dominated by "wet", 500 mb, synoptic types from October to April after Yarnel (1983). 3. Mean number of days dominated by "wet" surface synoptic types from October to April from Barry et al. (1982). 4. A low significance level was selected because of the exploratory nature of the analysis. Table 3.4. Summer Temperature and Synoptic Type Frequency. A. Significance test between a cool period and a warm period in Bella Coola basin Mean Summer temperature (mean departure) Warm Synoptic type frequency^ cool: 1948/49 to 1957/58 -1.7 ± 0.8 warm: 1959/60 to 1971/72 1.9 ± 0.4 20.5 ± 7.0 22.4 ± 8.0 One-tailed t-test (a = 0.10, df - 21) t = 4.62 significant t = 0.56 not significant B. Significance test between cool years and warm years cool years: 1975,57,56, 54,49 -2.3 ± 1.0 warm years: 1972,70,69, 67,61 1.9 ± 0.7 14.5 ± 6.1 26.5 ± 3.9 One-tailed t-test (a = 0.10, df - 8) t = 5.31 significant t - 3.00 significant 1. Warm synoptic types from Yarnel (1983). 68 group of synoptic types between a warm interval (1959/60 to 1971/72) and cool period (1948/49 to 1957/58) (table 3.4a), but for groups of years characterized by extreme departures a significant difference is evident (table 3.*tb) . Unlike winter precipitation, extreme departures in summer temperature tend to fall within periods of the same tendency. This would suggest that an interval of persistent departures in summer temperature has a higher probability of extreme occurrences and it is these extreme, within-period, years which are characterized by significantly different frequencies of synoptic types. Conversely, anomalous winter precipitation does not appear to be restricted to these longer more persistent intervals of above or below average departures. One or two exceptional years may occur during an interval of otherwise average conditions (e.g. 197*+. 1976). in addition to significant changes in synoptic type frequency positive anomalies in winter precipitation can be distinguished by the within-season persistence of synoptic types. For example, during the very wet winter season of 1963/64, 36 days between mid-October and mid-November (with the exception of one day) were classified as "wet" 500 mb synoptic types - clearly the most exceptional signal in the entire 33 year record. During that same winter season and for others between i960 and 1969 intervals of between 10 and 12 days were not uncommon whereas during wetter winters within dry periods these within-season runs were mostly less than 5 to 8 days. The conclusion drawn here is that the synoptic signature between drier and wetter intervals is related to the within-season persistence of "wet" synoptic types whereas anomalously wet years are characterized by both a higher frequency and greater persistence of these same types. Extreme departures in precipitation do not appear to be restricted to a 69 period of given tendency. This distinction is an important one particularly if there are well-defined response thresholds in the biogeophysical environment below which the impact of changing weather is not significant. (3.2) Responses of Hydrologic Variables to Climate Fluctuations Large scale motions of the atmosphere are manifested not only in regional precipitation and temperature indices, but also in hydrological phenomena such as winter snowpack, river runoff and glacier fluctuations. Linkages between these components of the terrestrial phase of the hydrological cycle and atmospheric circulation can be spatially non-homogeneous and lagged in time (Barry, 198l). This occurs because the atmosphere, a system with low heat and moisture capacities, adjusts rapidly to changing energy levels (Bradley, 1985)- The high water storage capacity of the terrestrial environment (e.g. snowcover, glaciers, lakes, groundwater), spatial variability in actual storage and the existence of response thresholds all tend to dampen the response to climatic change. For this reason it is necessary to understand how feedback mechanisms might be important in the system. The spatial and temporal trends in water storage and runoff along the southwestern British Columbia coast are examined here. Regional Winter Snowpack Snowpack data have been collected for a number of sites in southern British Columbia but few have records longer than kO years. The majority of solid precipitation falls between early November and late March. Thus April 1 snow water equivalents were compiled from a number of stations (see figure 2.1 for locations) and grouped into homogeneous regions using factor analysis (table 3-5)• Two statistically distinct groups exhibiting similar long term trends but minor interannual differences were identified: a Table 3.5. Winter Snow Courses in southwestern British Columbia. Station Latitude Longitude Elevation Years of Record Rotated (m) Components 1 2 COASTAL Bella Coola 52 31* 126°38' 1380 1953-1954 Newcastle Rdg. 50 24' 126°03' 1170 1961-Present 0.746 0.290 Powell (upper) 50 16' 1040 1938-Present 0.753 0.541 Dog Mountain 49 23' 1080 1945-Present 0.823 0.525 INTERIOR Precipice 52 26' 125 38' 1220 1972-1973 1970-1971 Tatlayoko Lake 51°36' 124°20' 1710 1952-Present 0.280 0.848 1300 1952-Present 0.319 0.872 Tenquille Lake 50 32' 122 56' 1680 1953-Present 0.332 0.895 1. See figure 2.1 for locations. 2. Rotated component loadings for snow course stations. Stations are grouped according to factor scores. 71 coastal zone to the west of the Coast Mountain crest and a zone in the lee of the Coast Mountains towards the drier interior. Secular variations in snowpack are plotted as adjusted partial sums in figure J>.k. Again, positive slope segments represent above average departures and negative slopes below average departures. A greater interannual variability is evident for the regional snowpack index representative of the interior but both series exhibit three periods of persistent departures: normal to below normal snowpack between 1957 and 1963f well above normal snowpack between 1964 and 1976, and finally strongly negative departures after 1976. The occurrence of below normal spring snowpacks after 1976 corresponds well with below normal winter precipitation along the central and southern coast of B.C. coupled with normal to above normal winter temperatures (figure 3-2). Between the mid-1960s and 1975 normal to above normal winter precipitation along this same coastal zone is reflected as persistent positive departures of winter snowpack for both coastal and interior regions. Winter temperatures were on average below normal during this same interval. Below normal winter precipitation and above normal winter temperatures dominated the southern coast between 1957 and 1964 resulting in below normal snowpacks. The association between regional snowpack volume and frequency of "wet" winter circulation patterns for periods and groups of years characterized by anomalous departures is tested using difference of means. Unlike the strong association between positive snowpack departures and above normal precipitation, four of the six extreme low snowpack years were not the same years with low winter precipitation. For three of these four years winter temperatures were above average which probably led to greater melting near the beginning and end of each season. Hence, average precipitation and above normal temperatures resulted in low snowpack 72 —I 1 1 1 r 1950 1960 1970 1980 1990 Year (AD) Figure 3-*» Adjusted partial sums for April 1st snowpack (water equivalent). Normal period is 1951-1980. Coastal winter precipitation taken from figure 3-1 is included for comparison. Positive slope segments represent persistent above average snowpack years whereas negative slopes represent below average snowpack. 73 volumes. Even with these complications the results in table 3-6 show that variations in spring snowpack are positively linked to the mean number of days dominated by "wet" synoptic types for groups of anomalous years and also for the two periods characterized by above average (1966-1976) and below average (1977~1980) departures. The latter result is partly related to the limited number of data during 1977~1980 and the above temperatures which occurred then, but does indicate that seasonally integrated hydrologic parameters such as snowpack are sensitive to changes in the frequency of specific circulation patterns not only for extreme years but also for sequences of years in which a tendency for persistent departures occur. Spring Runoff Spring runoff is likely to be greater following winters characterized by anomalously high precipitation and snowpack accumulation but spring and summer temperature conditions may enhance or depress actual water yield. Furthermore, spring runoff should be less spatially homogeneous than snowpack distributions because of the large spatial variability in runoff generating mechanisms and storage conditions within and between basins. However, as basin size increases, water yield becomes less responsive to isolated storms or to snowmelt periods of short duration. Thus, large scale synoptic trends may be reflected more directly in temporal runoff variability within larger basins (Bruce, 197*+). Most hydrometric records for streams draining the west slope of the southern Coast Mountains are shorter than kO years and the number of medium 2 sized basins (3,000 - 6,000 km ) for which discharge data exist is small. Time series of spring snowmelt runoff for the Bella Coola, Homathko and Squamish Rivers, three moderate size catchments draining partly glacierized 74 Table 3.6. Winter Snowpack and Synoptic Type Frequency. A. Significance between a high snowpack period and a low snowpack period Mean April 1st water equivalent (mean departure) Wet synoptic Wet synoptic 2 3 type frequency type frequency high: 1966-1976 low: 1977-1980 One-tailed t-test (a = 0.10, df = 13) 0.69 ± 1.26 -1.01 ± 1.22 t = 2.94 significant catalogues only to 1978 40.3 ± 8.9 32.3 ± 12.1 t = 1.61 significant B. Significance test between high snowpack years and low snowpack years high snowpack years: 1.32 ± 0.93 91.5 ± 4.13 46.5+8.8 (1976,74,68,67,64,54) low snowpack years: -1.28 ± 0.75 79.2 ± 6.3 29.8 ± 15.2 (1978,77,75,70,62,60) One-tailed t-test (a = 0.10, df = 10) t = 5.22 significant t = 4.01 significant t = 2.32 significant 1. Calculated as the average departure for the three lee side snow courses: Tashta, Tatlayoko and Tenquille. 2. Frequencies of 500 mb patterns after Yarnel (1983). 3. Frequencies of surface pressure patterns after Barry et al. (1982). 75 mountains of southwestern British Columbia, provide some evidence of secular runoff trends during the post-19*t5 period (see figure 2.1 for locations). Results are plotted as adjusted partial sums in figure 3«5> The interval May to mid-July was defined as the spring runoff period because hydrographs from all three rivers indicate that discharge peaks related to snowmelt occurred during this time (see figure 2.6). While there are minor interannual differences between each catchment, similar long-term secular trends, are evident. Fluctuating, but near average discharges occurred between 1947 and 1958 in the Bella Coola River. Following well above average spring runoff during 1958, below average discharges in Bella Coola and Squamish Rivers prevailed until 1966. Homathko River discharges were more variable through this period. After 1966, discharges for most of the coast were well above average until 1972 when this trend was sharply reversed for the remainder of the decade. Post-1980 flows are near average. A comparison of runoff trends with snowpack water equivalents and winter precipitation (figures 3-1. 3-** and 3-5) reveals some important differences. During the interval 1973-1976, when wet winter conditions prevailed, spring runoff in all three basins was below average. This divergence can be explained by the occurrence of well below average winter and summer temperatures during all four of these years resulting in lower runoff and increased water storage. Similarly, wet but cool conditions between 1963 and 1966 may explain the below average spring runoff, particularly in the Squamish River. Spring runoff data from the north coast were not examined but Royer (1982) has modeled secular trends in annual freshwater discharges in southeast Alaska. The estimates are based on seasonal trends in precipitation, temperature, and water storage and are calibrated against various catchment runoff data. The annual trends appear to fit quite well 76 T 1 1 1 1 1 r 1 1— 1940 1950 1960 1970 1980 Years (AD) Figure 3-5 Adjusted partial sums of spring runoff (May - mid July) for three coastal rivers. Normal periods for each station correspond to the length of record. Positive slope segments represent persistent above normal runoff. Note the the regional response of below average runoff after 1972. 77 the variations in winter precipitation along the north coast (figure 3.2a), suggesting that precipitation is the dominant factor in determining the temporal variation of runoff. Above average runoff occurred during the interval 1958 to 19&3 and after 1976. and below average runoff between 1964 and 1976. Below average runoff and winter precipitation prior to 1958 have been documented for Prince of Wales and Revi1 lagigedo Islands in the Alaska Panhandle (Karanka, 1986) which contrast with the above average annual runoff identified by Royer (1982). These differences for the earlier period may be due to variations in the definition of the water year, the much more limited area of the latter study and the sensitivity of the runoff model used by Royer. In general, the spatial coherence of winter precipitation and spring runoff is strongly evident but spring and early summer temperatures also contribute to runoff variability. Spring and Summer Runoff - Bella Coola River Factors other than winter precipitation have been incorporated into a number of empirical and physically based models for estimating spring flood discharges and high summer water yields (Ostrem et aj.,1967; Tangborn and Rasmussen, 1976; Power and Young, 1979; Collins and Young, 198l; Tangborn, 1984; Young 1985)- The most important are the distribution and magnitude of temperature departures and the contributions of glacier melt to total runoff. Glacier melt generated runoff can be as high as 50% of the total water yield even when glacial cover is relatively limited (Young, 1985). Estimates of glacier melt contributions to summer discharge in the Homathko River (21% glacial covered) range as high as 21% (British Columbia Hydro, 1984). While accumulation season precipitation and ablation season temperatures are important model components, these studies of ice-covered 78 basins have shown the importance of more spatially variable parameters such as relative humidity, incident solar radiation and wind in estimating summer runoff from individual glaciers. Direct measurement, or even indirect estimates, of these additional parameters are generally not possible for paleohydrological modeling. Models based on primary hydrometeorological variables such as temperature and precipitation are likely to explain less of the year-to-year variance in runoff. An assessment of simple runoff models for the Bella Coola River is made here in order to document the level of explained variance using primary meteorological variables only. Analysis of ablation season runoff in all four gauged tributaries of the Bella Coola River reveals some interesting differences. Interannual variability of early season runoff (May-mid-July) is similar throughout the catchment, presumably due to the contribution of water from a basin-wide melting snowpack. As the melt season progresses and icemelt contributions increase, two distinct runoff regimes become apparent: (1) a strongly diurnal runoff regime for the southern catchment (e.g. Talchako/Nusatsum Rivers) and (2) more uniform, but steadily declining flows, in the eastern and northern segments of the catchment (e.g. Atnarko/Salloomt Rivers) . During high snowpack years, and normal to below-normal snowpack years in which spring temperatures (April - May) are depressed, high seasonal runoff is common to all tributaries. It is low snowpack years coupled with high summer temperatures which accentuate the differences in runoff regimes. Since this latter situation occurs relatively infrequently, a summer runoff index was computed for the Bella Coola River below Burnt Bridge Cr., which integrates discharges from both glacierized and unglacierized sources. Although this may reduce the climate-runoff relationship by mixing populations, the index incorporates a basin-wide response to changing 79 hydroclimates and more properly estimates total summer discharge through the lower Bella Coola River. Several empirically-based linear statistical models, incorporating seasonally and regionally averaged climate data, were evaluated using an interactive version of all-possible-subsets regression. Indices of seasonal and monthly temperature, precipitation and spring snowpack for both coastal and interior regions comprise a group of independent variables used to estimate spring runoff. Preliminary models were constructed using a 33 year data set (1952-1984), limited by the length of spring snowpack records. Early season (May to mid-July), late season (mid-July to August) and summer (May to August) runoff indices were considered separately. Results are presented in table 3-7' Scatter plots of bivariate relationships, probability plots of residuals and a test for serial correlation for each model indicate that assumptions of the regression technique have not been vi olated. All three preliminary models relating Bella Coola River runoff to 2 climate are significant but the strongest is summer runoff (Ra = O.58) in which winter precipitation along the coast, spring snowpack volumes measured in the lee of the Coast Mountains and summer temperature indices derived for the interior, enter the equation. This combination of variables suggests that summer runoff is controlled by snowpack conditions and melt energy in the headwaters of the southern and eastern catchment. The three preliminary models were tested by using independent data from the Nusatsum and Salloomt Rivers. Runoff data from these two downstream tributaries were combined and compared with predicted runoff in Bella Coola River. The use of Nusatsum/Salloomt data was possible because of the high correlation between these data and observed discharge in the Bella Coola River (r > 0.79)- Verification procedures for regression models 80 Table 3.7. Significant variables and regression coefficients for climate-runoff models. Runoff Period Preliminary Models Final Model Climate Indices May to mid-July mid-Jul to Aug Coefficients Summer Summer WP(I) WP(C) SS(I) SS(C) AT(I) AT(C) ST(I) ST(C) Intercept 0.62 0.23 -0.09 1.21 0.11 0.49 0.49 0.01 0.56 0.18 0.78 0.01 0.68 0.78 0.01 Summary Statistics: „2 F SE = 0.52 = 0.45 = 1.50 = 7.15* = 0.45 0.29 0.24 2.22 5.8* 0.12 = 0.62 = 0.58 = 1.51 14.8* = 0.09 = 0.55 = 0.52 = 1.94 17.16* = 0.10 Confirmatory Statistics: r RE 0.74 0.19 0.41 0.05 0.89 0.41 0.84 0.36 WP = winter precipitation (Oct-Ap) SS = April 1st snowpack (water equivalent) AT = April temperature ST = summer temperature (May-Aug) (I) = Interior Region (C) = Coastal Region R = coefficient of determination 2 2 R = adjusted R a D = Durbin-Watson d (reject H - no o significant autocorrelation) F = F ratio (* significant; a = 0.025) SE = standard error RE = reduction of error statistic r = correlation coefficient The coefficient of determination (R ) is adjusted (yielding Ra ) to account for the number of predictor variables in each equation. 81 evaluate how well a model predicts a new value of the independent variable. Two statistics reported in table 3-7 (confirmatory statistics) are used to test the significance of the predicted runoff values: the correlation coefficient (r) and RE statistic. In all three models the correlation coefficients are significant suggesting a reasonable correspondence between actual and predicted runoff. The reduction of error statistic (RE) (Fritts, 1976) has theoretical limits of -<» and 1.0 where a positive value indicates that the predictions made by the model are better than a prediction based on the mean of the observed data. Again all three models have positive and significant RE values suggesting reasonable predictive ability of the models. The model which incorporates spring snowpack, a variable with the lowest partial correlation coefficient, was removed from further consideration because snowpack data prior to 1952 are not available. The calibration period was extended to 19^8 (earliest available runoff) and all outliers which might have had a significant influence on the regression coefficients were tested using Cook's D statistic (Draper and Smith, 1981) . None were found. The final model is used to reconstruct spring runoff prior to 1948 and these results are discussed in chapter 6. Analysis of spring runoff hydrographs shows that extreme runoff events in May or June are related to years with persistent positive departures in monthly winter precipitation and negative spring temperature anomalies. For this reason, the final model presented in table 3-7 may approximate the potential for high-magnitude spring runoff but, because there is no spring temperature index in the final model, is apt not to provide a good estimate of peak spring discharge. Hence, residual variance in the final model is highest during years characterized by monthly climatic anomalies which significantly influence runoff (e.g. depressed 82 April temperatures) but are not incorporated into any of the seasonal climate indices. Autumn Floods As noted in chapter 2 the autumn runoff period is hydrologically dissimilar to the summer discharge regime. Steadily declining freezing levels promote reduced contributions of glacier meltwater as the season progresses. Superimposed on these declining flows are short-duration, high-magnitude floods which have two distinct sources: (1) runoff due to heavy precipitation only and (2) runoff due to high precipitation combined with rapidly rising freezing levels and significant snowmelt at middle elevations. Typically, the former events occur early in the autumn (August to October) while the latter may occur as late as January following accumulation of snow throughout much of the basin. Analysis of daily precipitation records prior to autumn flood events in the Bella Coola River suggests that for most high-magnitude flows, antecedent precipitation is important. Flood generating storms are often preceded by a sequence of one or two day rainfall events which saturate the snowpack or soil so that subsequent runoff is maximized. Figure 3*6 demonstrates a mean separation of k to 5 days between storms, beyond which the potential for heavy storm runoff is greatly reduced. This separation of storm peaks appears to be valid for peak 2h hour precipitation events with recurrence intervals of k years or less; that is, flood generating storms with one-day recurrence intervals greater than k years generally produce enough precipitation to yield significant runoff and are not necessarily preceded by hydrologically (climatologically) distinct events (e.g. autumn floods of 1950, 1973) . 83 Figure 3-6 Antecedent autumn precipitation for the six largest period-of-record floods on Bella Coola River. 24 hour precipitation recurrence intervals are calculated using 80 years of record at the Bella Coola climate station. Mean separation of storm precipitation peaks is 4 to 4.5 days. 84 Analysis of synoptic charts for ten-day sequences prior to each major flood on the Bella Coola River suggests a number of synoptic conditions must be met in order to produce significant runoff. The most important is the rapid reduction in the velocity of eastward moving low pressure systems and in several cases a switch to a more northerly trajectory into the Gulf of Alaska. This is facilitated by high pressure ridging over the interior of the province. Warm and cold fronts become fully occluded as the centre of low remains over the coastal boundary. The warm sector aloft produces heavy rainfall and snowmelt at higher elevations. With the exception of the smaller 1982 flood, cyclogenesis is almost always centered south of the Aleutian Low between 155° and 170° W and between 45° and 55° N. Storm precipitation plots in figure 3-6 also reveal a major difference in synoptic climatology for floods within a sequence of years of above average precipitation (e.g. 1965. 1968). These events are characterized by a closely spaced sequence of lows intersecting the coast under a synoptic regime which persists for at least ten days prior to the storm and in some cases longer. Floods during other intervals do not have strong antecedents and are generally singular events (e.g. 1950, 1973). The spatial variability of autumn floods requires consideration if inferences regarding regional hydrology drawn from site specific evidence are to be valid. The intensity or size of cyclonic disturbances, storm track orientation and source areas for cyclogenesis may significantly affect one area without any appreciable effect in another. Evidence presented earlier suggests different winter precipitation trends along the north and south coasts, where the central coast is a transitional zone alternately exhibiting characteristics of one region or the other. For this reason the magnitude and timing of fall flood events in the Bella Coola basin are compared with Homathko and Squamish Rivers, 85 catchments of similar size and glacier cover in the central and southern coastal environments, respectively (see figure 2.1 for locations). Figure 3.7 is a plot of peak autumn discharges above a base discharge in each river for period-of-record data. It is apparent from figure 3-7 that autumn flood magnitudes for a given storm vary spatially. For example, the period-of-record maximum flood in October 1957 on Homathko and Squamish Rivers was only the sixth largest on the Bella Coola River. Although magnitudes may vary, the significance of autumn storms can be seen in the coast-wide response to rainfall and rain-on-snow events. The major autumn flood of late October, 19&5 on Bella Coola River corresponds with significant runoff events in the other coastal rivers, particularly Squamish River. The less precisely matched record with Homathko River may be due to the higher baseflow for that river and differences in physiographic or glaciologic effects. Runoff events in the Bella Coola basin during 1967» 1980 and 1982 were also recorded in these other rivers. This suggests that there is potential for using hydrological records of any one of these rivers as an index of major synoptic-scale storms for southwestern British Columbia. These synoptic-scale storms then become 'key' events in the longer term departure of climate along the central and southern coasts. Another feature common to all three rivers is the increased number of floods during the period 1957~1968. This is coincident with above average October - April precipitation along the coast associated with an increased frequency of ridging along the Pacific Northwest specifically, and the northern Hemisphere in general (Treidl et al., 1981; Knox and Hay, 1985). The evidence indicates that positive SST anomalies for the northeast Pacific Ocean and an intensified Aleutian Low were major factors. 86 OR Qb Qp_ Qb Qp Qb 5.0 -4.0-3.0-2.0-Bella Coola River near Hagensborg Ad=4,040 km2 GC=9% R=3533 m Qb=200 cms | • l I I | I l l I | I I 1 I | "I "1*1 I *| I I I I | F I I I | "I I I >»| 1945 1950 1955 1960 1965 1970 1975 1980 1985 6.0 -3.0-2.0-Homathko River at mouth r Ad=5,720 km2 R=3994 m GC=21% 4 Qb=400 cms In II I I I | I I ' | I I "» "I | "I "l "V* 'I 11 ' 'l '| I I I "I '| "I "l 'l I | 11 |"| i H 1945 1950 1955 1960 1965 1970 1975 1980 1985 5.0 -4.0 3.0 " 2.0 -Squamish River Bracitendale Ad=2,330 km2 R=2890 GC=16« Qb=350 cms I I I I I l I l l I 'l-l 1»"11 '< Hrr M-r-rl-r r-r 'il I r*i-rJr-r 1945 1950 1955 1960 1965 1970 1975 1980 1985 Figure 3«7 Frequency of autumn (August to January) rainstorm generated floods of the Bella Coola, Homathko and Squamish Rivers. Q is the maximum mean daily flow above a base discharge Q^. was ' selected as the mean August-September daily discharge to emphasize flood peaks above the maximum glacier melt contribution. Statistics include drainage area (A,), relative relief (R) and % glacial cover (%GC) . d 87 In summary, flooding on the Bella Coola River is related to several seasonally-averaged climate conditions as well as singular anomalies in the general circulation. High magnitude spring runoff is a function of accumulated snowfall throughout the winter season and spring temperatures: a short-term response to seasonally integrated climate. A higher probability of spring flooding occurs during intervals of increased winter precipitation (e.g. 1964-1969). Peak autumn floods, which generally produce the largest instantaneous discharges, are not restricted to these same intervals, although there is a higher frequency of flows above a given base discharge during years of above average winter precipitation. The major implication for inferences regarding seasonally-averaged hydroclimates is that certain flood-related evidence may not be exclusively associated with a particular runoff regime or interval of persistent climatic departures. Flood Frequency Five different runoff regimes can be identified for the Bella Coola River: spring snowmelt, glacier melt, rainstorm only, and spring and autumn rain-on-snow. Rain-on-snow during the autumn produces the highest discharges while spring snowmelt and rainstorm augmented glacier melt are probably the next most important in terms of peak discharge and sustained high runoff. In hydrological terms each could be treated as a separate population with distinct distributions and individual statistical properties (cf. Waylen and Woo, 1983)• In terms of impact on river morphology, channel and bank erosion, sediment transport and related alluvial processes, distinguishing between these several distributions is less important. All flows above the threshold for significant sediment motion should be considered. Therefore, the partial duration series is used here because it is based on all flows 88 above a given discharge regardless of their ordering in time. The partial duration series of all four gauged catchments and inferred flows of the Talchako River are shown in figure 3'8. The curves in figure 3-8 are ordered on the basis of catchment size with the exception of the Talchako River which produces higher-magnitude floods compared with the larger Atnarko River. Those basins with higher percentages of ice-cover have proportionately higher flood magnitudes with increasing recurrence intervals (;'.e. the ratio of Q20/Q2 'S ^ar9er *or t'ie Talchako, Nusatsum and Salloomt Rivers). At recurrence intervals less than 1.5 all plots exhibit similar curvature. The partial duration series for Bella Coola R. begins to depart from the extreme value series at flows near Q_2 which corresponds with the mean flood calculated using the partial duration series. Thus, the frequency of flows with moderate magnitudes on 3 -1 Bella Coola River (approximately 400 - 600 m s ) is notably higher on the partial duration series. This distinction remains an arbitrary one if only high-magnitude floods (Q^ or perhaps even as high as Q^Q) produce significant changes in the alluvial system. However, if in addition to high-magnitude floods sequences of moderate floods have a noticeable impact on the alluvial system, then an accurate estimate of their frequency is important. Changes in hydroclimatology which favor a higher frequency of flood flows in both autumn and spring seasons are apt to be better represented by the partial duration sequence. Possible shifts in the long-term flood frequency curves in response to changes in the hydroclimatic regime are considered below and in chapter seven. 89 10* within „ ,, „ , channel flows Bella Coola R. Partial Duration Series overbank flows <r> 3 OS < X Sio2 (n=39) alchako R.~ Atnarko R: Nusatsum -R. Salloomt R. (n=21) 10J <2. 1 mean flood (evs series) = mean flood(partial duration series) 1.05 1.25 2 5 10 RECURRENCE INTERVAL (YEARS) 20 50 100 Figure 3-8 Partial duration series for runoff from all five gauges in the Bella Coola River watershed. Within-channel and overbank flows are defined for the stable reach in which the main gauge is located (Bella Coola River above Burnt Bridge Creek). 90 (3•3) Longer-Term, High-Resolution Climate Analysis The frequency of both extreme and persistent departures in climate during the 20th century can be inferred from only a small number of climate stations in the study area. Although inferences of basin-wide climate will be biased towards these spatially-restricted records, comparison with other regional records and notable hemispheric changes in dominant circulation patterns provide a basis for assessing the utility of the records as sources of regional climate information. Monthly data are examined here for three reasons: 1) to augment and extend the short-term, regional records; 2) to provide a basis for comparison with biological and geophysical responses in the basin and; 3) to identify the occurrence of extreme intra-and interannual variations in the longer-term climate record. Neilson (1986) has displayed records of precipitation and temperature as contoured, two-dimensional surfaces. This technique has been adopted here for the two longer-term stations in the vicinity of the study area. The Bella Coola climate station is assumed to be representative of environmental conditions towards the west side of the basin specifically, and coastal conditions in general. Big Creek climate data are used to characterize trends in the eastern basin or Fraser Plateau area (refer to figure 2.1 for locations and table 3-1 for factor groupings). Contour surfaces of total monthly precipitation and monthly averages of mean daily temperature for the period 1904 to 1983 are illustrated in figures 3-9 and 3-10. Fifteen months of data are presented along the horizontal axes representing the current year and the last three months of the previous year. The most obvious features of figure 3-9 are the winter maximum in precipitation for Bella Coola and a trend towards above average precipitation in the latter half of the century (with the exception of high October to November precipitation between 193** and 19*t0) . There is also 91 MONTH Figure 3*9 Contoured surfaces of intra- and interannual temperature and precipitation for Bella Coola cl imate. stat ion, 1904-1983 (see figure 2.2 for station location). October to December values on the left are from the previous year. Temperature contour interval is 5° C and precipitation contour interval is 40 mm. The shaded areas on the temperature plot represent extreme departures. MONTH Figure 3-10 Contoured surfaces of intra- and interannual temperature and precipitation for Big Creek climate station, 1904-1983 (see figure 2.1 for station location). October to December values on the left are from the previous year. Temperature contour interval is 5° C and precipitation contour interval is 20 mm. The shaded areas on the temperature plot represent extreme departures. 93 evidence for higher and more persistent October to January precipitation after 1953- Summer precipitation was lowest between 1924 and approximately 1945 and during the interval 1963 to 1976. Mean December to February temperatures were consistently below 0°C until 1920 after which gradual warming occurred until at least 19^5 (note the frequency of closed contours < -5°C prior to 1920). Post-ig'tS winter temperatures, particularly 1948 to 1958 and 1962 to 1976, were more similar to those recorded prior to 1920. A tendency towards warmer springs in the latter half of the century is also discernible (see the slope of the 5 and 10°C contours near March and April). Cooler summers prevailed until approximately 1930 after which higher temperatures were evident with the exception of 19^5-1957 and 1968-1976. Some interesting differences and similarities are associated with climate trends at Big Creek (figure 3-10). A summer maximum in precipitation dominates there with lowest summer totals occurring between 1920 and 1940. Summer and winter precipitation is generally higher after 1940 compared to the first half of the century. Like Bella Coola, pre-1920 winter temperatures were amongst the lowest in the century, although unlike Bella Coola warming was not so evident until after 1940. The cooler winters of 19^8-1955 are most similar to the pre-1920 temperature departures. Early century summer temperatures at Big Creek were comparatively warm but maximum temperatures were not recorded until after 1920, ending around 1930 with a cooling trend which lasted for the next decade. Summer temperatures after 1958 are the lowest of the century with the exceptions of 1978 and 1979-With the exception of the interval 1930 to 19^0, the pre-1950 period at Bella Coola is characterized by years with less persistent departures of intra-annual monthly precipitation when compared with the post-1950 record 9** (figure 3-9°). This is particularly true for the winter months. At Big Creek, summer and early autumn months exhibit persistent positive departures in the pre-1920, 1940-1965 and post-1980 records. Extremes in monthly precipitation tend to cluster within these periods of positive departure but are not exclusively found there. The contoured surfaces would suggest a non-random ordering of climate departures through the 20th century. It has been widely recognized that three distinct periods of characteristically different circulation regimes have occurred over the last 80 years related to transitions from meridional flow patterns (pre-1920 and post-1945) to zonal flow patterns (1920-1945; see table 3-8). When the energy gradient between the pole and equator is steep, a series of high amplitude waves generate stronger longitudinal contrasts in climate than would be experienced under a zonal flow regime (Makrogannis et al., 1982). Persistent and often extreme departures occur when the meridional wave pattern remains entrenched. Even small longitudinal shifts in the mean position of blocking ridges and associated troughs can lead to abrupt changes in the direction of departure. Under a zonal flow regime, year-to-year variance is enhanced and the frequency of persistent departures is greatly reduced. Biasing and Lofgren (1980) have analyzed monthly sea-level pressure data for the N.E. Pacific sector covering the period 1899 to 1971' Using the map correlation method, they were able to classify objectively dominant circulation types characteristic of each period. During winters of the first two decades of this century above normal pressures in northern Canada resulted in an anomalous flow of cold arctic air, presumably associated with a southward displacement of the arctic front, resulting in cooler and moister conditions along the outer British Columbia coast. A strong 95 Table 3.8. Recognized Trends in 20th Century Synoptic Conditions in the N.E. Pacific Sector. 1900 -1920's SYNOPTIC CONDITIONS - meridional flow (some zonal) 1920's -1950's - zonal flow post 1950*5 meridional flow METHODOLOGY classification of N. hemisphere surface pressure maps by (1) and eigenvector analysis by (2) SOURCE 41) Dzerdzeevski (1966, 69) (2) Balinicky (1974) 1900 -1920's 1920's - 1950's 1950's -1970 decreased intensity of the January Aleutian Low intermediate conditions increased intensity of the January Aleutian Low and an eastward extension eigenvector analysis of N. hemisphere sea level pressure maps for both January and July Kutzbach (1970) 1899 - 1920's - above normal pressure in high latitudes of the Pacific (cool) map correlation method using 1920's - 1940's - below normal pressure in N. sea level pressure data Biasing and Lofgren (1980) Pacific sector (warm/dry) post 1940'a - more average conditions 1893 - 1920 - cooler/wetter in U.S. West 1920 - 1954 - warmer/drier 1955 - 1977 - cooler/wetter areally averaged monthly mean temperature and precipitation Diaz and Quayle (1960) for the contiguous U.S. 1895 - 1920's 1920's - 1945 increased frequency of cool/wet synoptic types increased frequency of warm/wet synoptic types map correlation method of surface pressure data using Kirchhofar cutoff scores Barry et al. (1981) post 1945 - more.variable conditions but warm and wet 96 1875 - 1905 - higher precipitation in the Pacific N.W. 1905 - 1945 - reduced variability in and totals of precipitation 1945 - 1975 - higher precipitation and greater variability eigenvector analysis of precipitation records along McQuirk (1982) U.S. west coast 1872-91 - wet periods in northern 1902-11 California associated 1932-41 with a northward shift in eigenvector analysis of 1962-71 the Pacific High precipitation records in the Granger C1979) U.S. West Wahl and Lawson (1971) 1892-1901 - drier periods associated 1912-31 with a southward shift in 1942-61 the Pacific High early 20th C - southward displacement of Aleutian Low and Pacific High - cooler/wetter analysis of Pacific sector pressure data for the 20th Angell and Korshover middle 20th C - northward displacement of century (1974,82) Aleutian Low and Pacific High - greater circulation intensity 1947 - 1966 - higher. 700 mb heights in Gulf of Alaska and Aleutians analysis of 700 mb pressure data and SST anomalies Douglas et al. (1982) 1966 - 1980 - lower 700 mb heights - increased zonal index 1951-54 - maximum blocking frequency 1966-71 in N. hemisphere analysis of 500 mb pressure Treidl et al. (1981) anomalies for N. hemisphere Knox and Hay (1985) 1960-65 - minimum blocking frequency 1974-78 in N. hemisphere 1948 - 1965 - slightly higher 500 mb heights for N.E. Pacific upper air pressure data grouped into several synoptic Yarnel (1985) 1965 - 1978 - slightly lower 500 mb types heights for N.E. Pacific 97 negative pressure anomaly in the Aleutians led to an increased southwesterly flow aloft and above normal summer precipitation. Reduced precipitation and higher temperatures after 1920 followed a northward displacement of cyclonic activity as positive pressure anomalies dominated the central Pacific, particularly during the summer. Circulation patterns between 193^ and 1937 best exemplify this period (Biasing and Lofgren, 1980: 717) especially in the sequence of spring and summer surface pressure anomalies; a feature which is most evident as warmer summer temperatures recorded at Bella Coola (figure 3-9) and significant because two of the largest autumn floods on the Bella Coola River occurred in 193** and 1936- After 1945 a return to meridional flow, although with higher amplitudes in the Rossby wave pattern, was associated with positive pressure anomalies in the western United States (Biasing and Lofgren, 1980). Monthly and seasonal variance of temperature and precipitation at Bella Coola and Big Creek were evaluated for these three periods (pre-1920, 1920-19^5. post-1945) along with an index of year to year variability of monthly values. Although some seasonal differences between stations exist, within-period variances of summer and autumn temperatures were consistently higher (average of 40% greater) in the earlier and later periods. In addition, variance of summer and winter temperatures was lowest between 1920 and 19^5• Similarly, autumn and spring precipitation exhibit highest within period variances (average of 65% greater) for the post-1945 period. The differences in between-period variances are partly a function of the frequency of extreme departures. Maximum 24-hour precipitation events of 80 mm or more at Bella Coola have a recurrence probability of less than 20% between 1920 and 1945 compared with 25-30% probability during the other two intervals. 98 Mean sensitivity is used here as an index of year to year 2 fluctuations within each period. A high mean sensitivity indicates the propensity for changes in the sign of departure between successive years within a given interval, and is thought to represent a lack of persistence. Zonal circulation regimes characterize these intervals. Precipitation at both stations for most months of- the year between 1920 and 19^5 exhibits the greatest year-to-year fluctuation (15 to 25% more compared with other intervals). Autumn and winter temperatures also exhibited higher mean sensitivities between 1920 and 19^5; however, summer temperature, particularly at Big Creek, shows a higher mean sensitivity between 19^5 and I983. With the exception of this latter result, mean sensitivity values give some support to the secular variations in circulation regime shown in figures 3.9 and 3-10. Comparisons With Other Regions Although the number of long-term stations ( > 80 years record) in coastal British Columbia is limited, there are some obvious similarities for the entire coastal region. Three-year running mean curves presented by Powell (1965) show a trend towards average and above average annual precipitation prior to 1920 for the outer coast (e.g. Clayoquot and Quatsino). This trend is less pronounced or reversed for the central and southern coasts suggesting that the non-synchronous response in winter precipitation for coastal regions prior to 1920 was driven by synoptic controls similar to those identified for the post-1945 period. Data 2. Mean sensitivity is calculated as follows: n-1 Msx =_U/n - 1) E I [2(xt + 1 - xt)]/(xt + { + xt) | x = value of climate variable in year t; n = number of years 99 presented by both Powell (19&5) anc' Karanka (1986) show a higher mean sensitivity during the intervening period (1920-1940) with no obvious or persistent departures in precipitation. Cooler than normal winter temperatures between 1904 and 1920 at both Bella Coola and Big Creek are also well documented in Karanka's Queen Charlotte Islands temperature index and the outer and inner coast climate stations analyzed by Powell. Although complete records in Bella Coola do not extend beyond 1904, evidence from other climate stations operating in the late 19th century (Queen Charlotte Islands, Clayoquot, Steveston) demonstrates that from approximately 1890 to 1904 winter temperature departures were mostly positive. The tendency for above average winter temperatures in Bella Coola between 1920 and 1945 is also evident for stations along the North Coast. (3.4) Conclusions Temperature and precipitation variability along the central British Columbia coast exhibit patterns which show some correspondence with noted 20th century changes in circulation regime. Climatic anomalies within these periods are short and not well coordinated between temperature and precipitation. The recent record shows that the synoptic signature for anomalously wet winters is both a higher frequency and greater seasonal persistence of specific synoptic types. Certain hydrologic variables such as snowpack and spring flooding respond to persistent climatic departures. Extreme departures in precipitation, which produce key hydrologic events, (e.g. autumn floods) are more randomly distributed across periods but have a higher probability of occurrence under a meridional flow regime. There is a definable regional signal of flood anomalies along the southwestern coast of British Columbia. Sedimentary deposits within drainage basins of this 100 region are more likely to reflect these extreme departures. However, other elements of the biophysical systems are known to have memory. The following chapters will explore the actual patterns of response. 101 CHAPTER IV Geomorphological Evidence of Response of Geophysical Systems To Environmental ChangerTests Within the Historical Record (4.0) Introduction In response to hydrological changes, certain effects are apt to occur within biological and geophysical subsystems. The purpose here is to examine several of these subsystem responses within the short term, using the period of instrumental record in the Bella Coola basin, and to test the utility of biogeophysical data for drawing inferences about local and regional hydrological changes. The responses examined include: glacier fluctuations, upland sediment transfers, and alluvial activity of the Bella Coola River. (4.1) Glacier Fluctuations and Recent Climate Change A valuable indicator of environmental change in mid and high latitude glacierized regions is the fluctuation in water storage reflected as changes in mass and ice-front positions of glaciers. Persistence in the rate of snow accumulation and glacier ablation over several seasons or more, can lead to substantial changes in the position of the ice-front and the deposition of ice-eroded materials which can then be used to infer the rate of ice-front movements. The purpose here is to establish the nature of the relationship between glacier fluctuations and prevailing climate so as to provide a basis for interpretation of glacier deposits indicative of former glacier configurations and, ultimately, associated hydroclimatic cond i t i ons. Several factors complicate the relationship between glacier fluctuation and hydroclimate. One group of factors is related to annual changes in the water (snow/ice) mass balance of a particular glacier 102 system. A second group is associated with glacier responses, such as fluctuations in movement integrated over the whole ice mass or indexed by changes in glacier snout position. In the first group, climatic factors such as winter precipitation and summer temperature generally account for much of the variance in the net ice-mass balance. However, they remain imperfect indices because of the importance of other parameters such as the redistribution of snow by wind and avalanching, ice-calving, and sampling and measurement errors in determining actual mass-balance for a given season (Fogarasi and Mokievsky-Zubok, 1978; Mayo and Trabant, 1984). Therefore, it is unlikely that single or even several point estimates of temperature and precipitation can fully account for fluctuations in net mass balance from year to year. The second group of factors is related to changes in ice-front position. Several studies have demonstrated that snout retreat may not be synchronous with overall losses in total glacier volume and that glacier sensitivity to changing climate is related to glacier size and physiographic boundary conditions (Ahlmann, 1953; Hoinkes, 1968; Smith and Budd, 198l). The actual rate of movement will depend on site-specific factors such as slope, degree of confinement, aspect, total ice thickness and sub-glacial drainage features. Several empirical and physically based models have been proposed to explain the variation in annual ice mass balance for temperate valley and cirque glaciers. Kuhn (1981) has developed a physical model which relates changes in mass to perturbations in parameters such as free atmospheric temperature near the glacier, total annual accumulation (from all sources) and the radiation balance. Alternatively, Tangborn (1980) developed an empirical model which requires estimates of temperature and precipitation measured at nearby stations to estimate fluctuations in the annual mass 103 balance. Although derived from different assumptions, the two climate-based models of glacier fluctuation appear to yield similar results regarding changes in temperature and precipitation corresponding to specific fluctuations in mass balance. In both studies cited above net mass balance appears to be most sensitive to changes in winter precipitation. This is in contrast with the simulation model results of Smith and Budd (198l) who found temperature to be the more important variable. This difference is probably related to climate variations at both the global scale (latitude/longitude and maritime or continental influences) and local site scale (windward/leeward effects). Hence, the relative significance of precipitation and temperature as indices of glacier fluctuations may not be consistent between regions. Mass Balance Fluctuations of Glaciers in Southwestern B.C. Several empirical models are examined in order to assess the relationship between climate fluctuations (temperature and precipitation) and glacier response in southwestern British Columbia. Glacier mass balance and equilibrium line altitude (ELA) data are available for three glaciers in the Squamish/Li1looet watersheds beginning in 19&5. and shorter series commencing in 197& for three glaciers in the headwaters of Bridge River. All these sites are located to the south of the Bella Coola River basin (see f i gure 2.1). Monthly and seasonally averaged climate data (temperature and precipitation) beginning in 1966 were transformed into two sets of regional climate indices representing the wetter south coast (Alta Lake, Garibaldi, Whistler roundhouse stations) and the drier interior (Tatlayoko Lake, Tatla Lake, Kleena Kleene and Big Creek). The latter set of indices is that used to represent the central coast climates in section 3-1 and was selected 104 because of the proximity of these stations to the Bridge River glaciers. Although there are other stations closer to the Bridge River basin, these were selected because of the higher spatial and temporal coherence of temperature measurements amongst interior climate stations and the greater regional significance of any derived relationships. Independent variables consisted of winter precipitation and summer temperature for each of two regions (coastal and interior) and winter mass balance determinations for each glacier. Unlike the Tangborn model, temperature range (an indirect measure of cloud cover and radiative fluxes) was not incorporated here since this variable proved to be of low significance in his model and because such a parameter is not easily constructed using indirect or proxy data. Table 4.1 is a summary of multivariate exploratory models for each glacier. In all three cases the strongest response model (model [1]) incorporates winter mass balance and summer temperature, accounting for 73 to 85% of the explained variance in annual net mass balance. When winter mass balance is excluded (model [2]) the explained variance remains significant for Sentinel and Place Glaciers but declines to between 52 and 59%. A lower degrees of freedom for Bridge Glacier yields a non-significant model when tested under more restrictive confidence levels ('a= 0.025). In each model [2] case, winter precipitation contributes more to the explained variance than summer temperature, especially for Sentinel and Place Glaciers which are located in a wetter, more maritime, region to the southwest. The region from which climate indices are derived appears to be important. For example, coastal precipitation is the more significant variable in explaining annual mass balance changes on Sentinel Glacier whereas interior precipitation is of greater significance for the other two 105 Table 4.1 Empirical response models for glacier net mass balance and climate Glacier Independent Variable (contribution to explained variance) Model Winter Mass Winter Summer d.f. Ra (2) (2) Balance Precipitation Temperature Sentinel [1] [2] 76% 65% (c) 24% (i) 35% (i) 18 73% 25.7 18 59% 11.4 Place [1] [2] 88% 79% (i) 12% (i) 21% (i) 18 73% 21.0 18 56% 10.4 Bridge [1] [2] 81% 54% (i) 19% (i) 46% (i) 85% 15. 52% 4.8* 1. Model [1] includes all significant variables; model [2] excludes winter net mass balance from the list of independent variables 2. Indices generated using coastal stations are marked (c) and those generated from interior stations are marked (i). 3. * not significant at a =• 0.025 106 sites. This was not unexpected since Sentinel Glacier is situated in a belt of heavy precipitation centered on Mount Garibaldi. Temperature indices generated from interior stations are more significant for all three glaciers than are coastal indices. This may be a result of the lower overall variance in summer temperatures for the coast due to persistent cloud cover. While only a limited number of independent variables was considered here, and the degrees of freedom are generally low, a significant portion of net annual mass balance in coastal glaciers can be explained using readily derived temperature and precipitation indices. The three glaciers examined represent a range in characteristics such as size, slope, aspect, accumulation area morphology and local hydroclimatology and it is likely that these factors contribute to the residual variance. Yet, regionally derived climate indices form a set of significant independent variables, albeit of varying importance between sites. Primary controls of net annual balance are winter snowpack measured on the glacier (regional snow course data were not strongly significant variables although they were tested) and summer temperature. Substituting regionally estimated winter precipitation, which will usually be available, reduces the strength of the relationship but the model remains significant. Data plotted in figures 3-1 through 3-5 indicate that the interval 1966 to 1984 was characterized by some of the greatest departures in temperature, precipitation, winter snowpack and runoff. Although this extreme variability extends the calibration range of the Sentinel and Place Glacier models, a limitation of linear models is that they represent mean responses only. The level of explained variance in these glacier-climate regression models is as high as that found in the spring runoff models. The 40 to 50% 107 unexplained variance using model [2] is related to at least three factors: the relative importance of temperature and precipitation at each site in determining net mass balance, the significance of extreme departures in controlling variables, and the representativeness of the regional climate i nd i ces. Mass Balance and Climate Since temperature and precipitation are only index climatological parameters, the linkages between glacier mass balance and synoptic scale circulation patterns were examined. Yarnel (1982) found that a significant proportion of mass balance variability for two Cordilleran glaciers during the interval 1966 to 197^ could be explained by the frequency of certain synoptic regimes. The frequency of synoptic types which produce above or below average winter precipitation and summer temperature for coastal British Columbia is tested here against groups of individual years in which winter and summer mass balances were above or below average. Because of the limited length of the data series and high interannual variance, periods of persistent positive or negative mass balance departures could not be defined with confidence. Results are presented in tables h.2 and 4.3-There are, on average, 11 more winter days dominated by "wet" 500 mb synoptic types and 17 more winter days characterized by moisture producing surface pressure types during strongly positive mass balance years compared with low mass balance years. In both cases these differences are significant. There is no significant difference in the frequency of "warm/dry" synoptic circulation types between years of high and low summer mass balance. This latter result is related to the difficulties in objectively identifying circulation regimes which might produce climate departures (e.g. warm or dry) during the summer season when the overall 108 Table 4.2 Winter Mass Balance on B.C. South Coastal Glaciers and Synoptic Type Frequency Significance test between high winter mass balance years and low winter mass balance years Mass balance years consistent over all glaciers Mean winter mass balance Wet synoptic (mean departure) type frequency Wet synoptic 3 type frequency high winter mass balance :(1976, 74, 1971,68,67) 1.48 ± 0.79 90 ± 4.1 46.6 ± 7.4 low winter mass balance :(1978, 77, 1975,70,72) -0.74 ± 0.58 79 ± 5.4 30.4 ± 6.1 One-tailed t-test (a = 0.10, df = 8) t = 4.42 significant t = 3.63 significant t = 3.78 significant 1. Calculated as the mean departure in winter mass balance for two (1966-1975) and five (1976-1984) coastal glaciers. 2. 500 mb synoptic types as defined by Yarnel (1983). 3. Surface pressure types as defined by Barry et al. (1982). 109 Table 4.3 Summer Mass Balance on B.C. Coastal glaciers and Synoptic Type Frequency. Significance test between high summer mass balance years and low summer mass balance years Mean summer mass balance (mean departure) 500 mb Dry/warm synoptic type frequency high summer mass balance years: (1977,71,70,69,67) 0.18 ± 0.17 23.6 ± 6.5 low summer mass balance years: (1978,76,75,73,72) -0.18 ± 0.07 20.0 ± 5.2 One-tailed t-test t = 4.36 (a = 0.10, df = 8) significant t = 0.97 not significant 110 circulation intensity is low and day to day circulation variability is reduced. It may also reflect the lower significance of summer temperatures in the mass balance models due to the dominance of maritime conditions in the lee of the Coast Mountains. In general, high mass balance years are characterized by an increased frequency of circulations with strong cyclonic curvature favoring the accumulation of snow along the south coast. Summer ablation does not appear to be simply related to the frequency of 500 mb synoptic types which are thought to favor warm/dry conditions. Hence, glaciers form on the coast only where winter precipitation is so high as to dominate the mass balance. Inland, where summer temperatures may dominate, fluctuations in mass balance are less predictable. Glacier Snout Fluctuations An alternative method for assessing the response characteristics of glaciers to climate change is to examine the record of fluctuations in the mean frontal, or in some cases lateral, position of several glaciers representative of a particular area. Paterson (198l) has outlined many of the key variables controlling glacier snout changes which ultimately respond to perturbations in mass balance. The most important geophysical factors are glacier size, bed slope, ice velocity, ice thickness and the geometry of accumulation and ablation areas. Smith and Budd (1981) have attempted to model the response of both small and large valley glaciers (3 2 to 30 km ) to assumed functional changes in accumulation and ablation. This study, and others like it, have shown that glacier response in the form of length changes to some anomalous period of changing temperature or precipitation, can lag by as much as 150 years depending on the magnitude of the perturbation and glacier geometry. Ill Aerial photographs offer the only practical means for estimating the temporal variability in glacier snout changes over large remote areas. Photographs taken in 1947, 1954, 1964 and 1978/79 are available for the Bella Coola basin in addition to field-based measurements made between 1983 and 1985• While the coverage is not annual, the data do allow for at least a decennial resolution of changes. From an inventory of ice lowering and frontal retreat of 50 glaciers, six were selected as representative of the size and geometry of glaciers present in the basin. The selection was made on the basis of measurements available from those glaciers which were investigated in the field. Relative base line distances were established using prominent terrain features near each ice front and were measured using stereopairs correctly scaled and oriented on an Wild A6 stereo plotter. For simple and very regular ice fronts, recessional distances are the average of several measured line segments between two ice front positions. Where the ice margin geometry is more complex, surface area changes were computed and divided by the length of the ice margin to yield a mean rate of advance or retreat. The data were then standardized to reflect the magnitude and direction of ice front movements for each glacier. Figure 4.1 is a summary of ice front fluctuations for six Bella Coola glaciers during each period of measurement. The major trend in the post-1945 period for all six glaciers is the switch from regionally persistent recession prior to 1964, followed by either a reduction in the rate of retreat or actual advances between 1964 and 1979- Examination of the 1964 photography did not show evidence for ice advance at that time (e.g. sediment free lobes with steep, crevassed fronts) suggesting that advances between 1964 and 1979 commenced sometime after 1964. For the larger valley glaciers such as the Fyles, Jacobsen and Talchako (the latter two of which are not shown in figure 4.1) recession 112 + o -Glacier Area Elevation Aspect Ape Glacier ( 4.6 km2, 1460 m, NE) + o -Fyles Glacier (20.0 km , 1320 m, NE) ~\ -H =F west margin north margin o -+ o -I Noeick Glacier ( 8.9 km , 1070 m, NE) n—TT 1 T + Borealis Glacier ( 2.9 km2, 1680 m, NE) " 1 1 1 1 rr ,r,,,i , =n 2 West Saugstad Glacier (1.3 km , 1600 m, NW) ••-•-•-•-•-1 o -Arjuna Glacier ( 1.5 km , 1200 m, N) 1 T 1 1 1 1 1 1 1 1 [ 1945 1950 1955 1960 1965 1970 1975 1980 1985 YEARS (AO) Figure 4.1 Relative rates of advance (+) and recession (-) of selected glaciers in the Bella Coola basin between 1945 and 1984. 113 was continuous but slower after 1964. Between 1954 and 1964 recession rates were average to above average for many glaciers, but for the two larger glaciers actual rates of retreat were increasing up to at least the early 1960s. Some of the smaller glaciers show declining recession rates throughout the 1947 to 1964 period. Between 1979 and 1983-85 the rates of advance declined or recession once again prevailed (figure 4.2). Although boundaries separating the periods of different frontal movement rates are arbitrarily fixed by the dates of air photography, some general similarities between ice front fluctuations and prevailing climate are evident. It appears that for glaciers in this study area larger than 2 approximately 10 km , continuous recession has occurred throughout the post-19^5 period. This is a continuation of 20th century recession from Little Ice Age maxima documented for many parts of the northern hemisphere (Grove, 1979; Porter, 1981) . The decline in recessional rates for these large ice masses and actual advances of some glaciers smaller than 2 approximately 10 km between 1964 and 1979 appear to be associated with the persistent above average winter precipitation along the coast, commencing in 1956 and lasting until at least 1968 (followed by several individual higher snowfall years, e.g. 1974, 1976). Winter precipitation was 7 to 10% above the post-1945 average during this interval. Increased recession rates or declines in the rate of advance between 1979 and 1983-85 are most likely related to the decreased winter precipitation and increased summer temperatures after 1976 for both the central coastal and central interior regions (see figure 3-D-The above evidence indicates that there is an element of sensitivity in both large and small glaciers related to the magnitude of climate changes in the post-1945 period. An interval of between 5 and 10 years is an estimate of glacier snout response time to departures in regional 114 Figure 4.2 West Saugstad Glacier in late July of 1985 • Note the lack of ablation debris on the ice front and the comparatively steep surface along the margin of the snout indicative of forward ice-movement . 115 precipitation and temperature which persist for at least 8 to 12 years. It is difficult to evaluate what magnitude of climate departure is necessary for a given interval to ensure a response in the glacial system but the evidence here seems to indicate not more than an average of 7 to 10% in winter precipitation for a decade. Interpretations of temperature and precipitation records prior to 19*+5 (see chapter 3) do not yield any evidence which would support the contention that the post-1964 ice front advances are due to climatic perturbations prior to 19^5- Thus, the magnitude and persistence of climatic departures in the post-1945 period have resulted in significant responses in glacier systems of the Bella Coola basin and past climate changes of similar magnitude and duration would have affected rates of ice movement. (4.2) Recent Fluctuations in Upland Sediment Transfers Several methods are available to assess the magnitude and frequency of clastic sediment transfer from upland sources to the higher order fluvial network. The most reliable is repeated surveys of aggradation or degradation in natural hillslope gullies or tributary alluvial channels (Dietrich et al., 1982) or in artificially constructed sediment traps or debris basins (Leopold and Emmett, 1976; Gardner, 1979). Unfortunately, several years or decades of monitoring may be required to produce meaningful results because of the lower sensitivity of hillslope elements to climatic perturbations (Brunsden and Thornes, 1979). Alternatively, measurements from air photographs taken at several different times can yield data on the frequency of mass wasting and, when the airphotos are of a large enough scale and some ground based measurements are available, reliable estimates of magnitude also can be made. Finally, indirect 116 evidence in the form of damage to vegetation or detailed historical records may be useful. Two methods were adopted to ascertain if there was any significant response of clastic sediment yields to changes in recent climate and runoff. These include (1) airphoto measurements of changes in channel width on tributary alluvial fans and (2) damage or burial of vegetation on these tributary fans which also provide some chronological data. The morphological analysis here is restricted to the post-1945 interval because aerial photographs were not available until 1946. Aerial photographs taken in 1946, 1948, 1954, 1961, 1974 and 1982 were used to evaluate the temporal variability in channel widths on major tributary alluvial fans. These dates represent the end or beginning of shifts in hydrological conditions in the Bella Coola area (see chapter 3)-Active channel areas on alluvial fans were computed for eight tributaries of the Bella Coola River. These include Thorsen, Snootli, Nooklikonnik, Noosgulch, Cacoohtin and Burnt Bridge Creeks and Salloomt and Nusatsum Rivers. Thorsen, Nooklikonnik and Salloomt basins are characterized by an absence of long-term storage sites. Most of the headwater reaches of Burnt Bridge Creek and the east fork of the Nusatsum River are long-term storage areas, whereas the remaining tributaries have isolated short-term storage sites (see figure 2.8). Other tributary alluvial fans, particularly those in headwater sediment-source areas (e.g. Talchako River) were either too confined or of limited longitudinal extent to reveal consistent trends. Each stereo model was reduced to a common scale using a zoom-transfer stereoscope. Although not corrected for tilt or swing, base lengths for each model were calibrated against at least two distances measured in the field which were identifiable on each set of air photographs. Linear measurements are accurate to within ± 5 m. Channel 117 areas were computed for the region which extends from the apex to the toe of the fan, excluding stable (vegetated) mid-channel islands. These areas are divided by channel length along the thalweg to yield an estimate of mean channel width. Figure 4.3a is a plot of standardized secular changes in mean channel widths between 1948 and 1982 for eight tributary fans. Increases in channel width occurred during the interval 1948 to 1954 and I96I to 1974 while decreases are noted during 1954 to 1961 and 1974 to 1982. Major exceptions to this overall trend are Snootli Creek which had continuously decreasing widths after 1954 and Salloomt River where widths continued to increase after 1974. Increasing widths are associated with bank erosion due to increased peak runoff and persistence of flooding as well as channel braiding which is related to increased sediment accumulations within the channel zone. Decreasing widths (channel recovery) occur as a single thread channel becomes dominant and channel margins are stabilized by vegetation. Channel widening on most tributary fans between 1948 and 1954 appears to be related to the November 1950 flood, therefore, channel recovery following the flood probably began prior to 1954. In four of the eight sites channel widths in 1961 were less than those in 1948 suggesting a continuation of a longer term trend towards channel stability. The response of these largely unconfined alluvial channels to high magnitude autumn floods in 1965, 1968 and 1973 and spring floods in 1964 and 1969 is evident as a trend towards increasing widths until 1974. Reduced or less variable widths occurred in most channels by 1982. Part of the recovery during this latter interval is related to training works (dykes) along the middle and lower reaches of four of the tributaries. A similar trend can be identified using damage to vegetation along channel margins as an indirect indication of flooding and sediment 118 Figure 4.3 Indirect evidence for secular changes in sediment delivery to outlet tributary channels of Bella Coola River. A) Mean width of alluvial channels on several tributary fans. B) Number of damaged or buried trees sampled along the the margins of the same tributary channels shown in A. Dates of spring (O) and autumn (•) floods and major channel training works are also plotted. Damage from the I968 flood was the most widespread. 119 transport (Yanosky, 1982,1983). Approximately 200 trees including alder, cottonwood, spruce and fir were examined for the presence of reaction wood indicative of wood stress due to bending or scarring. A sample of 55 trees from all eight sites showing significant damage is plotted in figure 4.3b. The histogram suggests a higher frequency of damage and burial after 1964. There is some bias in the data due to preservation problems of trees prior to the high magnitude floods of 1965 and 1968 but most of the sampled surfaces were occupied by mature vegetation older than 40 years and generally at elevations low enough to have been inundated by most post-1945 peak flows. In addition to channel morphology and tributary fans, the frequency and magnitude of new landslides throughout the period were estimated from photographs for the lower catchment areas of all eight tributaries. While there was a notable increase in the number of small slides (less than 300 m^) in hillslope materials between 1961 and 1982, most seem to have occurred in recently logged areas below access roads, particularly in Noosgulch and Cacoohtin Creeks. These small slides may have contributed locally to sediment transfers onto the outlet fans but are not considered as major sediment sources. Changes in headwater sediment sources were difficult to evaluate because basin-wide airphoto coverage was not available for each interval and access is restricted to only a few sites. A qualitative assessment of the degree of channel braiding downstream from glacial sources (e.g. west fork of Nusatsum River, lower Talchako River) and below major composite debris slopes indicates less temporal variability in channel conditions because in many instances transport limited conditions prevailed. Additional contributions of sediment from persistent flooding would be less 120 noticeable because of the already unstable (braided, shallow, wide) nature of these reaches. The results presented here provide some insight into the connection between runoff variability and sediment yield from major tributaries of the Bella Coola River over the last 40 years. Evidence presented in section 2.6 demonstrates that a significant amount of material is stored in the floodplain and channel zones of major tributaries. Therefore, clastic sediment yield over the short-term is linked to available sediment in higher-order channel segments of these tributaries and influenced less by the frequency and distribution of headwater sediment sources. Aerial photographs support the contention that increased runoff and frequency of flooding resulted in the movement of this sediment and destabi1ization of alluvial fan channels between the early 1960s and mid-1970s. Periods of channel recovery preceded and followed this interval. The aerial photographs are temporally discrete samples and therefore these data do not clearly distinguish between the proportional impact of single flood events versus a period of increased flood frequency. It is likely that the threshold for channel recovery (cf. Wolman and Gerson, 1978) is more closely tied to the highest magnitude events which introduce substantial volumes of clastic material. In addition, damage and burial of vegetation is the consequence of discrete events only. However, the distribution identified here supports the observed changes in flood frequency over the post-1945 interval. Sampling was restricted to a small number of surfaces along channel margins some of which may have been inundated preferentially (i.e. well-defined overflow channels), therefore the sample and interpretations remain imperfect. However, the methodologies appear to be valid for making inferences regarding long-term flow variability in tributary basins. 121 (4.3) Sediment Sources and Processes of Floodplain Development Inferences about environmental change from geomorphic evidence, such as channel pattern, hydraulic geometry and flood deposits, require an understanding of the processes involved in the transportation and deposition of alluvial material and primary sources of sediment. The purpose of this section is to document the processes of floodplain development and characterize the alluvial sediment sources for Bella Coola River. Processes of Floodplain Development The Bella Coola River exhibits an irregularly sinuous channel, sometimes split about channel islands and in some places braided: seasonal or perennial 'side' channels - subordinate anabranches of the river - are common. This river type has been referred to as a "wandering" gravel-bed system (Neill, 1973; Church, 1983) . The alluvial sediments consist of 0 to 3 meters of sandy channel-fill or overbank silty sands deposited above several meters of channel cobble-gravel. Similar sequences have been described for sinuous gravel channels by Bluck (1971. 1976), Jackson (1978) and Forbes (1983). Lateral accretion is the dominant mode of deposition. This river type is widespread in stable mountain valley and foreland settings and is typical of several rivers draining the west slope of the southern Coast Mountains of British Columbia. Church (1983) has considered recent river changes in some detail. Details of sediments comprising the alluvial portion of the Bella Coola River have been described elsewhere (Desloges and Church, 1987) so only the most salient points are given here. Storage of sediment in the contemporary channel zone varies along the river. In several reaches, large 122 volumes of resident gravel influence channel pattern producing multiple channels and associated gravel bar development. Intervening reaches have greater lateral stability and occur where the valley width is restricted by bedrock or tributary alluvial fans. Hence, sediment accumulation and resulting facies associations within the active channel zone can be divided into those within unstable and those within stable reaches. Medial bars and islands dominate the channel within unstable reaches. Medial bar complexes are composed of several mid-channel bars forming clusters upstream from confluences with major tributary channels (figures 4.4a and b). Islands typically are wooded and gravel is exposed only on the upstream side of the island. Channel shifting occurs by avulsion. The islands represent old floodplain surfaces which have been isolated by avulsion. Approximately 40 percent (22 km) of Bella Coola River is dominated by channel macrofbrms typical of laterally unstable channels. Bank-attached lateral bars or point-bars dominate the more sinuous stable reaches of the river (figures 4.4c and d). A low-flow back channel or chute, active only at high flow, is almost always present containing a cobble-gravel bed (probably crudely stratified) and structureless sand levees on both flanks. These subordinate channels are important sites for the preservation of sediments indicative of varying flow conditions through time. Abandonment and infilling of back channels lead to floodplain development. The overall importance of the sandy facies in terms of contemporary floodplain sediment volumes can be determined by assuming that the depth of scour (thalweg depth) in the active channel represents the bottom of the gravel basement or 'bar platform'. This assumption can be made because of the apparent vertical stability of the river (Church, 1983). Thicknesses of the sandy floodplain deposits range from a few Figure k.k Morphology and channel zone sediments of Bella Coola River. (A) An unstable reach of the river with medial bar deposits. Massive gravel (Gm) or planar stratified sediments (Gp) dominate the channel zone gravels. (B) Bar topography in unstable reaches is subdued and finer grained sediment wedges cover the gravel surface. Note the two year old growth of alder in finer sediments. (C) A sinuous reach of the river and accumulation of organic debris on a point bar surface. (D) Point bar deposits some times exhibit gravel deltas (a) which migrate into finer back channel deposits (b) . 124 125 centimeters to over 3 m of coarse gravelly sand to fine sandy silt and represent 31 to 36% of the total 'active' sediment pile (see table k.h). Nowhere does clay constitute more than 15% of any sample: most frequently it is much less. The sedimentary sequences in the alluvium reflect the integrated processes of floodplain development over time. In unstable reaches of the river, rapid channel avulsions, the occurrence of multiple channels with varying capacities for sediment and water discharge, and more variable flow conditions result in a diversified set of facies units within the channel and on the the vegetated floodplain surface. With increasing stability, the spatial variation in flow conditions declines and the probability of transitions between facies units becomes lower than in less stable reaches of the river. Progressive lateral accretion and the vertical stability of the river appear to favor the simpler facies sequence, although infilled back channels may still exhibit the wider variety of sedimentary structures. The greater incidence of overbank flooding in stable reaches produces vertical accretion of the floodplain here and development of the finer-grained facies (Desloges and Church, 1987)' However, lateral accretion deposits dominate the overall sequence in just the manner described by Wolman and Leopold (1957)• Channel and point bar development is accompanied by the cross-valley migration of the channel, resulting in the formation of a bar platform and, in addition to the main channel, several active slough channels or a single, poorly connected back channel. Sediment volumes (see table k.k) indicate that lateral accretion of gravels accounts for approximately 65% of sedimentation during floodplain construction. As the vegetation develops, sands are trapped on bar tops and overbank, so the floodplain rises by vertical accretion to its final level. Table 4.4. Volume estimates of alluvial-fill for the Bella Coola River. 1 (2) (3) Reach Type Valley Length Floodplain Area Overbank (sand) Basement Gravel Proportion of fill 2 7 3 7 3 (km) (km ) Volumes X 10 m Volumes X 10 m as overbank sands Unstable 22.1 (39%) 32.6 (42*) 4.2 (60%) 7.6 (59%) 36% Transitional 23.1 (40%) 30.6 (39%) 2.4 (34%) 4.3 (33%) 36% Stable 12.2 (21%) 14.4 (19%) 0.4 ( 6%) 0.9 ( 8%) 31% Total 57.4 (100%) 77.6 (100%) 7.0 (100%) 12.8 (100%) 35% 1. Reach type classification is based on the frequency of channels and characteristics of floodplain and channel sediments. Unstable reaches are high-gradient sediment storage zones with numerous sloughs, back channels and medial bars. Stable reaches are characterized by more sinuous single-thread channels. Transitional reaches exhibit a combination of morphological features. 2. Volume of fine-grained (sand and silt) vertical and lateral accretion deposits. 3. Volume of coarse-grained (gravels and cobbles) lateral accretion sediments above maximum thalweg depth. 127 Sands are very mobile in this high-energy, cobble-gravel river, moving predominantly in suspension. Once re-entrained, the sand may be flushed from the system or redeposited a substantial distance downstream. Much of the sand probably moves through the system in only one or two stages. The actual modes of channel migration in unstable reaches are progressive erosion on the cutbank opposite a growing channel bar, and rapid channel avulsion, or reduction of flow in the channel around one side of a medial bar, leaving behind a series of active, "subordinate" sloughs. Sloughs along the inner margins of point bars may remain active, sediment transporting channels for periods in excess of 50 to 80 years following subordination. Complete infilling is likely to occur only if the slough entrance becomes blocked by large organic debris. Sedimentation would be episodic and indicative of flood events only. In sinuous reaches of lower gradient, sloughs become poorly connected at most flow stages and a back channel develops with deposition of fine sands, silts and organic material. Infilling may never be complete as some of the older sloughs and back channels are preferentially occupied during high magnitude flows. As a result, channel avulsions frequently lead to the occupation of older channels rather than the development of new ones. This leads to a reduction in the potential number of sites with sequences representative of long-term sedimentation. Although avulsion produces abrupt realignment of the channel, progressive lateral accretion and erosion may be substantially more significant in the history of floodplain development and sediment exchange. Alluvial Sediment Sources The principal sources for alluvial Bella Coola valley are as follows: sediments deposited within the 128 (a) headwater tributaries introducing sediment to the active channel; (b) downstream tributaries entering along the river reach contributing sediment directly to the floodplain surface or to the active channel; (c) bank erosion and re-deposition of previously deposited alluvial material; and (d) fluvial erosion of non-alluvial deposits resident within the zone of channel activity. Sources (a) and (b) are directly connected with the uplands. Sources (c) and (d) represent remobi1ization of sediment previously introduced into the valley. Since less than 13% of the channel of Bella Coola River is in contact with non-alluvial deposits, the majority of which are either well-vegetated, stable colluvial slopes, bedrock or deposits of glaciomarine silt and clay, they are not important sources for the sand and gravel that dominate the alluvial valley-fill. Klages and Hsieh (1975). Mazzullo (1986) and Statteger et al. (1987) have demonstrated that the mineralogy of alluvial and littoral sand deposits may reflect provenance provided that source areas yield sediments dominated by distinctive lithologies, characteristic proportions of certain minerals, or contain unique minerals not found in other source areas; and that these properties are not appreciably attenuated by mixing during transport to a particular depositional site. To determine which source areas are most important for delivery of material to the Bella Coola River, the composition of sediments from several major tributaries was compared with channel zone and floodplain sediments sampled within the Bella Coola valley. Samples extracted from the active channel zone and floodplain were classified on the basis of discriminant functions derived for samples taken at the outlet of major 129 tributaries. The sampling design and data analysis are discussed in Appendix B and the results are plotted in figure 4.5-These results indicate that fine sediment (sands) along the active channel and in the floodplain derives mainly from the glacierized volcanic terrain of upper Talchako River. Downstream, there is in the channel a detectable admixture of sands from lateral tributary basins with plutonic sources. However, sands dominated by lateral sources are restricted to the vicinity of the tributary junction. Mixed sediments originate from tributaries which produce material that truly reflects a mixed bedrock source {e.g. Nusatsum River) and from mixing of plutonic and volcanic sediments on the floodplain. Larger clasts appear to reflect more uniformly the overall distribution of source terrain in the basin. About this several things need be said: certain plutonic lithologies - though by no means all of those in the basin - resist attrition and may become over represented in lag deposits; plutonic clasts may be derived disproportionately from eroded Quaternary deposits; finally, size-related effects may influence the results (size was not controlled in the surface clast sample). Sediments are delivered from volcanic headwater sources to the downstream channel and floodplain during basin-wide high flows, and during summer glacial runoff. Sediments originating from plutonic terrain are delivered to the margins of the floodplain during basin-wide floods, and also during storms in which the contribution of runoff from the lower basin is much greater than from the headwater areas. This is not an unusual occurrence because of the eastward decline of freezing levels during late autumn storms, and because of the marked gradient in precipitation across the basin. This pattern of sediment delivery influences channel morphology and the character of alluvial deposits of Bella Coola River. 130 Figure 4.5 Alluvial sediment sources for the Bella Coola River floodplain. Grouping corresponds to proportions of minerals in the sediments which reflect bedrock source conditions. Mixed sediments can be from truly mixed sources (1) (i.e. basins underlain by volcanic and plutonic sources) or may represent mixing once introduced into the higher order alluvial system (2). See the text and Appendix B for discussion. 131 (4.4) Equilibrium Conditions and Inferences from Channel Morphology Morphological adjustments of a stream channel depend on the magnitude and timing of water and sediment discharge to the river and the relaxation time of the system (Mackin, 1948; Brunsden and Thornes, 1979)- Alluvial channel morphology therefore can be used to infer fluctuations in the discharge of sediment and water provided that the equilibrium status of the river can be evaluated. Howard (1982) has defined equilibrium as a "single-valued, temporally invariant functional relationship between the values of an output variable and the values of the input variable (s) in a geomorphic system." Disequilibrium occurs when the output variable no longer resembles that predicted by the functional relationship previously observed. If the interval of time required to achieve a new equilibrium is longer than the frequency of significant perturbations then a complex sequence of morphological adjustment may occur and the functional relationship is no longer easily definable. Such a system is intransitive and the possibility of continuous disequilibrium makes inferences about the independent variables from morphological evidence unreliable. However, as Schumm and Lichty (19&3) point out, equilibrium is scale dependent. Hickin (1983) and Howard (1982) give several examples of this scale dependency but three divisions seem appropriate: -2 0 1) for very short time scales (10 to 10 years) and individual reaches or cross-sections of the river, channel gradient becomes a fixed value and sediment discharge is a dependent variable; 0 2 2) over longer time scales (10 to 10 years) and several reaches of the river, slope adjustments become dependent on the sediment and discharge regime; and 132 3 5 3) for geologic time scales (10 to 10 years) and entire fluvial networks, both slope and sediment discharge become dependent on various geological processes such as isostatic or tectonic adjustments. The first and third divisions are mostly for engineering and geological time scales, respectively, while the second is best suited for the geomorphic (and engineering) time scales of interest here. The generalized functional relation between variables which influence river morphology have been identified by Lane (1955) and re-interpreted by Schumm (1977) as: Qs « b,A,s/d,P and °= X,b,d/s Channel width (b) and meander wavelength (X) are directly proportional to discharges of water (Q^) and sediment (Qg)• Channel gradient (s) is inversely related to water discharge and directly proportional to sediment discharge. The opposite association is true for channel depth (d). Sinuosity (P) decreases as sediment discharge increases. Several variables have been selected to evaluate the equilibrium status of Bella Coola River and the impact of changing water and sediment discharge. These include indicators of vertical and lateral channel stability, channel width, width/depth ratio, meander wavelength, sinuosity and braid index. Vertical and Lateral Stability of Bella Coola River River bed and water surface profiles for a moderate flow in the Bella Coola River between the Water Survey of Canada gauge at kilometer 56 and river mouth are plotted in figure k.G. Below the Nusatsum River confluence a concave-up profile dominates with average bed-gradients of O.OO38 above Nooklikonnik Creek, declining to an average of 0.0018 near the mouth of the 133 river. Locally steep gradients can be found in proximity to several tributary confluences yielding a stepped profile. From Nusatsum River up to Noosgulch Creek gradients are comparatively steep (0.0032) and then become much lower until the Burnt Bridge Creek confluence (0.0016) . The overall upstream profile is convex-up. Changing bed-gradients in the vicinity of tributary alluvial fans suggests that constriction of the river by these features is an important factor in determining the river profile. Sediments accumulate in storage or "sedimentation zones" (cf. Church, 1983) above the alluvial fan constrictions and gradients adjust to accommodate the increased volume and calibre of sediment stored there (see Appendix B for evidence of grain-size variations). This promotes increased lateral instability in many of these reaches. Vertical stability of the river has been inferred for the interval 1948-1983 using hydrometric surveys from the two upstream gauging sites. Vertical changes in bed-elevation at each site have not exceeded 1 m between any particular flood of magnitude Q or larger, and absolute 5 elevation changes over the entire interval are less than 0.5 m. In other reaches, vegetation and stable, cobble-paved sediments also indicate long-term vertical stability. Thus, the overall vertical stability and lack of evidence for significant aggradation or degradation suggest an equilibrium between water and sediment supply on the river scale for at least the last 35 years and perhaps much longer. However, localized adjustments to increased sediment storage are made by lateral channel shifting and an increase in the number of subordinate anabranches. Figure k.G Water surface and bed longitudinal profiles of Bella Coola River between Water Survey of Canada gauge above Burnt Bridge Creek and river mouth. Arrows indicate the position of tributary creeks. Water profile survey for the lower river was done on July 31, 1981 and for the upper river on July i*. 1986 Source: Water Management Branch, Ministry of the Environment, British Columbia (unpublished data). 135 Analysis of maps, aerial photographs and field measurements indicates that lateral channel instability coincides with zones of increased sedimentation. Unstable reaches consist of several channels separated by mid-channel bars and the surrounding floodplain is dissected by back-channels and sloughs. Stable reaches exhibit a single, well-defined channel course. Channel morphology data were compiled for a number of reaches and grouped into laterally stable or unstable categories. Group assignments were made on the basis of observed channel pattern, floodplain sediment facies and vegetation characteristics. Thalweg slope, river gradient, channel width, width/depth ratio and hydraulic radius were measured at a number of sections in each reach and then averaged across each group. Results presented in table 4.5 demonstrate that the cross-sectional data are significantly different between stable and unstable reaches, which would confirm that the processes of channel development (e.g. bank erosion, bar development, channel migration) vary significantly in different reaches of the river. Lateral channel stability for several reaches was evaluated by considering the correlation of reach-averaged slope and channel forming discharges. The results of this analysis (not shown here) and the above discussion revealed several things: 1) that the river must be viewed as a sequence of hydraulically distinct reaches which have very different morphological and sedimentological characteristics; 2) no one reach is likely to be indicative of river or basin scale fluctuations in sediment or water discharge; and 3) significant changes in runoff and sediment flux are likely to influence first those reaches near the threshold of change (i.e. transition from stable to unstable pattern) and leave others less affected. 136 Table 4.5. Morphological data for selected reaches of Bella Coola River Reach1 N Width Width/Depth Hydraulic Thalweg Valley (m) Ratio Radius(m) Gradient Gradient Stable 7 104 ±13 41 ± 14 2.12 ± 0.39 0.0019 ± 0.0008 0.0026 ± 0.0010 Unstable 6 171 ± 39 109 ± 59 1.38 ± 0.41 0.0033 ± 0.0010 0.0034 ± 0.0012 one-tailed t-test t - 4.60 t - 3.17 t = 3.89 t = 3.22 t =1.31 (d.f. = 11, a = 0.10) significant significant significant significant not significant 1. Reaches vary in length and were sampled between the Antnarko/Talchako confluence and river mouth 137 Inferences from Channel Morphology The above results suggest that variations in morphological features of a river are potentially useful indicators of changing sediment and water discharge. Width, slope and channel pattern are functionally related to the discharge regime and adjustments therein may demonstrate the magnitude of change. Three indices were selected to assess the degree of association between secular changes in Bella Coola River morphology and changes in runoff and/or sediment supply: regime equations applied to width fluctuations in stable/unstable reaches, channel sinuosity and channel braid index. Sinuosity is the ratio of channel length to valley length while braid index is the ratio of total channel length, including all back channels and sloughs, to main stem length. The propensity for a river to occupy single, well-defined meandering channels when runoff magnitude, runoff variability and sediment discharge are reduced is reflected in increasing sinuosities and decreasing braid index. Secular variations in sinuosity and braid index, measured from several sets of air photographs, are plotted in figure 4.7a. Sinuosities are highly variable (range 1.05 to 1.27) for the lower Bella Coola River. No strong trend is evident but there is a tendency for decreasing channel curvature in both upstream and downstream river segments over the last 10 to 15 years in contrast to steady or slightly increasing sinuosities prior to 1968. The braid index is similar for both segments showing decreases up until 1968 followed by increases to 1982, particularly for the upstream reach. Although long-term trends are evident, significant channel changes occurred following major floods in 1950, 1968, 1973 and 1980 suggesting the importance of flood frequency and magnitude in determining channel pattern. 138 1.30 1.20 co O 1.10 3 C CO CD k. 3 k. CO a © a CO J3 "O w CO "O e CO CO 1.00 • upstream + downstream \ CO j» a. 3" Q. CD X 1-6 ^ i i 2.0 -1.8 1.4 Observed (3-yr weighted average) ——° Estimated 1950 1955 1960 1965 1970 1975 1980 1985 Years (AD) Figure U.l A) Secular changes in sinuosity and braid index for Bella Coola River above and below the Nusatsum River confluence. Decline in the braid index (dashed line) suggests increasing lateral stability. Sinuosity (solid line) shows less overall trend. B) Actual and estimated secular variations in annual maximum discharge at the Bella Coola gauge above Burnt Bridge Creek. Solid line is a three year moving average and dashed line is estimated flow using regime equations (Bray, 1982). 139 Channel modifications within individual reaches can be important; for example a loop extension cutoff in the autumn flood of 1973 accounts for much of the decline in sinuosity of the lower Bella Coola after 1973• However, it is usually adjustments within several reaches that control the variance. River training works installed since 1968, mostly in the form of dykes, have probably influenced the less dramatic changes than might have been expected following the 1973 and 1980 floods. However, their limited distribution does not appear to have affected the overall trend of increasing channel stability to I968, followed by reduced lateral stability after several floods in the late 1960s and early 1970s as sediment from upstream and tributary fans was deposited in the main stem of the river. Measurements of channel width in the three stable/unstable reaches were taken from several sets of aerial photographs available for selected years between 19*+8 and 1982. The highest probability of adjustments to changing discharge is in these threshold reaches. Bray (1982) has developed a set of regime equations from a sample of a Alberta gravel-bed river which relate hydraulic features such as bankfull width to some dominant discharge (in this case Q^) . In order to test the strength of inferences drawn from changing channel widths, annual peak, mean-daily discharges measured at the upstream gauge were compared with estimated departures in discharge using the Bray regime equations. Since the original equations were computed with reformulated with width (w) as an independent variable, an exercise which more appropriately estimates the error associated with the dependent variable (cf. Williams, 1984). The resulting equation takes the form: water discharge as the independent variable, the relations were log = -1.51 + 1.7971og w r2 = 0.97 s.e. = 0.19 log-log 140 The results plotted in figure 4.7b indicate some similarity between measured (3 year weighted mean departures) and estimated flows using channel widths in these reaches. Widths appear to have been increasing during the interval of higher-magnitude flows between 19&4 and 1976. Decreasing maximum mean-daily discharges (with the exception of 1980) have resulted in decreased channel widths since the early or middle 1970s as vegetation begins to colonize lateral bar surfaces. It should be noted that the relative proportion of width changes due to flood events and the less significant inter-event bank erosion cannot be distinguished with these data. However, air photographs taken only four months after a late January flood in 1968 indicate that flood-related channel changes are the most significant. The legal survey of 1889-93 fixed channel boundaries fairly precisely although features within the channel were estimated only. Estimates of channel width in the vicinity of the gauges taken from 1:31.680 scale maps are between 20 and 30% greater than recent measurements. The early 20th century trend then, is towards declining channel widths and discharge. (4.5) Recent Floodplain Sedimentation The distribution and composition of sedimentary facies in the Bella Coola River floodplain indicate that rates of sedimentation during and between flood events are highly variable (see Desloges and Church, 1987)• This variance is related to the frequency of channels in a particular reach of the river and how well connected a depositional site is to primary sources of alluvial sediment. Inferences regarding temporal variability in rates of sedimentation from alluvial sedimentary sequences require that these conditions be known with some confidence. 141 Rarely is it possible to calibrate sedimentation processes against recent alluvial activity because of low contemporary rates of sedimentation and the high variability in boundary conditions. Stratigraphical interpretations from a number of sites usually provide the only means of assessing temporal variability (cf. Brakenridge, 1985). Common difficulties in achieving accurate inferences about the changing alluvial environment have been enumerated by Butzer (1980) who demonstrates that chronological control and between-site correlation are most problematic. An attempt was made to determine alluvial sedimentation rates in the Bella Coola River in order to assess the linkage between recent runoff variability and sedimentation. Sediments in several active or recently active sloughs, back channels and ponds were sampled for this purpose. From an initial set of 16 sampling sites only six yielded sediments with sufficient textural, organic and structural variability to warrant consideration. Pollen, organic matter content and textural characteristics were used as primary indicators of sedimentation variability. Sampling locations, field methods and laboratory techniques are given in Appendix C. Interpretations from one particularly interesting site are given here along with a summary of results from all sites investigated. Four cores were recovered from standing water located in a former back channel area of the Bella Coola River approximately 500 m downstream from the contemporary stream gauging site above Burnt Bridge Creek (figure 4.8). Air photographs taken in 1946 indicate that the back channel was well connected to the Bella Coola River prior to this time. Over the last several decades the backwater has become an isolated series of ponds in which water and sediment are derived from two sources. During non-flood conditions water enters via a stream situated in a back eddy area on the concave bend of the meander loop. Much of the channel is presently 142 Figure 4-8 Topographic setting for the McCall Flats backwater ponds just below the Water Survey of Canada Gauge near Burnt Bridge Creek. Note the location of the overflow depression between the river and ponds. Vegetation and the legal survey of l889~93 both indicate that the general curvature of the river in this reach has been stable for at least the last 125 years. Backwater areas and the size of the concave bench opposite the point bar have changed significantly. 143 overgrown with aquatic plants and low-flow sediment delivery is minimal (suspended sediment concentrations in the ponds were as low as 0.2 mg/1). During flood, water and sediment are delivered through the same stream entrance as well as an overflow channel which becomes active only during high water stages when sediments are deposited directly into the ponds (figure 4.8). Concentrations of suspended particulate matter sediment measured during a moderate flood in the fall of 1984 were as high as 5 mg/1, although the overflow channel was not active. Core stratigraphy shown in figure 4.9a indicates the presence of several coarse layers (50 to 75% sand) which apparently relate to flood events over the last several decades. To test this hypothesis fine-grained layers in core MF-2 were sampled and analyzed for pollen type and concentration (see Appendix C for analytical procedures). Airphotos indicated that much of the surrounding forest vegetation {Pseudotsuga menziesii, Thuja plicata) was removed by commercial logging operations during the summers of 1969 to 1970 leading to the potential for establishing a chronostratigraphic marker using pollen. Pollen percentages, sediment texture, organic matter content and stratigraphy in core MF-2 are shown in figure 4.9b. Three distinct zones can be identified on the basis of changes in the percentage of pollen (4.9b). In zone I sediments, Lodgepole pine and the firs are dominant accompanied by low percentages of alder and grasses. With the exception of Douglas fir these species were not well represented in the surrounding vegetation community prior to logging suggesting that the pollen was derived from riverine sources. Lodgepole pine is a co-dominant species in middle elevation areas of the eastern catchment and its pollen is known for long distance airborne transport (Mannion, 1980) but based on analyses of pollen types and concentrations in Bella Coola River water 144 MF-4 MF-1 Non Arboreal Arboreal Pollen (AP) 1 Pollen (NAP) 6 ' 20'40 60% Figure 4.9 (A) Stratigraphy of McCall Flats cores. See figure 4.8 for sampling locations. (B) Pollen concentrations and inferred chronology for flood deposits in core MF-2. See text for discussion. 145 (normal summer discharge) and pond water (during the October 1984 high water event) significant waterborne contributions are also likely (see Append i x C) . Above 70 cm in zone il, a very sharp transition to alder-dominated sediments is evident along with relative decreases in Douglas fir and cedar in the arboreal component. These quite dramatic shifts are related to the influx of alder following logging in 1969 and 1970. Alder pollen is less prone to long distance transport thus the percentages shown in figure 4.10b probably represent the importance of alder in the local vegetation community. Alder percentages remain very high until the 15 cm level above which an increase in some of the conifers and grasses occurs (zone III). Absolute concentrations of pollen are highest in zone II which further reflects the importance of alder, a species which is a prolific producer of pollen grains. Based on water level data and cross-sectional morphology of the Bella Coola River adjacent to the overflow channel, flows in excess of (500 3 -1 m s ) would have sufficient depth to occupy the overflow channel. Although the configuration and height of channel divide will have been modified after each flood, the age and structure of vegetation adjacent to the overflow channel suggest that the modifications have not been significant. The airphoto evidence indicates that the back water area was not decoupled from primary river flows until sometime before 1946, so it is assumed that the basal gravels represent deposition from the 1934/36 sequence of floods. Between these basal gravels and the stratigraphic marker, which is assumed to represent 1970. two coarse "flood" units can be identified (figure 4.9a and b). The lowest unit may represent the 1950 or 1965 flood, although the latter is more likely since all aerial photographs taken prior to 1961 show that the ponds were very shallow and occupied by gravels in some places. 146 The second coarse unit is assigned to the 1968 flood, the largest measured flow over the last 40 years. Since the change from pine to alder occurs at the top of the fine-grained material which caps this flood unit, the whole sequence from 65~90 cm is probably related to the 1968 event (basal flood sands and post-flood siltation). Above the chronostratigraphic marker three prominent units can be identified and correspond with known high flows in the autumns of 1973 and 1980 and the spring of 1976. A fourth thinner unit, composed of finer material than the other three, might be associated with high water in 1982 but it is more likely material deposited during the waning stages of the I98O flood. The flood units assigned in core MF-2 correlate with layers in core MF~3 but are more difficult to cross-correlate with sediments taken in an adjacent pond which is not directly connected with the overflow channel (figure 4.9a). Evidence from the lower Bella Coola River is not as conclusive as the interpretations made from this site but two locations in particular (East Hagensborg Slough and Snootli Slough; see Appendix C) give clear evidence of above average sedimentation in the 1960s and early 1970s followed by lower depositional rates. Dykes constructed after the 1965/68 series of floods in parts of the lower Bella Coola River has reduced the sedimentation effects of floods during the 1970s and in 1980 hence, most of the infilling sediment is related to flooding during the 1960s (e.g. Rambeau Slough). Sediment accumulations in the Mclellan Rd. slough exceeds 120 cm and is probably related to a combination of lateral and longitudinal infilling. At least the top 30 cm of sediment (25%) is the result of post 1965 flooding. A similar trend is inferred for the Walker Island slough (see Appendix C). The evidence presented here suggests that, while most backwater or sloughs selected for this investigation probably became effective sediment 147 traps sometime after the two high magnitude floods in 1934 and 1936. significant accumulations of sediment did not occur until 1968 (average between 1968 and 1976 is 6.25 cm/yr). Sedimentation is highly episodic and mostly related to flood events greater than Qc. 5 (4.6) Conclusions Glaciological evidence provides a low-resolution record of climatic variability over the last several decades. The turning of glaciers in the Bella Coola basin was a response to persistently positive precipitation anomalies in the middle 1960s and early 1970s. Glacier size, basin aspect and other morphological parameters complicate actual responses and are independent of hydrological variations. Individual years characterized by extreme departures in glacier mass balance correspond to significant changes in the seasonal frequency of "wet" synoptic regimes. Variability of channel morphology on alluvial fans reveals a secular pattern similar to that for the basin glaciers. Above average precipitation and increased runoff between 1964 and 1974 lead to increasing channel width and promoted channel instability. However, closer examination shows that individual flood events are responsible for most of the channel changes. Floodplain sedimentation is clearly episodic and related to extreme events only. The 1957-1976 period was clearly exceptional in terms of sedimentation in back channels and sloughs. 148 CHAPTER V Sedimentological Evidence of Environmental Change: Tests Using Lake Sedimentation Rates (5.0) Introduction Long-term lake sedimentation rates can provide an indirect measure of environmental change. While the functional relationship between source area sediment yield and rates of deposition in lake settings may be complex, there is an increasing understanding of linkages and controlling variables in both glacio-lacustrine (Church and Gilbert, 1975) and non-glacial lake systems (Oldfield, 1981; O'Sullivan, 1983). The presence of proglacial, ice-marginal and ice-dammed lakes in the Bella Coola basin provided the opportunity to extend the record of recent environmental change. There are three important categories of variables which control the erosion, transportation and deposition of material in a glacio-lacustrine system: erodibility of terrestrial materials (i.e. sediment sources), hydroglaciological factors and limnological factors. Sediment flux rates depend on the character of sediment sources and the timing, frequency and magnitude of sediment transport processes (Karlen, 1981; Grunell and Clarke, 1987)- The interannual and seasonal variability of temperature and precipitation in mid-to-high latitude alpine regions provide for discontinuous fluxes of water and sediment of varying magnitudes from the surrounding terrain. Ablation season temperatures, winter and summer precipitation, spring snowpack volumes and near surface winds are important hydroclimatic controls of the rate of sediment yield from the glaciated terrain to the lake (Granar, 1956; Rainwater and Guy, 196l; Shaw et al., 1982; Simola and Tolonen, I98I; Ringberg, 1984; Leonard, 1985). Certain morphological, chemical, biological and dynamic components of the lake will control the distribution and deposition of material. Flat, extensive lake bottoms, free from internal disturbances such as slumps or 149 other sub-aqueous slope failures, are most likely to preserve annual layers or varved sediments (Hakanson, 1983)' Thermal stratification, wind-generated lake currents and seasonal biological activity interact to provide a range of depositional conditions each yielding different structural characteristics (Sturm, 1979; O'Sullivan, 1983) • Finally, disturbance by physical or biological processes and post-depositional diagenesis of both clastic and organic materials can alter pre-existing structures. Despite all these physical and chemical interactions annually laminated lake sediments indicative of prevailing hydrological and glaciological conditions are found frequently in glacio-lacustrine environments and provide geochronological evidence for environmental change on time scales ranging from days to millennia (Renberg, 1982; Morner, 1984). Reconnaissance of several lakes in the Bella Coola basin revealed the potential for the formation and preservation of annually-laminated sediments. Due to limitations of time and accessibility only Ape Lake, an ice-dammed, proglacial lake in the southern catchment, was selected for examination. A complete summary of field and laboratory techniques is given in Appendix E. The following is a discussion of the analytical results of this investigation. (5-1) Patterns of Sedimentation in Ape Lake Physical Setting Ape Lake, situated at an elevation of 1395 m.a.s.l., is located in the southwestern portion of the catchment straddling the divide between the Noeick and Talchako-Bella Coola watersheds (figure 5-1)• The lake occupies 2 an area of 2.47 km and has a contributing watershed of approximately 38 2 2 km of which 20 km (53%) is covered by valley glaciers and upland ice. 150 Figure 5.1 Location and bathymetry of Ape Lake. Fyles Glacier forms the ice dam to the west. Dates for glacier margins are taken from air photographs which were available beginning in 1947. Terminal moraine dates are derived from tree-ring evidence (see chapter 7). Bathymetry data were collected during the summer of ig84 with a 50 kHz echo sounder. 151 Fyles Glacier, a northeast flowing valley glacier which originates in the basin circumscribed by West Jacobsen Mtn., Mongol Mtn. and Mt. Fyles, forms an ice dam along the western margin of Ape Lake. Water levels have been controlled by an outlet cut into bedrock at the east end of the lake where waters discharge into Ape Creek, a tributary of Talchako River. Failure of the Fyles Glacier ice dam in October 1984 released approximately one half of the lake volume into the Noeick watershed (Jones et al., 1985) allowing for examination of bottom sediments in limited portions of the basin. Ape Lake can be divided into four sub-basins (figure 5-l» table 5«1)-Maximum water depths of 70 m occur in the larger, bowl-shaped, east basin which is separated from the rest of the lake by a sub-aqueous sill forming a continuous arch across the mid-basin (figure 5-2a and b) . The small embayment at the foot of Ape Glacier (Ape Glacier Bay) has maximum depths of just under 25 m. A shallow depression (£ 23 m) in the middle of the lake is bounded to the east by the sub-aqueous sill and to the west by lake delta sediments which originate from tributaries on both the north and south lake margins. Lake floor gradients then increase towards the Fyles Glacier ice dam where depths at the ice-front vary between 16 and 54 m. Ape Lake is monomictic to polymictic with weak thermal stratification developing during the summer. Although no measurements were made, overturning of the water column in the east basin is not likely to occur because the vertical thermal gradient is steep and the epilimnion is comparatively shallow (Gilbert and Desloges, 1987)- Surface sediments in the lake are dominantly allochthonous clastic material with organic contents, estimated by weight loss on.ignition, generally below 1 percent. Concentrations of suspended particulate matter in lake waters vary between 10 and 30 mg/1 with higher concentrations found closer to the Fyles Glacier ice dam and at depth. These concentrations are half to one sixth those Table 5.1. Ape Lake Morphology. 1 2 Sub-basin Area Max Depth Volume ,2 ,„6 3 km m x 10 m full drained West E Shoal 0.17 ca. 30 2.077 0.033 W Basin 0.61 57.6 15.752 0.372 Subtotal 0.78 — 17.829 0.405 East 1.50 70.6 64.947 38.768 Ape Glacier Bay 0.19 ca. 30 2.808 0.632 Totals 2.47 — 85.584 39.805 Percent of Totals 100 100 46.5 Surface Area 2.47 1.27 2 km 1. For locations of sub-basins see figure 5.1. 2. Volumes determined using bathymetric data collected August, 1984. Drained volume is amount released during the jokulhlaups of October 20, 1984, and August 2, 1986. 153 Figure 5-2 A) Ape Lake basin two days after the October 20, 1984 outflow. Emergent climax moraines of Ape Glacier (left) and Fyles Glacier (center) divide the half-emptied lake into four sub-basins. Water levels in the larger east basin (foreground) have been lowered by 20 m. Photo courtesy R. Lenci. B) A breached climax end moraine (foreground) separates the east basin (-20 m from full lake level) from the west basin. Sand and gravel deltas are prograding into a shallow mid-basin pond (-22m) immediately behind the terminal moraine. Standing water adjacent to Fyles Glacier is at -38 m. 154 measured in inflowing waters and are low in comparison to those measured in other glacial lakes (cf. Campbell, 1973; Gustavson, 1975; Pickrill and Irwin, 1983)- The majority of the water and sediment are derived from four ice-marginal and proglacial streams in the west basin (figure 5-3)• All four of these streams enter behind the mid-basin sill. Sediment Facies Four lacustrine facies can be recognized based on surface (Ekman) sediment samples taken from all sub-basins of Ape Lake. The distribution of these facies is shown in figure 5-3 and a complete description is given in Gilbert and Desloges (1987)• Facies I is light, medium-to-fine sand regularly interlayered with darker silt and is found in proximity to Ape Glacier (Ape Glacier Bay) and Fyles Glacier ice dam. Regular bedding in the sediments and the occurrence of isothermal water temperatures in this area of the lake suggest that deposition of the finer material is from suspension in a well mixed water column. Continuous, within-layer laminae of fine sand indicate the possible influence of underflow currents for sediment dispersal. Sediments for facies I are primarily derived from ice marginal and proglacial streams which enter directly into proximal sub-basins where sediments are effectively trapped behind the mid-basin sill. Within the shallow depression of the middle lake coarse sands and fine gravels of facies II accumulate. Sand and fine gravel show poorly defined layers and occur at the front of deltas that have been building out onto this portion of the lake for at least the last 50 years (figure 5-2b). Facies III deposits are associated with sedimentation processes in the deeper and larger eastern basin (figure 5-3)• Silty-clay laminae, separated from overlying silty-fine sand layers by a well-defined contact, alternate regularly with silty layers forming parallel sediment couplets which have a 155 Figure 5-3 Major water and sediment sources, facies distribution and sampling locations for ekman and percussion cores in Ape Lake. The four lacustrine facies are described in the text. Large arrows represent runoff and suspended sediment from predominantly glacial sources. Small arrows represent contributions from mostly nonglacial sources. Dashes lines are ephemeral stream sources. 156 true varve appearance (figure 5«4) . Multiple sub-layers or micro-laminations within the lighter, coarse fraction are not common but several coarser laminae (coarse silt - fine sand) can be found towards the base of some of the silty layers. Smith (1978) and Leonard (1981) found structures similar to facies III in Hector Lake, Alberta, and suggested that for annual rhythmites these coarse basal layers would represent an influx of sediment during spring snowmelt and summer icemelt peaks respectively. Like facies I and II no current-related structures are found. With the exception of one very small delta being formed along the north shore, direct sediment sources for the east basin consist only of small ephemeral streams. Since most underflow sediments are trapped behind the sills, material is delivered primarily in homopycnal flows. Correlation of laminae amongst all 1 east-basin Ekman cores is obvious even across distances of up to 1 km (figure 5-4). In the most distal portions of the lake and on benches which flank the east basin where water depths are considerably shallower (5-10 m), more massive clay-rich facies IV sediments are deposited (figure 5-3)• Core 84E14 shows 2-3 cm of massive to weakly layered clay overlying thin, regularly layered silty-clay indicative of reduced sedimentation rates. Deposition is primarily from suspension. Recent and Long-Term Sedimentation Rates Estimated sedimentation rates, using measurements from all Ekman cores, vary according to the dominance of specific sediment sources. In the west basin and Ape Glacier Bay, both of which are behind the mid-lake sill and are fed by major meltwater streams, sedimentation rates average 2.56 ± O.65 mm per couplet (n = 4 cores). In contrast, the average rate of deposition in the east basin is lower (1.8l ± 0.45 mm per couplet; n = 7 157 84E2 84E10 Figure 5-4 Bottom sediments of the east Ape Lake basin. Sampling locations are shown in figure 5-3- Distance between the more distal sample (84E2) and proximal sample (84E10) is approximately 225 m. Correlation between laminae is indicated by the lines. 158 cores) due to the larger lake bottom area and more distal location from primary sediment sources. A plot of laminae thickness with depth in figure 5.5 indicates that there are no strong spatial trends in sedimentation rate within the eastern basin itself suggesting sediments are derived from suspended load. Seven percussion cores were extracted in order to assess longer term sedimentary processes and rates of sedimentation (see figure 5-3 for sampling locations). Core lengths vary between 30.5 cm and 73 cm. Due to some distortion (wavy, sub-parallel laminae) interpretations are based on the comparatively undisturbed segments from the four deep water cores of the east basin. Layering in the upper 15~20 cm of all four cores is similar to that in the Ekman samples, showing light-coloured, clayey-silt material alternating regularly with darker silt-clay laminae which are bounded on the top by sharp horizontal contacts. Below 20 cm the light-coloured laminae become coarser (sandy-silt) and thicker, and several contain microlaminae of medium-to-fine sand. Structures and grain sizes are similar to recently deposited sediments in the west basin. Comparatively thick sedimentary layers also appear at several levels in the lower sedimentary sequence (figure 5.6)- It is possible that part, or all, of these massive layers formed by the redistribution of sediments from sub-aqueous slope failures, although they lack strong grading which is typical of turbidity currents (Harrison, 1975; Gilbert and Shaw, 198l) . Rates of sedimentation in the eastern basin are substantially higher in the lower core below 20 cm (cf. figure 5-6). An average rate of deposition is h.Sl ± 1.07 mm per couplet compared with a rate of 1.81 ± 0.45 mm for surface sediments. This higher rate is comparable with contemporary sedimentation rates in front of the Fyles Glacier ice dam which average over 3-00 mm per couplet. 159 Thickness (mm) J—i—i—i—i—i—i i—i i i i i i i i i i i i i i I i i i i i i i Figure 5.5 Couplet thickness versus depth in selected Ekman cores from eastern Ape Lake. Correlated laminae are connected by lines. Cores are plotted with increasing distance from water and sediment sources (see figure 5-3). 160 Thickness (ran) u Figure 5.6 Couplet thickness versus depth for percussion cores extracted from eastern Ape Lake. Correlated laminae are connected by dashed lines. Cores are plotted with increasing distance from water and sediment sources (see figure 5.3). 161 Grain size variation in core 84C2 was determined using 57 samples taken as representative of grain sizes observed under macroscopic and microscopic examination of the core (Gilbert and Desloges, 1987)• In the upper 20 cm there are only moderate differences in grain size of the lighter "summer" layers (63% silt/37% clay; silt-clay boundary at 4 m) and darker "winter" layers (46% silt/54% clay). Both indicate a low-energy depositional environment. Significant differences in grain size between light layers of the upper and lower core are evident. Below 20 cm the lighter layers are coarser (fine sand) and are characterized by silt/clay ratios that are four times those of the upper core. These significant textural variations suggest different sources for sediment and processes of deposition between the more recent and older sediments. Thicker, more massive appearing layers in the upper core contain higher percentages of silt when compared with other light laminae from the same level. This coarser texture is similar to that found in lighter layers of the lower core and indicates that more energetic processes were involved in the transport of this sediment. Internal grading was not visible in thin sections but the anomalous thickness of these layers may be related to different processes than those of the surrounding laminae. On the basis of changes in grain size and structure three types of depositional conditions might prevail throughout the year if these rhythmites are annual. During winter, following lake freeze-over, fine silt and clay are deposited from suspension to form the dark "winter" layers. Commencing with spring snowmelt runoff, coarser sediments are delivered to the lake and, later in the summer, icemelt derived materials contribute to total accumulation in the "summer" layer. Finally an influx of sediment from the surrounding basin and redistribution of sediment on submerged 162 slopes may contribute to the sediment pile following a high magnitude runoff event (e.g. autumn or early winter storm). Down core changes in grain size, sedimentary structures and rates of deposition indicate a change in the processes of sediment delivery during the interval represented by cores 8UC2 to 84C5. During the Neoglacial maximum Fyles and Ape glaciers coalesced and terminated in what is now mid-Ape Lake forming a terminal moraine (figure 5«1)• At the time of maximum ice advance Ape Lake would have consisted of the eastern basin with water and sediment derived directly from ice marginal and subglacial sources of the two glaciers. It is probable that facies I sediments found today in Ape Glacier Bay and in the west basin can serve as analogs for past conditions in the eastern basin, a hypothesis which is supported by similarities in grain size and structures found in the older east-basin sediments. Each of these streams would carry coarse sediments into the lake. The occurrence of multiple sandy layers within the "summer" unit, higher sedimentation rates for both "summer" and "winter" layers and a higher variance in laminae thickness all suggest that sediment delivery was dominated by hyperpycnal flows (underflow currents) during the interval of increased ice extent. A plot of mean grain size versus sorting in figure 5.7 demonstrates the distinctiveness of these sediments. Following retreat of the glacier ice fronts, ice marginal streams would have entered at positions behind the terminal moraine. Initially, for the larger Fyles Glacier, it is likely that the glacier remained grounded in the shallow mid-lake basin and that streams along the ice margin still entered directly into the east basin. After substantial vertical thinning and lateral retreat a greater proportion of sediments transported by meltwater streams and associated underflow currents would have become trapped behind the terminal moraine. Mixing interflow and overflow currents 163 2.6-, 2.4-2.2-2.0-1.8-1.6-\ \ Distal Dmo a i popyc flow \ hom  and x^ hypopycnal \ Proximal hyperpycnal flow \ \ A \o *\ \ \ A\ * A • • o Ape Lake Core 84E2 • "summer" lower core O "summer" upper core • massive • "winter" V \ \ o o * Cv o • : o \ \ \ \ "i—r 5.0 i—i—i—i—i—r*—i—i—i—i—i—i—i—i i r~ 6.0 7.0 8.0 9.0 10.0 Mean Grain Size (phi) Figure 5-7 Standard deviation versus mean grain size for samples taken from core 84C2. Coarser "summer" sediments (lower core) were deposited in proximity to Fyles/Ape ice fronts when these two glaciers drained directly into the east basin. Hyperpycnal flows (underflow) dominated. Finer "summer" sediments (upper core) are deposited after retreat of the two glaciers via homo- or hypopycnal flows. 164 (homopycnal and hypopycnal flows) would deliver sediment to the now distal eastern basin where deposition from suspension would occur (figure 5*7) • The implications for interpretations of hydrologic/climatic controls of sedimentation rates are two-fold. First, sediments delivered by meltwater streams directly are controlled partly by those factors governing icemelt rates and subglacial drainage, partly by sediment exchange and loss in the proglacial area and partly by the frequency of subaqueous slope failures at the lake margin where streams enter the basin. Single high-discharge events can be important and therefore the sediments may not represent the overall seasonal trend in runoff and sediment delivery. Secondly, for sediments transported in homopycnal flows the rates of sedimentation may be directly linked to the degree of lake water mixing at the point of stream entry and, because sedimentation rates are lower, more easily influenced by contributions from other non-glacial sources such as turbidity currents generated by isolated slope-failures. Thus careful consideration must be given to the possible origin of individual laminae and associated hydrologic inferences. Inferences regarding changing hydrologic conditions from changes in sedimentation rates can be made only if absolute dates can be associated with each lamina. If it can be shown that the sediment couplets referred to here are of an annual nature then the record of sedimentation rates can provide a potentially useful means for reconstructing the longer term hydroclimatology of the Bella Coola basin. (5•2) Periodicity of Sedimentation in Ape Lake ;s Extremely regular layering and textural contrasts of Ape Lake sediment couplets in both east and west basins provided the first 137 indication that these might be annual. To test this hypothesis Cs 165 was selected to verify the annual periodicity of these sediments. Cesium is particularly attractive because of its reasonably short half-life (30.5 years), precisely known fallout history and global distribution (McHendry et al., 1973)• Cesium has a particularly high affinity for the fine grained, organic-rich components of micaceous sediments; therefore, textures of silt and clay, such as those found in lake sediments, are most 137 likely to mirror the Cs atmospheric flux rate (Francis and Brinkley, 1976) . Several factors may influence the actual relationship between 137 atmospheric fallout and observed Cs concentrations in lake sediments. Fallout at any given latitude is influenced by the frequency and amount of rainfall (Davis, 1963), therefore, a strong gradient in isotope concentration might be expected across the Coast Mountains. Apart from the 137 flux rate to the lake surface itself, Cs may be temporarily stored in soils and vegetation in the surrounding catchment (McHendry et aj., 1973). Although data are limited, Wise (I98O) indicates a lag time of 2 to k years between peak fallout concentrations and the transport of cesium-enriched sediments to the lake bottom. Actual lag times would reflect the frequency and efficiency of storage sites in the routing system. Leonard (198l) suggests that there may be a lag of several years or more if cesium is absorbed onto the finer (<2 cm), unflocculated sediment, thereby remaining in suspension and possibly being transported out of the system. In either case it is probable that for a partly ice and snow covered catchment, such as that which surrounds Ape Lake, a lag of several years can be expected. 137 Other factors controlling the distribution of Cs are related to physical, biological and chemical characteristics of the lake (see Appendix E for the implications of these factors). 166 Despite these problems cesium has been used successfully to establish the periodicity of sediments in Pitt Lake, B.C. (Ashley and Moritz, 1979) » Bow Lake, Alberta (Leonard, 198l) and Hector Lake, Alberta (Leonard, 1986). Samples collected from Ape Lake were submitted for cesium determination to establish the periodicity of sedimentation and to assist in assigning calendar dates to rhythmites. Results Sampling was restricted to one core only in the east basin (84E2) (figure 5«3)- The coarse-grained nature of sediments in the west basin makes cesium dating difficult and thus reliable calendar dates cannot be assigned. Although it would have been advantageous to combine laminae from a number of cores in the east basin, initial inspection of the samples indicated potential problems with intra-annual layers and cross-correlation of units. The relatively thin laminae of core 84E2 precluded sampling on a couplet-by-couplet basis, hence, three or four couplets were included in each sample to ensure sufficient sample sizes even though a lower temporal resolution in ^^Cs activity would result. The sediment was trimmed to exclude contaminated material along the edge of the core and samples were then shaved using a sharp thin knife. Boundaries separating each sample were sharp and usually free of distortion so that mixing of sediment between samples was unlikely. A summary of sample depths, sample sizes and measurement procedures is included in Appendix E. 137 Concentrations of Cs in core 84E2 are plotted and compared with the atmospheric flux recorded at Lake Michigan in figure 5-8. The smoothed profile recorded in core 8'4E2 is similar to the atmospheric fallout rate suggesting that laminae in 84E2 are annual in nature and therefore are true 137 varves. Very low level concentrations of Cs (0.11 pCi/g) are detected in 167 137Cs(pCi/g) 0 10 20 137Cs(pCi/cm2/a) 0 1 2 1980-j Atmospheric flux Lake Michigan 1975-B Figure 5-8 A) 137Cs profile in core 84E2 from eastern Ape Lake (see figure 5-3 for sampling location). Concentrations were determined on a 2-k couplet basis. B) Estimated annual atmospheric flux of cesium-137 in eastern North America (solid line is from Robbins and £ddington, 1975). Dashed line is superimposed profile from A. 168 couplets 31-33 indicating that this is the lower boundary of the cesium profile. Peak concentrations are found in couplets 19~21 and then decline rapidly higher in the profile but do not return to the low levels measured 137 at the bottom of the core (couplets 1-2 were not measured for Cs activity). The peak values found in couplets 19~21 suggest that the high atmospheric fallout rates recorded in 1963 are incorporated in this level of the core. Under the assumption that each couplet represents a single year of deposition and that the surface layer recorded in core 84E2 represents sedimentation in 1984 then the sample containing couplet number 22 should correspond with the 1963 peak. Since this is not the case a lag of at least one year and as many as four years may be present which is a reasonable estimate given the evidence presented earlier. 137 High'values of Cs activity shown in figure 5-8 several years after the sharp decline in atmospheric fallout suggest one of two things. First, storage of cesium in the surrounding watershed is high and continuing 137 contributions of Cs from snow, ice and sediments to the lacustrine environment are significant. Evidence from both Ashley and Moritz (1979) and Leonard (1986) suggest that this is possible but lake bottom sediments from both these studies show a more rapid decline in cesium activity following the reduction in atmospheric fallout. The lag and smoothing of the profile is more pronounced in glaciated terrain which would suggest significant storage in snow and ice. A second possibility is that recycling of *^Cs is occurring at the lake bottom, thus maintaining these elevated values (cf. Albert et al., 1979)-Cesium isotope activity in Ape Lake sediments, then, provides good evidence for the annual periodicity of the rhythmite couplets. Though 137 absolute dates based on Cs activity cannot be assigned, the results do indicate that counting of well-defined couplets is a reasonable means of 169 establishing a history of sedimentation. The surface layer of most Ekman cores sampled in August 1984 consisted of thin light-coloured silt with no evidence of overlying clay deposits suggesting that this layer is comprised of sediments derived from spring snowmelt and summer ice melt during 1984. Hence, for the varve chronologies discussed in the next section it is assumed that surface layers, found in all cores were deposited during the summer of 1984 and that couplet number 22 was deposited in 1963. (5•3) Chronology and Controls of Recent Sedimentation Rates Measurements from Ekman cores taken in Ape Lake were used to construct a time series of recent (post-1948) sedimentation rates. To ensure that the chronology reflected lake-wide changes and to smooth anomalously high or low differences in sedimentation rates several cores were combined to produce a composite series. Samples from the east basin only were selected because of the morphological similarity of the lake bottom from which the cores were extracted and because sediments in this part of the lake are derived from sources which are representative of changes integrated over the entire catchment. Additional factors include high frequency of samples and the potentially longer record preserved in the east half of the lake. Typically, Ekman cores covered a period of 30 years occasionally extending beyond 35 years (1948) and all but one (84E1) contained sediments deposited during the summer of 1984. Problems associated with the presence of possible intra-annual summer layers in three Ekman cores reduced the number of samples used in the composite chronology to four, the same number of samples available for construction of the longer chronology. 170 Recent Sedimentation Chronology Total varve thicknesses in the four Ekman samples (84E1, E2, E7, E10) were standardized in order to reduce first order autocorrelation and give equal weight to each core. Each varve series was standardized by dividing the actual varve thickness by the sequence mean which yields a mean of 1.0 for the transformed data set below which sedimentation rates are considered low and above which sedimentation rates are high. Evidence presented earlier showed that declining rates of sedimentation occurred after the mid-1940s as the shallow middle-basin formed and then again after the early 1960s as the deeper west basin became fully developed. Thus, two sequence means were used for generating indices of secular variations in sedimentation rates, one for the earlier period (1948-1960) and one for the latter period (1961-1984) . All four standardized series were then combined as an unweighted average to form the composite chronology. Estimated lake-wide sedimentation prior to 1953 is based on fewer than four samples. The composite sedimentation chronology is plotted in figure 5-9« Sedimentation rates show some variance about the record means with well above average values in five individual years (1952, 1953. 1959. 1976, 1982) and below average sedimentation in the middle 50s, early 60s and 70s. For comparison, post-1948 generalized departures in winter precipitation for interior stations are also plotted in figure 5-9- While there is some agreement, particularly in the late 1960s and early 1950s when both precipitation and sedimentation rates were above average, the two series are not closely matched. This suggests other mechanisms may be important determinants of sediment delivery to the lake. It was thought that the influence of early autumn storms might be reflected in the sedimentary sequence. Significant rainfall and snowmelt runoff from both glaciated and non-glaciated terrain is apt to occur when 171 Standardized Departure 0.0 1.0 2.0 3.0 1980 1970 1960 1950 Figure 5-9 Composite sedimentation chronology for cores 84E1, 84E2, 84E7 and 84E10. Original curves shown in figure 5-5 were standardized by dividing observed sedimentation rates by the respective period mean. Two period means were used to reflect known fluctuations in the position of Fyles Glacier: 1948-1960 and 1961-1984. The standardized curves were then combined as an unweighted average. For comparison generalized departures of winter precipitation for interior stations are shown on the right. 172 the freezing level rises during autumn storms. Heavy rainfalls observed in late August of 1984 produced visible evidence of underflow currents and sediment redistribution in proximity to proglacial and ice-marginal stream outlets. However, careful inspection of prepared thin sections for east-basin varves associated with years in which several high magnitude autumn storms occurred revealed no systematic departures in sedimentation rates. Controls of Glaciolacustrine Sedimentation Rates in Ape Lake Inferences regarding changing environmental conditions from lake sedimentation rates require that the linkages between sediment yield and controlling glacio-hydroclimatological variables be known. Standardized values of varve thickness for the post-1948 series (figure 5-9) are compared with annually and seasonally measured hydroclimatic variables using correlation analysis. A set of 13 independent variables includes: winter precipitation (Oct.-Apr.) for coastal and interior regions; mean annual temperature; summer (May-Sept.) and late season (Aug.-Sept.) temperature; and runoff in the Talchako and Bella Coola Rivers (total summer, early summer (May-June) and late summer (Jul.-Aug.)). Results are presented in table 5-2. In general, most comparisons using the entire 32 year data set yield low and non-significant correlation coefficients with the exception of winter snowpack measured in the interior region (r = +0.39). late summer season temperatures (r = +0.44) and runoff in the Talchako River (r = +0.45). Scattergrams of varve thickness with precipitation, snowpack and runoff indices revealed a substantial amount of "noise" during years characterized by low sedimentation rates. This effect was much less pronounced in comparisons with temperature indices. Examination of multiple regression models (sedimentation rate as a function of hydroclimatic 173 Table 5.2. Pearson's r for Varve Thickness and Hydroclimatic Variables Winter Precip. Snowpack Temperature (i) (c) (i) (c) Annual Summer Late Summer [1] 0.03 0.10 0.39* 0.21 0.15 -0.12 0.44* (df = 32) [1] Bella Coola R. (df = 32) [2] Talchako R. (df = 19) Runoff Summer Early Late 0.31 0.21 0.30 0.39* 0.29 0.45* * significant at a = 0.10. (i) refers to index derived for interior stations and (c) for coastal stations [1] Entire post-1945 data set [2] Post-1965 data set 174 variability) indicated that confounding effects between temperature and precipitation exist. For example, removing years with anomalously low summer temperatures improves the relation with snowpack (r = +0.51) and with summer runoff in the Talchako River (r = +0.55)• Perkins and Sims (1983) and Leonard (1985) have found a positive and significant association between annual or summer temperature and sedimentation in proglacial lakes. Correlations were much higher than those for Ape lake and confounding effects were not reported. With increasing ablation season temperatures a greater proportion of ice melt and discharge into the lake is thought to lead to increasing concentrations of suspended particulate matter. A similar conclusion cannot be made from the results shown here. Low and even negative correlations between annual or summer temperature and sedimentation rates suggest that seasonally averaged temperatures from nearby stations are poorly related to conditions found at Ape Lake. However, during the late summer season (August-September), when the maximum snow-free glacier surface is exposed and glacier melt is at a peak, temperature and sedimentation exhibit a strongly positive relationship. April 1st snowpack and varve thickness also show a strong positive association particularly during years of average or above average late season temperatures. Increased snow cover (and temperatures) would lead to a higher discharge volume to the lake from the surrounding watershed and potentially higher concentrations of sediment in the lake if this inflow carries a significant suspended sediment load. Perkins and Sims (1983) found a strong negative correlation between snowpack and varve thickness in a southern Alaska lake and suggested that reduced glacier melt during years of heavy snowfall (increased snow cover and albedo of the glacier surface) produces lower sediment yield and thus reduced rates of deposition. 175 The opposite relationship found for recent Ape Lake sediments may relate to sources and transfer of sediment. A substantial amount of fine grained material (silt, fine-sand) can be found in the proglacial areas of Fyles and Ape Glaciers. During years of high snowpack, runoff from these areas may contribute more sediment than during low snowpack years. This hypothesis, however, would dictate that these areas be at least co-dominant sources of the suspended sediment found in Ape Lake. Inspection of non-glacial streams in July of 1985 and 1986 indicates that sediment concentration in these inflowing streams is small in relation to those draining Fyles and Ape Glaciers directly. Although correlations with regional runoff indices are generally low, a higher and statistically significant association occurs in years when temperature and snowpack conditions favour heavy runoff. This may suggest that certain minimum concentrations of suspended sediment are introduced to the lake each year but vary significantly only during moderate and high runoff years. Gilbert (1975) found that sedimentation in Lillooet Lake, British Columbia was significantly related to discharge during higher discharge years only, further supporting the contention that snowmelt generated runoff may be an important mechanism in delivery of sediment to the east basin. The higher correlation with Talchako River flows would support this conclusion. Recent depositional processes then, are related to several morphological and hydroclimatic factors which promote increased sediment concentration in Ape Lake. The potential for these proxy data to reveal a record of long-term hydrologic change is limited by the fact that no simple hydroclimate-sedimentation model can be specified. These complications are associated with several conditions: the strong maritime influence on runoff processes in coastal British Columbia; changing basin morphology, sediment 176 sources and glacier configurations over the last century; and the processes of sediment dispersal in the lake. Although a quantitative model cannot be explicitly specified for the period of detailed instrumental data, above average sedimentation in Ape Lake is associated with years characterized by above average snowpack, late summer season temperatures and regional runoff. The long-term sedimentation record was investigated in an attempt to verify hypotheses regarding controls of glacio-lacustrine deposition and to possibly make qualitative inferences about earlier hydrological environments. (5•4) Inferences from Long-term Variations in Glacio-Lacustrine Deposition The four deep-water, east-basin percussion cores (84C2, C3» C4, C5) were used to construct a standardized long-term chronology of sedimentation rates in Ape Lake prior to 1948. Due to increasing sedimentation rates lower in the sequence, the much longer percussion cores spanned a proportionally shorter interval when compared with the Ekman cores, the longest being 143 years (84C5). Thus, a chronology extending back to 1836 A.D. was available. Distorted sections of 84C3 and 84C5 were excluded so the number of cores used in the composite chronology varies between two and three in the period prior to 1904 (see figure 5-9). Prior to 1865 only one core (84C5) was used; therefore, there is a greater potential for bias in the earliest part of the record. Each sequence was standardized by dividing actual sedimentation rates by the l843-1940 mean. The mean for this interval was selected because prior to 1940 sediment delivery was directly into the east basin. The unweighted series were then combined to produce an average chronology (figure 5«10a). To enhance the lower frequency variation and to facilitate interpretations of longer-term trends the composite 177 Composite Sedimentation Index 0 12 0 12 Figure 5-10 Composite sedimentation chronology for cores 84C2, 84C3. 8kCk and QkZS- Original curves shown in figure 5-7 were standardized by dividing observed sedimentation rates by the 1836"" 19^+0 period mean. The standardized curves were then combined as an unweighted average. Unsmoothed data in A and a 5_year weighted moving average in B. 178 series was filtered with a five year weighted moving average (figure 5.10b) . The long-term trend shows declining sedimentation rates from the early 19th century through the middle of the 20th century. The decreased rate of deposition after 1940 is related to positional changes of Fyles Glacier rather than actual sediment yield but is included in figure 5«10 for comparison. Below average rates of sedimentation are evident after 1905 until 1940 and local minima occur between 1873 and 1884 and between 1845 and 1862. Comparatively high annual depositional rates are found in 1843. I863, I865, 1869. 1885, 1917, 1924, 1934 and 1940. The intervals 1836-1844, 1862-1885 and 1917-1945 are characterized by relatively high year to year variance in depositional rates. Mean sensitivity is approximately 20-40% greater than in other intervals. Key Events Examination of sedimentary structure and varve thickness in the record prior to 1940 indicates a more complex sequence of intra-annual sedimentation than the more recent sequence. Micro-laminations of silt and clay alternate with fine sand in several layers suggesting a sequence of discrete events {e.g. turbidity currents) might be responsible for the transfer of some sediment within the eastern basin. It was thought that sediment yield would reflect variation in discharge of meltwater streams as they entered directly into the east basin in addition to high-magnitude runoff events, particularly autumn storms which produce the greatest discharges in the regional drainage network. To facilitate comparison an annually calculated index of cumulative 3 -1 autumn discharge (summation of discharges above 250 m s in the Bella Coola river between late August and January) is plotted against positive 179 sedimentation anomalies for Ape Lake in figure 5.11» The less detailed record of flows prior to 1940 reflects the limited data available especially regarding flood events of low to moderate magnitude. The records do not match particularly well after 1940 when much of the sediment derived from Fyles Glacier was trapped behind the terminal moraine. A comparatiyely good association exists in the early 20th century record, particularly in the years I896, 1917. 1924, 1934, 1936. Major floods were recorded in the Bella Coola River during each of these years. The correspondence indicates the significance of autumn rainstorm-generated runoff in the earlier sedimentary record. The timing of autumn rainstorms also appears to be an important factor governing rates of sedimentation. Lacustrine deposition associated with flooding on November 19, 1936. the largest estimated flow in the Bella Coola River of this century, is lower when compared with other sedimentary units all of which correspond to runoff events earlier in the season (August to mid-October). It is probable that once lake freeze-over has commenced (November-December) rainstorm runoff has less effect on contributions of sediment to the lake. A greater proportion of precipitation which falls as snow at higher elevations later in the season will also result in lower runoff contributions to the lake. In addition to positive sedimentation anomalies, the frequency of microlaminations measured within each varve is plotted in figure 5•11 - Each sub-unit may represent sedimentation associated with material derived from spring runoff, summer glacier melt or autumn rainstorms. The number of layers within any "summer" unit varies from zero (homogeneous deposit) to four. As a result of major meltwater streams entering the lake behind the mid-basin sill, the post-1945 record contains the fewest intra-varve laminations. To confirm the temporal variability in frequencies of 180 Autumn Discharge Index Frequency of Microlaminations T-2 3 -i 5 1980-T—i—i—r—i 1 2 3 4 5 1970-1960-1950-1940-1930-Q 1920-« 1910-LU >-1900-1890-1880-1870-1860-1850-1840-i i i 2.5 2.0 1.5 1.0 Positive Sedimentation Anomalies Figure 5-11 Sedimentation anomalies and frequency of microlaminations in sediments of eastern Ape Lake. For comparison an autumn discharge index between the years 1896-1984 for the Bella Coola River is included. The discharge index prior to 1948 is based on historical observations and field evidence only. Note the correspondence of flooding and sedimentation anomalies. 181 microlaminations a one-tailed runs test was applied under the null hypothesis that there is a random distribution of these frequencies. The null hypothesis was rejected at the 90 percent confidence level (Z = 1.68) demonstrating that the higher frequency of intra-annual layering noted between 1863 and 1878 and between 1905 and 1918 is significant in the long-term record. Based on the higher frequency of intra-annual sedimentation and higher rates of sedimentation during the former interval it can be inferred that some combination of above average spring runoff, longer summers and greater numbers of autumn rainstorms occurred then. The inferred early 20th century departures can be tested qualitatively by examining climate trends at Bella Coola and Big Creek. It was noted that between 1904 (limit of early 20th century records) and approximately 1920 a persistent meridional flow regime dominated the general circulation. Although summer temperatures at Bella Coola were notably below average during this time (figure 3*9) » summer temperatures at Big Creek and regional winter precipitation were above average (especially 1906 to 1912 and 1918 to 1920). This is coincident with a higher frequency of intra-varve laminations. Hydroclimatic Inferences Further quantitative tests of the relation between hydroclimatic factors and sedimentation rates prior to 1940 (1904-1940) were considered using coastal and interior climate indices. Summer and winter precipitation, summer temperature (May-Aug.), early season temperature (May-Jun.), mid season temperature (Jul.-Aug.) and late season temperature (Aug.-Sept.) were standardized for this interval and compared with sedimentation rates in the east basin. Because of the unknown contribution of isolated autumn storms those varves which are suspected of containing 182 sediment derived from these events (1905, 1909, 1917, 1924, 1934, 1936, 1940) and which contribute a disproportionate amount to the overall variance in the series, were removed. Much like the recent sedimentation sequence, precipitation and summer temperature indices contribute little to the explained variance of pre-1940 sedimentation rates. Again, late season (Aug.-Sept.) temperature seems to be the most important and significant variable (r = +0.51) for the filtered series (which represents a substantial improvement over the unfiltered series where r = +O.36). It is probable that during the maximum advance of Fyles and Ape Glaciers a small or insignificant proglacial area would have been exposed, thus snowmelt-generated runoff would not have been an important mechanism for sediment delivery. The above results suggest that controls of recent (post-1945) sedimentation in ice-dammed Ape Lake vary from those found in proglacial lakes fed by one or two major meltwater streams. A combination of stream flow from the surrounding ice-free basin and glacier melt appear to be important elements in the redistribution and delivery of sediment in the basin. The strength of the relations does not permit precise long-term reconstructions to be made but qualitative inferences regarding environmental change may be possible. (5•5) Summary and Conclusions Several factors appear to be important in determining the processes of sediment distribution and deposition in Ape Lake. A fairly abrupt shift in the character of deposits from microlaminated and thick summer varves containing layers of fine sand to thinner, uniform, silty summer varves is associated with changes from an underflow/interflow depositional environment to one dominated by overflow and fallout of suspended material. 183 In the latter case, snowpack volumes and late season ablation temperature appear to control depositional rates. This association can be only partly verified in the pre-1940 record where late summer temperature and the frequency of autumn rainstorms influence rates of deposition. This shift in depositional regimes coincided with the opening of the west basin sometime between 1940 and 1945. Inferences regarding changes in environmental conditions over the longer-term using rates of sediment deposition as proxy evidence clearly depend upon a stable model (s) relating the two. No satisfactory quantitative model could be specified. Grain size, structure and depositional rates in the east basin appear to be consistent within the pre-1940 period suggesting that summer temperatures and the frequency of high-magnitude rainfall events are important explanatory variables for this interval. It is probable that above average sedimentation rates in 1840, 1843, 1863/64/65, 1871/72, 1885, I887 and 1896 are associated with unusually warm summers and/or major flood-producing storms. Distinguishing between the two mechanisms remains problematic. 184 CHAPTER VI Biological Evidence of Environmental Change: Tests Using Tree-Ring Evidence (6.0) Introduction Landforms and sediments commonly provide only limited resolution of hydroclimatic variability. Variations in tree-growth are known to be associated with climatic factors integrated over the course of a season or throughout a particular year. In order to utilize this form of proxy data it is necessary to consider response characteristics and potential sources of error involved in drawing inferences about long-term changes. The purpose here is to identify these responses, assess the utility of transfer functions developed using tree-ring data and compare inferred hydrologic changes with observed fluctuations over the 20th century. (6.1) Tree-Growth and Climate Annual wood increments consist of low-density, large-celled earlywood (spring) and high-density, small-celled latewood (late summer or early fall). Each year's increment is controlled by several factors, the most important of which are climate and site characteristics. The major regulators are temperature and precipitation because they control the water balance in and around the tree and regulate the metabolic growth rate (Kramer and Kozlowski, 1979). Other less direct factors include competition between species and between units of the same species, slope stresses, physical damage and depletion of soil nutrients. Growth rates from year to year will respond to all these factors but are generally dominated by the most limiting (Meyer et al., 1973)-Moisture or heat stress during a significant portion of the growing season will result in reduced photosynthesis and a lower production of biomass. Actual thresholds for stress are partly species dependent and 185 partly influenced by site conditions such as local drainage, soil depth and sources for nutrients. Between some lower and upper thresholds, each characterized by an integrated set of climatic, physiological and pedological conditions, growth rates are optimal from year to year and will be limited only if some other factor is in short supply. This situation generally produces little growth variation through time and thus a strong climatic signal is not present. In order to isolate the effects of climate it is necessary to select sites which are host to trees that are near or at their limit of tolerance for moisture or energy. It is not uncommon for a climate factor to be limiting for only part of the growing season. For example, above average temperatures in the early spring may initiate bud opening and produce copious amounts of earlywood. However, if conditions deteriorate as the growing season progresses below average biomass production may occur and the resulting wood increment may not reflect the extremes in growing conditions. Because it is the available energy that determines the timing of bud opening, temperature is initially a more important factor (Kramer and Kozlowski, 1979)- Later in the season, when the demand for water and nutrient intake is at a maximum, the availability of water is more important. Long-term variation in tree growth is a function of: (1) a quasi-deterministic component related to the size of the tree and dynamics of the forest stand and soil environment and (2) a stochastic component related to the persistence of climatic effects in one year on the physiological status of the tree in ensuing years (Graumlich, 1985). Hence, the way in which a tree responds to climate in any give year may depend on growing conditions in previous years. The actual memory of the system may be several years but the effects may damp out quickly if above and below average growing 186 conditions occur in close succession. Therefore, tree-ring growth indices tend to exhibit an autocorrelative structure of at least a one year lag. Tree growth response may be compared directly with temperature and precipitation or indirectly via other hydrological parameters which are known to be sensitive to climatic change. Stockton and Fritts (1973) have shown that tree-growth may be influenced by groundwater conditions, in particular the level of the groundwater table and degree of soil saturation. Other attempts have been made at reconstructing river runoff variations using tree-rings (Stockton, 1976; Holmes et al., 1979? Campbell, 1982; Jones et al., 1984). The comparisons remain approximate since runoff and tree growth are both controlled by climate variations rather than there being a direct functional association between them. Conifers of the Pacific Northwest take advantage of the warm, moist climate by continued, although reduced, photosynthesis and nutrient uptake throughout much of the autumn and early winter seasons. Increased moisture stress and greater continentality of temperatures towards the interior can lead to a differential seasonal response in growth characteristics for the same species (Waring and Franklin, 1979) • Tree-growth responses to climatic change have been investigated for a number of sites and species in northwestern North America. Keen (1937). Schulman (1956) and Stokes et al. (1973) performed much of the earlier work on drought sensitive Douglas firs in the drier regions of eastern Oregon, Washington and south-central British Columbia. More recently Fritts and Lough (1985) and Graumlich and Brubaker (1986) considered long-term temperature variations of the Pacific Northwest based on tree-growth variations for species at or near tree line. In addition, Graumlich (1985) investigated the response of drought sensitive species in the Pacific Northwest. While long-term temperature and precipitation reconstructions 187 have been undertaken for areas further north in the Yukon (Jacoby and Cook, 198l) and further east in the southern Canadian Rockies (Josza and Oguss, 1985). little information is available for coastal and western Fraser Plateau areas of central British Columbia. Methods Selection of suitable sampling sites is the most important step in obtaining tree-ring data which are likely to contain a climate signal of regional significance. Graybill (1982), Graumlich (1985) and others have demonstrated that the width of a tree-ring (R) at any time t will depend upon: a biological growth trend (G) as a function of increasing age, soil dynamics and a short-term physiological memory; a macrocl imatic signal (r\) common to all trees at a site which can be decomposed into several frequencies; disturbance factors (D) such as fire, insect damage or slope stress which may be unique to one tree or common to all trees at a site; and some random growth signal (E) unique to each specimen; i.e., Rt " Gt + \ + Dt + Et Although the breakdown of the model is not entirely realistic since some factors may become confounded at certain frequencies (e.g. soil nutrient status and macroclimate) it does provide a framework for decomposition of frequencies within the ring-width time series. In order to maximize the climatic signal in this study a series of cores were extracted from several trees at a number of sites which showed no adverse reaction to the disturbance factors cited above. Usually two increment cores were extracted from each tree at breast height 188 perpendicular to slope orientation; a sampling scheme which minimizes the possibility of sampling growth trends related to slope stresses. Sites (trees) characterized by over-steepened slopes, poor drainage, frequent fire damage, extensive scarring, intense competition and over mature (decaying) forest cover were avoided. Entire tree disks were taken at some sites in order to assess more completely the within-tree and between-tree variance of growth patterns. Details of sampling and acquisition of ring-width data are given in Appendix D. A total of eight sites were selected for sampling, seven of these within the basin and one towards the drier interior of the province (see figures 6.1 and 2.1). Four of the eight sites were at or near tree line within forest communities of subalpine fir/lodgepole pine/western hemlock but chronologies were developed only for subalpine fir {Abies lasiocarpa). Tree line sites are thought to be primarily temperature stressed. The four other sites were in mature stands of predominantly Douglas fir (Pseudotsuga menziesi i) located in the drier valley bottom of the eastern Bella Coola basin (three sites) and near the confluence of the Chilko and Chilcotin Rivers on the Fraser Plateau (one site). In addition to the four Douglas fir chronologies produced here, tree-ring data for several sites centered around Williams Lake (Schulman, 1956) and elsewhere in the middle Fraser River (Stokes et al., 1973) were also considered. Douglas fir trees within these regions are thought to be moisture stressed. A summary of site names, sampling frequency, temporal coverage of each chronology and signal-to-noise ratios is given in table 6.1. Full data series and summary statistics are in Appendix D. A site averaged chronology was produced by first removing the deterministic portion of the long-period variance using negative exponential or low order polynomial curves fit to each series. Selection of Figure 6.1 Locations of tree sampling sites in Bella Coola basin. Sites outside the basin are shown in figure 2.1. Douglas fir sites are at valley bottom locations while subalpine firs sites are near treeli ne. 190 Table 6.1. Summary of Tree Ring-Width Chronologies Used in this Study. A. This Study Alpine Fir Rainbow Range Ape Lake Noohalk Lakes East Nusatsum # of trees # of cores # of tre # of cores Period Signal-to-noise % autocorr. sampled sampled in chronology in chronology A.D. 45 34 21 16 83 61 34 21 11 12 15 20 18 10 1725-1984 1740-1984 1871-1983 1711-1983 ratio 4.38 5.03 3.25 2.98 reduction 3% 6% * 2% 55! * Douglas Fir Stuie (Tweedsmuir Lodge) 25 Burnt Bridge Terrace 8 Burnt Bridge Slope 6 Bull Canyon 15 46 16 12 30 15 12 18 1579-1984 2.54 12 1704-1983 3.18 15 1705-1983 1.67 16 1650-1983 6.21 2% 3% 1% 4Z B. Other Studies 3 Williams Lake 4 Pavilion 44 20 1420-1944 1480-1965 5.50 6.31 3Z Kamloops 20 1420-1965 4.25 4% 1. Calculated as the ratio between the best fit polynomial and high frequency year-to-year variance. 2. Z reduction in autocorrelation using autoregressive-integrated-moving-average (ARIMA) model of Cook and Holmes (1985). The * indicates significant improvement from the original series. 3. From Schulman (1956). 4. From Stokes et al. (1973). . 191 an appropriate growth detrending function was facilitated by plotting each series and identifying common patterns of low frequency variance. This procedure maximizes the trends shared at a particular site and reduces first-order autocorrelation. Standardized tree-ring indices were then generated for each sample by dividing observed growth by expected growth (detrending function) and then combining all index series as an unweighted average to form the master site chronology. In an attempt to assess and remove the stochastic portion of the long-period variance in each chronology, autoregressive modeling procedures developed by Cook and Holmes (1985) were applied. The autoregressive model is calibrated using characteristics of the chronology during the period of observed climate data (1930-1983) and is then applied to the whole series assuming that the form of the persistence during the calibration interval has not changed in the past. When the modeling resulted in significant improvement (i.e. reduction in variance due to autocorrelation) the new chronology was used (see table 6.1). Full details of the model are given in Cook and Holmes (1985) • In effect all raw data series used in the following analyses have been pre-whitened (cf. Graumlich, 1985) • (6.2) Response Characteristics Although it is assumed initially that moisture and temperature stresses are the most important factors influencing yearly growth, other hydrological and micro-site factors may be significant. Graumlich and Brubaker (1986) have shown that sites near tree line may respond not only to fluctuations in available energy but also to the accumulated depth and duration of snowpack during the early part of the growing season. With increasing snowdepths below average soil temperatures are maintained for a 192 greater proportion of the growing season resulting in stomata closure and decreased carbon assimilation (Tranqui11ini 1979; Tesky et aj ., 1984). Several seasonally and annually averaged climatic and hydrologic indices given in Chapter 3 were compared with the response signal contained in the chronologies listed in table 6.1. The strength of correlation between ring-width and individual climate variables was evaluated using partial correlation coefficients, a procedure which holds constant the covariance of other variables and more accurately portrays singular response characteristics. Partial correlation coefficients for winter (Oct.-Apr.), summer (May- Aug.) and annual temperature and precipitation indices (interior stations) are plotted in figure 6.2. In addition, spring snowpack (interior stations) and Bella Coola River summer runoff indices are also correlated with each chronology (figure 6.2d) . The general pattern for Douglas fir trees is positive correlations between ring-width and precipitation (especially winter precipitation) and negative correlations with temperature. The association would suggest that years, particularly winter seasons, characterized by above average precipitation lead to greater moisture availability during the growing season and thus increased biomass production. Above average winter and summer temperatures increase potential evapotranspiration and reduce available moisture leading to negative correlations between growth and summer temperature. Chronologies developed for valley bottom locations in the eastern basin (Stuie, Burnt Bridge) are somewhat anomalous in that correlations with winter and annual temperatures are positive. The more transitional climatic conditions of these sites (warmer winters and higher annual precipitation which falls partly as rain during the winter months) might lead to positive growth responses during warm winter years. Unlike sites further to the east, Douglas firs in the valley-bottom locations 193 A , Winter Temperature B Annual Temperature Tal -0.2 .3 -0.1 i i i 0.1 0 1 3 , o bb -e-bc -0 2 - pv -0 4-al • • rr en^ , -0;1 nl -0.2 obb °bt 0.1 -0.2 --0.4 --0.3 Summer Temperature ^1 -0.1 -0.1 --0.3 -0.6 0.3 0.1 0.1 0SH obc D Snowpack •0.4 e ro 13 ro i i__ 1 1 1 in 1 1— o a ^4 -0.2 -0.4-1 0.2 Q bb O bc^h kl 90 C . , -3 0.1 0.3 H, Opv °bt o rr tl al Figure 6.2 Partial correlation coefficients between ring-width for several master tree-ring chronologies and selected hydroc1imatic variables. Tree line sites (•) include: Rainbow Range (rr), Ape Lake (al) , East Nusatsum (en) and Noohalk Lakes (nl) . Interior sites (O) are Pavilion (pv), Kamloops (kl) , Williams Lake (sh) , Bull Canyon (be), Stuie-Tweedsmuir Lodge (tl), Burnt Bridge Terrace (bt) and Burnt Bridge Slope (bb). Partial correlation coefficients significant at a = 0.10 are shown by vertical and horizontal lines. 194 appear to be more responsive to summer precipitation conditions (figure 6.2c) . Subalpine fir generally show an insignificant response to precipitation variability and winter or annual temperatures. However, summer temperature is an important factor positively related to growth particularly for the Ape Lake site (figure 6.2c). Summer runoff indices are positively correlated with all chronologies. There are several possible causes for this association but it is most likely a combination of above average summer temperatures and above average water storage in the soil. Spring snowpack shows a negative correlation with growth in subalpine fir, which is particularly significant at the Ape Lake site. High snowpack years may retard growth at some sites involving mechanisms much like those proposed by Graumlich and Brubaker (1986). Most of the Douglas fir chronologies show a positive relationship with snowpack which is probably also related to a high soil moisture status. Although partial correlations are generally low the results in figure 6.2 demonstrate that Douglas fir trees, particularly towards the drier interior, are most directly related to annual or winter precipitation variability. Growth at tree line sites, especially farther to the east, is closely associated with summer temperature and winter snowpack conditions. (6.3) Growth Variations and Climate Interior Douglas Firs Three of the seven Douglas fir chronologies in table 6.1 exhibited similar low and high-frequency departures in growth (Bull Canyon, Pavilion, Williams Lake), hence all three were combined as an unweighted average. This produces a stronger regional signal and smooths out between-site anomalies. Since the chronologies were constructed at different times the 195 period 1966-1983 is based on only one chronology, 1945-1965 on two and the pre-1945 interval on all three. Growth variations of Douglas fir between 1900 and 1983 are plotted in figure 6.3a. Above average growth occurs between 1940 and 1970 whereas negative departures dominate the interval prior to 1940 and after 1970. Shorter periods of near average growth are found between 1900 and 1915 and between 1934 and 1940. Notably persistent trends in growth occur in the post-1960 interval and between 1915 and 1934. The highest year to year variance appears to cluster around the middle of the century commencing in the early 1930s and ending in the late 1950s. Also plotted in figure 6.3a are generalized departures in annual precipitation at Big Creek. These patterns correspond fairly well with growth variations in the averaged Douglas Fir chronology suggesting that available moisture is strongly associated with biomass production in this species. Although there is a strong association between generalized precipitation trends and tree growth in figure 6.3a there are notable extreme departures. These anomalies are important to consider for two reasons: (1) they represent extreme growth variations about a long term mean which may be due to a combination of environmental conditions not indexed by one or two climate variables and (2) these extremes can significantly influence linear statistical models. To assess the temporal variability of these extremes the high frequency component of the series in figure 6.3a was extracted by first smoothing with a 13-year weighted filter. The observed growth was then divided by the smoothing function for each year to produce the high frequency series plotted in figure 6.3b. The upper and lower threshold lines shown in this figure are set at ±1 standard deviation from the mean. The frequency of crossings above the upper threshold and below the lower threshold were tabulated as runs and tested under the null hypothesis that 196 DOUGLAS FIR 0.4-1 I 1 1 1 1 1 1 1 1 " 1 1900 1950 2000 YEAR (AD) Figure 6.3 Regional Douglas fir chronology derived using three site chronologies. A) Combined low and high frequency tree-growth signals compared with generalized departures in precipitation at Big Creek (above A). B) High frequency tree-growth signal from A. Threshold lines are plus and minus one standard deviation from the mean. 197 the extremes are randomly distributed. Application of a one-tailed runs test resulted in rejection of the null hypothesis (z = -2.62,a = 0.10). Thus, the clustering of extreme positive and/or negative departures during the intervals 1900-1910, 1936-1945 and 1952-1961 is statistically s i gn i f i cant. The higher mean sensitivity of growth during these three periods implies a higher year-to-year variance in factors controlling growth. The response relationships in figure 6.2 suggest winter or annual precipitation as a dominant control of growth at interior sites. Analyses presented in chapter 3 indicated that mean sensitivites in precipitation for most months of the year were highest between 1920 and 19^+5. A similar correspondence does not exist for the earlier period (1900-1910) but the period 1952 to 1961 also shows a high between year variability in precipitation departures. The implied relationship then is towards a higher mean sensitivity in Douglas fir growth during periods of less persistent departures in precipitation, a characteristic of zonal circulation regimes. Subalpine Fir Three of the four subalpine fir chronologies in table 6.1 exhibited similar long-term and high frequency departures as well as higher signal to noise ratios (Rainbow Range, Ape Lake, Noohalk Lakes). In order to maximize the regional signal of growth variations at tree-line the three chronologies were summed and then averaged. A plot of the averaged chronologies for a common interval 1900 to 1983 is given in figure 6.4a. The most visible trends are average to below average growth after 1950 and above average growth prior to 1950, particularly 1900-1920. The most persistent negative departures occur in the 1970s whereas strongly positive departures occur between 1916 and 1920. The interval 1925 to 1945 198 Figure 6.if Regional subalpine fir chronology derived using three site chronologies. A) Combined low and high frequency tree-growth signals compared with generalized departures in temperature at interior stations (above A). B) High frequency tree-growth signal from A. Threshold lines are plus and minus one standard deviation from the mean. 199 is characterized by a high year to year variance in growth. The observed variations in growth show some correspondence with the generalized 20th century trends in summer temperature for Interior stations (figure 3.1c). The only notable divergence is above average growth during the first two decades when summer temperatures at Big Creek were mostly average. As with the Douglas fir chronologies the temporal variability in extreme departures was evaluated (figure 6.4b). Although not as strongly clustered there is a statistically significant grouping of positive and negative departures during the intervals 1925 to 1945, 1955 to 1964 and 1976 to 1983 (z = -1.79.a = 0.10). The first two periods overlap with those identified for the Douglas fir chronologies although there is not an exact correspondence. This similarity would suggest a common factor controlling extreme growth responses. Examination of temperature, precipitation and snowpack records for several of the individual extreme years indicates that snowpack conditions may be important. For example, well above average winter precipitation (and snowpack) in 1957 and 1976 corresponds with vigorous growth in all Douglas firs sampled for this project while at the same time well below average growth was noted for subalpine fir. Summer temperature during these same years was low resulting in well above average water availability for Douglas firs (vigorous growth) and poor growth for subalpine fir due to the low soil and air temperatures at tree line sites. These qualitative comparisons indicate that the long-period variance of growth variations in Douglas and subalpine fir trees are associated with departures in annual precipitation and summer temperature, respectively. High-frequency departures in growth have a tendency to cluster within the same intervals for both species and are related to periods of high year-to-year variance in climatic departures. Seasonal temperature variations 200 complicate the responses of Douglas fir and spring snowpack conditions influence the early season growth in subalpine firs. (6.4) Development of Transfer Functions The similarity of generalized growth departures for both species of trees and significant partial correlation with certain climate variables (figure 6.2) suggest that there may be some basis for developing transfer functions using tree-ring data. Transfer functions are constructed for the period of observed climate in order to incorporate the mutual variance between the dependent (temperature or precipitation) and independent (ring-width) variables. Linear statistical models have been used for this purpose and are applied here to assess the quality of inferences drawn from this form of modeling technique and proxy data. The implication of the preceding analysis is that these linear models will reveal only the averaged trend in climatic departures. Each chronology in table 6.1 represents site-averaged tree growth. While visual inspection of each series demonstrates a reasonably high spatial and temporal coherence amongst species, precise agreement between all chronologies during all intervals is not present nor is it reasonable to expect such agreement. In order to extract the regional signal common to all series objectively, principal component analysis was undertaken. The principal components (PCs) represent weighted mean departures which emphasize the extreme years and thus should 'track' climate more precisely (Peters et al., I981; Josza and Oguss, 1985)• Each PC represents a different weighted combination of the input series and they are formulated so that they are mutually independent, or orthogonal. PCs were calculated separately for each species using the six chronologies considered above. PC scores are based on the entire 334 year 201 tree-ring data set for Douglas fir and 245 year data set for subalpine fir rather than using just the interval for which climate data were available (1930-1983)• This procedure incorporates the long-period variance associated with tree growth and allows for estimates of climate variations prior to the instrumental record. A separate analysis revealed that correlations between climate data and PC scores calculated for the shorter interval were only marginally stronger. Correlations between PC scores and climate data (table 6.2) indicate the following: (1) the first PC for Douglas fir is most highly correlated with winter precipitation (October-April) whereas the second PC is associated with winter temperature and the third with summer precipitation; (2) the first PC for subalpine fir is significantly correlated with summer temperature and the second PC with spring snowpack; and (3) in both cases the first PC accounts for 65% or more of the overall variance in the original data. Since the evidence points to the significance of annual precipitation in controlling Douglas fir growth the three PC scores for this species were used as independent variables in a stepwise multiple regression analysis against annual (Sept.-Aug.) precipitation. Data for interior stations between 1930 and 1983 were used in the calibration. Selection of the most significant principal components was made using a low inclusion level (F = 2.0). Addition of all three components yields a normal multiple-regression function based on the original independent variable set. Preliminary models were constructed using one half of the 1930-1983 data set (randomly selected points) and each model was tested for predictive ability (table 6.3). The entire data set was then used to develop a full model in which significant outliers were tested for. The final model in table 6.3 is the Table 6.2. Correlation Matrix Between PC Scores and Hydroclimatic Variables Precipitation Temperature Summer Spring Winter Summer Winter Summer Runoff Snowpack Douglas fir %VR3 PC 1 (683!) 0. .38* 0. .19 0. .21 -0, .09 0. . 18 0. .07 PC 2 (21%) -0. .09 0. ,04 -0. ,25 0, .02 0. .15 0. .15 PC 3 (11%) 0, .10 0, .34* 0. .14 -0, .13 0, .06 -0, .08 Subalpine fir ZVR3 PC 1 (65%) 0. . 19 -0. 15 0. .09 0. .45* 0. ,18 0. .09 PC 2 (26%) -0, .04 0. .11 -0. .16 -0, .21 -0. , 11 0. .39* PC 3 ( 9%) -0. . 10 0. 12 0. ,03 0, .06 0. ,10 0, .10 1. For interior stations. 2. Bella Coola River above Burnt Bridge Creek. 3. - %VR is percent variance reduction * - indicates significant at the 0.90 confidence level 203 Table 6.3. Coefficients of Transfer Functions for Annual (Sept.-Aug.) Precipitation Indices. Principal Preliminary Models Full Final Component 12 3 Model Model (n=25) (n=23) (n=21) (n=51) (n=48) 1st 0.43 0.38 0.17 0.31 0.30 2nd -0.16 3rd 0.34 0.36 0.37 0.38 Intercepts 0.19 -0.21 0.22 0.06 0.04 Summary Statistics: R2 =0.25 R 2 = 0.21 a D - 1.81 F =3.93 SE =0.75 Confirmatory Statistics: RE = 0.35 r =0.52 = 0. .37 = 0. , 19 = 0, .33 = 0. .16 = 2. .10 = 1. .64 6. ,20* = 2. ,29 0, .67 0. .72 0. .43 0. .10 0. .69 0. .41 =0.22 =0.29 =0.18 =0.25 = 1.78 = 1.94 = 7.17* = 9.48* =0.71 =0.62 R = coefficient of determination 2 2 R = adjusted R a D = Durbin-Watson d (reject H - no significant autocorrelation) o F = F ratio (* significant; a = 0.025) SE = standard error RE = reduction of error statistic r = correlation coefficient 204 full model minus significant outliers and represents the best calibrated transfer function using the available data. Explained variance of the dependent variable (annual precipitation) using restricted portions of the data set (preliminary models) varies between 16 and 33 percent. With the exception of the weakest model [3], reduction of error statistics (RE) confirm the reasonable predictive ability of each when tested against the withheld portion of the data set. The first and third principal components are the most significant variables in all but the first model. Using the full data set only 21% of the variance is explained but 3 outliers significantly influence the regression equation. When removed, the proportion of the explained variance increases to 25% (final model). The three estimates which yield the highest explained variance share regression coefficients which do not differ significantly and exhibit the lowest standard errors. The final model is used to estimate annual precipitation for the interior between 1900 and 1983. Plots of the reconstructed values and filtered series are given in figure 6.5- The trend retains much of the variance in the original growth series. Below average precipitation is inferred between 1900 and 1945. fluctuating but average to above average conditions to 1970 and, with the exception of 1976 and 1977. below average precipitation to 1983- A distinct change to higher year-to-year variance occurs after approximately 1918, a tendency which ends in the mid-1960s. With only 25% of the overall variance in the dependent variable explained, the final model in table 6.3 and reconstructed September to August precipitation series in figure 6.5 should be considered approximates at best. Extreme departures in growth do not always correspond with extreme climate which produces a high standard error. Analysis of residual variance demonstrates that the poorest fit corresponds with periods of high year-to-205 1.6 -i 1.4 H Vi W U S i.2 H 2 2 i.o-H * 0.8 -u 0.6 -0.4 -Vi — a i.2 A a z M 2 O M H 2 1.0 — S 0.8 H H 0.6 -1 1900 Figure 6.5 Tree-ring derived estimates of annual precipitation using Douglas fir chronologies. Low and high frequency components are shown in A. Low frequency or response of the mean in B is generated by a five-year weighted moving average of the series in A. 206 year variance in growth (1920-1945); a result which further supports the fact that the mean and not extreme departures are being reconstructed. Interpretations of the sequence in figure 6.5 should be made in recognition of this limitation. A stepwise regression analysis was also undertaken using the three PCs generated from the subalpine fir chronologies. With the same inclusion level (F = 2.0) only the first principal component entered the equation explaining 45% of the dependent variable (summer temperature for interior stations). Since only one PC was important the trace of this component adequately portrays tree-ring responses and therefore is used directly as an index of temperature variability between 1900 and 1983 (figure 6.6). The much higher level of explained variance leads to greater confidence about inferences drawn from the series in figure 6.6. The first principal component of growth variations for subalpine fir trees in the Bella Coola basin matches quite well with the overall growth patterns of mountain hemlock in the Cascade Mountains 500 km to the south (c.f. Graumlich and Brubaker, 1986). In addition, there is a close correspondence between the PC time series and estimates of July heating degree days reconstructed for Alaska and northwestern Canada (Jacoby et al., 1985)-Although there are common secular trends Graumlich and Brubaker (1986) have placed different environmental interpretations on the growth variability using information from other species. While estimates of summer temperature variations using other species were not made in this study it is recognized that the effect on growth patterns of interaction between snowfall accumulation and summer temperature is complex and therefore the interpretations given here remain approximate until more data from a variety of species are acquired. 207 1.8 -1.6 -w 14-a z U M Q Z 3 s Cd Cd H 1.2 -1.0— 3 3 td H 0.8 -0.6 -0.4 -1.4 -i 1.2 -1.0— 0.8 -0.6 -0.4 J I 1 1 1 1 1 ' 1 1 1 I 1900 1950 2000 Year (AD) Figure 6.6 Tree-ring derived estimates of summer temperature using subalpine fir. Low and high frequency components are shown in A. Low frequency or response of the mean in B is generated by a five-year weighted moving average of the series in A. 208 (6.5) Inferences About 20th Century Hydrology from Tree-Rings The temperature and precipitation trends inferred from tree-rings in the previous section are tested further using 20th century variations in observed and modeled runoff of the Bella Coola River. Two separate comparisons are made with the tree-ring data: summer runoff and autumn storm-generated runoff. Summer Runoff Summer runoff models for the Bella Coola River were evaluated in chapter 3 and are used here to estimate variations in summer runoff for the period 1904 to 1983 using measured temperature and precipitation data as input parameters. The final model presented in table 3-5. which includes winter precipitation and summer temperature as independent variables, accounts for 55% of explained variance in summer (May - August) runoff. This model was shown to predict reasonably well independent runoff data from Salloomt and Nusatsum Rivers. Predictions from empirically-based models are assuredly valid only if the new input variables fall within the range of the original data set used for model calibration. Extrapolation occurs when the independent variables lie beyond this range. A prediction may be classified as an extrapolation even if each input variable is within the calibration range but they have not been observed in the combination in which they are entered into the reconstruction. These extrapolated values can be isolated by constructing the minimum covering ellipsoid (MCE) which circumscribes the observed values (Cook and Weisberg, 1978). The MCE was derived from a program given in Graumlich (1985) and applied using temperature and precipitation from the coastal and interior regions. Since indices based on multiple station data could be calculated only for the post 1930 interval (limited by the 209 length of the shortest record) an MCE was constructed for the 1904-1930 interval separately using the more limited data base. Actual and estimated summer runoff for the Bella Coola River is given in figure 6.7a. Prior to 1918 estimated summer runoff was well below average although three of the seasonal estimates from this period are extrapolated values. Following 1918 average to above average summer runoff characterized by higher year to year variance dominates until the mid-1930s after which there was more average summer runoff, with the exception of 1939 and 1940. Extrapolations are also noted for the years 1922 and 1934. As with most empirical regression techniques the extreme runoff years are poorly modeled (e.g. 1958, 1973)- Thus, extremes in summer runoff are apt not to be well represented in the reconstructed record. Differences between actual and estimated runoff for extreme runoff years during the calibration interval indicate predictions of runoff vary between -15 and +30%. However, the series does indicate periods in which the potential for high magnitude runoff is highest and appears to reflect the various within-period variances of the independent variables. For comparison, a runoff index was constructed using the tree-ring derived estimates of precipitation and temperature. Since observed precipitation accounts for 75% of the explained variance in summer runoff (figure 6.7a), a weighting ratio of 3:1 was adopted for averaging the precipitation series derived from Douglas fir responses (figure 6.5a) with the temperature series inferred from subalpine fir (figure 6.6a). The runoff index is plotted in figure 6.7b. Although some similarities between the two series in figure 6.7 exist, the general correspondence is poor. Above average runoff in the late 1950s and 1960s can be inferred from the tree-ring reconstructions which is also documented in the observed runoff series. However, below average 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 Year (AD) Figure 6.7 Observed, estimated and reconstructed summer runoff indices for Bella Coola River above Burnt Bridge Creek. A) Observed runoff (1948-1983 - solid line) is used to construct the equation for generating the estimated discharge (dashed line). Circles show extrapolated values. (B) Reconstructed summer runoff using precipitation and temperature transfer functions derived from tree-r i ngs. o 211 inferred runoff during the 1970s and between 1920 and 19^0 does not appear in the observed record or in the indirectly estimated series of figure 6.7a. In addition, average runoff conditions would be assigned to the first two decades of this century based on the tree-ring derived index when it appears summer runoff was below average. Part of this discrepancy is based in methodology and part is due to response characteristics of the trees. Modeling of the precipitation index using tree-data results in a reduction of the importance of extreme departures in the original record and hence, average departures only are documented in that series. This bias extends back to the original detrending functions used to standardize each tree specimen. Detrending and chronology development are based on the entire ring series and both are oriented towards major long-term departures. Since runoff is related to spatial as well as temporal variations in snowcover and runoff-generating mechanisms during each year the regionally-averaged climate variables remain imperfect indices and the tree-ring estimates even more so. In addition to methodological difficulties, the trees themselves appear to be associated with strongly persistent trends (cf. figure 6.3) which can only partly be explained by climate fluctuations. When both species show strong positive departures in growth (e.g. 1960s) the system also appears to respond in a predictable manner (e.g. high summer runoff) but mixed responses, such as average or opposite growth trends, do not always relate to specific system effects of a predictable nature. Additional variance related to site factors or changes in temperature and precipitation over the long-term probably influence the growth patterns substantially. The modeling techniques adopted here, which are also customarily used be others, are only basic steps towards understanding a more complex interaction between tree growth and climate. 212 Rainstorm and Snowmelt Flooding Analysis of climatic departures in the post-1945 period (chapter 3-0) indicated that the interval of above average winter precipitation and higher frequency of autumn and spring floods (1958-1976) was also associated with summers characterized by positive temperature anomalies. Synoptic circulation regimes which yield increased winter season precipitation and thus above average spring snowpacks are also associated with a greater potential for high-magnitude autumn storms. Above-average summer temperatures can have two effects: (1) positive, early-season temperatures (May-June) control snowmelt rates and the potential for spring flooding and (2) late-season temperatures (July-August) control ice melt rates which can contribute significantly to rainfall storm-runoff in late August and early September. Therefore, the intraannual persistence of above average winter precipitation and summer temperatures appears to be associated with an increased potential for more frequent and higher magnitude flooding. An assessment of the 20th century variability of these intraannual trends was made by combining summer temperature and winter precipitation data from the Bella Coola climate station and contouring positive climatic departures. The contoured surface in figure 6.8 demonstrates that the most persistent winter/summer departures occurred after 1958. This same intervals is characterized by much higher frequencies of flood runoff during both the autumn and spring (recall that pre-1948 flows were not measured systematically so the threshold for considering an event as a flood is rather high). The high summer temperatures and wet winters are particularly evident between 1964 and 1968, an interval with four of the largest spring and fall flood events. Prior to 1933 the intra- and 213 ooniDmzD-ozii c_ ID cn o ^ a cz c m ci o m r" CD ~U —I < o DOUGLAS FIR ALPINE FIR Figure 6.8 Inter- and intraannual persistence of wet winters and warm summers at Bella Coola. Winter months (September to April) are contoured positive departures in monthly precipitation (>0.50cr). Summer months (May to August) are positive departures in monthly temperature (>0.50a). Standardized departures were calculated using the 1904-1983 mean for each month. Contouring interval is 0.50a . Horizontal lines are associated with measured and reported spring snowmelt (4 ) , autumn rainfall (< ) and autumn rain-on-snow floods (<£=). Vertical boxes on the right indicate intervals of persistent positive departures (dark shading is strongly postive and weak shading is marginally positive) in alpine fir and Douglas fir tree-growth. 214 interannual persistence of positive departures is lower and is reflected in the reduced frequency of reported (pre-instrumental) flooding along the r iver (f igure 6.8). It appears that the intraannual persistence of climatic anomalies which maximize water yield from the basin are closely related to the potential for flooding. Well above average summer temperatures or winter precipitation alone do not necessarily lead to significant water yield or a high potential for flooding. Prior to the 1930s persistent climatic anomalies were rare, flood occurrences low and thus the potential for important geomorphological changes reduced. Although the pre-1958 period is characterized by less persistent climatic trends the years between 1933 and 1940 exhibit greater continuity in climate departures and more significant occurrences of flood events. This corresponds with a peak in global warming (McGuirk, 1982). In terms of 20th century climate the interval between 1958 and present appears to be the most significant in terms of interannual variability and intraannual persistence of flood generating conditions. The tree-ring record again provides some corroborating evidence for these trends (figure 6.8). The comparison is not exact because only a single climate station is shown (Bella Coola) but the interval of greatest flood frequency corresponds with vigorous growth in both Douglas fir and subalpine fir. There is also a significant correspondence between late winter and spring flood years and extreme growth responses in the two spec i es. The patterns revealed in figures 6.7a and 6.8 suggest that the flood frequency curves presented in chapter 3-0 would probably be shifted downwards if the same family of curves could be constructed for the interval 1904-1948. This is due to the lower frequency of reported and 215 inferred floods of moderate and high-magnitude for the interval prior to 1948. (6.6) Summary and Conclusions The annual record of growth in Douglas fir and subalpine fir reveal the following. Douglas fir growth shows a strong regional signal related to moisture availability and can be used as an index of low frequency variance in annual precipitation representative of the interior or eastern portion of the Bella Coola basin. Transfer functions cannot be constructed with strong statistical confidence because of the high standard error involved in calibration. Subalpine fir growth is related more strongly to summer temperature variations but is also influenced by the degree of snowcover. A better statistical association exists between subalpine fir and temperature than between Douglas fir and precipitation. There are positive responses in growth of both species during years characterized by strong intraannual persistence of warm/wet conditions. Extreme growth responses tend to cluster and are shown here to be associated with the intraannual persistence of extreme climate. The combined records of inferred temperature and precipitation do not provide a good index of runoff variations over the 20th century. The techniques employed here are in wide use today and the evidence indicates that both extreme responses and a longer term persistence are characteristic of these species. As a result, working backward form the record to the hydrometeorologcial antecedents becomes difficult. Inferences about climate near the Bella Coola basin using this form of proxy data can provide qualitative estimates of low-frequency variations in temperature and precipitation. Evidence from other studies indicates that more than one 216 species needs to be considered if more precise interpretations temperature variability are to be made. 217 CHAPTER VI I Inferences About Late Neoglacial Environments (7.0) Introduction Several of the potentially useful methods for assessing environmental change are extended here for the purpose of identifying Late Neoglacial environments of the Bella Coola basin. Specifically, biogeophysical data are considered as major indirect data sources. These include paleo-glacial evidence, tree-ring data, lake sedimentation rates and floodplain development. In order to utilize these data it is necessary to consider further the response characteristics of each attribute and potential errors involved with the application of quantitative or semi-quantitative statistical models. These sources for error are apt to have an important impact on the quality of inferences made. Hence, the chapter will demonstrate the resolution of reconstruction methods on the relatively short time scale of only a few centuries. (7-1) Inferences of Environmental Change from Former Glacier Extent Glacier mass balance and recent glacier front fluctuations indicate that the extent of glaciers represent prevailing temperature and precipitation regimes and, therefore, are potentially useful indicators of environmental change during the late Neoglacial period. Since direct measurement of glacier mass balance in the past is not possible, several indirect measures are evaluated here using glaciological and geomorphological evidence preserved in the Bella Coola River basin. All of these methods are associated with estimating the altitude of former and present climatic snowlines, the difference of which is generally considered to be representative of climatic change. Difficulties that arise concern 218 errors of estimating elevations of contemporary and former snowlines and in determining the relevant climatic controls. Glacial equilibrium line altitude (ELA), or the mean elevation of the zone which separates the accumulation and ablation areas on a glacier, represents the local climatic snowline and can be estimated easily for both contemporary and former glacier positions (Bradley, 1975)• Persistent trends in temperature and precipitation are known to affect the mean position of the equilibrium line (Dorrer and Wendler, 1976; Porter, 1977; Whillans, 198l) and the spatial distribution of former ELAs can be used to infer the magnitude and sources of precipitation (Miller et aj. 1975; Leonard, 1984). In addition to ELA estimates, other methods for determining the position of the climatic snowline during glacial maxima can be tested. These include: (1) the altitude of cirque floors, (2) the glaciation threshold and (3) the altitude of lateral moraine deposits. Prior to a discussion of these measures and sources of error the climatic implications of a changing snowline are considered. Snowline Fluctuations and Climate Change ELA data from Sentinel, Place and Bridge Glaciers were used to estimate annual changes in the climatic snowline for a transect across the Coast Mountains of southwestern British Columbia. Summer temperature and winter precipitation were noted in chapter h as important factors influencing the glacier net mass balance. ELA data were used as independent variables against winter precipitation and summer temperature indices in order to construct a set of linear models relating climate and snowline fluctuations. Although the true functional form of these relationships has been reversed (i.e. ELA is actually dependent on climatic variations) the 219 resulting transfer functions provide a means for assessing the magnitude of departure in temperature and precipitation. Linear statistical models appear to perform as well as certain physically based models (cf. Tangborn, 1980; Kuhn, 1981). Plots of relative ELAs versus temperature and precipitation indices for the three glaciers are given in figure 7.1. In all three cases a significant relationship exists between ELA and winter precipitation while only at Bridge Glacier does summer temperature show a significant association with equilibrium line altitude. Hence, during low snowpack years with above average summer temperature ELAs are significantly higher. Sentinel Glacier is located in a zone of extremely high annual snowfall and is the only glacier to exhibit a persistent positive mass balance for most of the years in which measurements were made (1965-present)• Summer temperature indices seem to explain only a minor amount of the variance related to fluctuations of the ELA. Towards the drier eastern side of the Coast Mountains where winter precipitation is 20-80% lower and summer cloud cover is less persistent, summer temperatures appear to have a greater influence. For a 100 m change in ELA a 20% change in winter precipitation occurs at Sentinel Glacier while only 12 and 11% differences in winter precipitation are required at Bridge and Place Glaciers respectively. The apparent higher sensitivity at the latter two sites may be related to two factors: (1) significance of extreme departures in summer temperature and (2) glacier hypsometry. When ELA data from summers characterized by extreme positive and negative departures in temperature are removed, the sensitivity of the ELA to changes in precipitation declines to 22% for Bridge and 13% for Place Glacier. Unlike Sentinel Glacier, which exhibits a fairly uniform hypsometry through the zone in which the equilibrium line fluctuates (1600-2080 m), Bridge and Place Glacier ELA zones are characterized by a fairly 220 RELATIVE ELA (m) Figure 7-1 Relation between relative ELA on three glaciers in southwestern British Columbia and winter precipitation indices (# and solid line) / summer temperature indices (O and dashed line). Significant relationships only are shown. Climate indices are standardized departures from grouped station means (chapter 3)-221 low gradient icefield and a steep icefall, respectively. Hence, the position of the ELA (calculated by identifying the area of accumulation as a proportion of the entire glacier area) is more sensitive to a given change in precipitation on Place Glacier (ice-fall) than Bridge Glacier (ice-field). This would suggest that the Bridge Glacier relationship is "under-sensitive" and the Place Glacier function is "over-sensitive". Given the stability of the relationship at Sentinel Glacier, and the fact that the slope of the linear function falls in between the two estimates derived from Place and Bridge, a value of 20% is adopted as an estimate of the change in winter precipitation required for a 100 m shift in the ELA. During summers characterized by extreme temperature departures and for glaciers in which the accumulation/ablation zone boundary occurs in a steep ice-fall region this value is likely to be larger - i.e. more than a 20% change in precipitation is required. For an average winter precipitation of 1200-1500 mm (eg. Garibaldi) a 20% change equates to a precipitation increase or decrease of between 240 and 300 mm. The difficulties encountered in this analysis suggest that glaciers with odd hypsometries (cf. Bridge and Place Glaciers), particularly in the zone within which the ELA fluctuates, should not be used for reconstructions. Glacier aspect is also likely to be an important factor in determining site-specific fluctuations of the ELA. For a consistent basinwide or regional indication, presumably only one aspect should be used. Estimates of the Contemporary Climatic Snowline Since mass balance measurements have not been made in the Bella Coola basin, several indirect lines of evidence must be used to derive estimates of changes in the climatic snowline. Three methods are used here to 222 determine the contemporary snowline: (1) mapping of the late ablation season snowline from vertical and oblique air photographs, (2) using the height of the mean annual 0°C isotherm and (3) identifying the ELA on several contemporary glaciers using an inferred value of the accumulation area ratio (AAR). Airphoto coverage of the Bella Coola basin for the late ablation season (Sept.-Oct.) is limited, but government photography taken in September 1944 and 1961, August 1979. and oblique photos taken in September and October 1984 provide some indication. Although there is substantial spatial variability, east and north facing slopes of moderate steepness yield a mean estimate of 1870 ± 75 m. The mean value declines to less than 1800 m on the windward side of the Coast Mountains. The contemporary climatic snowline can also be estimated by the height of the 0°C isotherm averaged throughout the accumulation and ablation seasons. The nearest station with upper atmospheric measurements is Port Hardy on the north tip of Vancouver Island, 210 km to the southwest (figure 2.1). The 30 year annual mean (1951-1980) height of the 0°C isotherm is 1812 m. It was anticipated that the annual 0°C isotherm would be biased downwards if winter temperatures were substantially below zero. However, mean daily surface temperatures during the winter months at most central coast stations are above zero. The snowline estimate from Port Hardy is a considerably higher than the value of 1550 m obtained using the moist adiabatic lapse rate of 0.6°C/100 m and station elevations from the western Chilcotin. The uncertainty of the annual average lapse rate suggests that the former estimate is more representative. The ELA of contemporary glaciers can be estimated by identifying the elevation of a line which places a given proportion of the glacier in the accumulation area. An accumulation area ratio (AAR) of O.65 has been used as an initial estimate (Meir and Post, 19&2; Miller et al., 1975; 223 Meierding, 1982; Hawkins, 1985), but as Leonard (1984) indicates the hypsometry of the glacier surface can have an important influence on ELA location. AARs on icefield-outlet type glaciers tend to be higher because of the increased proportional area at higher elevations while piedmont glaciers have a greater proportion of ice in the ablation area. The hypsometry of Bridge, Place and Sentinel glaciers yield period-of-record average AARs of 0.50, 0.29 and 0.72, respectively. In recognition of these limitations two groups of glaciers were selected for measurement. The analysis presented in Appendix F.l yields a mean ELA estimate of 1836 ± 206 m for glaciers of regular hypsometry in the eastern catchment. In summary the three methods for estimating the modern climatic snowline for the eastern portion of the catchment yield relatively similar results (1870 ± 75 m, 18.12 ± I85 m and 1836 ± 206 m) , although considerable spatial and temporal variance is involved depending on the technique and sampling strategy. Estimates of Former Snowlines and Inferences of Hydrologic Change Several methods were used to determine the elevation of the climatic snowline in the Bella Coola basin during the late Neoglacial maximum. These include the glaciation limit, cirque floor altitudes, the elevation of the highest point on lateral moraine deposits, and reconstruction of paleo-ELAs during former glacier advances (cf. Meierding, 1982; Leonard, 1984; Sutherland, 1984). A description of the sampling design and errors associated with each method is given in Appendix F.2 and the results from the analysis are illustrated in figure 7*2. Ostrem (1966) proposed that the average elevation between the summit of the lowest mountain with a glacier and the highest peak capable of snow (ice) accumulation without a glacier in the region (;'.e. glaciation limit) 2400 -2200-(SS) (•) glaciation limit • cirque floor altitude • lateral moraine altitude • equilibrium line altitude 1 - present 2 - former 2000- CS) AP ELA 1800-" 1600 -1400-E W n s east basin west basin north facing south facing 1200-Figure 7»2 Comparison of estimates for contemporary snowline (horizontal dashed lines) and paleoclimatic snowline in the Bella Coola basin. Modern snowline estimates are from air photographs (AP), equilibrium line altitudes (ELA) and isotherm elevation (1). Points are mean estimates of former snowline made from morphological evidence (see text for discussion). Vertical lines are one standard error about the mean. Measurements were made in the east basin (E), west basin (W) and in north (n) and south (s) facing catchments. 225 was climatically controlled. Glaciation limit values were derived for the Bella Coola area and compared with the small scale regional map for southwestern British Columbia (Ostrem, 1966) . Values plotted in figure ~].2 range from 2000 m in the western most basin to 23^0 m in the lee of the Coast Mountains, somewhat higher than those suggested by Ostrem. It has been documented that the glaciation limit is typically 200 to 300 m higher than the inferred climatic snowline (Miller et al., 1975; Porter, 1977)' Hence, the strong east-west gradient revealed here and the subjectivity involved in computing the snowline index result in a large margin of error. Cirque floor altitudes are thought to represent the lowest level of glacial erosion during maximum ice advance. All cirques with characteristics indicative of recent ice-occupation and subsequent withdrawal (e.g. fresh trim line, little or no soil development, lichen free debris) in the eastern Bella Coola basin were sampled (n = 25). Aspects between 270° and 90° only were considered. An estimate of the mean cirque floor elevation is 1610 ± 110 m (figure 1.2). Since it is not possible to ensure that cirques formed during the late Pleistocene are excluded from the sample, the mean cirque floor altitude may be biased towards a value lower than the actual Little Ice Age climatic snowline. The maximum altitude of lateral moraine deposits is known to coincide with the boundary between accumulation and ablation zones of the glacier system and is thought to be an approximation of the climatic snowline (Meierding, 1982) . From a total sample size of 36 sites, the maximum altitude of lateral moraine deposits varies from 1320 ± 150 m in the west to 1980 ± 100 m in south-facing basins of the eastern catchment (figure 7'2). North facing catchments in the eastern basin yield the most reasonable estimate of 1620 ± 110 m. 226 Estimates of former ELAs were made for the same set of glaciers used to estimate contemporary equilibrium line heights. The same steady-state AAR values were also used for each glacier. Moraines and erosion trim lines mark former glacier boundaries and were used for contouring paleo-glacier surfaces. From these analyses a mean ELA of 1733 ± 75 m was determined (figure 7-2). This is 103 m lower than the average contemporary ELA - a significant difference when tested at a significance level of 0.10 (t = 2.16) . The estimates of former ELA indicate a high spatial variability in the climatic snowline. The geographically and physically consistent method used for assessing ELA depression during the Little Ice Age suggests a decline of approximately 100 m. However, even in an attempt to control for the strong east-west climate gradient and the influence of basin aspect, the change may have been as much as 250 m (figure 7-2) . Results given in Kuhn (1981) and Tangborn (I98O) would indicate that the former value is the more realistic estimate of possible snowline depression between the contemporary and Little Ice Age environments in the Bella Coola basin. Using the relationships developed for glaciers in southwestern British Columbia the climatic implication is an increase in winter precipitation during the Little Ice Age of around 20% given that precipitation is the only major factor controlling snowline variations. Concurrent decreases in summer temperature would reduce these estimates but the evidence indicates fluctuations in summer temperature, at least for the moister southwest region, are not as important. A 0.6 C° change in mean summer temperature only (1965-1984 average) would have a comparable effect on snowline elevation for sites in the drier interior. However, temperatures appear to be less important in the fluctuations of mass 227 balance and glacier snout changes at most sites in southwestern British Columbia (see figure 7-1) • Standard errors in the climate-snowline relationships and in the determination of former snowlines are sufficiently high to make these conclusions tentative at best. In addition, there is some risk in assuming that the short-term climate-ELA relationship is applicable to changes in the mean or steady state characteristics of the system. However, the methodologies do provide some indication of the direction and magnitude of departure and are in rough agreement with those given in Smith and Budd (1981) and Tangborn (1980). Little Ice Age Glacial Chronology Glacier fluctuations in the basin were examined in order to document responses of glaciers to changing hydrological conditions over the last several centuries and to provide additional evidence for the timing of these changes. In total 17 glaciers were investigated ranging in size from 2 small cirque glaciers (< 2.0 km ) to major outlet valley glaciers (> 2 35km )• Dating methods were primarily based on dendrochronology with lichenometry and soil development used as additional indicators to verify tree-ring derived ages. Details of site selection, sampling locations and methods are given in Appendix F.3 A major limitation to accurate tree-ring derived age estimates of moraines is associated with the colonization period, or ecesis interval. This is the elapsed time between stabilization of the moraine surface and establishment of tree seedlings. To document the probable colonization period, aerial photographs taken at different times between 1948 and 1982 were used to delimit recently exposed ice-marginal areas at several sites. A ground survey for the frequency and size of tree-seedlings in these areas 228 was then conducted to assess the lag time. Except where the material comprising moraines was unusually coarse, the ecesis interval at these sites averaged between 15 and 20 years (Appendix F.3). Hence, for most sites an ecesis value of 20 years was used. Where moraine materials were unusually coarse, a value of 30 years was assumed. Additional sources of error are related to sampling methods which are discussed in Appendix F.3-An average estimate of error associated with the dendro-derived ages is ± 10 years but it may be larger at higher elevation sites, where very blocky debris is characteristic of the moraine and where valley winds are persistently strong. In most instances, a single terminal or lateral moraine was the dominant depositional feature. Recessional moraines are generally not very well developed or are entirely absent, indicative of rapid and uniform retreat from Little Ice Age maxima. The frequency versus age distribution of terminal and recessional moraines from all 17 sites investigated hence is illustrated in figure 7-3a. The majority of terminal moraines were colonized between i860 and 1895 while most recessional moraines formed after 1885• The dating methods adopted here indicate that no moraines formed prior to the first half of the 19th century. The comparatively limited lichen growth on the older surfaces and the absence of soil development on all but the lowest elevation sites, would support this observat i on. Two sets of recessional moraines appear to have formed during the overall retreat of moderate and large sized glaciers: one set in the first decade of the 20th century, and a second set between 1930 and 19^0. It is most likely that these moraines are indicative of brief pauses in ice retreat caused by a return to wetter and/or cooler climates. Analyses by McGuirk (1982) and Karanka (I986) indicate that cooler and wetter 229 Study Area ^Coast™ Southern Canadian Mountains Rocky Mountaxns Number of Moraines ,„ „ 0 2 4 6 8 10 2 4 6 8 10 0 5 10 15 1950 -i—i—i—i—i—i rrp—|—i—i—i 1900 1850 < OC < 1800 -1750-1700-1650 A J I L B J I l I l • maximum moraines recessional or other moraines Figure 7-3 Dendro-dated, late Neoglacial terminal and recessional moraines in selected regions of the Canadian Cordillera. A) this study (see Appendix F.2 for locations). B) data given in Ryder et al. (unpublished). C) data from Luckman (1986) for Main Ranges of the Southern Canadian Rocky Mountains. 230 conditions prevailed from I89O to 1905 for most of northern Pacific coast, which may have slowed retreat from mid and late 19th century maxima. Similarly, evidence presented in chapter 3 demonstrates a return to slightly cooler temperatures after 1930 in the interior region and wetter conditions in the coastal region. The chronology in figure 1.3a agrees reasonably well with the Little Ice Age chronology derived for several sites in the Coast Mountains to the south of the Bella Coola basin (figure 7-4b) (Ryder et al., unpublished; Mathews, 1951)' Ryder et al. found a small number of l8th century moraines at Tiedemann and Franklin Glaciers, while Mathews (1951) found evidence for older moraines at Helm, Lava, Sphinx and Warren Glaciers in Garibaldi Park. These few sites are the only ones which indicate retreat from Little Ice Age maxima prior to the 19th century. By comparison, there is a higher frequency of moraines in the Canadian Rocky Mountains which were stabilized in the 17th and l8th centuries (figure 7 .he) (Luckman, 1986) . In part, the distribution of moraines shown in figure 7-3a will reflect the timing of linear changes in glacier margins as a function of glacier size and valley position rather than a synchronous response to a common change in climate. The absence of moraines older than the early 19th century demonstrate that the Little Ice Age advance was the most severe excursion of glaciers during the whole of the Holocene Epoch in the basin. Attendant changes in other biogeophysical attributes would also be expected to have responded significantly to these anomalous perturbations. (7•2) Long-Term Low and High Resolution Biogeophysical Evidence Long-term records of growth variations in Douglas and subalpine fir were compiled using the tree-ring chronologies given in Appendix E. These records are compared here with sedimentation rates in Ape Lake and glacial 231 moraine chronologies discussed previously. The shortest site-chronology for each of the tree species limits the interval of consideration which in the case of Douglas fir is post-1650 and for subalpine fir post-1740. Neither of these records, nor that of lake sedimentation rates, is very long but they appear to overlap with the interval of substantial glacial changes in the Bella Coola watershed. Fluctuations in Average Temperature and Precipitation The final tree-ring climate model presented in table 6.3 (see chapter 6) for Douglas fir is used to estimate long-term variability in annual precipitation for the central interior. Since the linear transfer function for Douglas fir does not incorporate extreme climatic departures and all modeling attempts yielded relatively high standard errors, inferences regarding changes in annual precipitation are necessarily restricted to identifying period-average trends about a long-term mean. A plot of inferred trends, smoothed using a five-year weighted moving average, is presented in figure 7'4a. With the exception of the interval 1850-1875 it appears that most of the 19th century was characterized by above average precipitation particularly in the first and last decades. The record prior to 1800 indicates less persistent trend, fluctuating above and below the mean although the variance during most of the l8th century was lower than that inferred for the last half of the 17th century and first half of the 19th century. Variance in tree-growth is generally higher than that observed during the calibration period (post-1930), except between 1700 and 1800 AD, leading to extrapolations of estimated precipitation. While there is less statistical confidence in the extrapolated values we may suppose that they are representative of generalized trends in annual precipitation. I—i—i—I—I—T—I—I—I—i—|—i—i—i—i—|—i—i—i—i—|—i—i—i—i—|—i—I—i—i—|—«—i—i—i—| 1650 1700 1750 1800 1850 1900 1950 2000 YEAR (AD) Figure J.k Long-term departures in reconstructed winter precipitation (A) and the first PC score of subalpine fir (B). Original data are smoothed with a five-year weighted mean. Generalized departures in precipitation for the lower Columbia basin in Washington State (c) are from Graumlich (1987) and for temperature departures inferred from various high elevation sites in the Cascade Mountains of Washington State (d) are from Graumlich and Brubaker (1986) . 233 Using growth variations from four drought sensitive species Graumlich (1987) identified several wet and dry periods for the Columbia River basin east of the Cascade divide in Washington State. Prominent drought periods were from I87O to 1890, the 1840s, 1770 to 1780 and around 1680. Three dominant wet periods were inferred from 1810 to 18^5• 1780 to 1790 and 1690 to 1715- Comparison of this record with the reconstruction given in figure 7.4a reveals a lack of correspondence between the two records. This is consistent with the observations made in chapter 3 on spatial variations of precipitation for the north and south coastal areas of British Columbia and with the conclusions given by Namias (1978) and Yarnel (I985) in their analysis of recent precipitation variability. Over the long-term then, precipitation in the Bella Coola basin may be more closely linked with trends along the north coast. Fluctuations in summer temperature inferred from growth variations in subalpine fir show more persistent departures than precipitation over the last 240 years (figure 7.4b). Below average temperatures dominated the last half of the 19th century preceded by average temperatures from I83O to I85O. Above average temperatures characterized the interval I76O to 1800 while average to below average summer temperatures were inferred between 1730 and 1750. Again, variance in growth over the calibration period (post-1930) does not cover the entire range of variance throughout the whole record. In particular, positive departures in the late l8th and early 20th centuries were higher. Comparison between long-term temperature reconstructions for Washington State (Graumlich and Brubaker, 1986) and the record given here reveals many similarities (see figure 7>4d) . Most evident is the below average summer energy conditions for much of the 19th century followed by substantial warming which began around the the turn of the century. Unlike 234 the Washington State record, pronounced cooling is inferred for the the Bella Coola region after 1950. The longer tree-ring record from further south shows well below average temperatures were likely between 1695 and 1765: a trend which is evident in the early record from Bella Coola. From the combined record of inferred precipitation and temperature departures several, intervals of persistently warm/dry and cool/wet conditions can be identified. These two hydroclimatic settings lead to glacier recession or expansion respectively, and were shown to be coincident with phases of decreased or increased flood activity in southwestern British Columbia. The inferred (and observed) warmer and drier climate between 1900 and 1950 is the most prevalent departure in the last 330 years. In contrast cooler and wetter conditions dominated much of the 19th century. Near average to below average precipitation was associated with above average temperatures between 1765 and 1810. There are cooler and drier periods, separated by warmer and near average precipitation prior to 1760. Recession of some coastal glaciers from Little Ice Age maxima at the beginning of the 19th century may have been a response to the inferred warming which commenced in 1765- However, increased precipitation and cooling after 1810 produced maximum stillstands among most glaciers in the Bella Coola watershed. Below average precipitation between I85O and 1875 signaled the end of the Little Ice Age and induced the first real trend towards glacier retreat. The response was slowed by continued cool temperatures and a return to above average precipitation until the turn of the century. Brief pauses in the retreat of ice occurred in the early 20th century in response to small increases in precipitation. 235 Inferences From Lake Sedimentation Although the record of variation in sediment yield to Ape Lake is comparatively short, there is strong evidence for declining sedimentation rates over the last 150 years (see figure 5«H)« Comparison between sedimentation rates for the period of direct and indirect drainage to the east basin of Ape Lake can be misleading because of changes in process (hyperpycnal versus homopycnal flow) and changes in the storage capacity of newly uncovered lake area in front of the calving Fyles Glacier. One possibility is to assume that the most recent sedimentation rates in Ape Glacier Bay and the west basin are reasonable analogs for the character of sedimentation in the east basin prior to 1940 - an assumption supported by similarities in texture between the two sets of samples. Comparisons are made here by estimating total annual sediment volumes delivered by meltwater streams using observed mean accumulation rates and 2 associated lake areas. For a combined total area of 0.81 km and a mean 7. deposition rate of 2.55 mm/yr (1952-1984), a total of approximately 2000 m of sediment is deposited in Ape Glacier Bay and the west basin each year. In comparison, average depositional rates of 3-86 mm/yr can be established during the period 1836-1940 when Fyles and Ape Glaciers were discharging 2 directly into the east basin. For an area of 1.3 km this corresponds with a total load of 5020 m^/yr during this earlier period. Since the two glaciers are the primary source of sediment and runoff for both periods, this two and a half times difference is large enough to suggest that declining rates of sedimentation in the east-basin are related not only to changes in intermediate storage capacity and depositional processes, but also to real declines in total sediment yield from a set of retreating glac i ers. 236 Leonard (198l, 1986) found evidence for above average lake sedimentation rates during phases of increased ice extent and initial ice retreat, although the association was not always spatially or temporally consistent. The interval covered by the Ape Lake sedimentation chronology is too short to represent the entire Little Ice Age glacial cycle (approximately 1500-1900 A.D.), but the evidence does tend to support the findings of higher sediment yields during maximum ice extent and initial retreat, followed by a substantial decline. Average and below average rates of sedimentation are superimposed on this long-term trend of declining rates. Above average sedimentation during 1885 to I890, 1863 to 1877 and I836 to 1846 shows some correspondence to inferred above average or increasing summer temperatures and inferred winter precipitation. A similar sequence of above average sedimentation rates in the late 19th century was noted by Leonard (1985) for Hector Lake, Alberta, while the data given in Perkins and Sims (1983) for Skilak Lake, Alaska, also show local increases during these periods. One other piece of evidence sheds some light on changes over the last several centuries. Core 84C6 from the shallow lake bottom at the extreme east end of Ape Lake is capped by 6 cm of massive, fine silt and clay over approximately 170, regularly laminated, silt and clay couplets punctuated with small dropstones. The couplets have the characteristic appearance of true varves and although there is distortion in the core they can be cross-dated, in isolated sections, with the varve chronology from the deeper basin. Based on this correlation, the massive material represents deposition following retreat of Fyles Glacier in 1940-45 to a position behind the terminal morai,ne. A higher than normal sedimentation rate and frequency of dropstones within the sediment occurred during an estimated 25 year period at the end of the l8th century. This suggests extensive calving 237 of Fyles Glacier and increased suspended sediment loads to the lake and is coincident with a short period of increased sedimentation in both Hector Lake and Skilak Lake. High-Frequency Variations in Tree Growth, Lake Sedimentation Rates and Related Hydroclimatic inferences In addition to the long-period trends shown in figure 7.k, the high frequency or interannual variability in the first principal components for Douglas and subalpine firs was also examined. Notable year-to-year variance in Douglas fir growth occurs during the intervals 1950-1960, 1925-1945» 1825-1870, 1790-1800 and 1710-1745, (figure 7-5a). Although not as frequent, similar clusterings of extreme growth departures for subalpine fir occur during 1925-1940, 1860-1875 and 1740-1755 (figure 7-5b). Intervals of reduced annual variability in growth in both species occurred during I96O-I983, 1870-1920, 1800-1820, and 1750-1780. Schulman (1956) and Stokes et al. (1973) have also shown these more persistent but lower variance periods in Douglas fir growth to have occurred between 1660 and I69O AD and between 1470 and 1490 AD . The apparent clustering of extreme positive and negative departures in tree-growth was tested under the null hypothesis that the extremes are randomly distributed. A one-tailed runs test was conducted for both series plotted in figure 7-5' The z-scores for Douglas and subalpine fir were 3-64 and 2.10, respectively. With a = 0.10 the null hypothesis was rejected in both tests indicating a non-random ordering of extreme growth response. Based on the relationships established in chapters 3 and 6 it is most probable that intervals characterized by persistent above or below average departures in growth are dominated by an atmospheric circulation regime with a strong meridional component. High year-to-year variance occurs during periods of pronounced zonal flow when the mean position of the jet £: |—i—i—i—i—|—i—i—i—i—|—i—i—i—i—|—i—i—i—i—|—i—i—i—i—j—i—'—i—i—|—i—i—i—i—| 1650 1700 1750 1800 1850 1900 1950 2000 YEAR (AD) Figure 7.5 Long-term high frequency signals in the first PC score of Douglas fir (A) and subalpine fir (B) . Departures are from a weighted 13-year filtered series. Horizontal line are plus and minus one standard deviation from the mean. 239 stream fluctuates north and south. This would yield more variable precipitation and temperature trends along the central coast of British Columbia. Lamb (1979) has demonstrated that there was minimal strength in zonal circulation between 1790 and 1829 during January for the North Atlantic area which appears to correspond with enhanced winter precipitation recorded in the Interior of British Columbia and declining summer temperatures around the study site. Biasing and Fritts (1976) have inferred the variability of several winter pressure types for the period commencing in 1700 using some of the same tree-ring data analyzed here. They demonstrate that persistent departures of pressure leading to increased precipitation and below average temperatures for the Pacific Northwest occurred between 17^+5 and 1770 and the first and last decades of the 19th century. This further supports the contention that persistent departures in growth are coincident with anomalously high pressure over Alaska and northwest Canada which promotes strong ridging to the west and movement of cyclonic disturbances either into or away from the study basin. High year-to-year variance in growth is demonstrative of zonal circulation. Comparison of sedimentation rates with growth anomalies indicates that extreme positive departures in sedimentation are coincident with extreme negative departures in Douglas fir growth. Low winter precipitation and high summer temperatures appear to be the common factors in these years, particularly in the pre-1885 record. Although some similarities are evident (see figure 7'5)» the geophysical and biological indices retain certain unique characteristics demonstrative of an aggregate climate effect - over individual seasons for sedimentation rates and over one or more growing years for tree-rings. 240 (7-3) Effects on the Bella Coola River Fluctuations in temperature and precipitation as well as the inferred frequency of extreme events are apt to have had an effect on the Bella Coola River. The purpose here is to assess the possibility of extending the long-term record using evidence preserved in the Bella Coola valley. Responses of floodplain vegetation and changes in sedimentation patterns over the last several centuries are used to make these inferences. Floodplain Age The distribution of floodplain surfaces of different ages can provide some indication of the frequency with which floodplain sediments are being recycled and the temporal variability of river channel changes. A combination of data sources was used to derive estimates of floodplain age between the mouth of the Bella Coola River and the Atnarko/Talchako confluence. In order of decreasing importance these include: 1) air photographs and early survey maps, 2) age of resident vegetation, 3) personal communication with long-term residents, 4) historical records, 5) pollen and radiocarbon dating of floodplain sediments and 6) archaeological art i facts. Details of the available maps were augmented and extended by sampling vegetation on several disparate surfaces. Two major difficulties for inferring age using vegetation arise in ensuring that the resident forest cover is first generation growth and in establishing the ecesis interval or colonization period. Some surfaces had been cleared by early settlers (or commercially logged) and then abandoned. Many areas of the floodplain have been logged but most occurred after 1945. and therefore careful cross checking with air photographs assisted in identifying disturbed sites. To facilitate dating, each areal vegetation unit sampled in the field was 241 surveyed for signs of old growth (stumps, deadfall, rootwad hollows, etc.) and then verified, where possible, by checking with long-term residents (or former owners) of the property. Increment cores, tree-disks and ring counts on stumps provided the major sources of information. Alders colonize recently flooded channel zone and floodplain areas, often in the first growing season following the flood. Seedling germination requires only a thin layer of silty-sand (see figure 4.4b). Alders and black cottonwood dominate older and less frequently inundated surfaces in which the ecesis interval for forest communities of this composition is apt to be close to zero. On the oldest surfaces, decadent black cottonwood (100+ yrs), western red cedar, Douglas fir and Sitka spruce prevai1. Determining the age of cottonwood trees was most problematic. Ring boundaries are diffuse and porous, hence difficult to distinguish and the rings which can be identified are not necessarily annual in nature (Everitt, 1968). For these reasons alder and coniferous vegetation were mostly sampled. Where cottonwood was the dominant species, wood stains were applied to the cores to enhance the ring structure. Other species were dated to verify inferred ages. Surveys of alder/cottonwood forests at several sites showed that conifers, of similar age to the surrounding deciduous vegetation, were 2 present, albeit in much lower numbers (average of one tree per 1000 m ) and much smaller sizes. The implication is that the oldest members of a predominantly coniferous forest cover may be original colonizers, thus ecesis intervals may be short. Since the probability of sampling the oldest member decreases with the size of the area to be sampled, the ecesis interval is probably significantly different than zero. The sample surveys made here suggest a maximum colonization period of 25~30 years. However, complete forest succession from deciduous to coniferous may take as long as 242 60-80 years. Figure 1.6 illustrates the percentages of floodplain surfaces of different ages for the Bella Coola River. Four age categories are used in figure ,.G: areas older than 1700 A.D., those areas between 1700-1900 A.D., 1901-1938 and 1938-1982. Approximately 27% of the total floodplain area (excluding the area of the active channel zone) is older than 1700 A.D. and almost 80% of this occurs below the Nusatsum confluence (marked as downstream on figure 7-6).* Approximately half of the floodplain area is estimated to be in the range 1700 to 1900. The data indicate that the younger areas in this category (1800 to 1900) are more frequent in the downstream reaches while the older areas are more frequent upstream. Almost 24% of the floodplain has developed since the beginning of the 20th century. This distribution, in conjunction with other evidence reveals several things. First, inferences about rates of floodplain development over the last several centuries are possible because of the long-term vertical stability of the river. The absence of significant aggradation or degradation makes the available floodplain area approximately constant. Secondly, areas older than 1700 A.D. are in some cases significantly older. Age estimates of buried organics within some shallow overbank deposits are in excess of 1700 years B.P. (see Appendix A; samples S-2698 and S-2730). Over the last 280 years, a proportionally larger area of the floodplain has been constructed during the post-1900 interval. This is manifested in two trends; a decline in the number of anabranches or sloughs occupied by the river and a decrease in the width of the dominant active 1 This estimate is thought to be reasonable because of offsetting errors. In the first instance it may be an underestimate because of the problem of determining if the growth is first generation. Secondly, it may be an overestimate because of difficulties in delimiting the floodplain boundaries. A fairly generous boundary was used where the flood limit was not know precisely. 243 50 -, na < 40-Q. 1 30 o Lf) u a. 20-10 -Upstream Downstream < 1700 1701-1900 1901-1938 1939-1982 Age Class (Year AD) Figure 7-6 Frequency distribution of floodplain age based primarily on estimates from surface vegetation below the Water Survey of Canada gauge above Burnt Bridge Creek. Note that age class sizes are not equal. Upstream and downstream segments are separated by the Nusatsum River confluence. Total Q^QQ floodplain area is approximately 30 km of which 38% is upstream of the Nusatsum River and 62% below. Area estimates are subject to greater error for the two older age classes but errors are probably no larger than ± 5%> 244 channel in at least the stable reaches. Approximately 14% of the floodplain was constructed between 1900 to 1938 and 10% between 1939 and 1979. Differences between the two periods cannot be considered significant but during the latter interval channel avulsion has been less frequent resulting in continued stabilization of early 20th century surfaces and the development of no new channels. Once established, vegetation type and age provide a chronology for channel migration. The succession of vegetation on point bar segments in laterally stable reaches of the river indicates the mean rate of channel migration to be in the range of 3.0 to 3.5 m/yr (Desloges and Church, I987). Channel migration rates in sinuous reaches upstream of the Nusatsum confluence indicate that the replacement or turnover rate for sediment along a unit length of the 0.5 km wide floodplain would be between 145 and 165 years. Once deposited there, floodplain sediments remain undisturbed for periods of the order one to a few centuries. Downstream of the Nusatsum confluence the turnover rate is longer because of the increased valley width but migration rates in stable reaches are comparable. While regular cross-valley migration of the channel is important, the presence of floodplain remnants which date to the previous millenia indicate that over the long-term regular lateral progression does not reach all areas of the floodplain. From the generalized interpretations given here it appears that a significant amount of sediment turnover and floodplain development which began around the turn of the century was commensurate with predominant changes in prevailing hydrology and related sediment transfers. Although a decrease in precipitation and an increase in summer temperature, which became fully apparent by 1920, promoted increased channel stability, the river and associated sediment transfers still respond in a significant way 245 to the higher-magnitude flood events. These events appear not to be exclusively linked to long-term persistent climatic trends. The apparent higher frequency of surfaces which date to the 19th and 20th centuries suggests that the most significant changes to the alluvial environment occurred during and after the peak of the inferred Little Ice Age interval. Evidence for a Long-Term Flood History Sampling of the vegetation in several areas of the floodplain provided the opportunity to assess the effects of flooding on the structure and growth of woody plants. This exercise was undertaken because damage and burial of vegetation along the margins of contemporary alluvial fans provided good evidence of recent (instrumental period) flood events (see chapter 4). There are several effects flooding might have on woody vegetation which survives inundation. A juvenile may be bent and partially buried yielding differential stress on the trunk thereby producing cell damage or reaction wood (Yanosky, 1983)• On older individuals damage to the upstream side of the trunk may produce scarring and localized distortion to the ring-pattern. Finally, burial of trunks or branches may produce new stems which can then be used to date the event. Damage patterns then will depend on the age and species of tree and the position of the site with respect to flooding. In addition to the cores and stumps used to date the floodplain surface, tree disks and 2 or 3 increment cores/tree (radii) were sampled from other vegetation which was thought to have been affected by flooding. Vegetated channel banks in stable reaches of the river are the preferred sites because the discharge-stage relationship would likely have been constant during the recent past. The sampling design and locations produced 246 mixed results in which only a very small number of consistent dates could be assigned. During the 20th century the effects of the 1936 flood were the most widespread and dramatic. Vegetation sampled on alluvial fans, several areas of the floodplain and even on several hillslopes contained some evidence of scarring or reaction wood from this autumn flood event. Less frequent, but equally obvious at certain sites, was damage during 1924. The only other major event recorded at several sites was a flood either in the autumn of 1805 or the spring of 1806. Less significant damage occurred in I896, 1886, 1826 and 1785- Evidence from earlier in the l8th century is more scarce because of the smaller number of older trees and the frequency of heartrot in the older individuals. Coniferous trees, in particular spruce and cedar, were the most valuable data sources. There was some indication of severe flooding at some sites, particularly from the I98O and 1968 floods, but earlier dates could not be confirmed because either very few individuals responded or adjacent lower-elevation sites, which should have been more susceptible to damage, showed no evidence of flood damage. The somewhat singular response is not so surprising given that flooding can occur locally as a result of log jams and channel avulsion. Some damage was also the product of disturbance from adjacent individuals (falling trees). The attempt to reconstruct a long-term flood chronology using woody vegetation cannot be considered successful but there is an indication that a more rigorous sampling design in several stable reaches of the river might yield more consistent results. Long-Term Floodplain Sedimentation Examination of recent floodplain sediments in back channels and sloughs was undertaken to assist in interpretations of older floodplain 247 sequences. Results presented earlier demonstrate that floodplain sedimentation is episodic, dependent on channel configuration near the site and associated with a variety of depositional processes. Pollen and spores are incorporated in a variety of ways and represent airborne, riverborne and local sources. For these reasons only a selected number of sites, in which vertical accretion was thought to be the dominant mode of deposition, were examined for variations in stratigraphy, mineralogy, grain size variations and pollen concentrations. Site selection was based on the distribution of inferred floodplain age (see figure 7-6) and inspection of sediments from boreholes in each of the cross-valley transects. Seven sites were selected but after examination of grain size and pollen data four were thought to have formed by lateral accretion of the river channel and thus were less likely to be of any significance. All three remaining sites were in stable reaches of the river (locations are plotted in Appendix figure C.1). Only one of the remaining three sites is discussed here but reference is made to two others. The Brekke site is on the upstream side of the Salloomt River fan in a confined portion of the Bella Coola valley (figure C.1). Due to river training works on the south side of the valley, Bella Coola River is currently undercutting a portion of the floodplain which has been stable for at least the last 70 years. Bankfull width of the river here (256 m) accounts for approximately 37% of the maximum valley width and, given that the height of the bank above mean summer flow is only 1.1 m, it is probable that overbank flows are frequent. A monolith was extracted from just behind the currently eroded bank to determine the physical and organic properties of the sediment. Figure 7-7 is a summary of these characteristics. Grain sizes are mostly fine sand and silt with a minor amount of clay indicative of a low energy depositional environment. Charcoal is found in • Arboreal 1 Non Arboreal Brekke m g Figure 7-7 content and the Brekke s Stratigraphy, grain size variations, organic matter pollen concentrations in vertical accretion sediments ite on the Bella Coola floodplain. of NJ 00 2k3 three distinct horizons and the lowest is the thickest containing whole log fragments. The ring structure is that of cedar. Below 110 cm there is medium sand with faint bedding structures which in turn overlies gravel at 190 cm. The summer water table is coincident with the transition between fine and medium sand. The original surface vegetation of mature cedar and spruce was logged in the 1920s. From ring counts on standing stumps an estimate of the age of the surface is 220 ± 20 yrs B.P. and a radiocarbon date on the lowest charcoal layer in the sequence was 955 ± 120 radiocarbon yrs B.P. Based on pollen concentrations four zones can be assigned (figure 1.1). Zone I is lateral accretion sediment and contains no pollen. From the micro topography of the site it is possible that this was a former slough or back channel and the basal sands represent the bar top facies. Zone II is layered silts and fine sands with three distinct charcoal horizons. The lowest horizon can be traced laterally for at least 15 m in all directions indicating that it was an in situ forest burn. Fragments of charcoal constitute the other two horizons but they are also found at consistent depths in the surrounding terrain. Organic matter content between the organic layers is low. Pollen in zone II is dominantly alder, grasses and ferns with most of the floodplain conifers represented as well. The dominance of alder pollen in the lower 10 cm of this zone suggests that the site was a recently active slough channel or else very near to frequently flooded surfaces of the main channel. Above this zone the three charcoal horizons and abundance of successional forest pollen types (spruce/cedar) indicate that overbank flooding (vertical accretion) occurred episodically. Although textures do not change, zone III is marked by a substantial increase in spruce and other conifer pollen and a decrease in the non-250 arboreals. This shift may represent accretion of the floodplain to its final level. The increased absolute and relative importance of the grasses, ferns and several other species not found in the lower horizons is demonstrative of recent disturbance in zone IV. Human disturbances to this site in the form of logging and cultivation began as early as 1920. Evidence from this site suggests that vertical accretion of the floodplain has proceeded slowly over the last millenium. Even though a persistent channel course along the south side of the valley would reduce the rate of sedimentation, the river is sufficiently wide in relation to the valley width that it is unlikely that the larger flows did not go overbank at this site. There are no structural or mineralogical indications that the site has undergone continual scour and fill. A recent accumulation of 3-5 cm of medium sands can be attributed to post-1960 overbank flows. Inferences about long-term rates of sedimentation can be made given the following assumptions: 1) the top 3-5 cm of deposition is related to floods between 1965 and 1980; 2) significant changes in the pollen assemblage above 15 cm is related to forest clearance around 1920 AD: 3) sediments deposited above the upper most charcoal layer (i.e. top hO cm of sediment) represent flood events which post-date the establishment of the mature, 220 year old, cedar/spruce forest (rooting depths of the trees generally correspond with this horizon which further supports the claim); h) the principle channel of Bella Coola was not persistently entrenched on the south side of the valley and 5) no erosional hiatuses exist between dated horizons. Under these assumptions the rate of vertical accretion averages 18 mm/yr over the last 20 years, 19 mm/yr over the last 80 years and around 18 mm/yr over the last 220. The radiocarbon dated horizon of 955 ± 120 years (figure 7«7) corresponds to a wider range in calendar dates due to increased 251 fluctuations in atmospheric carbon content around this period (cf. Stuiver, 1982). A minimum estimate for the age of the lower horizon is 1210 AD or approximately 550 years older than the upper most organic layer. For a total accumulation of 43 cm between charcoal horizons this relates to a maximum estimate of the mean depositional rate of 8 mm/yr or half the rate of the recent sedimentation. Using mean and maximum age estimates for the lower horizon of 1030 AD and 910 AD the depositional rate is reduced to 6 and 5 mm/yr, respectively. Dating of older organic horizons at two other sites (Canoe Crossing and Firvale) could not be made with any confidence but again it appears that recent depositional rates, based on pollen changes following land clearance and recent flood sediments, are significantly higher in relation to the total amount of accumulated sediment. Within the limitations of the techniques adopted here a tentative inference would be that floodplain development over the last few centuries, or Little Ice Age interval, has proceeded more rapidly than over the last millenium. This conclusion is supported by inferences made from the distribution of dated floodplain surfaces. Clearly then, the resolution of long-term changes in alluvial sedimentation in the Bella Coola valley is low and limited by the availability of suitable sites. Substantial reworking of the floodplain over the last three centuries has removed much of the evidence and difficulties in establishing long-term dating control are pervasive. Charcoal is an abundant constituent of vertical and lateral accretion sediments but the available fragments have a high potential of being reworked and thus may yield anomalous estimates of age. An even higher frequency of sampling than that adopted here may be required to verify and extend further these limited conclusions. 252 Chapter VI I I SYNTHESIS, CONCLUSIONS AND FUTURE DIRECTIONS The major objectives of this study were: 1) to characterize the response of selected biogeophysical attributes to recent climatic change; 2) to determine the resolution and length of paleoenvironmental records in the study area and; 3) to ascertain the regional significance of inferred pre-instrument environments. The overriding goal is to establish the basis for inferences about past environments. Paleohydrological studies of individual drainage basins in coastal British Columbia have not been made, yet there is increasing evidence to suggest that such studies provide a better understanding of integrated physical processes (cf. Helley and LaMarche (1973) and Coghlan (1984) for California studies). This study was an attempt to determine the feasibility of systematically documenting recent environmental change on a spatial scale commensurate with factors that are apt to control regional climate and hydrology. It was recognized that the nature of results from the first two objectives would influence what could be said about long-term environmental history. The following discussion is divided into three sections: a summary with conclusions about the methodological implications of the study and characteristics of the response records; a discussion of inferred Little Ice Age environments of the study area; and directions for future research. (8.1) Methodological Implications and Biogeophysical Response Records Paleohydrological models are based commonly on interpretation of cause and effect relationships in an environmental system containing both hydroclimatic and biogeophysical components, such as that defined in figure 1.1. The attempt here was to consider how functional relations between the two components would influence inferences about the hydroclimatic system 253 using indirect and proxy data sources from the biogeophysical system. The strategy was to test for the type and strength of linkages within the entire system over recent time scales - in particular, the post-1945 interval. The intention was to quantify the linkages for retrospective estimation of pre-instrument climate and hydrology using the longer response record. A selection of biological and geophysical attributes was made in recognition of the fact that any single attribute would not provide incontrovertible evidence of changing hydroclimatology. This is because the actual temporal resolution of these elements varies from seasonal to approximately decennial and because the extent to which any attribute provides evidence of regional changes in hydrology was not initially known. It is clear from this study that inferences about hydroclimate are twice removed from the available biogeophysical evidence; one set of relations intervenes between the biogeophysical elements and hydrological controls, and a second between the system hydrology and climatological controls. Severa