SEDIMENTOLOGY AND CHRONOLOGY OF NEOGLACIAL LAKE ALSEK, YUKON TERRITORY by JEFFREY PETER SCHMOK B.A. (Hons.) The University of Ottawa, 1983 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES (Geography) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA September 1986 © JEFFREY PETER SCHMOK, 1986 In p r e s e n t i n g t h i s t h e s i s i n p a r t i a l f u l f i l m e n t of the requirements f o r an advanced degree a t the U n i v e r s i t y o f B r i t i s h Columbia, I agree t h a t the L i b r a r y s h a l l make i t f r e e l y a v a i l a b l e f o r r e f e r e n c e and study. I f u r t h e r agree t h a t p e r m i s s i o n f o r e x t e n s i v e copying o f t h i s t h e s i s f o r s c h o l a r l y purposes may be granted by the head of my department or by h i s or her r e p r e s e n t a t i v e s . I t i s understood t h a t copying or p u b l i c a t i o n of t h i s t h e s i s f o r f i n a n c i a l g a i n s h a l l not be allowed without my w r i t t e n p e r m i s s i o n . Department of The U n i v e r s i t y of B r i t i s h Columbia 1956 Main Mall Vancouver, Canada V6T 1Y3 Date )E-6 (3/81) ABSTRACT Neoglacial Lake Alsek has formed many times during the last 3000 years when Lowell Glacier has blocked the flow of Alsek River. Although the basin is now empty, evidence for former lakes includes: valley-side beaches, driftwood strandlines, ice-rafted erratics, vegetation discontinuities, and unusually large (outburst-flood related) bedforms. The largest fillings covered approximately 463.0 km of the valley bottom and were approximately 28.37 km in volume. The associated outburst floods produce instantaneous peak discharges estimated at 1.09 x 105 m 3 s"1 . Alsek valley sediments were examined for a record of past fillings and drainings. Physical, stratigraphic, and lithofacies analyses were conducted on sediment cores. Five facies assemblages are defined and used to interpret depositional environments: (1) matrix supported diamicton interpreted as a deposit of iceberg-rafted sediment, (2) sands interpreted as tractive current deposits, (3) massive muds interpreted as rapidly deposited lake margin derived sediment, (4) laminated muds interpreted as distal glaciolacustrine deposits, and (5) carbonaceous muds interpreted as eutrophic pond organic detritus deposits. Facies sequence analysis indicates non-randomly ordered sedimentary sequences. Cyclic sedimentation is not indicated. Many occurrences of erosional unconformities indicate either depositional hiatuses or an unknown amount of missing sediment. A single radiocarbon date of 2840 ^ C years B.P. underlies 1.04 m of Lake Alsek sediment. This dated material overlies approximately 1.5 m of sediment presumably associated with Lake Alsek. The potential for absolute dating of the Lake Alsek stratigraphy has been shown to be quite high. ii iii TABLE OF CONTENTS page Abstract ii List of Tables vi List of Figures vii Acknowledgements ix Chapter 1 INTRODUCTION 1 1.1 Study area 1 1.1.1 Physiography 3 1.1.2 Drainage 3 1.1.3 Pre-Quaternary geology 4 1.1.4 Surficial materials 5 1.1.5 Soils 6 1.2 Late Quaternary climatic and glacial history 7 1.3 Ice-dammed lakes 10 1.4 Previous work on Neoglacial Lake Alsek 12 1.5 Outstanding questions 17 1.6 Scope and objectives of this thesis 18 Chapter 2 GEOMETRY OF THE NEOGLACIAL LAKE ALSEK BASIN 19 2.1 Existing information 20 2.2 Extent of Lake Alsek 21 2.3 Alsek Valley levelling survey 23 2.4 Valley cross-sections 25 2.5 Barometric altimetry 25 2.6 Lake Alsek bathymetry 28 iv 2.7 Filling times and outburst flood magnitudes 31 Chapter 3 STRATIGRAPHY AND SEDIMENTOLOGY 37 3.1 Introduction 37 3.1.1 Sediment sources 38 3.1.2 Sediment sinks 41 3.2 Field and laboratory procedures 42 3.2.1 Sampling strategy 43 3.2.2 Sample site descriptions 46 3.2.3 Laboratory procedures 47 3.3 Sedimentary structures 49 3.4 Core lithostratigraphy 54 3.4.1 Lithofacies type description scheme 55 3.4.2 Lithostratigraphy of Lake Alsek cores 57 3.5 Sediment properties 70 3.5.1 Densities, organic matter and moisture content 71 3.5.2 Grain size analysis 73 3.6 Facies assemblages 85 3.6.1 Matrix supported diamicton (facies D) 85 3.6.2 Sands (facies S) 86 3.6.3 Massive muds (facies Fm) 88 3.6.4 Laminated muds (facies FI) 89 3.6.5 Carbonaceous mud (facies O) 90 Chapter 4 FACIES SEQUENCE ANALYSIS 92 4.1 Introduction 92 4.2 Method 93 V 4.3 Results of analysis of Lake Alsek sequences 97 4.4 Summary sequence for Alsek Valley sediments 106 4.5 Correlation and spatial variability 107 4.5.1 Stratigraphic correlations 107 4.5.2 Spatial variability 110 4.6 Discussion of facies sequences 110 Chapter 5 CHRONOLOGY OF NEOGLACIAL LAKE ALSEK EVENTS 113 5.1 Available Holocene dating techniques 113 5.2 Relative age estimates 115 5.3 Radiocarbon result 116 5.4 White River ash 117 5.5 Historical information 118 5.6 Further chronological work 119 Chapter 6 SUMMARY AND CONCLUDING REMARKS 120 6.1 Summary of results 120 6.2 Concluding remarks 121 References Cited 123 Appendices Appendix I Levelling survey results 130 Appendix II Grain size distribution results 134 Appendix III A review of Markov chain analysis 141 Appendix IV MARKOV.S: a facies sequence analysis program 161 Appendix V Stratigraphy of soil pits 169 Appendix VI Modified Livingston piston corer 175 vi LIST OF TABLES page Table 1.1 Recognized phases of Lake Alsek. 16 Table 2.1 Altitude determinations by barometric altimeter. 28 Table 2.2 Lake Alsek morphometric data. 33 Table 2.3 Lake Alsek filling times and outburst flood magnitudes. 33 Table 3.1 Suspended sediment load of Alsek River. 39 Table 3.2 Characteristics of the population of Lake Alsek basin small lakes. 44 Table 3.3 Four part diamict lithofacies code and symbols. 56 Table 3.4 Physical properties of Lake Alsek sediments. 72 Table 3.5 Characteristics of lithofacies groups. 74 Table 3.6 Statistical grain size parameters for 30 Lake Alsek samples. 76 Table 3.7 Folk (1980) verbal equivalents for grain size parameters. 77 Table 3.8 Grain size data summarized by major lithofacies type. 78 Table 4.1 Transition count matrix and expected count matrix for KM4.4. 98 Table 4.2 Summed transition count and expected count matrices for KM13.4C. 102 Table 4.3 Summed transition count and expected count matrices for KM13.4D. 104 Table 4.4 Transition count matrix and expected count matrix for KM50.4. 105 Table 5.1 Available techniques for dating Holocene events. 114 vii LIST OF FIGURES page Figure 1.1 Index map of the study area. 2 Figure 1.2 Quaternary chronology for Kluane National Park. 8 Figure 1.3 Sedimentary sequences exposed in pits at (a) 8 km W of Haines Junction, and (b) near Pine Creek. 14 Figure 2.1 Elevation data of features in the Lake Alsek basin. 22 Figure 2.2 Alsek River longitudinal profile. 26 Figure 2.3 Valley cross-sections looking downriver at (a) KM13.0, and (b) KM53.0. 27 Figure 2.4 Digitized contour map of the Lake Alsek basin. 30 Figure 2.5 Hypsographic curve for the Lake Alsek Basin. 32 Figure 2.6 Reservoir volume for the Lake Alsek basin as a function of elevation. 32 Figure 3.1 Bathymetry and sediment sample locations of the eight sample sites. 45 Figure 3.2 Sedimentary structures: (a) parallel laminations, (b) deformed beds. 50 Figure 3.3 Sedimentary structures: (a) cross-laminations (b) erosional unconformities. 51 Figure 3.4 Massive beds and other sedimentary characteristics. 52 Figure 3.5 Lithostratigraphy of sediment sequences extracted from eight sample sites. 58 Figure 3.6 Sand-silt-clay ternary diagram of Lake Alsek sediments. 75 Figure 3.7 Plot of mean grain size against sorting. 80 Figure 3.8 Plot of mean grain size against skewness. 80 Figure 3.9 Plot of mean grain size against kurtosis. 81 Figure 3.10 Plot of skewness against kurtosis. 81 Figure 3.11 CM diagram of Lake Alsek sediment samples. 84 Figure 4.1 Observed and expected facies transitions for core KM13.4C (34-120). 94 Figure 4.2 Facies relationship diagrams (a) KM4.4 sequence, (b) KM13.4C sequence, (c) KM13.4D sequence, and (d) KM50.4 sequence. 100 VU1 Figure 4.3 Schematic representation of the extent and sedimentary record of Lake Alsek inundations over time. Figure III. 1 Classification of Markov chains according to various combinations of dependence, order, and step length. ix ACKNOWLEDGEMENTS The assistance of many people is required for the success of field oriented graduate research. I wish to gratefully acknowledge Dr. Garry Clarke, Dr. Michael Bovis and Dr. Michael Church for their very capable guidance as my supervisory committee. In particular, I thank my research supervisor, Garry Clarke, for providing categorical support and encouragement in every aspect of this project throughout its duration. The people who shared in the many tasks of field work with energy and bonhomie are especially appreciated: Monty Alford, Sam Collins, Marc Gerin, Francis Jones, Alice Kenney, Paul Langevin, George Mobley, Kelly Nordin, and Tenzing Norquay. I am indebted to Gary Barrett, Bruce Dagg, Joe Desloges, Mike Maxwell, Scott Robeson, Clarence Schmok, and Cathy Souch for many valuable discussions throughout the preparation of this report. For unqualified support and reassurance in other matters, I owe a great deal to Susan Lees and, of course, my parents. Special thanks also go to: Andy and Carol Williams, and the Arctic Institute of North America for invaluable logistical support; Dieter Schreiber for technical advice and construction of equipment; Dr. Les Cwynar for advice on coring procedures and equipment loans; and Ray Frey of Parks Canada for permission to undertake this project in Kluane National Park. This project was supported by a Natural Sciences and Engineering Research Council postgraduate scholarship, and grants from the University of British Columbia Committee on Arctic and Alpine Research. 1 Chapter 1 INTRODUCTION The aim of this study is to investigate the characteristics and chronology of Neoglacial Lake Alsek. The study strategy is to observe and interpret the Alsek Valley sedimentary record so that the Neoglacial history of Lake Alsek can be inferred. This entails a stratigraphically based investigation of the lacustrine sediments deposited within the former glacier-dammed Neoglacial Lake Alsek basin. Reconstruction of the depositional environment is attempted and related to the spatial and temporal variability observed in the sedimentary record. The information gained through this investigation may also provide additional insight into the late Holocene glacial chronology of the region. 1.1 Study Area The study area lies within the Alsek River basin of southwestern Yukon Territory (Fig. 1.1). This study is concerned only with the Neoglacial Lake Alsek basin itself and does not include the area downstream of Lowell Glacier. The area is covered by NTS topographic map sheets Dezadeash (115A) and the eastern portion of Mount St Elias (115B-115C E/2). Portions of the following basins are included in the study area (the bracketed codes are Water Survey of Canada hierarchical basin subdivisions): Lowell Glacier(4*8ABG), Dusty Glacier(4*8ABH), Felsite Glacier(4*8ABI), Disappointment Glacier(4*8ABJ), Maxwell Glacier(4*8ABK), Jarvis River(4*8ABL), East side of Alsek River(4*8ABM), Pine Lake(4*8AAA), Kluhini River(4*8AAG), Mt. Bratnober(4*8AAH), Kathleen River(4*8AAI), and Kaskawulsh Glacier(4*9CAL). At present, Lake Alsek is empty, "but has formed many times during the last 3000 years when Lowell Glacier, a large surging glacier in the St. Elias Mountains, advanced across Alsek Valley and blocked south-flowing Alsek River....The lake extended east from the St. Elias Mountains into Shakwak Valley....At its maximum, Lake Alsek was almost 200 m deep at the glacier dam and over 100 km long, and thus was the largest 2 Figure 1.1 Index map of the study area. Located on this map are the place names referred to in the text, sample site locations, benchmarks established during 1984, and soil pit sites. The position of the Lowell Glacier terminus changes almost yearly; the position shown is from 1974 air photographs. 3 Neoglacial ice-dammed lake in North America, if not in the world." (Clague et al. 1982, p. 3 01). Various filling levels of Lake Alsek during Neoglacial time are indicated by wave cut benches and beach gravels (in many instances with driftwood windrows) found on the valley sides. The most extensive pondings would have inundated approximately 463 km^ of the valley bottom with approximately 28.4 km 3 of water (Chapter 2). This would make it the largest lake in the Yukon Territory (the next largest being Kluane Lake at 409 km )^ and would submerge the Haines Junction townsite and at least 15 km of the Alaska Highway. 1.1.1 Physiography The physiography of the southwestern Yukon has been described in detail by Bostock (1952), Kindle (1952), and Rampton (1981). The major physiographic features of the area are the St. Elias Mountains and the Kluane Plateau, which are separated by a NW-SE trending structural depression termed the Shakwak Trench. Bordering the Shakwak Trench are the Kluane Ranges, which in this region comprise the northern edge of the St. Elias Mountains and which are separated from the rest of St. Elias Mountains by the Duke Depression. At least five crustal blocks, (i.e. tectonostratigraphic terranes) can be identified in the southwestern Yukon, each composed of rocks of greatly differing age and character. A concise chronological summary of the major geologic events associated with these mountains and their physiographic character can be found in Campbell and Rampton (1980, pp.3-17). 1.1.2 Drainage The southwestern Yukon has a complex history of drainage routing and discharge changes. These changes are primarily due to the physiographic evolution of the land surface and to the influence of glacier ice. The Tertiary drainage evolution is discussed by 4 Tempelman-Kluit (1980), and late Quaternary drainage patterns are discussed by Bostock (1969), Johnson (1986), and by Sticht (1951). Except for the Alsek River basin which drains into the Pacific Ocean at Dry Bay, Alaska, the present day drainage of the southwestern Yukon is mainly through the Yukon River basin and into the Bering Sea. In late Tertiary time, however, most of the drainage from the southwest Yukon was probably into the Pacific Ocean via the Alsek Valley. A more subdued regional physiography allowed drainage in this direction until major episodes of late Tertiary uplift (Tempelman-Kluit 1980). In late Quaternary time, the area drained by Alsek River may have been larger than present. Bostock (1969) has suggested that additional drainage from Kluane Lake basin (across the Slims-Kaskawulsh divide) contributed to the Alsek drainage at some time since late Holocene deglaciation. Present-day drainage patterns have existed since late Neoglacial, when an advance of Kaskawulsh Glacier forced the Kluane Lake drainage to be captured by the headwaters of Yukon River. Even if this hypothesis of a major shift in drainage patterns does not prove correct, there have been minor drainage changes at this divide in historical time (Bostock 1969, Johnson 1986, and S. Collins, personal communication). Changes in the internal plumbing of the Kaskawulsh Glacier can influence which side of the divide meltwater will flow. These changes have been observed in the last three decades, and probably occur frequently whenever the Kaskawulsh Glacier terminus is in an advanced position. 1.1.3 Pre-Quaternary Geology The bedrock geology of the area has been described in some detail by Kindle (1952), Bostock (1952), and Campbell and Dodds (1975). Recently, much of this information has been reviewed by Campbell and Rampton (1980, pp.3-17). The study area is underlain by thick sequences of detrital sedimentary rocks, ranging in age from Ordovician to late Cretaceous, which accumulated in deep-water marine environments. At 5 various locations, these rocks have been metamorphosed to varying degrees and intruded by granitic rocks. The entire sequence has undergone many episodes of uplift and erosion. Displacement of crustal blocks along tectonically active fault systems is still occurring. This entire assemblage is often overlain by Tertiary volcanic and sedimentary rocks. In the Alsek Valley, outcrops are predominantly dark coloured mafic rocks such as andesite, basalt and other volcanic rocks. Outcrops of lighter coloured granitic rocks are rare in the Alsek Valley itself, but are found in the Icefield Ranges to the west, where Lowell Glacier originates (Clague and Rampton 1982, p.98). 1.1.4 Surficial Materials In Kluane National Park, which encompasses part of the study area, surficial materials have recently been mapped and described by, Rampton (1981). The study area is overlain by mainly glacial and proglacial sediments deposited during, and since, the late Pleistocene Kluane glaciation. Late Pleistocene ice filled Alsek Valley to an elevation of about 1500 m and flowed north and east into Shakwak Valley. Lineated smooth bedrock ridges, striations, polished rock, and streamlining of virtually all exposed bedrock outcrops indicate glacial erosion. Drift exposures in Alsek Valley were not examined in any detail during this study. Glacial drift was observed in sections along Bear Creek, Archibald Creek, and several other unnamed creeks in the study area, and is probably thickest on the valley floor. Drift thicknesses in Shakwak Valley near Kluane Lake exceed 90 m (300 feet) in places (Denton and Stuiver 1967, p.490). The drift surface morphology is expressed in some locations as hummocky terrain with associated kettle fields, which are often found in areas of ice-contact, stratified drift. Kettle fields are found in the study area about 13 km upstream of Lowell Glacier, and in many places in Shakwak Valley from about 60 km to about 71 km upstream of the Lowell Glacier. Outwash deposits probably aggraded on the valley floors as glaciers ablated, but 6 surface expression of these deposits has been altered by Neoglacial outburst floods from Lake Alsek. During deglaciation, Shakwak Valley was free of ice prior to the valleys in the St Elias Mountains. Ice remaining in Alsek Valley blocked drainage to the Pacific and formed Glacial Lake Champagne, a lake that existed for tens or possibly hundreds of years and deposited thick sequences of glaciolacustrine sediments (Kindle 1952, p. 16). These parallel-bedded and laminated deposits are exposed in many Holocene streamcuts and can be tens of metres thick in Dezadeash Valley and close to 70 m thick in Aishihik Valley (Kindle 1952, p. 16; Hughes et al. 1972, p.9). During regional glacial retreat, gravity-driven geomorphic processes probably underwent a period of heightened activity (Church and Ryder 1972; Johnson 1986). Increased discharge and increased sediment yield during this period involved rapid deposition of alluvial fans, glacial lake deltas, and colluvial materials in the Alsek basin. By 8.7 ka BP, glaciers had retreated to present-day sizes (Rampton 1981, p.33) and Hypsithermal loess was deposited in large volumes throughout much of the southwestern Yukon. Since deglaciation, Holocene organic peat deposits have accumulated at some sites in the study area (Rampton 1981). The White River Ash, a Holocene volcanic tephra dated at approximately 1.23 ka BP (Denton and Karlen 1977), can be found in exposures on the northern margin of the study area (Lerbekmo et al. 1975). 1.1.5 Soils The soils of Dezadeash Valley have been mapped and described in a detailed reconnaissance soil survey by the Canada Department of Agriculture (Day 1962). Further information on soils in the Alsek Valley region, in the context of vegetation distribution and abundance, is given in Douglas (1974). Due to the youth of the parent material and the dry, cold climate soils are generally poorly developed in the Alsek basin. In general, brunisolic soils are the most common order, followed by regosols, chernozems, gleysols, and 7 others. Besides the general control of soil age, the distribution of soils on the valley floors is controlled more by the influence of parent material than any one of the other soil formation factors (Day 1962). The fact that present day soils are so poorly developed on Lake Alsek-related sediments suggests that buried paleosols with Ah horizons a few centimetres thick may have required several hundreds of years to form. Buried paleosols with a CaCC>3 depleted A horizon (such as the Hypsithermal "Slims Soil") may have required several thousands of years to form (C.A.S. Smith 1985, personal communication). 1.2 Late Quaternary Climatic and Glacial History The Quaternary glacial history of the southwestern Yukon has been studied by Denton and Stuiver (1967) in the northeastern St.Elias, by Denton and Karlen (1973) in a more general synthesis, by Denton (1974) in the White River valley, and by Rampton (1970, 1981) in the Natazhat and Klutlan glacier area and in Kluane National Park. The Holocene glacial history in particular has been studied by Denton and Karlen (1977) in the northern St Elias Mountains. Integrating and adding to these results, Rampton (1981) has compiled a Quaternary chronology of glacial and nonglacial events for Kluane National Park. This chronology places the Pleistocene/Holocene boundary at approximately 12.5 ka BP, the beginning of the Hypsithermal at approximately 8.7 ka BP, and the onset of neoglaciation at approximately 2.8 ka BP (Fig. 1.2). The early Postglacial period was probably a time of intense nonglacial redistribution of recently exposed or deposited glacial drift, and elsewhere has been called the "paraglacial" period (Church and Ryder 1972). Glacially oversteepened slopes experienced increased landsliding and other failures during this period of heightened geomorphic activity. In Kluane National Park for example, Rampton (1981, p.33) has mapped large landslide deposits that have no active source area today. Many glacial lakes appeared in ice-choked valleys at this time, and some parts of the study area were inundated to an elevation of approximately 853 m (2800 feet) by Glacial Lake Champagne 8 Figure 1.2 Quaternary chronology of glacial and nonglacial events for Kluane National Park. Time intervals follow Denton and Stuiver (1967), Denton and Karlen (1973), and Rampton (1971). After Rampton (1981). g l a c i e r expansion .450 a BP i g l a c i e r c o n t r a c t i o n •1050 a BP g l a c i e r expansion .1250 a BP g l a c i e r c o n t r a c t i o n -2100 a BP g l a c i e r expansion .2800 a BP g l a c i e r c o n t r a c t i o n •8700 a BP-Ne o g l a c i a l Hypsithermal as o (—* o o n> 3 re .o c d e g l a c i a t i o n E a r l y P o s t g l a c i a l .12.5 ka BP. g l a c i e r expansion Kluane G l a c i a t i o n 9 (Kindle 1952). Evidence for this lake includes valley-side wave-cut benches that can be traced from Haines Junction to Whitehorse, and thick sequences (up to 70 m) of glaciolacustrine deposits of rhythmically laminated sediment. Fine sediment exposed on floodplains of heavily loaded glacial streams was available for tranport by wind and resulted in widespread deposits of loess at this time. As deglaciation progressed, "the amount of loess produced far from the glacier termini in such areas as Kluane Lake, Haines Junction, and Dezadeash Lake diminished, although loess was still being produced in large amounts near glacier termini and their active valley trains" (Rampton 1981, p.32). The Hypsithermal interval (8.7 ka BP to 2.8 ka BP), was a time of climatic warming and of glacier retreat well back into the mountains. This interval of reduced geomorphic activity, and relative inactivity of erosional and depositional processes compared to pedogenic processes, allowed the widespread development of soils. In the study area these are known collectively as the Slims Soil (Denton and Stuiver 1967). The Neoglacial is generally defined as "the period of glacier expansion subsequent to maximum Hypsithermal shrinkage" (Porter and Denton 1967, reported in Rampton 1981, p.33). The widespread advance of glaciers at about 2.8 ka BP marked the beginning of neoglaciation in the study area (Rampton 1981; Denton and Karlen 1973). Other major periods of glacier advances occurred from approximately 1.25 ka BP to 1.05 ka BP, and again from about 0.45 ka BP to the present (Rampton 1981, p.34). Radiocarbon dates from Neoglacial loess and terminal moraines, along with lichenometric and dendrochronologic data, were used to define the regional Neoglacial period (see Denton and Karlen 1977). However, Rampton notes that recognizable early Neoglacial moraines are in general poorly preserved and few in number. This, along with the small number of absolute dates used to define the specific periods of Neoglacial expansion in the southern St. Elias Mountains, suggests that many details of the Neoglacial chronology in the study area still have to be established. 10 Elsewhere, the late Holocene climatic chronology has been clarified in greater detail (see for example: Burrows and Gellatly 1982; Ryder and Thomson 1986; Porter and Orombelli 1985). Records of the earlier glacial advances have been preserved in only a relatively small number of locations mainly because much of the earlier evidence is destroyed by subsequent advances. A more sensitive record of Neoglacial fluctuations may be stored in ice-dammed lake sediments, which have the potential to capture and preserve both long-term and high resolution information on Neoglacial events (section 3.1). 1.3 Ice-Dammed Lakes Ice-dammed lakes are broadly defined as lakes impounded behind an ice dam, or else under, within, or on top of the ice body itself. They occur typically under the following conditions (Young 1980, p.285): 1) lakes impounded in an ice-free tributary valley at the side of a major valley glacier; 2) lakes impounded when a major ice-free valley is blocked by the advance of a glacier from a tributary valley (i.e. Lake Alsek); 3) lakes impounded in the embayment between two converging glaciers; 4) lakes impounded within or under glaciers; 5) lakes formed on the surface of glaciers. Ice-dammed lakes are abundant in glacierized areas; for example, Post and Mayo identified 750 ice-dammed lakes in southeastern Alaska significant enough to be visible from the air. They note that "the number and size of individual lakes vary enormously during the seasons and from year to year" (Post and Mayo 1971, p.l). Ice-dammed lakes may exist for a period of a few years up to hundreds of years, and may go through several cycles of filling and draining (Young 1980, p.286). The drainings are often sudden and rapid, and even small reservoirs can cause dramatic floods which are commonly referred to as "jokulhlaups" or simply as glacier outburst floods (Clarke and Mathews 1981, p. 1452). These outburst floods are often orders of magnitude 11 greater than non-outburst floods from similar source areas and "may be significant geomorphic and hydrologic agents in shaping [the] landscape" (Gilbert 1972, p.l). Drainage can take place either over the ice dam or alternatively, through or underneath the ice dam. The initiation of drainage may be in one or more of the following ways (Post and Mayo 1971, p.2): 1) Slow plastic yielding of the ice due to hydrostatic pressure differences between the lake and the adjacent, less dense ice; 2) Raising of the ice dam by floating; 3) Crack propagation under combined shear stress due to glacier flow and high hydrostatic pressure; 4) Drainage through small, preexisting channels at the ice-rock interface or between crystals in the ice; 5) Water overflowing the ice dam, generally along the margin; 6) Subglacial melting by volcanic heat; 7) Weakening of the dam by earthquakes. Once initiated, the rate of drainage quickly increases as the conduit enlarges rapidly from the release of the gravitational and thermal energy of the lake water (Clarke 1982, Clarke et al. 1984, Clarke 1986). Outburst flood hydrographs typically have a steep rising limb and an even steeper falling limb, with the entire flood wave passing in a matter of hours or days, depending on the reservoir volume. Extremely large outburst floods with peak discharges reaching 2.74 x 10^ - 13.7 x 10^ m^ s"* from Glacial Lake Missoula may have lasted several weeks (Clarke et al. 1984, p.294). Spectacular evidence of large outburst floods in the Alsek Valley includes extremely large fluvial bedforms (which have morphologies similar to dunes, giant ripples, or sand waves), scour hollows, gravel pendant bars, eddy and expansion bars, and downstream of the ice dam, flood terraces. The magnitude and character of outburst floods from Lake Alsek may be reconstructed either by use of an empirical flood magnitude 12 formula (see for example: Clague and Mathews 1973; Beget 1986), by paleohydrological consideration of the channel morphology (see for example: Baker 1973; Williams 1984), or by physically-based numerical models incorporating site-specific geometric considerations (see for example: Clarke 1982; Clarke and Waldron 1984). 1.4 Previous Work on Neoglacial Lake Alsek The landforms and deposits which comprise the substantial body of evidence for the existence of Lake Alsek are so plentiful and of such youthful appearance, that even the earliest explorers noted them. One of these was A. H. Brooks, who reported "a series of small terraces" that "are undoubtedly of lacustrine origin" on the eastern side of the Alsek Valle3' near what is known today as Beachview Creek. Brooks conceived of a lake impounded by a rock dam located about 30 km upstream of Lowell Glacier, that eventually drained as the dam was eroded by flowing water (Brooks 1900, p.349). The first to propose a glacier dam for Lake Alsek was J.H. Sticht (1951), who noted that "the Alsek River was probably dammed by different glaciers at different times, and very likely on some occasions by several glaciers at the same time. As a slight advance of the large glaciers A (Lowell) and B (Tweedsmuir) would today block the Alsek Valley, it is most likely that damming of the river in the past was due to ice and not to landslides" (Sticht 1951, p.92). The hypothesis of many temporally distinct phases of Lake Alsek, each dammed by glacier ice, seems to have originated here. After noting at least eight separate beaches, Sticht wrote "these beaches probably represent a complex history of fluctuation of the glaciers in the St. Elias Mountains" (Sticht 1951, p.93). Sticht found beaches marked by lines of driftwood, and even the standing trunks of trees killed by the lake water. When cored in 1944 a white spruce "about 30 feet high and 9 inches in diameter 1 foot from the ground" near Bear Creek showed 46 annual rings, thus limiting the most recent ponding at this site to at least pre-1898 (Sticht 1951, p.93). 13 More detailed observations were recorded by Kindle (1952, pp.21-23, map 1019A) who made rough measurements of filling heights and attempted to attach dates to these. Elevations given by Kindle (1952, p.22) include: 2,240 feet (682.8 m asl) for the topmost beach encountered, 2,095 feet (638.6 m asl) for the highest driftwood littered beach, and 1,970 feet (600.5 m asl) for river level at the right-angle bend of Dezadeash River (note: the first two elevations are within a few metres of more recent surveys of the same features, but the Dezadeash River level at the right-angle bend is actually 573.81 m asl (1882.6 feet) as determined by the 1984 levelling survey (Appendix I)). Kindle hypothesized at least two fillings of Lake Alsek, the most recent lake forming around 1850, and an earlier filling sometime prior to 1725. Evidence for these dates was based on tree ring counts of living trees associated with vegetation discontinuities and the two major beaches mentioned above. The thirty-six wave-cut benches counted by Kindle on the Dezadeash Valley sides were perceived by him to each represent "a pause in the level of the lake waters" (Kindle 1952, p.22). Johnson and Raup (1964, pp. 18-34) present in some detail stratigraphic evidence for at least three phases of Lake Alsek. In deep pits in the Dezadeash Valley, they found organic paleosols interbedded with three unleached gray silt layers which they considered to be of lacustrine origin (Fig. 1.3a). The lowermost of these overlies a paleosol developed on the reddish brown silts of Lake Champagne, which was identifiable by being "leached of carbonates to depths of 6 to 12 inches" (Johnson and Raup 1964, p.30, "lake #1"). While their paper has diagrams of only two soil profiles showing these relationships, presumably these are representative of many such soil pits examined. The oldest and largest Lake Alsek ponding recognized by Johnson and Raup filled to an altitude of 646 m (2120 feet, lake no.3, Fig. 1.3a,b). This lake "lasted long enough to accumulate at least 10-12 inches of fine-textured gray silt on its floor" (Johnson and Raup 1964, p.30). The White River ash deposit (now dated at 1.23 ka BP) appears near the bottom of this silt layer (Fig. 1.3b). Overlying this and a 1.25 cm (0.5 inch) thick 14 Figure 1.3 Sedimentary sequences exposed in soil pits at (a) 8 km W of Haines Junction at mile 1019 Alaska Highway, and (b) near Pine Creek at an elevation of 626 m asl. After Johnson and Raup (1964). OT 40- Lake 14 80H Lake 13 a. a) a 6 T-l X o u a. a. < \20A Lake I I Dark brown silt filled with humas and roots of present prairie plants. Lake 13 Gray s i l l , partially leached of carbonates beneath the existing turf. H Organic layer (ca 1/2") Cray silt unleached. Lake 11 Organic layer (ca.1/2"). Cray silt, with irregularly layered reddish brown stain. Slightly leached at top, otherwise unleached. Organic layer (ca 1/2"). Reddish brown silt, leached toward the base indicating former leaching throughout. Yellowish gray clay, unleached with about 2" of gray, very sandy clay at the top. Crumby gray clay, with lenses (1-4" thick) of yellowish gray, unl«ach«d. Dark brown silt with humus. Gray silt, partially leached of carbonates. Volcanic ash layer (discontin-uous, 1/4 - 1/2" thick). Organic layer (1/4 - 1/2" thick. Reddish brown silt, faint ef-fervescence with dilute acid. Yellowish gray silt, faint ef-fervescence with dilute acid. Cray slit, finely stratified, merges with varied mater-ials at depth of about 8 0 " below ground surface; un-leached. 15 organic layer, another gray silt deposit about 25-30 cm (10-12 inches) thick, has been associated with a lake level of about 622 m asl (2040 feet, Lake No.4, Fig. 1.3a). Above this is another thin organic layer in turn overlain by another gray silt layer about 12.5-15 cm (5-6 inches) thick, corresponding to the most recent inundation above a level of about 600m asl (1970 feet, Lake No. 5, Fig. 1.3 a). This paper neglects to provide any sound sedimentological evidence for a lacustrine origin of the fine-textured gray silt, which could have a number of origins, including wind blown silt, subaerial slopewash, or extreme overbank flooding of the Dezadeash River. Oh the basis of lichen and vegetation discontinuities on the valley sides, and on the occurrence of major driftwood strandlines, Clague (1979, p.63) proposed that Lake Alsek may have had many more ponding phases. The ponding elevations of each phase of Lake Alsek recognized in this paper are 678 + , 640, 637, 623, and 595 m asl. Approximate ages were assigned on the basis of lichen growth curves and radiocarbon dates associated with major driftwood lines. These findings are supported by Rampton (1981), who also reexamined the soil profiles reported by Johnson and Raup (1964) at a site approximately 8 km West of Haines Junction on the Alaska Highway (Appendix VI). He found additional evidence of older lake phases in the form of lacustrine sediments interbedded with the White River Ash and with an underlying "Hypsithermal" paleosol (radiocarbon dated at 2820 + 60 1 4 C years BP), which served to substantiate the idea of 5 separate lake phases. The most recent work on the chronology of Lake Alsek is a comprehensive and detailed paper by Clague and Rampton (1982), which attempts to sort out the chronology of the lake using a multi-faceted approach. Their approach "consisted of a review of oral and written accounts of inhabitants and early explorers in the region, radiocarbon dating of driftwood and paleosols, direct dating of living trees at various levels within the Lake Alsek basin and on Lowell Glacier moraines, dendrochronological dating of driftwood, and lichenometric dating of beaches, flood terraces, and moraines" (Clague and Rampton 16 1982, p. 103). The recognized phases of Lake Alsek and probable age of each phase are reproduced in Table 1.1. Table 1.1 Recognized phases of Neoglacial Lake Alsek (Clague and Rampton 1982). This chronology of filling and draining events uses a number of criteria to date and define these lake phases, but is probably incomplete. Datum for ages is 1982. Lake level Lake depth Probable age (m asl) (m) (a BP) 561 57.3 73 595 91.3 130 623 119.3 180-220 637 133.3 250-350 640 136.3 400-500 >645 - ->610 - ->609 _ In this chronology, five lake phases have strong supportive evidence (the 595, 623, 637, 640, and >645 m asl levels) and two of these have been dated using at least two reliable independent techniques (the 595 and 623 m asl levels). However by compiling a body of less conclusive, but nonetheless significant evidence, they concluded that a total of eight separate lake phases can be consistently recognized. Nevertheless, Clague and Rampton acknowledge that "our record of Lake Alsek events probably is incomplete" due in large part to removal of the evidence of small pondings by subsequent and higher pondings. It is quite possible that the observed nearly linearly decreasing filling levels of recognized phases of Lake Alsek through time is merely a consequence of this problem (Table 1.1). 17 The absolute dating of only two Lake Alsek phases can be done with a high degree of confidence and with reasonably good precision (the 595 level at 130 a BP, and the 623 m asl level at > 180 <220 a BP). The remaining lake phases can be assigned reasonable but not very precise age ranges, that require substantiation by additional and independent techniques. Some problems in dating Lake Alsek events are discussed in detail by Clague and Rampton (1982, pp. 112-115) and can be summarized as follows: 1) insensitivity of radiocarbon dating on materials less than about 500 years old, 2) dendrochronology limited by master chronologies dating back only about 320 years, 3) dating of substrates using lichenometry limited by a lack of local lichen growth curves for the Lake Alsek Basin, 4) destruction of beaches and driftwood layers by successive ponding phases, 5) varied provenance and age of wood in any single driftwood strandline. As well, the two oldest radiocarbon dates on buried paleosols exposed in a soil pit 8 km west of Haines Junction are interbedded with unweathered sediment of uncertain origin. The stratigraphic value of these dates is entirely contingent upon a lacustrine origin for the sediment, which is not at all certain (Clague and Rampton 1982). Indeed, the three earliest Lake Alsek pondings have very little substantive evidence beyond a possible lacustrine origin supposedly associated with these sediments. 1.5 Outstanding Questions In light of the previous work on Lake Alsek noted here, several fundamental questions regarding the Neoglacial history of Lake Alsek remain: 1) How many phases of Lake Alsek were there? 2) What was the maximum extent of each lake phase? 3) What is the age of each lake phase? 4) Is each lake phase associated with a single filling and draining event, or with a number of cyclic filling and draining events? 18 5) Can sedimentological principles be used to differentiate Lake Alsek depositional environments? 6) How does the Lake Alsek chronology relate to the Neoglacial climatic chronology? 1.6 Scope and Objectives of this Thesis The fundamental purpose of this project is to investigate the characteristics and chronology of Neoglacial Lake Alsek. More specifically, this study is concerned with documenting and interpreting the Neoglacial lacustrine sediments associated with Lake Alsek. The objectives of this study are, in rank order: 1) to determine if a resolvable sedimentary record associated with Lake Alsek exists in the lake basin (in areas of high depositional rates and high preservation potential); 2) to analyze the stratigraphy of sediments in the Lake Alsek basin using an explicitly defined lithofacies scheme; 3) to construct lithofacies assemblages and analyze vertical lithofacies sequences to attempt interpretation of depositional environments; 4) to correlate stratigraphic records and establish the extent and the timing of Lake Alsek ponding phases; and, 5) to add absolute chronologic information to the history of Lake Alsek. 19 Chapter 2 GEOMETRY OF THE NEOGLACIAL LAKE ALSEK BASIN In this study, it is clear that precise location of a stratigraphic unit of Lake Alsek sediment is essential to deduce where and when a Lake Alsek inundation took place. Knowledge of the landscape geometry onto which the stratigraphic record of Lake Alsek events has been imprinted is of fundamental importance to its interpretation. For example, horizontal location provides information about relationships to surrounding landforms; this can usually be satisfactorily determined with the aid of air photographs and maps. Vertical position provides information about relationships between lake filling heights and deposited sediments, and also underlies interrelationships among sediment sampling sites. This is more difficult to measure remotely, therefore accurate levelling surveys had to be undertaken in the Alsek Valley. In large part, the spatial variability of Lake Alsek sediments can be explained by the variety of lake sizes which have inundated the Alsek Valley through Neoglacial time. An extensive lake filling which inundates a larger area of the basin necessarily leaves a sediment imprint over a more widespread area than that from a smaller lake filling. Additional spatial variability can be explained by the differential influence of flood waters as the lake drains from the Lake Alsek basin. Those areas affected by erosion and transportation processes may have significant portions of the sedimentary record removed. Areas of deposition, on the other hand, may record the drainage event in distinctive sediments. The purpose of this chapter is to establish geometrical and related data (basin hypsometry, basin filling times, and outburst flood magnitudes) for the Lake Alsek basin. The basic geometric information is essential for locating the various sites in the basin so that simple spatial interrelationships can be determined. Furthermore, these data provide independent insights into the past geomorphological environment and therefore can aid in interpretations of the sedimentary record. 20 2.1 Existing Information Topographic Maps The entire region has been mapped at a scale of 1:250,000 and spans two NTS topographic map sheets: The Dezadeash, Yukon Territory (115A) sheet is a first edition map from air photography and ground surveys undertaken in 1948. The Mount Saint Elias (115B-115C E/2) sheet is a second edition map from air photography undertaken in 1951 and 1956 in the Alsek River area and includes a portion of the intensively surveyed USA-Canada international boundary zone. For these reasons it is considered to be more reliable than the Dezadeash sheet. Mapping of the area at a scale of 1:50,000 is incomplete. The two half-sheets whose unavailability is most conspicuous are NTS 115 A/12 (West) and NTS 115 A/5 (West). These two sheets cover the Alsek Valley from Lowell Glacier to 53 km upriver, where Dezadeash River turns southward through a gap in the Kluane Ranges. To make matters worse, these missing 1:50,000 scale maps are on the western edge of the unreliable Dezadeash 1:250,000 sheet. ' An accurate map of the Lowell Glacier terminus at 1:50,000 was produced under the direction of G. Holdsworth (1976) by the Glaciology Division of the Department of Environment, based on air photography taken in August 1974. This was produced as a base map for high resolution monitoring of the movement of the Lowell Glacier terminus by remote sensing, but no elevation control for the map was surveyed. Geodetic Benchmarks Elevation control in the study area comprises a number of geodetic benchmarks emplaced along the Alaska Highway and the Haines Highway. In the Haines Junction area there are three independently surveyed sets of benchmarks: 1) United States Coastal and Geodetic Survey markers from 1943 with individual names such as "B8 (2191.7 ft.)" and "R8 (2063.103 ft.)"; 21 2) Canadian Lands Survey markers from 1962 with individual names such as "RW13 L56 G803"; and 3) Canadian Geodetic Survey markers probably emplaced in 1978, with individual names suchas"78Y 526". Presumably the 1943 Survey markers were used to fix elevations on the 1:250,000 maps. It remains unclear as to which BM's were used to survey the more recent 1:50,000 maps. Whether or not the older BM's were tied into the newer surveys is unknown. It is possible that different datum conventions were used for one or all of these surveys. However, it is expected that all surveys would have conformed to the 1927 North American Datum standard; therefore it is assumed that all the surveyed altitudes are compatible. 2.2 Extent of Lake Alsek The highest ponding phases of Lake Alsek have not yet been established accurately, nor dated precisely (section 1.4). Definite Neoglacial beaches can be observed at levels up to at least 668 m (Figure 2.1), and even higher than these is a beach at 678 m which is less definitely Neoglacial, but because of its "youthful appearance" has also been attributed to Lake Alsek rather than to Glacial Lake Champagne (Clague and Rampton 1982, p.95). Hughes et al. (1972, p.9) report that Lake Alsek beaches are present on the valley sides up to a level of 685 m asl (2240 ft asl), but do not indicate where these are nor the basis for their dissociation from Glacial Lake Champagne. The highest fillings are also the oldest of the established phases. At least one lake phase must have reached an elevation of at least 645 m sometime prior to about 800 a BP based on the evidence of an "isolated piece of driftwood or deadfall" dated at 910 ± 50 a BP (Clague and Rampton 1982, p.112). The lowest alternate drainage route from the Lake Alsek basin is across the divide near Champagne which separates the Dezadeash and Mendenhall rivers at an elevation of 22 Figure 2 . 1 Elevation data of features associated with the study of Lake Alsek. Five sample sites were inundated by all of the recognized fillings, seven sites were inundated by all but one of the recognized fillings, and the eighth site was inundated by only the highest fillings. See text for discussion of the errors associated with the surveyed markers and the sample sites. 700 --678 high f i l l i n g 675 -- 668 beaches 650 - 640 lichen bk. -637 lichen bk. 625 -600 -to (0 c o 5 575 w 550 525 -500 Clague and Rampton (1982) - 623 driftwood 610 paleosols 595 driftwood 628.9 USCGS BM "R8" .617.8 BM "Piton .561 driftwood Elevation Survey (1984) - 645 KM69.0 - 592.4 BM "BC .576.0 BM "KFE" - 529.7 BM "WB" 521.2 BM "PicPic" River at 503.7 Lowell Gl Sample Sites 575 KM53.0 - 574 KM50.4 552 KM28.5 545 KM13.4C 536 KM13.4D 535 KM1.2 - 514.4 KM4.4 23 between 701.0 and 716.3 m. Scrutiny of air photographs from this area reveals no evidence of an extensive outlet channel cut into the moraines at this divide. Strandlines in the Lake Alsek basin at or above this divide can therefore be safely ruled out as products of Lake Alsek, and probably should be assigned to Glacial Lake Champagne. A drainage outlet at the head of the Kaskawulsh River down the Slims River Valley is unlikely since Bostock (1969) reports, in the context of Kluane Lake drainage, that the bedrock divide here is at an estimated elevation of 768.1 m, which puts it above the drainage divide into the Mendenhall basin. Less extensive than the maximum are a number of temporally distinct Lake Alsek phases of lesser magnitude (Clague and Rampton 1982). In order to establish the range of lacustrine and outburst flood environments which may have existed in the Alsek Valley at various times in the past, a list of seven maximum filling levels have been extracted from Clague and Rampton (1982 Table 3). These filling levels are all associated with wave-cut benches with evidence of different ages at elevations of 561, 595, 623, 637, 640, 668, and 678 m asl (Table 1.1). 2.3 Alsek Valley Levelling Survey The purpose of this survey was to carry the vertical control from Geodetic BM's on the Alaska Highway into the Dezadeash and Alsek Valleys as far south as Lowell Glacier. The instruments used were a Wild T2 (grad) optical theodolite on which was mounted an AGA Geodimeter 122 infrared laser ranger (910 nm wavelength), and the rod target was an AGA reflecting prism mounted 1.50 m above rod base. To survey most of the Alsek Valley, inflatable boats were the only feasible mode of transportation. Because upstream travel is not possible on account of excessively shallow and fast flow, the survey was limited to one direction, and could not be closed in the usual manner. To allow some form of closure check, double-poling was used over the length of the survey line. The survey instruments were set up at a distance of less than 500 m 24 downstream from BM "BC" located on the right bank of the Dezadeash River near Beachview Creek. Two independent shots of slope distance, horizontal angle and azimuth angle were made on the prism positioned here. Then, the instrument position was maintained while the prism was floated about 500 m downstream where two (later three) temporary rod positions about a metre or so apart were found and marked. After taking shots on each of these rod positions, the instrument was then floated downstream about 500 m or so, where it was set up again and back shots of the two temporary rod positions were taken. At this point measurements were checked by a field computer for significant closure error, so that shots could be taken again if necessary. When none were found the rod crew floated downstream of the instrument position about 500 m and the process was repeated. In this leap-frog manner, a one-directional 51 km traverse of the Dezadeash and Alsek valleys to Lowell Glacier was completed. To determine survey precision, elevations were computed as though two separate lines had been surveyed over the entire traverse so the cumulative RMS variance could be determined for permanent elevation markers. The cumulative RMS variance at the end of the entire traverse at marker "WB" is RMSE = 0.0157 (the standard deviation is 0.125), therefore the altitude determination ( ± 2 standard deviations) for the final marker "WB" can be given as 529.64 ± 0.250 m, (Figure 2.1). Elevations and cumulative horizontal distances for the levelling traverse (completed over several years under the direction of S.G. Collins) from the USCGS Benchmark R8 (2063.103 ft.) on the Alaska Highway to the marker WB near Lowell Glacier are given in Appendix I. Semi-permanent markers were installed at several points along the length of the traverse in order to retain the elevation control for future use. These markers, PITON, BC, KFE, PicPic, and WB, are described in detail in Appendix I (elevation marker descriptions) and located in Figure 1.1. In order to obtain a reasonably accurate longitudinal river profile, shots that were taken on targets within a few centimetres of river level were noted as such in the field book. These were later extracted from the data 25 so that a river profile representing daytime river level conditions over the period July 29 to August 3, 1984 could be constructed (Figure 2.2). 2.4 Valley Cross-Sections Accurate cross-sections of the Alsek valley were required for: (a) locating sediment sampling sites on the valley sides, and (b) determining channel cross-sectional area and wetted perimeter for paleohydrological analysis of outburst floods. This information may be directly obtained from topographic maps, but because the maps covering the Alsek Valley have a relatively coarse contour interval of 500 ft, valley cross-sections were measured directly at two key locations. Field measurements were undertaken using a Brunton compass sighting level for vertical angle and chaining to obtain slope distance. Where possible, a 30 m chain length was used and vertical angles were measured with the Brunton to 0.2 degrees. Speed and portability are the advantages of this technique, and the loss of precision is relatively minor. Slope distance can be measured to a precision of 0.10 m and vertical angle to a precision of 0.2 degrees. From the Alsek River longitudinal profile results the approximate river level at each cross-section could be determined. Barometric altitudes of the highest points of each cross-section were also noted. The results of two cross-sections, at KM13.0 and KM53.0, are graphed at 10 times vertical exaggeration in Figures 2.3a,b. 2.5 Barometric Altimetry For rapid determinations of elevations in vegetated and remote areas, one or more often, two American Paulin System barometric altimeters were employed. Both were used as roving instruments from a point of known elevation at river level and readings were corrected for temperature and a linear change in barometric pressure. In all such measurements, every effort was made to close the traverse as quickly as possible. Often this could be accomplished in two hours or less. Also, in order to reduce the error 26 Figure 2.2 Alsek River longitudinal profile. Surveyed section is from early Neoglacial moraines of Lowell Glacier to a point 57.4 km upstream. Four permanent elevation markers (BC, KFE, PicPic, and WB) were established during this part of the survey. 27 Figure 2.3 Valley cross-sections looking downriver at (a) KM13.0, and (b) KM53.0. Note the different vertical and horizontal scales and 10 times vertical exaggeration. The river levels are graphed as zero elevation, and width is estimated since no measurements were undertaken. Alsek Valley cross-section KM53.0 (10X V.E.) IRoadcut| 0.0 t 600.0 700.0 HORIZONTAL 01 SOO.O 900.0 STANCE IM) 1000.0 1100.0 1200.0 A l s e k V a l l e y c r o s s - s e c t i o n K M 1 3 . 0 ( 1 0 X V . E . ) p-§ 3 ISite KM13.4CI lAlsek River (523.3 m a s l ) | ,«,., m* «... «... a... "... «... "»•« '•»•• j^flfoKW-oisfflSft i f f ' ' 28 associated with spatial variability of atmospheric conditions, barometric altimetry was usually undertaken over very small distances and in the morning or evening (when atmospheric conditions are relatively stable). Where possible, several separate observations were made of the same site and always by the same observer. A table of results which includes site descriptions, number of observations and mean elevation is given in Table 2.1. Readings were recorded to the limit of scale resolution on the altimeters (0.5 ft) and are assumed to be accurate. Errors were only approximately estimated as ± 0.1 ft due to scale resolution, and ± 0.5 m to ± 2.0 m due to non-linear variability of atmospheric conditions. Table 2.1 Altitude Determinations by Barometric Altimeter Site Name Closure Time Elevation Error (min) (m asl) (m) KM1.2 17 hrs 535. ± 2.0 KM4.4 5 514.4 ± 0.5 KM13.4A 204 560. ± 1.0 KM13.4A 68 556.1 ± 0.5 KM13.4A 68 557.5 ± 0.5 KM13.4C 204 547. ± 1.0 KM13.4C 68 543.9 ± 0.5 KM13.4C 68 544.0 ± 0.5 KM13.4D 204 537. ± 1.0 KM13.4D 68 533.4 ± 0.5 KM13.4D 68 537.5 ± 0.5 KM69.0 167 644. ± 1.0 KM69.0 117 645. ± 1.0 2.6 Lake Alsek Bathymetry By combining existing topographic map data and the levelling survey results, an improved contour map of the Lake Alsek Basin was produced. This composite map is useful for analysis of lake basin hypsometry, rough approximations of valley cross-sections, mapping the areal extent of pondings, and for analysis of the confined valley geometry through which the draining lake water must flow (Figure 2.4). This map was digitized in such a way that available and forthcoming map information could be aggregated in a streamlined manner. The contour interval on the 29 1:50,000 maps is 100 ft, so to concatenate the contour data from the 1:250,000 maps (which have a 500 ft contour interval) a simple four point linear interpolation was done on photocopy enlargements of the maps. The complete contours are digitized at intervals of 100 ft over the range 1700 ft to 2300 ft. Digitization was done using a UBC software package called *DIGIT (Mair 1982) with output data files formatted (2F9.1, 12), and units of metres (conforming to the UTM grid). Each contour line was digitized clockwise if the area contributes volume to the basin and counterclockwise if the contour encloses a hill (at this resolution no hills were detected). Contours from different map sheets are spliced together in the data file. Splices can be identified by an 12 value of + 1 in place of a normal 12 value of + 2. New contours are identified with an 12 value of 0, and an alphanumeric title. When the 1:50,000 scale maps of the rest of the Alsek basin become available, their integration into this data base should be straightforward. The main channels of the Alsek, Kaskawulsh, Dusty, Jarvis, Aishihik, and Dezadeash Rivers were also identified and digitized. In each case straight line segments were carefully chosen along the center of the mapped blue river line, ignoring bars that would be inundated by a normal river freshet. In braided sections the center of the braided width of the river floodplain was approximated as the river line. This river line is used to identify sites in the basin as approximate upstream distances from Lowell Glacier, and for locating the lowest points on the valley floors. In this study, all sites are designated by their distance from Lowell Glacier along this river line (using the distance formula). For example, site KM69.0 is perpendicular to a point on this line approximately 69 km upstream of Lowell Glacier. Note that the map-derived river line and the surveyed river line are completely independent; therefore the computed distances along these two lines may be in some disagreement. This discrepancy is minor nearer to Lowell Glacier, but increases with upstream distance. For example, site KM53.0 is 57.4 km upstream of Lowell Glacier along the survey line. The longitudinal river profile of the Dezadeash and 30 Figure 2.4 Digitized contour map of the Lake Alsek basin. Contour interval is 100 feet. Contours range from 1700 feet asl (518.2 m asl) to 2300 feet asl (701.0 m asl). Straight line channels of the major rivers are taken from map sheets to represent valley bottom. Fiducial marks are intersections of a 10 km UTM grid. See Figure 1.1 for reference names and section 2.6 for explanation. 31 Alsek Rivers is used as the elevation datum for further altitude determinations in the Alsek Valley, while the digitized river line is used for site names and for locating the valley bottom. From this contour map the area enclosed by each contour was integrated using a one-directional trapezoidal rule method and graphed against elevation to produce a discrete value hypsographic curve over the domain of interest (Fig. 2.5). In order to determine areas at intermediate elevations (which correspond to filling levels) simple linear interpolation is used because uncertainty associated with the contour map itself is much greater than the relatively minor underestimation of volume associated with curve fitting. The basis for this decision is the relatively large magnitude of the inaccuracy associated with the contour lines used to compute areas. To obtain reservoir volumes, this hypsographic curve was then integrated over the area of each filling height (Fig. 2.6). Integration of the hypsographic curve was undertaken by a cumulative procedure for increasing areas. For this reason Figure 2.6 is composed of linear segments instead of parabolic arcs as would be expected if trapezoidal integration were used. From these curves various morphometric parameters can be compiled. Using the morphometric definitions given by Hakanson (1981) these data are tabulated for various filling levels of Lake Alsek (Table 2.2). Lake depth is the maximum depth assuming lake bottom is the river elevation at the early Neoglacial moraines of Lowell Glacier (503.7 m asl). Mean depth D m e a n of the various lake levels is equal to V/A, where V is the lake volume and A is the lake surface area (Table 2.2). 2.7 Filling Times and Outburst Flood Magnitudes Filling Times The time required for complete filling of a lake of any given extent can easily be obtained once the reservoir volume and water influx are known. Assuming no leakage, the 32 F i g u r e 2 .6 R e s e r v o i r v o l u m e for the L a k e A l s e k b a s i n a s a f u n c t i o n of e l e v a t i o n . V o l u m e s w e r e ob ta ined b y i n t e g r a t i n g the h y p s o g r a p h i c c u r v e fo r e a c h f i l l i n g he ight . T h e c u r v e w a s ob ta ined b y connect ing these v o l u m e s w i t h l i n e a r s e g m e n t s . L a k e v o l u m e s m a y be s l i g h t l y u n d e r e s t i m a t e d due to l i n e a r i n t e r p o l a t i o n b e t w e e n d a t a po in ts on the h y p s o g r a p h i c c u r v e . 690-Volume (km**3) F i g u r e 2 .5 H y p s o g r a p h i c c u r v e for the L a k e A l s e k B a s i n . A r e a s enc losed b y e a c h con tou r w e r e n u m e r i c a l l y i n t e g r a t e d f r o m the d ig i t i zed c o n t o u r m a p . Z e r o d e p t h is a t 5 0 3 . 7 m a s l w h i c h is the A l s e k R i v e r e leva t ion a t the e a r l y n e o g l a c i a l m o r a i n e s o f L o w e l l G l a c i e r . 710-1 : : ; : : • 1 1 , ! , ! ! , 1 0 100 200 300 400 500 600 700 Area (km**2) 33 Table 2.2 Lake Alsek Morphometric Data Lake Level Depth3 ^mean Area Volume (m asl) (m) (m) (km2) (km3) 561 57.3 23.2 54.65 1.268 595 91.3 32.0 146.35 4.687 623 119.3 43.1 230.76 9.947 637 133.3 48.6 272.78 13.255 640 136.3 49.6 282.03 13.983 668 164.3 59.5 396.83 23.607 678 174.3 61.3 462.98 28.367 a) To calculate lake depth, the lake bottom is assumed to be the river elevation (503.7 m asl) at the early Neoglacial moraines of Lowell Glacier on August 3, 1984 at approximately 1800 PDT. Table 2.3 Lake Alsek Filling Times and Flood Magnitudes Lake Height Lake Filling Times (days) Qmaxd (m asl) Alseka No Lowellh No Kaskc (mV 1) 561 71.6 95.3 142.5 1.28xl04 595 264.6 352.3 526.7 3.15xl04 623 561.6 747.6 1117.8 5.30xl04 637 748.4 996.2 1489.5 6.46xl04 640 789.5 1050.9 1571.3 6.71xl04 668 1332.8 1774.2 2652.7 9.63xl04 678 1601.6 2132.0 3187.6 1.09xl05 a) entire Alsek as gauged (1975-1984) above Bates River is 205 m s b) no Lowell Glacier input, flux to Lake Alsek is 154 m3s" .^ c) no Kaskawulsh River input, flux to Lake Alsek is 103 m3s"*. d) flood magnitudes calculated using empirical formula, see section 2.7. 34 filling times to attain various lake levels were calculated using 3 plausible input fluxes (Table 2.3). The first is simply the measured annual mean discharge (Water Survey of Canada 1975-1984) of the Alsek River as gauged above Bates River (205 m3s"1). There remains a possibility that input fluxes were significantly larger than this because of additional input from the Kluane Lake basin prior to about 450 a BP (Bostock 1969). The second input flux does not include any runoff from Lowell Glacier itself, -estimated at 1/4 of the above discharge- leaving an input flux of 154 m^ s"* into the basin. The third input O -I flux (103 m s ) does not include either the Lowell contribution nor the Kaskawulsh River, the latter being excluded because of historical routing changes which have left the river virtually empty at times (Johnson 1986). Outburst Flood Magnitudes Outburst flood events from Lake Alsek are responsible for radical changes in sedimentary environment so it is desirable to obtain independent estimates of the magnitude of these events. Outburst flood modelling (e.g. Clarke and Waldron 1984; Nye 1976) and detailed paleohydrologic analysis (e.g. Atwater 1984; Baker 1973; Waitt 1984; Williams 1984) are both beyond the scope of this thesis; hence only the Clague and Mathews (1973) empirical formula for estimating peak discharge is used here. The Clague and Mathews (1973) empirical formula is simply an exponential regression of peak discharge (Q m a x ) on total lake volume (V). In its most recent form (Beget 1986, p. 137), the regression model is; Qmax = ° - 0 0 6 5 V ° - 6 9 with a coefficient of determination of r = 0.86. The data set used in this model includes 21 historic outburst floods covering a range of Q m a x values from 50 to 5 x 10^ m s^"*, and a range of lake volumes from 2.6 x 10^ to 7.0 x 10^ m 3. 35 For all but the smallest Lake Alsek fillings, use of this method to determine outburst flood magnitudes would involve extrapolation beyond the range of data used in the regression model. Nevertheless, this is a useful method of estimating peak discharges of outburst floods from the basin since it only requires that reservoir volumes for the various levels of the lake be known. These can be estimated by first generating the hypsographic curve for the reservoir and then integrating this function for each filling height (Table 2.2). The calculated instantaneous peak discharges (Q m a x ) range in magnitude from 1.28 x 104 m3s"1 for the 561 m asl filling, to 1.09 x 105 HIV1 for the most extensive filling (Table 2.3). These results are useful as order of magnitude estimates of Q m a x associated with the range of filling levels. Estimates of outburst flood magnitude are also possible by other methods (section 1.5). Paleohydraulic Environment When order of magnitude estimates of Q m a x are combined with the data on basin geometry, information on the hydraulic conditions at specific sites in the lake basin (for a given flood magnitude) may be obtained. Knowledge of current velocity, Froude number, Reynolds number, and bed shear, for example, would be of great value in the interpretation of sediments and bedforms found at various points in the basin. Application of paleohydraulic techniques (Baker 1973; Williams 1984) to Alsek Valley was not undertaken in this study. More detailed analysis is required for realistic reconstruction and the following example will illustrate this. Consider the simple discharge relation, Q = AV, where Q is discharge, V is average current velocity, and A is channel cross-sectional area. The peak discharge (Q m a x ) can be estimated, and the cross-sectional area (A) of the channel can be measured, therefore an order of magnitude estimate of maximum velocity ( V m a x ) during lake drainage should be possible. For an example of this approach, recall that for the 640 m asl filling, the channel cross-section at KM13.0 is about 2.0 x 105 m 2 (Fig.2.3a), and Q m a x for this filling level is about 6.71 x 104 m 3 s"1 (Table 2.3). Applying the discharge relation, the maximum flood velocity may be 36 approximated at V m a x = 0.34 m s"^ , averaged through the cross-section. In application to cross-sections in the lake basin, however, this approach is faulty. Because Q(t) increases to maximum as lake depth and cross-section area A decrease, A is a function of time as well. An additional problem is that specific sites located on the valley sides may not remain inundated when maximum velocity V m a x occurs. Thorough analysis requires that a detailed outburst flood hydrograph (i.e. Q(t)) and A(t) be computed at various points in the lake basin itself using a physically-based outburst flood model (Clarke et al. 1984). The work necessary to do this is beyond the scope of the thesis, but nonetheless is seen as a valuable adjunct to the sedimentologic results presented here and to paleoenvironmental reconstruction of the Lake Alsek basin. 37 Chapter 3 STRATIGRAPHY AND SEDIMENTOLOGY 3.1 Introduction The Alsek Valley has undergone extreme changes in geomorphological conditions many times during the Neoglacial period. Church (1980, p.24) states that "the most obvious source of information about geomorphological processes (at least about depositional processes) lies in the sediments themselves". In particular, a lacustrine sedimentary environment may record both long term and high resolution information on Holocene geomorphological events. Paleoenvironmental information may be captured from the sediment cascade and stored in sediment sinks but the recovery of reliable records can be problematic due to long term instability of almost all such environments (Church 1980, p.26) Attempts at stratigraphically based investigations of the Lake Alsek sedimentary record have been frustrated in previous studies. Specifically, Clague and Rampton (1982, p. 99) state that Lake Alsek bottom sediments are relatively uncommon and are thin or absent entirely above about 625 m elevation. Furthermore, episodic erosion of Lake Alsek sediments in the Alsek Valley is indicated by the widespread occurrence of catastrophic flood deposits (see Clague 1979, p.65). Clearly, the magnitude of drainage events has an important influence on the preservation of deposited sediments. Several studies of the characteristics and chronology of Glacial Lake Missoula (another large ice-dammed lake) have been based on the lake basin sedimentary record (Chambers 1971; Baker and Bunker 1985, p.24). Interpretation of the Lake Missoula sediments suggests that many extensive fillings and drainings occurred.and that a detailed chronology of events is provided by beds of varved sediments separated by distinctive weathered and eroded unconformities (Chambers 1971). Because similar paleoenvironmental information is desired for Lake Alsek, it is reasonable to undertake a similar systematic examination of the Lake Alsek basin sedimentary record. 38 Identification of suitable depositional environments is possible from site location and site reconnaissance. Even if sediment sinks can be identified, the relatively short-lived fillings of Lake Alsek may mean that the local sedimentation rates were too low for a resolvable record of individual events to be read. Finding suitable sites in the Lake Alsek basin is complicated by characteristically episodic depositional processes associated with inundations and drainings (Chapter 2). In the ice-dammed Lake Alsek basin, sedimentological events may entail quite different geometric and hydraulic conditions from one episode to the next. Because of this, a single locale at one time may be a depositional environment and at another an erosional one, resulting in spatial heterogeneity of the depositional record. 3.1.1 Sediment Sources Drift The Alsek Valley was occupied by large valley glaciers (to an altitude of approximately 1800 m near Lowell Glacier, sloping to less than 1520 m in the vicinity of Haines Junction) during the Kluane Glaciation ending about 12.5 ka BP, and to a lesser extent during the period of deglaciation between 12.5 ka BP and 8.7 ka BP (Rampton 1981, p.30). Drift deposited by these glaciers is extensive on the valley floors, and resulted in a suite of glacial landforms (e.g., ground moraines, lateral and end moraines, kettle lake fields, eskers, and flutings). Drift thicknesses in the Alsek Valley are unknown, but drift distribution has been mapped (Rampton 1981, map 13-1979). Glacial sediments and their weathering products were available for erosion and transportation by lake water as Lake Alsek filled and drained. No evidence of a stable lake level has been found (section 2.2) so it appears that Lake Alsek was always either filling or draining and that lake levels were transient throughout the Neoglacial. Because of this, the valley sides were probably progressively subjected to the action of waves. This is mostly an erosional process and is an important source of sediment that has been observed at a contemporary 39 ice-dammed lake (Gilbert and Desloges 1986). After many Lake Alsek phases, these sediments may become winnowed as finer grained material is progressively removed. Fluvial Sediments Heavily laden glacial streams such as the Kaskawulsh and Dusty rivers carry allochthonous sediments into the Lake Alsek basin. These streams issue directly from a glacier source and transport large volumes of suspended and bedload sediment. Holocene valley fill is most extensive along the Kaskawulsh River. Suspended sediment loads measured at a point about 20 km downstream of the Lowell Glacier at Bates River indicate a peak observed load of 392 mg l" 1 on July 20, 1978 (Q = 571 m3s"1) and an average load for the seven observations of 217 mg 1"* (Table 3.1). Unfortunately, it is not known how much these relatively high loads have changed through time. Table 3.1 Suspended sediment load of Alsek River (above Bates River) Source: unpublished data, Water Resources Branch, W.S.C. Date of Samples N Mean Cone. S.D. Q (mgr1) (m3s-1) July 20, 1978 5 392. 37. 571. August 23, 1978 5 209. 12. 471. July 12, 1979 5 295. 23. 743. August 25, 1979 5 271. 34. 629. October 26, 1979 5 73. 2. 137. July 16, 1981 5 193. 12. 640. October 15, 1982 6 85. 8. 140. 40 Loess Deposition of loess in the Alsek Valley is presently occurring at greatly reduced rates when compared with estimated late Pleistocene rates (Rampton 1981). Nevertheless, some contemporary loess deposition occurs locally near glacier termini, and to a lesser degree in the Alsek Valley. Loess may be a significant clastic sediment source for some of the small lakes near the Lowell Glacier terminus. No loess has been deposited in the Dezadeash Valley since the early Postglacial. Organic Materials During a Lake Alsek phase, organic materials are probably deposited at lesser rates than clastic materials, as in modern glaciolacustrine systems (Gilbert and Shaw 1981). During the intervening periods, small ponds in the Alsek basin (which are probably oligotrophic immediately after a Lake Alsek phase) eventually become productive enough to support lake bottom vegetation. While it is not known how long colonization takes, nor what species this vegetation is, organic detritus is presently accumulating in most of the ponds which were visited. This is currently occurring at rates high enough to produce a carbonaceous deposit of detrital gyttja (Lowe and Walker 1984, p. 131). This deposit is characteristically black in colour, has no internal structure, and is not much denser than water. Mineral sediments incorporated in this deposit are probably transported to the pond by wind or by slopewash, since channelized inflow is minimal. Marl At a number of sites in the Lake Alsek basin, pale yellow-grey clay-rich marl deposits can be observed. In Pine Lake for example, marl is apparently precipitated by the aquatic plant 'Chara' (Kindle 1952, p.23). Dissolved calcium enters the lake in the streamwater of Marl Creek at the west end of the lake. Marl deposits around Pine Lake and in other lakes are not being buried by other sediments, which suggests that there must be relatively little allochthonous sediment input to these lakes. It is not known whether or 41 not this process has been operative throughout the Holocene (i.e. between Lake Alsek phases). 3.1.2 Sediment Sinks Throughout the Neoglacial, a strongly contrasting range of erosional and depositional processes have been active in the Lake Alsek basin. Storage sites in the sediment cascade are likely to exist at only a very few locations in the basin. Discussion of sediment sinks will be in terms of those which existed during Lake Alsek phases, and those which existed during the non-inundated phases. Lake Alsek Phases During a Lake Alsek phase, the entire basin acts as a deep-water glaciolacustrine sediment trap. Sediments deposited from suspension are relatively well distributed throughout the lake and reflect processes characteristic of a glaciolacustrine environment., During an outburst flood, turbulent and tractive forces probably increase dramatically for at least some part of the drainage as the lake water quickly drains from the narrowly confined basin (section 2.7), and the processes of deposition may reflect this. The development of currents associated with lake drawdown means that transported sediments are possibly better sorted and coarser than most of the glaciolacustrine sediments, and the spatial variability of sediment characteristics would be much higher. Depositional sites are normally associated with regions of flow expansion and flow eddies at scales corresponding to channel width, and on a smaller scale, with recessed hollows (i.e. negative steps) in the channel bottom and with the lee side of flow obstructions (Leeder 1982, p.62). Non-inundated Phases During intervening non-inundated phases, which are probably relatively long term (Clague and Rampton 1982), most of the basin is a valley-bottom, terrestrial environment lying above local base level. Because of this, most of the empty Lake Alsek basin is no longer a sediment sink and is dominated by normal fluvial processes in the valley bottom. Exceptions to this are residual lakes and ponds in which lacustrine processes are still 42 important, although at a reduced scale. These water bodies are limited in large part to very small kettle-type ponds, flood-scoured ponds, and one glacially eroded lake, all of which combined occupy only a very small proportion of the empty Lake Alsek basin. Except for one deep lake ( D m a x > 21 m) which occupies a bedrock basin, most of these groundwater and snowmelt fed ponds are shallow ( D m a x < 3m) and eiitrophic with very little allochthonous sediment input. These small ponds also have the potential to preserve both lake phase and non-lake phase sediments. This is partly because they are recessed into the valley floor, providing some resistance to scouring during a lake draining event, and partly because they are not affected by contemporary riverine processes on the valley floor. In particular, the form of the kettle-type lakes has not been significantly altered since the end of the Kluane Glaciation. Additionally, pedogenic processes which apparently have been destructive to the fine structures in Lake Alsek sediments (Clague and Rampton 1982) are not active in these lacustrine environments. Completely sedimented kettles in close proximity to kettle lakes revealed sediments with most structures either completely destroyed or lacking in resolution when compared to their stratigraphic counterparts from the kettle lakes. Except for very limited rootlet penetration, bioturbation is not significant in any of the cores retained for analysis. 3.2 Field and Laboratory Procedures Field investigations during summer 1984 consisted of (a) locating suitable sites, (b) measurement of the site geometry, (c) sampling of the sediments, and (d) transportation of samples to the lab. Field investigations during summer 1985 consisted of (a) further sediment sampling at the choicest sites, and (b) obtaining stratigraphically controlled samples for radiocarbon dating. Laboratory investigation consisted of (a) cutting and preparation of the samples, (b) description and photography of the samples, and (c) measurement of various sediment properties. 43 3.2.1 Sampling Strategy Unlike the Lake Missoula basin, no natural or man-made exposures of Lake Alsek sediments are known to exist. To obtain a comprehensive record, air photographs were used to locate sites where deposition and preservation of Lake Alsek sediments was likely to occur throughout the most extensive area of Lake Alsek inundation. A total of 37 sites were identified from air photographs, ranging in size from very small ponds a few tens of meters across to one large lake 5.6 km long. The physical characteristics and associated information for each site are listed in Table 3.2. On the basis of accessibility, spatial context within the basin, and water depth, 14 of these were selected for further investigation. Water depth is an indicator of overall value as a sediment trap and to a lesser extent, of preservation potential. After a preliminary field visit to each of these sites in July and August 1984, 11 were selected for coring (two of the 14 were found to be empty and one was bedrock bottomed). Each of the sampled lakes (with the exception of KM69.0) was sounded with a lead line along one or more transects fixed against recognizable points along the shoreline. Using these soundings and the shoreline shape taken from air photographs, bathymetric maps for each sample site could be approximated (Fig. 3.1). Samples were removed by piston corer or crust freezer sampler from the deepest areas of each lake. Sampling locations at each site are marked on Figure 3.1. Where possible, cores were replicated, but for the following reasons complete replication was not always achieved: 1) the sampler was unable to penetrate sand layers and gravel layers thicker than about 20 cm. The gravel layers are apparently discontinuous, thus allowing coring to take place in some areas but not in others even a meter or two apart; 2) logistical constraints, such as lack of time, broken equipment, and grizzly bears. The square-rod piston corer, modified after the Livingston design, incorporates a Table 3.2 C h a r a c t e r i s t i c s of the p o p u l a t i o n of Lake Alsek b a s i n lakes and ponds. H Newname Bk H (ra a s l ) A1rphoto Status Lmax(m) Wmax(m) Dmax(m) Bas i n(ha) Notes 1 KM 1 . 2 L 535 0+2 0 A23001-162 Vstd cored usef 1 130+2. 25+1.0 6 +0.5 5.0+1.0 bedrock and grav e l b a s i n 2 KMO. 5 L 590 0+5 0 A23001-162 Vstd 3 KM4 . 4 R 514 4+0 5 A23001-162 Vstd cored usef 1 80+2.0 40+1.0 2 9+0.2 6.0+1 .0 no Inflow, spruce age = 34 a 4 KM 4 . S L 652 4+5 0 A23001-162 Vstd 5 KM6 . 2 R A23001-162 Vstd 6 KM7 . D L A23001-162 7 KM8 .8 R A23001-162 8 KM10 .0 R A23001-162 Vstd 2+0.5 shallow, creek Inflow 9 KM10 .2 L 556 0+5 0 A23001-162 Vstd 10 KM 13 .4A L 558 0+1 0 A23001-162 Vstd s o l 1 p i t s i t e 11 KM 13 .4B L 543 0+5 0 A23001-162 Vstd 12 KM 13 .4C L 545 0+1 0 A23793-180 Vstd cored usef 1 22+1.0 16+1.0 2 3+0.2 1.0+0.5 'Swimming Bear' spruce age=63 a 13 KM 13 .40 L 536 0+1 0 A23793-180 Vstd cored usef 1 55+1.0 35+1.0 2 4+0.2 2.0+0.5 'Al1ce' spruce age = 54 a 14 KM 18 .8A L A23793-183 sm a l l , marl, bedrock b a s i n 15 KM 18 .8B L A23793-183 16 KM18 . 8C L A23793-183 smal1,mar 1 17 KM24 .2 R A23793-193 Vstd r i v e r l e v e l 18 KM28 .4 R A23793-189 Vstd cored 400.+5. 100.+5. 6 5+0.2 60.+10.0 'Long Lake' coarse bottom seds 19 KM28 .5 R 552 0+2 0 A23793-189 Vstd cored usef 1 100.+5. 50.+5. 2 0+0. 1 70.+10. stream Inflow 1.0 m 3 s _ 1 20 KM28 .9 R 574 0+5 0 A23793-189 Vstd cored 500.+5. 100.+5. 3 5+0. 1 50.+10. beaver dammed 21 KM30 .4 L A23793-191 Vstd marl, s m a l l e r than KM30.6 22 KM30 .6 L A23793-191 Vstd 4 + 1 .0 marl, bedrock b a s i n 23 KM30 .8 L A23793-191 Vstd 24 KM32 . 2 R A23793-191 near BM 'KFE', sandy 25 KM32 . 4 L A23793-191 26 KM32 .5 L A23793-191 Vstd 27 KM44 .6 R A23793-200 near Kask/Dez c o n f l u e n c e 28 KM47 .2 R A23793-203 29 KM50 .4 R 574 0+0 1 A23793-204 Vstd cored usef 1 100+3. 60+3 . 1 2+0.2 10.+2.0 spruce age = 85 a 30 KM51 .6 L A23793-204 31 KM52 .0 L 573 3+0 1 A23793-204 Vstd cored 600+20. 375+1.0 9 6+0.2 25.+5.0 'B1g bend' sandy bottom 32 KM53 .0 R 575 0+2 0 A23793-207 Vstd cored 1 5+0.2 33 KM62 .6 R A23000-189 Vstd 'Ha1's lake' 34 KM69 .0 R 645 0+1 0 A23000-192 Vstd cored usef 1 5600+50 1200+50 18.0+3. 3800.+100. 'P1ne Lake' no bathymetry done 35 KM73 .2 R 666 0+5 0 A23000-192 Vstd 36 KM75 : 2 L A23000-192 37 KM83 . 6 R A23000-192 tot = 37 23 Vstd 11 cored 7 u s e f u l s i t e s • a) These names r e f e r to d i s t a n c e s from l a t e N e o g l a c i a l terminal moraines at Lowell G l a c i e r measured along the r i v e r f l o o d p l a i n In stra1ght-11ne segments. b) Bk r e f e r s to the h y d r o l o g i c l e f t or r i g h t bank of the r i v e r . c) A l t i t u d e s determinations are explai n e d 1n text. d) A i r p h o t o names are taken from the Federal government a l r p h o t o s . e) Status r e f e r s to the u s e f u l n e s s of the s i t e In t h i s study. f ) Pond dimensions are maximum length (Lmax), maximum width (Wmax), and maximum depth (Dmax). Bas i n area f o r each lake was estimated from a i r photos except f o r KM69.0 which was measured from NTS 115 A/14 East. 45 Figure 3.1 Bathymetry and sediment sample locations of the Lake Alsek basin sample sites. Outlines are from air photographs, and depth soundings and sample locations determined by relationships to recognizable features. 46 one metre stainless steel coring tube of 4.98 cm inside diameter. The crust-freezer corer, of a stainless steel wedge design, was filled with dry ice and ethanol, but was only used once because in most cases the uppermost sediments were successfully sampled using the piston corer. A complete description of the corer designs, the coring methods, including drill platform stabilization, core hole casing procedures, sediment-water interface sampling, core extrusion, and core handling and storage, is given in Appendix VTLI. A total of 26 cores, ranging in composite length from 0.08 to 4.91 m were collected from eight lakes during the summer of 1984. With the exception of one core from site KM69.0, none of the cores penetrated into the underlying Hypsithermal soil or Pleistocene till which presumably underlies the Neoglacial lacustrine sediments. Cores from the other three lakes sampled are excluded from the subsequent analysis for the following reasons: 1) KM28.4 was found to be beaver dammed. 2) KM28.9 is a deep, narrow lake bordered by steep cliffs of weathered Tertiary volcanic rocks which have fallen into the lake and made the lake bottom sediment too coarse to sample. 3) KM52.0 is a 9.6 m deep sandy bottomed lake with a surface elevation very close to river level elevation. There is no stream inflow (July 1984) and only a small outlet stream, so apparently this lake has groundwater inflow currents which may be responsible for removing fine grained sediment from the lake bottom making it too coarse to sample. 3.2.2 Sample Site Descriptions During filling and draining phases, sediments deposited at each site reflects its location in the Lake Alsek basin. During intervening phases the sediments only reflect the biologic productivity of the pond and ephemeral inputs of water and sediment. Site locations in the Lake Alsek basin are identified by their distance upstream of the Lowell Glacier ice dam. Only one site (KM69.0) lies in the Dezadeash Valley; the rest lie in the more constricted Alsek Valley portion of the Lake Alsek basin (in the Kluane Ranges). 47 The ponds sampled are mostly small, as are their catchments (Fig. 3.1, Table 3.2). All but two ponds have no permanent channelized input, although ephemeral channels were observed at some. The two exceptions are KM28.5, and KM69.0 (Pine Lake) which lies in an unglacierized basin of approximately 3800 ha. The morphological setting of some of these ponds and lakes is presumed to result from Lake Alsek drainings. Sites KM1.2, KM4.4, and KM53.0 are elongated lakes aligned with the valley axis found in scoured hollows possibly eroded by Lake Alsek flood waters during a major drainage event. Site KM28.5 and possibly KM50.4 are lakes impounded by outburst flood deposited gravel and cobble bars. Site KM50.4 may alternatively be an abandoned backwater channel of the Dezadeash River. Sites KM13.4A, KM13.4C and KM13.4D, along with all of the soil pit sites, are kettle-type lakes and are found very close together with several similar lakes in a kettle field. Site KM69.0 is an elongated glacially eroded bedrock basin with a long axis trending NW-SE in line with the Skakwak Valley. 3.2.3 Laboratory Procedures Upon arrival in the lab, all core boxes were opened and checked for longitudinal compaction or damage during transport. All cores were individually unwrapped and split longitudinally in an ABS core tray. One half was immediately rewrapped and replaced in the core box. The other half was scraped clean with a knife, photographed with B + W film, and described on a long sheet of paper placed beside the core. The observations include descriptions of stratigraphic units, boundary types, texture, structures, Munsell colours, and other distinctive features. Some cores were allowed to dry for several weeks before being photographed so that details of structure could be more easily seen (see Appendix VHI for discussion of methods). Samples were taken directly from the exposed core using a stainless steel cylindrical s.ubsampler with an internal volume of 2.850 ± 0.025 cm These samples were weighed to determine wet bulk density, then oven-dried overnight at 105 to 48 determine dry bulk density and moisture content. The weight loss after drying is assumed to be due to the loss of interstitial water. Organic matter content was determined by loss-on-ignition (L.O.I.) in a 450 oven for 6 hours. All samples were weighed at room temperature and at room humidity to the nearest 0.01 g (Hakanson and Jansson 1981). Particle size analysis was conducted following the standard procedures of Folk (1980; moisture replicate method) and of ASTM D422-72 (1972). Moist samples were removed from the core, thoroughly mixed with a dispersant, and allowed to soak for about 24 hours. Dispersants used were 50 ml of a 5% Calgon solution for the first 18 samples, then 50 ml of 5% sodium hexametaphosphate solution for the following 12 samples. In no case was sediment flocculation observed. The sample was mixed vigorously in a mechanical stirrer, then with a spray bottle was wet-sieved first through the 2 mm screen, then through the 0.063 mm screen. Gravel and sand components were oven dried and sieved at 1/2 phi intervals with the weight of each fraction recorded to the nearest 0.01 g. The < 0.063 mm material was washed into a one-litre sedimentation tube and topped up to one litre with distilled water. When six samples were prepared in this way, a bench top hydrometer analysis of settling was undertaken over either 24 or 48 hours using the methods of ASTM D422-72 (1972). Hydrometer sample sizes ranged in dry mass from 9 g to 16 g, except for three unexpectedly large samples of 55 g, 71 g, and 82 g. The sieving and hydrometer results were analyzed using the in-house computer programs GRSIZE.S and SEDIMENT. Results were plotted as frequency distribution histograms and as cumulative curves using standard arithmetic ordinate graph paper. The Folk (1980) inclusive graphic statistics were calculated for all samples which were analyzed down to Dg, and the Folk (1980) graphic statistics were calculated for all others. The computational forms of these statistics are those provided in Folk (1980). 49 3.3 Sedimentary Structures Sedimentary structures observed in some or all of the Lake Alsek basin cores include: parallel laminations, deformed beds, cross laminations, erosional contacts, and massive sediments. Severe limitations on the ability to observe and recognize sedimentary structures were imposed by the narrow (essentially one-dimensional) view of the sediment sequences offered by the small diameter cores. Parallel Laminated and Thinly Bedded Sediments Thinly laminated muds, usually consisting of darker-coloured clay laminations intercalated with lighter-coloured silt laminations, were observed in most cores (Fig. 3.2a). Individually laminated couplets range in thickness from less than 1 mm to more than 3 cm. Closer examination and several thin sections of some of these structures reveal distinct layering with no apparent gradation from one layer into another. All laminations in KM69.0 and most in the rest of the sampled lakes exhibit silt/clay thickness ratios of 1:1 to 1:10. These sediments have the appearance of varves (annually laminated rhythmites) and were most likely deposited in a glaciolacustrine environment, but no independent test (see for example Leonard 1986; Zhao et al. 1984) was undertaken to determine if they are in fact true annual varves. The variability in lamination thicknesses may be due to fluctuations of inflowing water which are directly related to sediment influx (Church and Gilbert 1975, p.88; Pickrill and Irwin 1983, p.73) or to different filling heights associated with various lake phases. Variability in clay layer thicknesses is unusual in true annual varves (Ashley 1975, p.318) and may be due to different filling heights associated with various lake levels. Deformed beds In many places, layers of sediment were observed to lie in orientations significantly different from layers both above and below. Disturbance due to coring is unlikely because of the preservation of fine laminations both above and below the deformed sections (Fig. 3.2b). Post-depositional disturbance of these layers may be due to a number of causes, 50 51 52 Figure 3.4 Massive beds, carbonaceous muds, and other sedimentary characteristics. KM13.4C (15-94) Sm SS Fl(3) Fm 3mm 'm n 53 such as subaqueous slumping, dewatering, compaction, and disturbance by tractive currents associated with outburst floods or with turbidity currents. Cross-laminations Cross-laminated sands were identified in sediments from KM13.4C and tentatively in KM53.0 (Fig. 3.3a). However, cross-laminations were not observable until the sample had been thin-sectioned so it is conceivable that this structure is more common in the Lake Alsek sequences than has yet been observed. The thin section is small but it appears that stoss side preservation of the ripples is absent, a diagnostic characteristic of the type A ripple lamination of Jopling and Walker (1968, p.973). Their environmental interpretation of this structure is a lower flow regime current transporting abundant sediment as traction load. Alternatively, it may be that cross-laminations were formed through gradual infilling of the pond. Sand avalanching down the sides of the pond would form similar structures without a causal relationship with ambient hydrodynamic conditions. The small size of the thin section makes these interpretations tenuous. Erosional Unconformities In many instances, erosional surfaces could be identified. Usually, these are sharp boundaries between an underlying mud deposit and an overlying sandy or silty deposit. Erosional surfaces are differentiated from depositional hiatuses by eroded surfaces of the more cohesive muds, and by mud rip-up clasts in the sand deposits, often with thin laminations still preserved (Fig. 3.3b). The erosional event associated with these unconformities is a significant aspect of past environments. Erosional unconformities are an important source of paleoenvironmental information in stratigraphic sequences (Ager 1981; Dott 1983; Hilton-Johnson 1982). Despite this significance, an erosional surface represents a time gap of unknown size in the sedimentary record and may thwart a complete paleoenvironmental reconstruction. 54 Massive sediments (sands and fines) In most of the Lake Alsek sequences, layers of massive sediments composed of particles ranging in size from sands to silts and clays were observed (Fig. 3.4). The sediments were examined both while wet and again after air drying, but it is nonetheless possible that structures were present but not observed. Indeed, cross-laminations in a sand layer were only detected after thin sectioning of the sample. Presuming the sediments are massive, then these layers were either deposited without internal structure, or the depositional structure has since been destroyed by bioturbation. No bioturbation structures were observed in any of the sequences, therefore the massive sediments must have arisen through rapid sedimentation (i.e. dumping), where there is ^sufficient time for bedforms to develop, and the sediments were transported in suspension (Passega 1964, p.842). 3.4 Core Lithostratigraphy Upon initial examination of the cores, it was apparent that similar facies were repeated many times over in a single sequence, and from core to core. On account of the complexity and number of stratigraphically distinct layers, laboratory characterization of each layer by textural and/or compositional analyses was not practical. Instead, a systematic characterization based on a rigidly defined lithofacies scheme was undertaken. Using this scheme, visual appraisal of grain size, bedding, and sedimentary structures is used to subdivide the sedimentary sequence. The lithofacies stratigraphy provides a framework for subsequent laboratory investigations. This objective approach has been used successfully for a variety of sedimentary deposits (see for example: Walker 1979; Eyles et al. 1983). 55 3.4.1 Lithofacies Type Description Scheme The descriptive scheme used for the study of Lake Alsek sediments is based on principles set out by Eyles et al. (1983) in their development of a lithofacies scheme for glacial diamict sequences. They note that this formally defined glacigenic lithofacies scheme is compatible with the fluvial and glaciofluvial lithofacies scheme described by Miall (1978, p.598). The essence of an objective descriptive scheme is the systematic documentation of grain size, bedding, and sedimentary structures based on visual appraisal of the sedimentary sequence. This is accomplished through the definition and application of a comprehensive lithofacies code. With slight modifications, the lithofacies scheme of Eyles et al. (1983, p.396) is sufficiently logical, consistent, and portable to be of value in the study of Lake Alsek sediments. The lithofacies code and associated symbols used in the analysis of Lake Alsek sediments is illustrated in Table 3.3. At the broadest hierarchical level, the code uses the capital letters D, S, and F to represent diamict, sand and fines, respectively (note: diamict is defined here as any poorly sorted clast-sand-mud admixture regardless of depositional environment, whether glacial, paraglacial, periglacial, or non-glacial, terrestrial, or aqueous (Eyles et al. 1983, p.394)). After this designation, one to three mnemonic lower case letters are used to identify combinations of internal characteristics. In diamict these are; m for matrix supported, c for clast supported, -m for massive, -s for stratified, and -g for graded. In sands these are; r for rippled, t for trough cross-bedded, h for horizontal lamination, m for massive, g for graded, and d for deformation. In fine-grained layers these are; 1 for laminated, m for massive, and -d for dropstones (Table 3.3). To ensure a comprehensive scheme, the lithofacies code was slightly modified as follows: 1) the FI code includes a count of the dark/light couplets enclosed in brackets; 2) grading is indicated on the vertical profile log with a vector to indicate the direction of fining and the length of the graded section; and 56 Table 3.3 Four part lithofacies code and symbols. Diamict clasts can be varyingly represented allowing clast size and its variation to be recorded. Laminations are counted so that the number of couplets can be recorded with the FI code. Grading is recorded with a vector indicating the direction of fining and length of graded section. After Eyles et al. (1983). FACIES CODE SYMBOLS Diamict, D: Dm : matrix supported Dc : c l a s t supported D-m : massive D-s : s t r a t i f i e d D-g : graded s i z e of symbol i s pro p o r t i o n a l to the c l a s t s i z e Sands, S: Sr : r i p p l e d St : trough cross-bedded Sh : h o r i z o n t a l lamination Sm : massive Sg : graded Sd : s o f t sediment deformation sand Fine-grained (mud), F; FI : laminated (// of couplets) Fm : massive F-d : with dropstones laminated massive Organic, 0: 0 : carbonaceous mud Contacts Er o s i o n a l : Conformable: 57 3) an O designation (mnemonic for organic) is added to describe unoxidized, black carbonaceous mud with plant structures sparse or non-existent. Schematic representation of a coded stratigraphy is accomplished through a set of lithofacies symbols (Table 3.3). These symbols have limited resolution, therefore lithostratigraphic diagrams of Lake Alsek cores include the designated codes. Once sedimentary sequences are rigorously described in this manner, "facies assemblages, facies sequences and lateral facies relationships may be defined, and are used to interpret depositional environments" (Eyles et al. 1983, p.394). It is emphasized that penological and textural analyses are secondary in this approach. Laboratory investigations of sediment properties is therefore presented after core lithostratigraphies. 3.4.2 Lithostratigraphy of Lake Alsek Cores Coded lithofacies diagrams for each core from the eight ponds that yielded useful sediment were constructed with the aid of full-scale photographs and the logged core descriptions. Cores are grouped by sample site and are designated by the sample site name and the vertical length (in centimetres) of the core below the sediment-water interface. Schematic representations of the cores from each site are illustrated in Figures 3.5a through 3.5h. Core lithostratigraphy is discussed briefly in this section. KM 1.2 This pond is bounded on two sides by bedrock and on a third by a gravel bar (Fig. 3.1). Although only 1.2 km from the maximum Neoglacial moraines of Lowell Glacier, it is more than 28 m above the surveyed river level. Upon drilling it was found that a coarse gravel layer prevented penetration of the corer beyond depths of 10-20 cm below the present day organic layer and, despite numerous attempts, only one core was recovered (Fig. 3.5a). Massive fine-grained (Fm and Fmd) lithofacies are interbedded with a single layer of massive diamict (Dmm) and a single layer of carbonaceous mud (O). A sandy lithofacies 58 Figure 3.5 Lithostratigraphy of the sediment sequences extracted from each of eight sample sites. Lithofacies coding and symbolic representations as in Table 3.3. Scale bars refer to depths below water-sediment interface. Site altitudes are of water surfaces, (a) Site KM1.2 (535 m asl). KM1.2 (535 m asl) Depth (cm) 10 Fm Fmd Fm Dmm 1-20 M M P M MHMMiMi ':::::-:\o::.. ;-;;;.• 30 F l ( l l Sm Fm Fl ( l Fm Fl(9; FI(IO: Fm Sm S S — " Fm Sm Fm fg Fm FldOt F i h j Fit? Fm FU4 Fm Fl(7; Fm Sg 0 Fm Sm Fl(3)| Fm «S Sg 0 Fm Fl(8) Sm Fl(6) Sm Fm Fl(3) 0 Fm Sg Fm » Sml \ Fid (6 Fm F l ( l 0 Fm Fl(7 Fm Fid (8 Fm Sg Fm Sm Fm Fl(2) Sg "0 ,SS^ " Fm Fl(6)| Fm Fl(3)| Sm Fm SS SS SS Dmg Fm Sg ,*-^ Fm Sg Fld(4: Fm F1(2J Fm . — £ ° 0 Fm FI(6: ft Fm Fmd Fl(7 0-92 «S 6S SS 0-100 77-115 62-114 61 Figure 3.5 (cont'd) (d) Site KM13.4D (536 m asl). o tssa&igj sg 15-94 62 Figure 3.5 (cont'd) (e) Site KM28.5 (552 m asl). r 10 Depth I (cm) Y 20 I 30 0 Sm Fm| Fmd Fm .SS 0-28 Fm t Sm Fm Sg Sm Fl(3) •SS 28-52 Fm Dm n'-'Pi-52-73 63 Figure 3.5 (cont'd) (f) Site KM50.4 (574 m asl). Depth (cm) 10 L20 •30 Sm Fm Sm Fl(2 Sm Fl(6 Sm Fm 0 F l ( l l Sm Fl(2 Sm F l ( 3 * Fm Sm 0 SS 114-208 •SS •SS Fm Sm Fm Sm ommm Sm Fm 3-46 SS 48-111 64 Figure 3.5 (cont'd) (g) Site KM53.0 (575 m asl). Depth (cm) 10 20 I 30 Sm Sm Sg Sm Sr 0 Sm 0-42 • SS !!Ill111 SS Sm 42-85 65 Figure 3.5 (cont'd) (h) Site KM69.0 (645 m asl). r o Depth (cm) 10 • 20 30 Fm 22-59 125-220 25-123 333-429 66 without any observable structure (Sm) at 33-35 cm stopped coring at the 35 cm depth. A layer of massive matrix supported diamict (Dmm) observed at 16-22 cm depth is probably the layer which stopped coring at other locations. Contacts appear to be conformable in this sequence. KM4.4 A steep bedrock slope on one side and gravels on the others impound this pond. Poor vertical and horizontal control of the core locations was due to choppy surface water conditions at the time of coring. At 4.4 km from the Lowell ice dam and at less than 5 m above river level, this site is probably the most sensitive to very small Lake Alsek pondings, but may be subject to scour by draining lake waters because of its location at a constriction in the valley sides (Fig. 2.4). Many layers of rhythmically laminated fines (FI), massive fines (Fm), and massive silty sands (Sm) characterize these sequences and are especially finely stratified at the deepest levels. Several occurrences of graded sandy lithofacies (Sg) are also found at depth. The deepest core (130-141) was retrieved from the same borehole as the next to deepest core (82-135) so, despite the depth measurement, must underlie the 82-135 sequence. The contorted boundary in this core is due to coring disturbance; coring was stopped by the massive silty sand (Sm) found at the base of this core. When the cores are stratigraphically correlated (section 4.5.1) and combined into a single sequence, seven separate carbonaceous mud layers (O), and eight sharp (erosional) contacts at the base of sandy lithofacies (S) can be identified (Fig. 3.5b). KM13.4C This is one of several kettle-type ponds in an area of hummocky stratified drift deposits. Of the 5 cores recovered from this site, the deepest reached a depth of 1.2 m but there is one core which is actually stratigraphically deeper by about 0.10 m (Fig. 3.5c). All of the cores contain a detailed and finely stratified sequence of lithofacies types. The massive silty sand (Sm) lithofacies predominates in terms of total length, followed by 67 the massive fine-grained (Fm) lithofacies. Many layers of rhythmically laminated fines (FI) and massive fines (Fm) occur, and in seven places contain dropstones; when the cores are correlated four of these are from stratigraphically equivalent layers. These sequences can be easily correlated using stratigraphic marker layers; the composite sequence has a total of six carbonaceous mud (0) layers and seven sharp erosional contacts. Also, a distinctive matrix supported diamict (Dm) occurs in four sequences at stratigraphically equivalent locations, but appears to be massive (Dmm) in two occurrences, stratified (Dms) in another, and graded (Dmg) in the fourth. Because stratigraphic correlations between cores at this site are reliable (section 4.5.1), it is likely that these are spurious differences. Instead, they reflect an inadequacy of narrow cores for detailed sedimentological analysis (Fig. 3.5c). . Field investigations during summer 1985 were made exclusively at this site. The pond was pumped dry and a pit dug 1.95 m into the sediments so that (a) greater exposure could be achieved, (b) sediments too coarse to core could be sampled, and (c) samples large enough for radiocarbon dating could be collected. Samples of the carbonaceous mud (O) layers at 30 cm depth and at 1.05 m depth were collected in whirlpak bags for radiocarbon dating. To assess the thickness of sediments at this site, a soil auger was used to probe beyond the pit floor. Several layers were encountered during augering; these are presumably composed of layers of mud and sand. At a depth of 2.25 m, a gravel layer at least 20 cm thick was encountered. Artesian groundwater flow from this layer filled the pit in a matter of minutes, making further sampling impossible. KM13.4D This site is another kettle-type pond in the same area as KM13.4C, but is approximately 10 m lower in elevation, and thus was presumably more susceptible to scour by draining lake water, but was also more sensitive to small Lake Alsek events. Reaching a depth of 1.15 m, four cores were recovered from this site. 68 These sequences exhibit similar lithofacies variability to those found at the KM13.4C site. However, the massive fine-grained (Fm) lithofacies predominates the length of these cores, and there are many more occurrences of the laminated fine-grained (FI) lithofacies. Distinctive stratigraphic layers permit unambiguous correlation between these cores. The lowest 20 cm of the composite sequence exhibits finely stratified layers of massive fine-grained (Fm) and laminated fine-grained (FI) lithofacies at a scale similar to the bottom-most sediments observed at KM4.4. Also, the composite sequence exhibits four carbonaceous mud (O) layers and seven erosional contacts, all of which underlie a sandy lithofacies which is either massive (Sm) or graded (Sg). Structure in the sandy lithofacies is difficult to detect in the narrow cores and may be present but unobservable. Matrix supported diamict (Dm) is massive (Dmm) in one occurrence and graded (Dmg) in two others in the composite sequence (Fig. 3.5d). KM28.5 This pond is impounded by bedrock on the west side and a gravel bar on the eastern side. This site had an inflowing stream of about 1 m3s"^ (on July 31 1984), and no outflowing stream; the basin may be leaking through the gravels of the bar (outburst flood related) which impounds it and possibly removing fine-grained sediment in this way. A single sequence 0.73 m long in three parts was recovered from this pond. Massive sand layers (Sm) made coring difficult at this site and coring was eventually stopped at a depth of 0.73 m by an impenetrable layer of massive matrix supported diamicton (Dmm). Massive fine-grained (Fm) lithofacies dominate this core, which also exhibits one layer of laminated fine-grained (FI) lithofacies and two layers of sandy (Sm and Sg) lithofacies. No erosional contacts or carbonaceous mud (O) lithofacies were observed (Fig. 3.5e). KM50.4 This site was less than 1 m above the Dezadeash River level measured in late July 1984, and may be inundated by overbank floods in the course of normal fluvial processes. 69 A single core 2.08 m long recovered from this site consists mainly of massive silty sands (Sm) and massive fines (Fm) between zero and 1.55 m depth. Beneath this depth are several more layers of these two lithofacies interstratified with fine-grained laminated (FI) lithofacies and two carbonaceous mud (O) lithofacies, one of which is at the core bottom. A thin layer of shells occurs at about 0.94 m depth (Fig. 3.5f). KM53.0 Similar to site KM50.4, this site was very close to the level of the Dezadeash River in late July 1984. The sediments recovered in 0.85 m of core are exclusively massive, graded, and in one case, rippled sandy lithofacies (Sm, Sg, Sr) and organic detritus. The organic detritus is identified on the core log as a carbonaceous mud (O) but is actually composed of large pieces of woody detritus and is not black (Fig. 3.5g). KM69.0 The largest lake investigated in this study, this site is also most distal and at the highest elevation and therefore is least sensitive to pondings of Lake Alsek. Because of this and its location in the wide valley of the Shakwak trench this site is considerably less susceptible to erosion from outburst flood related currents. Thus it is probably the only site with an undisturbed record of the complete Neoglacial period. A single continuous core 4.91 m long was recovered from this site. Rhythmically laminated fines (FI) with occasional dropstones (Fid) predominate in this sequence. In addition, massive fines (Fm and Fmd), and some matrix supported diamicton (Dm) lithofacies also occur as interbeds. In the massive fine-grained (Fm) lithofacies near the core top are several occurrences of mollusc shells. Sandy (S) lithofacies and the carbonaceous muds (O) are absent from this core and contacts are conformable except for the sharp contact which overlies the lithofacies at the core base (Fig. 3.5h). The base of this core is a massive fine-grained lithofacies with abundant dropstones (Fmd) which exhibits a marked and abrupt change in density, colour, and moisture content 70 from the rest of the core (section 3.5). In addition, the abundance of dropstones is much greater than in any other occurrence noted and may even be more aptly identified as a massive diamicton (Dm). Indeed, subsequent grain size analysis of this sediment (section 3.5.3) indicates that this is much coarser (containing several gravel sized clasts) than any of the other fine-grained lithofacies. Because of these characteristics, this particular lithofacies occurrence may be a Lake Champagne deposit (noted elsewhere as "stony silty clay"; Clague and Rampton 1982, p. 107) and therefore will not be considered as a Neoglacial deposit. 3 . 5 Sediment Properties To further characterize Lake Alsek sediments, a limited laboratory investigation was undertaken that included measurements of wet and dry bulk density, moisture content, loss-on-ignition, and particle-size analysis on a representative set of samples. Bulk density and water content are expressions of sediment compaction, and are also related to sediment texture. As unconsolidated sediments become progressively buried, density normally increases and water content decreases. This trend may not be apparent if the textural control over these parameters is strong. Loss-on-ignition (L.O.I.) is a surrogate measure of the organic content of a sediment (Hakanson and Jansson 1981, p.76). Organic contents of lacustrine sediments are higher in lakes with high productivity and with redox conditions that inhibit degradation of organic materials (Hakanson and Jansson 1981, p.80). Other factors which control the variability of organic content of lake sediments are the rate of sedimentation and the degree of sediment compaction (Hakanson and Jansson 1981, p.75). Also, the influx of allochthonous organic matter into the basin has some control. Particle-size characteristics of a deposit are used to aid in identification of the sedimentary process and environment at the time of deposition. While the use of texture alone to identify process or environment is not possible, strong interpretive statements 71 about process and environment can sometimes be made. Complicating factors in this kind of interpretation are many; for example, McLaren (1981) states that the source sediment is the chief control over the grain size distributions of most deposits. Sly et al. (1983) on the other hand, maintain that the main difficulty in using textural characteristics as a means of identifying depositional environment may be that equilibrium between the sediments and ambient hydrodynamic conditions cannot be assumed. Particle-size analyses were undertaken in the study of Lake Alsek sediments with two primary objectives: 1) to elicit quantitative data on the sedimentological properties of the various deposits, and 2) to attempt to distinguish fundamental differences between depositional environments associated with the various Lake Alsek deposits. The parameters listed above were measured on selected samples from some cores, but were mainly undertaken as a stratigraphic investigation of cores KM13.4C (35-120 cm), KM13.4C (15-94 cm), KM13.4C (0-77 cm) which are considered the longest and highest resolution records. Cores KM13.4D (0-92 cm), KM13.4D (0-100 cm) and KM4.4 (0-94.5 cm) were also analyzed to investigate spatial variability of the above properties. Core KM69.0 (423-491 cm) was analyzed to investigate temporal variability in the sediments near what may be the Neoglacial/Hypsithermal boundary. 3.5.1 Densities, Organic Matter and Moisture Content The results of these analyses are tabulated in Table 3.4 for 33 samples taken from three cores. Moisture content was calculated both as a ratio of the wet and dry mass difference (i.e. mass of the water) to the dry mass (gravimetric wetness), and as a ratio of the wet and dry mass difference to the total wet mass. The second method is seen as more rational since it is otherwise possible to obtain moisture contents of more than 100%, which is misleading (Hakanson and Jansson 1983, p.74). 72 Table 3.4 Wet and dry densities, moisture contents, and L.O.I. Sample Wet Rho Dry Rho M.C. a M.C. a LOI (depth) (gem-3) (gem"3) (Mw/Ms) (Mw/Mt) (%) KM 13.4C(34-120) 40 Fm 1.91 1.51 26.3 20.8 1.7 42 0 1.41 0.79 79.9 44.4 6.2 48 Fm 1.91 1.51 26.5 21.0 1.4 54 0 1.36 0.77 75.9 43.2 5.8 58 FI 1.72 1.36 26.2 20.8 2.4 65.5 S 1.23 1.18 4.2 4.0 1.8 72.5 S 1.22 1.19 2.6 2.6 0.9 82 Fm 1.96 1.56 26.1 20.7 1.4 86.5 FI 1.99 1.58 25.5 20.3 2.0 89 Fm 1.63 1.23 45.0 31.0 2.2 94 S 1.95 1.52 28.6 22.3 2.1 98.5 S 1.68 1.29 30.2 17.7 1.9 105.5 Fm 1.93 1.53 26.7 21.1 1.6 112 FI 1.81 1.40 29.4 22.7 • 0.0 116 S 1.60 1.23 29.6 22.9 1.7 118 FI 1.80 1.32 35.6 26.3 1.1 KM69.0(425-491) 426 FI 1.27 0.60 109.9 52.4 4.2 437 FI 1.25 0.55 125.9 55.8 3.3 454 FI 1.41 0.78 80.7 44.7 2.8 472 FI 1.53 0.95 61.9 38.2 1.5 479 Dm 2.22 1.81 22.9 18.6 1.2 485 Dm 1.85 1.44 28.6 22.2 1.2 490 Dm 1.67 1.58 44.5 30.8 2.5 491 Dm 1.95 1.56 24.4 19.6 1.4 KM13.4D(0-92) 12 0 1.54 0.92 67.6 40.3 2.7 17 Fm 1.89 1.31 45.2 31.1 1.1 37 Fm 1.88 1.35 39.0 28.0 1.6 47 FI 1.98 1.51 31.7 24.1 1.2 60 Fm 1.89 1.41 33.8 25.3 2.0 69.5 S 1.97 1.64 20.6 17.1 0.6 74 FI 1.87 1.43 30.6 23.5 2.3 82 FI 1.86 1.38 34.9 25.9 1.0 89 FI 1.90 1.43 33.2 24.9 1.5 a) Moisture content notation: Ms is the dry mass of the solids, Mt is the total mass for the entire sample, and Mw is the mass of the water (Mt - Ms). 73 Examination of the table of results does not suggest any temporal trend within each core, nor any significant differences between cores in density or water content (Table 3.4) . A down-core reduction in moisture content and increase in wet density might have been expected due to sedimentary compaction, but this is not apparent. Textural control of density and moisture content can be examined when the samples are summarized into texturally similar groups. Using the lithofacies assemblages discussed in section 3.6, large differences in water content and density become apparent (Table 3.5). Water content of the carbonaceous muds group (0) is the highest (42.6%), followed by the laminated fines (FI) group (31.6%), then the massive fines (Fm) group (24.9%), the diamict (D) group (22.8%), and finally, by the sands (S) group (14.4%). The bulk density variations are inversely related to the water content variations, as might be expected (Table 3.5). Loss-on-ignition values range from 0.0% in one sample of a.laminated fines (FI) to 6.2% in a sample of the carbonaceous mud (O) group. When the samples are grouped by major facies types (section 3.6) there are significant differences between L.O.I, data for the carbonaceous mud (O) lithofacies (average L.O.I, of 4.97%) and all other lithofacies (Table 3.5) . This result confirms the earlier affirmation of the carbonaceous mud (O) as a separate lithofacies on the basis of organic content. No measurements were undertaken to determine if the source of the organic material is autochthonous or if it is land derived detrital organic material. 3.5.2 Grain Size Analysis The grain size frequency histograms and cumulative plots of the 30 samples analyzed are presented in Appendix II. The results of grain size analysis confirm the visual appraisal of texture used to aid in lithofacies description in section 3.4. Classified on a textural ternary diagram (Folk 1980), the most common sediments are silty sands, sandy clays, silts, and muds (Fig. 3.6). Various statistical parameters were computed Table 3.5 C h a r a c t e r i s t i c s of Lithofacies Groups Facies Group Width (cm) Wet rho g/cm5 Dry rho g/cm3 L.D.I. (%) Mstur cnt (Mw/Mt) Structure Upper Boundary Lower Boundary Colour Other D (mean) (SD) 2.5 1 . 9 1 .92 0. 23 1 .60 0.15 1 . 58 0.62 22.8 5.5 diamicton di ffuse abrupt 1/2 grain gray mtrx varied c l a s t 1i thology S (mean) (SD) 4.8 4 . 2 1.61 0. 33 1 . 34 0.19 1 . 50 0.60 14 . 4 8.9 beddi ng graded abrupt eros i ve 5Y 4/0-2 dark gray occ. clay rip-up c1 asts F± (mean) (SD) 1 . 4 0.9 1 . 70 0. 27 1.19 0. 37 1 . 94 1.14 31.6 12.7 p a r a l l e i 1 am i ns sharp trncated graded abrup t 5Y 5/1-2 gray varves ? dropstones Fm (mean) (SD) 4.0 2 . 1 1 . 88 0. 10 1 .43 0.12 1 .63 0. 35 24 . 9 4.6 mass i ve di ffuse abrupt d i f fuse abrupt 5Y 6/0-1 1t gray homogeneous mud dropstones 0 (mean) (SD) 1 . 3 0.7 1 .44 0.09 0.83 0.08 4.97 1 .80 42 . 6 2. 1 trncated di ffuse abrupt bl ack oxidizes to orange 1) Boundary descriptions are from Gardiner and Dackombe (1983, p.103) 2) Widths measured on only a se l e c t i o n of facies occurrences. 3) Mt is the total mass, Mw is the mass of water. 75 Figure 3.6 Classification of Lake Alsek sediments according to their sand-silt-clay composition (Folk 1980). The sand/silt boundary is at 0.063 mm and the silt/clay boundary is at 4 microns. The gravel fraction of samples is included as part of the sand component. See Table 3.6 for numerical values of statistical grain size parameters for the 30 Lake Alsek samples. Sand Clay:Silt Ratio Table 3.6 Numerical values of s t a t i s t i c a l grain size parameters for 30 Lake Alsek samples Sample gravel sand S i l t clay Median Mean Sort 1ng Skewness Kurtos i s D99 Descr i pt i on (%) (%) (%) (%) (phi) (phi ) (phi ) (mm) KM13.4C(35-120) 38-40 65 0 35.0 7 .09 7 11 1 29 0.0 1 17 0.063 49-51 73 1 26 . 9 6 .26 6 39 1 57 -0. 18 0 92 0.044 54-56.5 73 7 26.3 6 .46 6 63 1 42 -0. 18 1 02 0 .063 91-93 54 8 45.2 7 . 26 7 37 1 45 -0.02 0 89 0.044 106-108 60 8 39 . 2 7 .06 7 19 1 47 -0.5 0 97 0.063 112-113 .5 66 3 33.7 6 .98 7 02 1 38 -0.02 1 00 0.063 103-106 42 . 3 22. 9 26 1 8.8 0 .81 1 29 4 16 -0. 20 0 59 11.3 90-99 16 . 8 78 4 4.9 4 .39 4 43 1 16 -0.21 1 33 0.17 62-77 85. 7 13 2 1 . 2 2 .54 2 61 0 94 -0.28 1 30 0.50 KM13.4C ( 15-94) 35-47 1 . 1 78 . 1 18 9 1 .9 2 . 72 2 78 1 03 -0. 19 1 22 0.50 65-69. 55-58 21 . 0 74 2 4.8 4 .40 4 37 1 27 -0.9 1 43 0.25 36-48 80. 0 17 9 2. 1 2 .69 2 77 1 08 -0. 28 1 40 0.50 72-74 15.5 39. 6 32 8 12 . 1 3 .03 3 08 3 65 0.0 0 80 5.7 87-91 34. 2 58 6 7 . 3 3 .80 4 24 1 60 -0. 53 1 72 0. 25 KM 13.4C (0-77) 17-47 85. 6 12 5 1 .9 2 . 57 2 65 0 86 -0. 27 1 28 0. 35 68-71 27 . 1 20. 9 32 9 12.7 3 .32 2 47 4 47 0.18 0 60 11.3 KM 13.4D (0-92) 37-39 52 0 48 .0 7 .47 7 43 1 29 0.07 1 06 0.063 39-42 51 1 48.9 7 .50 7 51 1 27 0.01 1 00 0.063 47-50 31.5 43 . 8 15. 5 9. 1 0 .97 1 34 3 58 -0. 22 0 95 8.0 80-83 55 3 44 . 7 7 . 32 7 43 1 25 -0.12 0 98 0.044 89-92 51 3 48 . 7 7 .46 7 54 1 45 0.01 1 07 0.063 65-71 28 . 1 66 7 5 . 2 4 . 16 4 30 1 39 -0. 26 1 10 0. 25 KM13.4D (0-100) 83-85. 62-65 53.0 13. 0 23 7 10. 3 2.51 -0 . 19 4 81 -0.64 0 56 22.0 KM13.4A ( 10-82) 75-82 19.2 34 . 9 36. 4 6.9 3 .28 2 10 3 91 0. 29 1 59 11.3 47-55 1 . 4 81 . 9 12 0 4 . 7 2 . 23 2 44 1 53 -0. 38 2 02 2.0 KM4.4 (0-94.5) 84-90 2 . 2 80 6 17.1 6 .06 6 22 1 35 -0. 19 1 13 0. 125 21-28 8.0 76 0 16.0 5 .70 5 78 1 73 -0.07 1 09 0.250 KM69.0 (423-491) 484-486 34 2 65.8 7 .99 7 96 1 31 0.08 0 88 0.044 424-426 32 2 67 .8 9 . 2 8 08 1 97 0.72 0 57 0.063 491-492 31.7 30. 3 23 1 7 . 1 1 . 18 1 47 4 05 -0.14 0 67 11.3 491-494 46 .4 26 8 20 4 6.3 -0.40 O 26 4 46 -0. 25 O 70 22 .6 Table 3.7 Folk (1980) verbal equivalents for s t a t i s t i c a l grain size parameters of 30 Lake Alsek samples Sample Median Size Mean size SD Sorting Inc1 us i ve Graphic Descr ipt ion Skewness Kurtosi s KM13.4C(35-120) 38-40 f i ne s i l t f i ne s i l t poorly sorted synimetr i ca 1 1eptokut ic 49-51 med i um s i l t med i um s i l t poorly sorted negatively skewed mesokurt ic 54-56.5 medium s i l t medium s i l t poorly sorted negaitvely skewed mesokurti c 91-93 f i ne s i l t f i ne s i l t poorly sorted symmetri ca1 piatykurt i c 106-108 f i ne s i l t f i ne s i l t poorly sorted symmetr i ca1 mesokurt ic 112-113.5 f ine s i l t f ine si 11 poorly sorted symmetr i ca1 mesokurt ic 103-106 coarse sand medium sand extremely poorly sorted negatively skewed very p l a t y k u r t i c 90-99 coarse s i l t coarse s i l t poorly sorted negatively skewed 1eptokurt i c 62-77 fin e sand fin e sand moderately sorted negatively skewed le p t o k u r t i c KM13.4C (15-94) 35-47 fine sand fin e sand poorly sorted negatively skewed 1eptokurt i c 65-69, 55-58 coarse s i l t coarse s i l t poorly sorted symmetr i ca1 1eptokurt i c 36-48 fin e sand fine sand poorly sorted negatively skewed 1eptokurt i c 72-74 very f i n e sand very fine sand very poorly sorted symmetrical piatykurt i c 87-91 very f i n e sand coarse s i l t poorly sorted very negatively skewed very l e p t o k u r t i c KM13.4C (0-77) 17-47 fine sand fin e sand moderately sorted negatively skewed 1eptokurt ic 68-7 1 very f i n e sand fin e sand extremely poorly sorted p o s i t i v e l y skewed very p l a t y k u r t i c KM13.4D (0-92) 37-39 f i ne s i l t f i ne s i l t poorly sorted symmetr i c a l mesokurt i c 39-42 f i ne s i l t f1ne s i l t poorly sorted symmetr i ca1 mesokurt i c 47-50 coarse sand medium sand very poorly sorted negatively skewed mesokurt i c 80-83 f i ne s i l t f i n e s i l t poorly sorted negatively skewed mesokurt i c 89-92 f i ne s i l t f i ne s i l t poorly sorted symmet r i ca1 mesokurtic 65-71 coarse s i l t coarse s i l t poorly sorted negatively skewed mesokurt ic KM13.4D (0-100) 83-85, 62-65 pebbles very coarse sand extremely poorly sorted very negatively skewed very p l a t y k u r t i c KM13.4A (10-82) 75-82 very f i ne sand fine sand very poorly sorted p o s i t i v e l y skewed very l e p t o k u r t i c 47-55 fine sand fine sand poorly sorted very negatively skewed very l e p t o k u r t i c KM4.4 (0-94.5) 84-90 med i um s i l t med i um s i l t poorly sorted negatively skewed 1eptokurt i c 21-28 medium s i l t med ium s i l t poorly sorted symmetr i ca1 mesokurt i c KM69.0 (423-491) 484-486 f1ne s i l t f i ne s i l t poorly sorted symmetr i ca1 p l a t y k u r t i c 424-426 c 1 ay very f i ne s i l t poorly sorted very p o s i t i v e l y skewed very p l a t y k u r t i c 491-492 medium sand medium sand extremely poorly sorted negatively skewed piatykurt i c 491-494 very coarse sand coarse sand extremely poorly sorted negatively skewed piatykurt i c 78 from the measured particle-size distributions and are summarized numerically in Table 3.6, and for purposes of discussion, by verbal equivalents in Table 3.7. In addition the graphic statistics are summarized using the facies assemblages discussed in the next section (Table 3.8). Preliminary interpretations of some of these results are presented in an attempt to investigate and compare Lake Alsek depositional environments. Table 3.8 Grain size data summarized by facies assemblage Facies n Md M Z Skj K G (phi) (phi) (phi) D 6 1.48 1.68 4.10 -0.10 0.85 S 11 3.75 3.87 1.47 -0.24 1.36 FI 7 7.71 7.57 1.42 -0.11 0.94 Fm 5 6.83 6.94 1.44 -0.09 0.99 When grain-size distributions are plotted as cumulative graphs using phi units on the abscissa, the resulting curves can be of value in the comparison of sediment samples (Folk 1964, p.77). The form of the curve contains information regarding the type and intensity of the various processes that existed during the time of sediment deposition, and statistical values are taken directly from this curve. The position of the curve on the phi grain size axis is a rough indication of the environmental energy available at the time of deposition - positively shifted curves indicating deposition under lower energy conditions. The histograms and cumulative curves representing the sediment samples taken from the Lake Alsek cores are graphed so that all sediment finer than the limit of analysis is included in the final histogram bin (Appendix II). The grain size range and the modality of a sample are an overall indication of the average energy conditions related to that sample (Folk 1980 p.41). Examination of the 79 grain size histograms suggests that most of the D lithofacies group may be polymodal, the S and Fm lithofacies groups may be unimodal and some of the FI lithofacies may be bimodal (Appendix II). The Graphic Mean Mg (average grain size) is the best graphic measure for determining overall grain size (Folk 1980, p.41). The geological significance of this parameter is based on the concept that Mg is a function of the size range of available materials, and of the energy environment (current velocity or turbulence) of the medium transporting the sediment (Folk 1980, p.4). Lake Alsek sediments do not exhibit any continuous temporal trend in this parameter because of the sparse sampling of the core sequences. When grouped into major facies assemblages significant differences are apparent (Table 3.8). The coarsest sediments are the diamict (D) lithofacies group (Mg = 1.68 phi) followed by the sands (S) group (Mg = 3.87 phi), while the finest sediments are the massive fines (Fm) group (Mg = 6.94 phi) and the laminated fines (FI) group (Mg = 7.57 phi). Interpretation of these results is left until the next section (section 3.6). The Inclusive Graphic Standard Deviation (Oj) is a measure of the spread of the distribution about the mean and is interpreted as a measure of sorting, with lower values indicating better sorting (Folk 1980, p.42). This parameter is a function of at least four variables; size range of available materials, depositional process, current characteristics, and time, with the first of these considered to be the most important (Folk 1980, p.4). Again, temporal or spatial trends in the Lake Alsek data cannot be evoked here, and even when grouped by major lithofacies types (Table 3.8), three poorly sorted groups are quite similar (sands group (S): 0j = 1.47, laminated fines group (FI): (Jj = 1.42, and massive fines group (Fm): (7T = 1.44), the exception being the diamict (D) group (OT = 4.10) which is considerably less sorted. However, Folk (1980, p.5) notes that sediments with Mg values in the range of 6 to 8 phi are probably mixtures of two better sorted subpopulations, which may be the case for the laminated fines (FI) lithofacies group (Mg = 7.57) and the massive fines (Fm) group (Mg = 6.94). WhenCTj is graphed against Mg 80 Figure 3.7 Plot of mean grain size against sorting. Lake Alsek Sediments: SD vs Mz CL cn c o 00 4 . 5 -4 -3 . 5 3 -~ 2 . 5 O 1.5 0 . 5 Al j f k . j A A i A X : X ^ : ; ¥ - 1 0 1 2 3 4 5 6 Mean Size (phi) Legend A D G r o u p X S G r o u p O FI G r o u p 0 F m G r o u p u K M 6 9 . 0 b a s e Figure 3.8 Plot of mean grain size against skewness. CO V) 0> c cu on Lake Alsek Sediments: Sk vs Mz 0 . 5 - 0 . 5 -A| ; A 3 A • a S : \ X x x X X o i X - 1 0 1 2 3 4 5 6 7 8 9 Mean Size (phi) Legend A D G r o u p X S G r o u p O FI G r o u p 0 Fm G r o u p S K M 6 9 . 0 b o s e 81 F i g u r e 3.9 P l o t o f m e a n g r a i n s i z e a g a i n s t k u r t o s i s . 3 -'in 2 -1 -Lake Alsek Sediments: KG vs Mz A X * X X X 8 „ .^.B.nJffl.. A a A" A A ) D - 1 0 1 2 3 4 5 6 7 3 9 Mean Size (phi) L e g e n d A D G r o u p X S G r o u p • FI G r o u p B F m G r o u p H K M 6 9 . 0 B o s e F i g u r e 3 .10 P l o t o f s k e w n e s s a g a i n s t k u r t o s i s . Lake Alsek Sediments: Sk vs KG 2 - • 'in O X X ^ X E a a - 0 . 5 o.o Skewness — i — 0 . 5 L e g e n d A 0 G r o u p X S G r o u p • FI G r o u p S Fm G r o u p H K M 6 9 . 0 b o s e 82 (Fig. 3.7), facies with similar characteristics can be identified, and some interpretation of the characteristics of the depositing current can be made (Sly et al. 1983, p.221; Folk 1980, p.5). This graph indicates that none of the diamict (D) group nor the two samples from the base of core KM69.0 can be considered as sediments in equilibrium with hydraulic conditions at the time of deposition, (based on a comparision with Fig. 1 of Sly et al. 1983). The Inclusive Graphic Skewness (Skj) is a dimensionless measure of the symmetry of the particle-size distribution. This has been interpreted as a sensitive environmental signal related to the degree of reworking and mixing of different sediment subpopulations (Folk 1980, p.7). Most of the Lake Alsek sediments are symmetrical or negatively skewed indicating more coarse particles than expected from a normal size distribution. Positively skewed exceptions to this are three samples from the diamict (D) group and one sample from the laminated fines (FI) group, which appear to be outliers from the trends exhibited in a plot of Skj against M z (Fig. 3.8). The Graphic Kurtosi6 (KQ) is a measure of the peakedness of the sediment size distribution determined by measuring the ratio between the sorting in the central portion of the size distribution curve, and the sorting in the tails of the curve (Folk 1980, p.44). Values of this dimensionless parameter below 0.90 indicate a platykurtic (flattened) distribution and values greater than 1.11 indicate a leptqkurtic (excessively peaked) distribution, while values between these extremes are mesokurtic (Folk 1980, p.45). Lake Alsek sediments range between K Q = 0.56 and K Q = 2.02 with no strong trend in any one facies group, but when graphed against M z , associations by facies group are strong (Fig. 3.9). Folk (1980, p.7) states that plots of skewness against kurtosis are promising clues to environmental differentiation, a concept used by Sly et al. (1983) to identify sediments that are in equilibrium with prevailing hydraulic conditions, making interpretations of the formative origin of sediments possible. A plot of Skj against K Q for Lake Alsek sediments 83 (Fig. 3.10) shows that all samples fall into rough groupings by facies assemblage except for the diamict (D) group and one sample from the laminated fines (FI) group. The result that samples can be grouped on this graph affirms the facies assemblages constructed in the following section. When compared to the boundary curves of Sly et al. (1983, p.228 Fig. 12; note that Sly et al. use computed moment measures rather than graphic measures, therefore the Lake Alsek results may not be compatible), it can be seen that the sand facies group (S) may be in hydraulic equilibrium with a high energy regime, while some of the laminated fines (FI) and massive fines (Fm) sediments are in hydraulic equilibrium with a lower flow regime (Fig. 3.10). These conclusions are tentative though, because of the differences in computational methods of kurtosis and skewness. The median particle-size Md is a measure of central tendency which corresponds to the midpoint ( D Q Q ) in the cumulative distribution. This parameter has been computed for use in CM plots, but otherwise is of little interpretive value (Folk 1980 p.41). To distinguish between different depositional environments on the basis of grain size data alone, Passega (1964) proposed the use of CM diagrams in which C is the size of the coarsest 1% by weight of the sample (Dgg), and M is the median size (Md). Reference plots of sediments with known source can be overlaid onto a plot of unknown samples and distinctions made between dominantly traction load deposits, turbulent suspension load deposits, and settling-from-suspension deposits (Passega 1964). Applying this technique to Lake Alsek sediments (Fig. 3.11) indicates that these distinctions can be made unambiguously. All samples in the laminated fmes (FI) and massive fines (Fm) facies groups fall into the region of quiet water deposits. Samples in the S group are less clustered but generally fall into the region of turbulent suspension load deposits transported near the bed. Within-group differences are probably related to small scale variability of hydraulic conditions near the bed. The D group samples all plot outside the defined regions, suggesting that these samples are not suitable for this type of sedimentological analysis. 84 Figure 3.11 CM diagram of Lake Alsek sediment samples. The regions plotted are from Passega (1964): the circle represents the region of pelagic deposition and the other represents the region of tractive current deposition. CM Diagram of Lake Alsek Sediments Legend A D G r o u p X S G r o u p • FI G r o u p B F m G r o u p M = Median Size (mm) 85 3.6 Facies Assemblages Ideally, a facies is a distinctive sediment deposit that forms under certain conditions of sedimentation, reflecting a particular process or environment (Reading 1978, p.4). The Lake Alsek sequences can be grouped into five facies assemblages on the basis of sedimentary structure, lithostratigraphy, sediment properties, and grain size analysis observations listed in the previous sections. Using these five facies assemblages as the limit of interpretive resolution in this study, interpretive statements regarding genetic origins of these facies can be made. Attempts at environmental synthesis are undertaken in a preliminary way in subsequent chapters (in particular section 4.6). The notation used in previous sections is maintained here. These facies assemblages will also be used in the facies sequence analysis of the next chapter (chapter 4) where the temporal (i.e. stratigraphic) distribution within the sequences will be examined. 3.6.1 Matrix supported diamicton (facies D) Description This group consists of lithofacies types Dm, Dmm, Dmg, and Dms and is called facies group D. All occurrences are massive (Dmm) except for two graded (Dmg) and one stratified (Dms) occurrence. Two observational caveats must be noted here, the first being the problem associated with narrow cores, and the second associated with the selective sampling of this deposit by the piston corer. In general, this group is an extremely poorly sorted clast-sand-mud admixture with a gray (5Y 5/1) matrix and clasts of various lithology. The mean grain size is fine sand, and is extremely poorly sorted, but this facies group displays more within-group variability than any other. In large part, the size distributions are platykurtic, and are negatively skewed. Again, within-group variability of kurtosis and skewness is greater than for any other facies assemblage, and three samples were positively skewed. The wet density is 1.92 g cm"3 (S.D. = 0.23), the L.O.I, is 1.58% (S.D. = 0.62) and moisture content is 22.8% (S.D. = 5.5). The upper boundary is most 86 often diffuse and the lower boundary is most often abrupt with clasts projecting downwards into the underlying facies. Interpretation On the basis of exotic clast lithology and extremely poor sorting, facies group D has tentatively been interpreted as a deposit of iceberg-rafted material. Gravel-sized clasts are mostly angular, and when examined with a hand lens, striations can be seen on some of these clasts. These characteristics seem more compatible with a glacial rather than a fluvial origin for the clasts. In addition, many clasts can be identified as phaneritic igneous rocks which were probably derived from outcrops in the Icefield Ranges to the west (section 1.1.3). Local tills may contain clasts of similar lithology, but this source is improbable since facies D is interstratified with glaciolacustrine sediments (sections 3.6.3 and 3.6.4). Large erratic blocks of similar lithology are scattered over the floor of the Alsek Valley and were floated into the Lake Alsek basin by icebergs derived from Lowell Glacier (Clague and Rampton 1982, p.98). Gustavson (1975, p.261) found a layer of poorly sorted clayey silt with sand sized clasts in varved glaciolacustrine sediments which he interpreted as material dropped from icebergs. This may be a modern counterpart to facies D, which is generally more poorly sorted and coarser than the deposits observed by Gustavson (1975). 3.6.2 Sands and coarse silts (facies S) Description Included in this group are the lithofacies types Sm, Sg, and a single occurrence of Sr. Most occurrences are massive (Sm) with a few occurrences of graded layers (Sg). In general, facies group S is a poorly sorted silty sand (Table 3.8)). Within-group variability of the various statistical parameters is the least of any facies assemblage. These sediments are negatively skewed, indicating an excess of coarse particles, and are leptokurtic, indicating an excessively peaked distribution. The wet density is 1.61 g cm 87 (S.D. = 0.33), the L.O.I, is 1.50% (S.D. = 0.60) and moisture content is 14.4% (S.D. = 8.9). Overall, this is the thickest of the facies groups (mean = 4.8 cm), and displays the most variability in thickness (Table 3.5). The upper boundary is often graded, while the lower boundary is often abrupt or erosive; laminated mud rip-up clasts are sometimes present. Cross-laminations have been tentatively identified in two thin-sections, but what appear to be massive deposits are most common. Unfortunately, without better exposures identification of this structure remains tenuous. Interpretation The deposits of this group are interpreted as tractive current deposits, an interpretation compatible with results of the CM pattern analysis. The vigorous hydraulic regime needed to transport these sediments may be related to either slump initiated turbidity currents in Lake Alsek, or draining lake water being channelled through the narrow Alsek Valley. There are two substantive problems with a turbidity current process: (a) any turbidity current generated up-valley of the sample sites would follow the valley gradient (i.e. the present day river course) and probably not affect the sample sites, (b) the particle-size distributions for facies S samples exhibit negative skewness, while turbidite sands usually display a positive skewness (Zhao and Hsu 1984, p.82). Depositional structures associated with the transporting current were not observed in most occurrences of this facies. As explained in section 3.3, this may simply be a result of observational difficulties. On the other hand, massive deposits of sand-sized material may arise through rapid sedimentation (i.e. dumping), where there is insufficient time for bedforms to develop. Rapid sedimentation may occur during the presumably catastrophic jokulhlaup associated with a draining of Lake Alsek. Modern counterparts from an ice-dammed lake basin of the S facies interpretation presented here are rare. Observations were made of the sediments of ice-dammed Ape Lake before and after catastrophic drainage during the summer of 1984 by Gilbert and Desloges (1986). A preliminary report on some of these observations indicates that sandy 88 sediments were deposited as a result of the jokulhlaup in those areas of the lake that had drained completely (Gilbert and Desloges 1986). 3.6.3 Massive silts and clays (facies Fm) Description Facies group Fm includes lithofacies types Fm and Fmd, which are generally deposits of poorly sorted silts and muds with grain size distributions which are mesokurtic and symmetrical. These sediments are massive, light gray in colour and range in thickness from less than 1 cm to about 5 cm. The upper boundary is diffuse or abrupt as is the lower boundary, and infrequent dropstones (Fmd) can be seen on the exposed core. This facies group and the FI facies group exhibit nearly identical textural characteristics but because of the fundamental differences in structure remain classified as a separate facies groups. The wet density is 1.88 g cm"3 (S.D. = 0.10), the L.O.I, is 1.63% (S.D. = 0.35) and moisture content is 24.9% (S.D. = 4.6). Interpretation These sediments display similar textural characteristics to the muds and dropstone muds deposited in a nearly continuously ice-covered lake described by Zhao and Hsu (1984). The presence of erratic pebbles in the Lake Alsek sediment supports a similar origin, possibly in a seasonally (i.e. eight months) ice-covered lake. The Lake Alsek sediments are better sorted, so an analogy along these lines may be flawed. At ice-dammed Ape Lake, fine grained sediments eroded from the lake margins during drainage were deposited in the deeper portions of the lake as a massive mud layer (Gilbert and Desloges 1986). Because the Lake Alsek sites were presumably continuously covered by water, a similar depositional process may have been active in the lakes and ponds studied in detail. Little evidence of bioturbation is present in any Lake Alsek sediments; therefore the massive structure of this facies probably is not due to bioturbation. Instead, the lack of depositional structures may be caused by rapid sedimentation or post-depositional dewatering. Sorting by lake marginal waves as the lake 89 filled or drained is a likely mechanism to explain the relatively well sorted characteristic of this facies. Interpretation of this facies assemblage is probably more valid by analogy with the Ape Lake sediments (i.e. rapid sedimentation) than on the basis of seemingly inconclusive textural analogy with pelagic sediments. 3.6.4 Laminated silt and clay sediments (facies FI) Description Lithofacies types FI and Fid are grouped into facies assemblage FI. This is a poorly sorted clayey silt with grain size distributions that are negatively skewed and mesokurtic. Negative skewness indicates an excess of coarse particles. Many samples have a bimodal distribution with clay and silt size modes (Appendix II). The mean particle-size of this facies (M^ = 7.57 phi) makes it the finest sediment of the Lake Alsek samples. The sediments exhibit thin parallel laminae, alternating between lighter clay and darker silt. Occasionally dropstones are observed. With few exceptions, clay layers are generally > 1 mm thick, while silt layers are generally < 1 mm thick. Close examination of several sequences indicates that individual laminae do not grade into one another, and are apparently internally structureless; however not all sequences were examined in detail. The wet density is 1.70 g cm"3 (S.D. = 0.27), the L.O.I, is 1.94% (S.D. = 1.14) and moisture content is 31.6% (S.D. = 12.7). The upper boundary is most often abrupt or even truncated, the lower boundary either graded or abrupt. Interpretation These sediments are strikingly similar to rhythmically laminated couplets of fine sediment found in glaciolacustrine environments (Ashley 1975; Leonard 1986). If these are glaciolacustrine deposits they may be annual laminae (varves). Alternatively, they may be composite varves, representative of some other temporal scale. This question remains unresolved. Lamina thickness is not an adequate guide to resolving this question because Lake Alsek water depths probably varied dramatically from phase to phase and even from one cyclic filling to the next, thereby obscuring any sediment influx variations. 90 Also, clay layer thicknesses, which are often roughly constant in annual varves (Ashley 1975), would not be constant in Lake Alsek because of different filling levels from one phase to another (section 4.5.1). A Lake Alsek filling to the 678 m level might take up to approximately 8 years (Table 2.3). If the varves were indeed annual, then this would be the upper limit on the number of varves associated with the lowest lakes. The varves might also be expected to become thinner upwards as the suspended sediment supply is spread over a larger area. A complete sedimentary sequence might include a deposit of jokulhlaup-related sediments between varved deposits of gradually decreasing thickness. The separation of clay-size particles from silts in a varve is considered to be a consequence of differential settling velocities. Silt particles settle within a few months after introduction into the water column, while clay-size particles remain in suspension until the lake surface becomes frozen and all wind-generated water movements have ceased (Zhao and Hsu 1984, p.83). The distinct light/dark layering observed in a few closely examined samples of facies FI is compatible with this thinking. No sedimentary structures were observed in facies FI laminae, suggesting that these sediments were deposited distally (Ashley 1975; Zhao et al. 1984), away from the influence of turbid underflows. 3.6.5 Carbonaceous mud (facies O) Description The carbonaceous mud facies comprises structureless sediments that are black in colour and that oxidize to orange when exposed to air. Loss-on-ignition of 4-8% by mass indicates much more organic matter than is found in any other facies group (Table 3.4). In addition, wet and dry densities are significantly lower than all the other facies groups, and moisture content is significantly higher. No trend in density was detected in the three samples analysed, but it is expected that density would increase with depth of burial. No .9! grain size analysis was undertaken, but all occurrences are homogeneous and the inorganic fraction is presumably composed of silt and clay-sized particles. Interpretation This carbonaceous mud layer has a modern counterpart that is currently being deposited on pond bottoms at the KM1.2, KM4.4, KM13.4D, KM13.4C, KM28.5, and KM53.0 sites. The vegetation which is apparently the primary source for organic detritus is photosynthetic, so growth is restricted to shallow ponds without turbid inflow. Root systems beneath the O facies are sparse or absent entirely. This is also the case with modern organic sediments. After burial by sediments associated with a Lake Alsek event, some period of time must have elapsed before re-colonization. If this were more than a few years (which seems likely), then organic deposits would only be expected in a non-lake phase and not between cyclic fillings and drainings of Lake Alsek. The homogeneity of this facies indicates little variability in sedimentary processes during accumulation. This means either rapid sedimentation or a relatively unchanging depositional environment. It has been noted elsewhere that even over Holocene time, sediment character in small ponds (with small basins and lacking sediment input from streams) fluctuates little (MacDonald 1982, p.30). The small ponds in Alsek Valley are in similar environments suggesting that, despite being relatively thin, facies O probably is deposited slowly over an extended period of time. 92 Chapter 4 FACIES SEQUENCE ANALYSIS 4.1 Introduction The object of studying a sedimentary record is to deduce, on the basis of the information contained in the record, what environmental and sedimentary conditions prevailed at the time of deposition. The logical path of information transfer can be summarized as follows: from Environment to Sedimentary Process to Stratigraphic Section to Observation to Interpretation (Schwarzacher 1975, p.3). Three fundamental difficulties impede the transfer of information along this idealized path. First, the integrity of the signal itself as recorded in the sedimentary sequence may be faulty, in that a particular sedimentary deposit (or facies) may not correspond uniquely to one environment. For example, glaciolacustrine varves are sometimes found to be an annual deposit, but what appear to be similar deposits can also range in temporal scale from days to decades (Gilbert and Shaw 1981). Secondly, there is a loss of information contained in the sedimentary record on account of noise. What is noise and what is signal is really a matter to be evaluated at the interpretation stage; however these distinctions invariably creep into the observations themselves. It may be that the resolution of observations is fundamentally more instrumental in discriminating noise from signal than the sedimentary interpretation is. Assessment of the signal quality as part of the interpretation process is nonetheless important. Thirdly, initial examination and description of the stratigraphic section is subject to observational errors. Correct observations can be distinguished from mistaken observations because a property of correct observations is that they are consistently reproducible. Systematic error in correct observations, on the other hand, simply introduces a degree of uncertainty into subsequent quantitative analysis. Perhaps the greatest difficulty in making useful observations on a stratigraphic sequence is determining 93 a priori what constitutes a complete and efficient set of observational standards (Schwarzacher 1975, p. 15). Facies sequence analysis uses a statistical technique to aid in interpretion of sedimentary sequences. In the Lake Alsek sequences, apparently similar facies are repeated many times, and clear evidence of sequential ordering greatly assists their environmental interpretation. The technique used here is a special case of contingency table analysis called Markov chain analysis, which is employed here to determine vertical facies relationships. Specifically, the observed Lake Alsek facies sequences are tested for significant ordering, and examined for possible cyclicity. Markov chain analysis has been used successfully elsewhere (see for example: Krumbein and Dacey 1969; Miall 1973; Walker 1979, 1984; Carr 1982; Powers and Easterling 1982). (See a review of the technique in Appendix ITI). 4.2 Method The vertical sedimentary sequence of each Lake Alsek core is first subdivided into discrete lithofacies types (Chapter 3). These are then recombined into a manageable number of facies groups (Walker 1984, p. 1). The five facies assemblages described in section 3.6 (D, S, Fm, FI, and O), are sufficiently detailed to describe adequately the facies changes in most of the Lake Alsek sequences. In addition, a "scoured surface" (SS) category is included so that major erosional events in the sequence could be analyzed (Walker 1984, p.4). In the cases of KM53.0, KM50.4, KM69.0, KM28.5, and KM1.5, one or more of these facies assemblages were not observed; this does not impede analysis of these sequences. Vertical facies sequences are structured into upward transition matrices by counting upward-occurring changes in facies types and ignoring facies thicknesses (Fig. 4.1). The principal diagonal contains structural zeroes and the matrix elements tally the transitions from the row i facies to the overlying column j facies. These matrices are 94 Figure 4.1 Observed and expected facies transitions for core KM13.4C (34-120). Each upward occurring facies change is tallied to produce the observed transitions matrix. Under the model of quasi-independence, expected values are computed. A Pearson statistic is used to check goodness-of-fit. ss as ss .ss OBSERVED TRANSITIONS 0 S F I Fm 0 SS X.J t. 8. 3. 13. 7. 7 . 39. TRANSITION PROBABILITY D S F 1 Fm 0 SS P(I) 0 0.0 O.C 1 .00 0 0 0.0 0.03 S 0.0 0. 22 0.78 0 0 •0.0 0.21 FI O.O 0.0 0.0 0 67 0.33 0.08 Fm 0.08 0.08 0.08 0 38 0.38 0.33 0 0.0 0.0 0.0 0.83 0. 17 0. 18 SS 0.0 1 .00 0.0 0,0 0 0 0. 18 Number of Iterations -Est for A 1 0. 20 Est. for Bj • 0. 13 Est for A 1 1 .80 Est. for BJ • 1. 33 Est for Al 0.60 Est. for BJ - 0. 42 Est for Al 2.60 Est. for BJ • 2 . SO Est for Al 1 .20 Est. for BJ 1 . 06 Est for Al t .40 Est. for BJ 1 . 09 Number of Iterations • 5 Est for Al 0. 15 Est. for BJ 0. 13 Est for Al 1 .68 Est. for BJ - 1. 26 Est for A 1 0.48 Est. for BJ • 0. 40 Est for Al 3.40 Est. for BJ > 2 . S1 Est for Al 1 .07 Est . for BJ • 1 . 00 Est for A 1 1 .25 Est. for BJ t . 03 QUASI - INDEPENDENCE MATRIX (011 • al • bl )• . S FI Fm • SS X 1 . 0 0. 19 0.06 0. 43 0 . IS 0 . 16 1 .OO S 0.2 1 0.67 4 . 7 1 1 .69 1 . 73 9 .00 FI 0.06 0.61 1 . 3S 0 .48 0 .50 3 .00 Fm 0.43 4 . 28 1 . 35 3 . 42 3 .51 12 . 99 • 0. 14 1 . 34 0.42 3. 00 1 . 10 6 .00 SS 0. 16 1 .58 0.50 3. 51 1 .26 7 .00 1 .00 8.00 3 .00 13 . 00 7 .00 7 .00 39 .00 STANDARDIZED DIFFERENCES D S FI Fm 0 SS D' ' -O 44 -0. 25 0.86 -0 39 -0.40 Si- -0 46 1 .63 1 .06 - 1 30 -1 .32 F'l -O 25 -0 78 - 1 . 16 2 18 0.71 Fm 0 87 - 1 59 -0.30 0 86 0.80 0 -0 37 -1 16 -0.65 1 . 16 -0. 10 SS -0 40 4 32 -0.71 - t .87 - 1 12 GOODNESS OF FIT OF Q-I MODEL CHI-SOUARE- 48.081 WITH D .F.« 19 Prob. of exceeding chl square • 0.000250 34-120 95 analyzed using the methods set out in Appendix DI, and written into the program MARKOV. S (Appendix IV). Markov Chain Analysis Row, column and grand totals are first computed for each upward transition matrix. If the grand total does not exceed 15 transitions then the sequence is considered to be too short and the analysis stops. Otherwise, row and column totals are used to compute a matrix of quasi-independent values (this matrix has also been called the random matrix). An iterative proportional fitting procedure is used to maintain row and column totals. The iterative procedure begins by assuming initial values for ey(l) that divide the grand total among row totals (Appendix IH). On the second iteration, a second matrix of expected values is computed using a proportional fitting algorithm. This adjusts the initial values so that marginal proportions are more closely approximated. The iterative process is continued until two successive matrices converge to differences of less than 1%. Structural zeroes do not affect this procedure and reappear in the resulting matrix of expected transition counts under the model of quasi-independence. The two matrices are then tested for goodness-of-fit using a standard Pearson X statistic. The validity of this statistic (and associated assumptions) is discussed in Appendix III. To avoid confusion, the chi-square distribution itself is referred to in words, 9 9 while the computed statistic is referred to as the Pearson X or simply as X . The degrees of freedom are adjusted for an embedded matrix by subtracting the number of structural zeroes from the total degrees of freedom. Using the appropriate chi-square distribution, the probability of exceeding the computed Pearson X is evaluated (Fig. 4.1). The choice of significance level for these statistical tests requires some comment, since detection of any sort of non-random ordering has important environmental implications. The null hypothesis ( H Q ) is that there is no significant difference between the observed transition matrix and the quasi-independence matrix. The alternative hypothesis ( H ^ ) is that there is a significant difference, and non-random ordering in the 96 observed sequence is indicated. The probability of a Type II error, where a false null hypothesis is incorrectly accepted, must therefore be minimized. For this reason, a significance level of 0 . 3 0 is used for hypothesis testing at this stage. If the probability of exceeding the computed Pearson X is greater than 0 . 3 0 , the null hypothesis ( HQ ) is accepted. Otherwise, the alternate hypothesis (H-^ ) is accepted, and the search for extreme cells which correspond to dominant transitions begins. Standardized differences between the observed transition matrix and the quasi-independence matrix are computed to provide an overall indication of the magnitude of individual cell residuals. Since these values are approximately normally distributed, residuals which vary significantly from zero are largely responsible for the non-random ordering (i.e. Markov property). These residuals are most likely to be identified in the stepwise cell selection procedure (Powers and Easterling 1982, p.922). The first cell to be selected as a dominant transition is the transition with the greatest value in the standardized difference matrix. Then, a stepwise selection procedure is begun to identify the most extreme of the remaining transitions. After a particular cell has been selected it is treated as a structural zero and the next extreme cell is identified. By computing the Pearson X^ for each remaining possibility, the cell that most reduces the matrix Pearson X value when it is deleted is found. This procedure is repeated until either a significance level of 0 . 3 0 is passed, or the degrees of freedom are reduced to one (Table 4.1). The list of extreme cells is then used to construct a facies relationship diagram (Walker 1979). This diagram is used to connect facies schematically with the dominant transitions into a depositional pattern. Observed transition sequences from different cores are tested for similarity using the computational procedures outlined in Appendix DI. Sequences from the same sampling site are tested for similarity so that these data sets might be safely combined into a larger data set while remaining representative of the sequences at a given site. Where possible, sequences from different sampling sites are compared against one another to test for 97 homogeneity. The null hypothesis (HQ) in this test is that there is no significant difference between different sequences. The alternative hypothesis ( H - ^ ) is that there is a significant difference. The test statistic used to test H Q here is a Pearson X statistic. A significance level of 0.10 is used for this test to avoid the danger of type I error, where the true null hypothesis of similarity might be incorrectly rejected. Once the quasi-independence model has been rejected for an observed transition sequence, it is then logically tested for double-dependence. In other words, the sequence is tested for the possibility that the occurrence of a state has a probabilistic dependence on the preceding two states in combination, rather than just the previous state alone. This test requires a total of at least 200 transitions in a sequence when using six facies types, otherwise there are too many sampling zeroes for a valid goodness-of-fit test (Appendix IH). For the sequences analyzed in the Lake Alsek study, this test cannot be undertaken satisfactorily because of data set size limitations. 4.3 Results of Analysis of Lake Alsek Sequences Results are discussed on the basis of sampling sites since sediments could be correlated easily at this scale, while between-lake correlations could not be established satisfactorily on the basis of stratigraphy alone (section 4.5.1). The observed transition matrices, the expected transition matrices under the model of quasi-independence, and the stepwise selection of dominant transitions for each sequence are in Tables 4.1, 4.2, 4.3, and 4.4. Results from KM1.2 The transition count matrix contains only seven transitions and there are 14 sampling zeroes; therefore, this sequence is too sparse to meet the goodness-of-fit test assumptions. Also, the core itself is of low quality because of sampling difficulties making initial lithofacies identification problematic (section 3.4.2). The results from this site therefore are deemed unsuitable for this analysis. 98 Table 4.1 Transition count matrix and expected transition count matrix for the sediment sequence from KM4.4. The Pearson X 2 statistic indicates that the fit is poor, and a Markov dependence exists in the sequence. The stepwise selection procedure extracted four dominant transitions prior to termination. OBSERVED TRANSIT IONS • s FI Fm 0 SS XI 0. 0. 1 . 0. 0. 0. 1 . - 0. 0. 10. 3. 1 . 0. 16. -\ i . 10. 0. 1 . 2. 2. 16. "« 0. 0. 3. 0. 3. 2. a. 0. 0. 2. 1 . 0. 2. s. 5S 0. s. 0. 1 . 0. 0. 6. <-J 1. 15. 16. 8. 6. 6. 32. TRANSITION PROBABILITY 0 S F1 Fm 0 SS P(I) 0 0.0 0.0 1 .00 0.0 0. 0 0. 0 0. 02 s 0.0 0.0 0.63 0.31 0. OS 0. 0 0. 29 F 1 0.06 0.63 0.0 0.O6 0. 13 0. 13 0. 31 Fm 0.0 0.0 0.38 0.0 0. 38 0. 25 0. . 15 0 0.0 0.0 0.40 0.20 0. 0 0. 40 0. . 12 SS 0.0 0.83 0.0 0. 17 0. 0 0. 0 0. . 12 Number of I t e r a t i o n s • 1 Est . f o r Al 0.20 Est. f o r B] • 0. 10 Est . f o r Al • 3.20 Est . f o r BJ • 2. 08 E s t . f o r Al • 3.20 Est . f o r BJ - 2. 22 Est . f o r Al - 1 .60 Est . f o r BJ - 0. 91 Est . f o r A1 - 1 .00 Est. f o r BJ • 0. 64 Est . f o r Al 1 .20 Eat. f o r BJ * 0. 65 Number of I t e r a t i o n s • 4 Est. f o r Al 0. 15 Eat. f o r BJ • 0. . 10 Est. f o r Al - 3.34 Eat. f o r BJ • 2 . 12 Est. . f o r A1 - 3.70 Est . f o r BJ • 2 .32 Est. . f o r A1 - 1 .39 Eat. f o r BJ • 0 .87 Est, , f o r A1 * 0.83 Eat. f o r BJ • 0 .61 Est. . f o r Al • 1.00 Est . f o r 8J • 0 .62 QUASI - INDEPENDENCE MATRIX ( O i l - a! * b i : 0 S FI Fm 0 SS XI . 0 0.0 0.32 0.33 0. 13 0 .09 0. 10 1 .00 s 0.34 0.0 8.21 3.07 2 . 17 2. 21 16 .00 FI 0.35 7.85 0.0 3. 21 2. .27 2. 31 15 .99 Fm 0. 13 2.94 3.21 0. 0 0 .83 0. 87 a .00 0 0.08 1.76 1.92 0. 72 0 .0 0. 52 5 .OO SS 0. 10 2.12 2.31 0. 87 0 .61 0. .0 6 .00 X.J 1 .00 15.00 16.00 8. OO 6 .00 S .00 52 .OO STANDARDIZED DIFFERENCES D S FI Fm 0 SS D 0.0 -0.57 1.09 -0 .36 -0 .31 -0.31 s -0.38 0.0 0.63 1. . 10 -0 .80 -1 .49 FI 1 .09 0.77 0.0 -1. .23 -0 . 18 -0.20 Fm -O.S6 -1.72 -0.12 O. .0 2. . 33 1 .22 • -0.28 -1.33 0.06 0. .33 0. .0 2.06 SS -0.31 1.98 -1.32 0. , 14 -0. . 78 0.0 GOODNESS OF FIT OF 0-1 MODEL CHI-SQUARE" 33.136 WITH O.F.- 19 ProC. of M C t f O I n g c h l square - 0.023187 Stepwise S e l e c t i o n of C t l U Step No.- 1 O i l No.* 4. S A f t e r d e l e t i o n of c e l l , new model f i t i s : Ch*square • 27.70 Prop -0.066727 OOF -18 A f t e r s e l e c t i o n of 1 c e l l s : Total t r a n s i t i o n count- 49.0 S t r u c t u r a l zeroes- 7 Step No.- 2 C e l l No.- S, 6 A f t e r d e l e t i o n of c e l l , new model f i t i s : Chi square - 23.71 Prop -0.127577 OOF -17 A f t e r s e l e c t i o n of 2 c e l l s : Total t r a n s i t i o n count- 47.0 S t r u c t u r a l zeroes- 8 Step No.- 3 C e l l No.- 4, 6 A f t e r d e l e t i o n of c e l l , new model f i t i s : Chi square - 19.33 ProP -0.250849 DOF -16 A f t e r s e l e c t i o n of 3 c e l l s : Total t r a n s i t i o n count- 49.0 S t r u c t u r a l zeroes- 9 Step No.- 4 C e l l No.- 6, 2 A f t e r d e l e t i o n of c e l l , new model f i t i s : Chi square - 16.94 Prod -0.322619 OOF -15 99 Results from KM4.4 The cores from this site constitute a single sequence. When subdivided into five facies (D, S, FI, Fm, O) and an erosional surface (SS) the sequence has 52 transitions. The transition count matrix (6 x 6) has 13 sampling zeroes and six structural zeroes (Table 4.1). The transition count matrix is significantly different from the quasi-independence model (at a significance level of 0.30) with a Pearson statistic of 33.1 with 19 degrees of freedom (prob. = 0.02). This indicates the presence of a first order Markov dependence between at least two facies types. Stepwise selection of cells (using the Pearson X^ statistic) reveals four dominant non-random transitions before the chosen level of significance (CX= 0.30) is exceeded. These transitions are: Fm->0, 0->SS, Fm->SS and SS->S (Table 4.1). The remaining transitions appear to be adequately described by a random transition process. The nature of the Markov process is explored by using these transitions to connect the corresponding facies in a facies relationship diagram (Walker 1984). In this depositional pattern, four of the six facies are involved in a Markov interaction, while the remaining two states are randomly dispersed throughout the sequence (Fig. 4.2a). The depositional pattern indicates that the massive muds (Fm) facies is overlain by either the carbonaceous mud (O) facies, (that presumably occurs after a final drainage of a Lake Alsek phase), or an erosional unconformity (SS), (that results from the larger drainages). Results from KM13.4C The five sediment cores recovered at this site are first analyzed separately, subdividing each sequence into five facies types and one erosional surface (SS). The results indicate that all five sequences are significantly different from the quasi-independence model (at a significance level of 0.30). In all five sequences, the most important non-random transition is SS->S. This transition was chosen as the first extreme cell at the exclusion of all others. This 1 0 0 Figure 4.2 Facies relationship diagrams (a) KM4.4 sequence, (b) KM13.4C sequence, (c) KM13.4D sequence, and (d) KM50.4 sequence. These diagrams show the depositional pattern indicated by Markov chain analysis. Connections between facies have been identified as the dominant transitions in each stratigraphic sequence. a)KM4.4 b)KMl3.4C c) KM13.4D d) KM50.4 101 strongly suggests that the erosional surface (SS) is always the result of a process that immediately precedes deposition of the poorly sorted sand (S) facies. The second most important transition in four sequences is the F1->SS transition. This suggests that the laminated mud (Fm) facies (presumably deposited during a Lake Alsek phase) is scoured by an erosional event, but is cohesive enough to remain the least affected. In one core (34-120) the second most important transition is the Fl->0 transition. This suggests a strong tendency for carbonaceous mud (0) facies (deposited in eutrophic ponds during empty Lake Alsek intervals) to be deposited immediately after the laminated muds (FI) facies (associated with a Lake Alsek ponding). The five cores were then tested for lateral homogeneity; with a statistic of 68.7 with 96 degrees of freedom (prob. = 0.98) the hypothesis of similarity must be accepted. The transition count matrices were summed into a single transition matrix, resulting in a total of 157 transitions with 16 sampling zeroes (Table 4.2). When this matrix is tested against the model of quasi-independence, the the fit is poor (X^ = 203.7, d.f. = 19), thus indicating a Markov dependence. Stepwise selection of extreme cells indicates that the most important transitions in this sequence are: SS->S, F1->SS, Fl->0, and 0->SS (Table 4.2). The selection of these four dominant transitions involves four of the possible six facies in the facies relationship diagram (Fig. 4.2b). The two randomly occuring states in this sequence are the diamict (D) facies and the massive muds (Fm) facies. Overall, this depositional pattern is very similar to that from site KM4.4, except that in the KM13.4C pattern the laminated muds (FI) facies is in place of the massive muds (Fm) facies. Like the pattern observed in the KM4.4 sequence, this pattern is not closed; meaning that there are end points in the depositional pattern (FI and S). Therefore this is not a cyclic sequence but rather a non-randomly ordered sequence intermittently interspersed with random facies transitions. 1 0 2 Table 4.2 Summed transition count matrix and expected transition count matrix for five cores from KM13.4C. The Pearson X 2 statistic indicates that the fit is poor, and a Markov dependence exists in the sequence. The stepwise selection procedure extracted four dominant transitions prior to termination. OBSERVED TRANSITIONS 0 S FI Fm 0 SS XI . • 0. 0. 1 . 10. 0. 0. 11. s 0. 0. 2. 33. 0. 0. 35. FI 0. 0. 0. 0. 2 . 7. 9. Fm 12. 4 . 6. 0. 17 . 17. 56. 0 0. 0. 0. 16. 0. 3. 19. .SS 0. 27. 0. 0. 0. 0. 27. X.j 12. 31 . 9. 59. 19. 27. 157. QUASI - INDEPENDENCE MATRIX (01 I » a l * bt D S FI Fm 0 SS XI . D 0.0 1 .92 0 .48 5 .96 1 . 06 1 . 59 1 1 .01 S 2.31 0. .0 1 .72 21 . 50 3. .82 5. .71 35 .06 FI 0.52 1 . 55 0 .0 4 .81 0. ,85 1 . 28 9 .01 Fm 6. 30 18 . 90 4 .67 0. ,0 10, , 4 1 15. ,57 55 .86 0 1 . 15 3 .46 0 .85 10 .71 0 .0 2 .85 19 .02 SS 1 .72 5. . 17 1 .28 16. .02 2. .85 0 .0 27 .04 X. j 12.00 31 , .00 9. .OO 59. .00 19. .00 27. .00 157 .00 STANDARDIZED DIFFERENCES D S FI Fm 0 SS D 0 .0 - 1 .39 0.76 1 .65 -1 .03 -1 .26 S -1 , .52 0. .0 0.22 2.48 - 1 .95 -2.39 FI -0. .72 - 1 . .25 0.0 -2 . 19 1 .24 5.06 Fm 2 . 27 -3 . 43 0.61 0.0 2 .04 0.36 0 -1 .07 - 1 . .86 -0.92 1 .62 0 .0 0.09 SS - 1 . .31 9. .60 -1.13 -4.00 - 1 .69 0.0 GOODNESS OF FIT OF Q-I MODEL CHI-SOUARE- 203.690 WITH D.F.- 19 Prob. of e x c e e d i n g c h l square • 0.000000 Stepwise S e l e c t i o n of C e l l s Step No.- 1 C e l l No.- 6, 2 A f t e r d e l e t i o n of c e l l , new model f i t Is: Ch1square • 111.48 Prob -0.000000 DOF -18 A f t e r s e l e c t i o n of 1 c e l l s : T o t a l t r a n s i t i o n count- 130.0 S t r u c t u r a l z e r o e s - 7 Step No.- 2 C e l l No.- 3. 6 A f t e r d e l e t i o n of c e l l , new model f i t Is: C h l s q u a r e - 27.38 Prob -0.052690 DOF -17 A f t e r s e l e c t i o n of 2 c e l l s : T o t a l t r a n s i t i o n count- 123.0 S t r u c t u r a l z e r o e s - 8 Step No.- 3 C e l l No.- 3, 5 A f t e r d e l e t i o n of c e l l , new model f i t 1s: C h l s q u a r e - 19.31 Prob -0.252788 DOF -16 A f t e r s e l e c t i o n of 3 c e l l s : T o t a l t r a n s i t i o n count- 121.0 S t r u c t u r a l z e r o e s - 9 Step No.- 4 C e l l No.- 5. 6 A f t e r d e l e t i o n of c e l l , new model f i t i s : C h l s q u a r e • 13.12 Prob -0.592726 DOF -15 103 Results from KM13.4D Four cores were recovered at this site and analyzed separately (as in the previous case), then tested for lateral homogeneity. In the analysis of separate cores, the fit against the model of quasi-independence could be rejected soundly in all cases. Testing for similarity between the four cores results in a Pearson X 2 of 59.6 with 72 degrees of freedom, and the probability of exceeding the chi-square value of 0.85 (Table 4.3). This result indicates that all four transition frequency matrices can be combined safely into a larger matrix, resulting in a total transition count of 138. When this data set is tested against the model of quasi-independence, the Pearson X 2 result is 158.1 with 19 degrees of freedom (prob. = 0.0), clearly rejecting goodness of fit. This implies that a Markovian interaction must exist between one or more facies pairs. Stepwise selection of cells identifies the following preferred transitions: SS->S, 0->D, SS->D, 0->SS, F1->S, and F1->D (Table 4.3). The facies relationship diagram constructed from these six dominant transitions includes five of the six facies (Fig. 4.2c). This depositional pattern shares only two transitions with the pattern constructed for KM13.4C. The most striking difference is that the diamict (D) facies non-randomly overlies each of the carbonaceous mud (O) facies, the laminated mud (FI) facies, and the erosional unconformity (SS). The lower boundary of the diamict (D) facies is often indistinct sometimes making it difficult to distinguish the underlying facies. Results from KM28.5 The cores taken at this site were treated as a single continuous sequence, but only 12 facies transitions between four facies could be tallied (Figure 3.5e). This data set is too small for the analysis to be undertaken satisfactorily. Results from KM50.4 Using four facies and one erosional surface, 30 transitions in the composite vertical sequence were tallied. The transition count matrix has 10 sampling zeroes and five structural zeroes (Table 4.4). The quasi-independence model is significantly different from 104 T a b l e 4 .3 S u m m e d t r a n s i t i o n c o u n t m a t r i x a n d e x p e c t e d t r a n s i t i o n c o u n t m a t r i x f r o m K M 1 3 . 4 D . T h e P e a r s o n s t a t i s t i c i n d i c a t e s t h a t t h e f i t i s poor , a n d a M a r k o v d e p e n d e n c e e x i s t s i n t h e s e q u e n c e . T h e s t e p w i s e s e l e c t i o n p r o c e d u r e e x t r a c t e d s i x d o m i n a n t t r a n s i t i o n s p r i o r to t e r m i n a t i o n . OBSERVED TRANSITIONS 0 S f 1 Fm 0 SS x t . 0 0. 0. 1. 8 . 0. 0. 9. s 0. 0. 5. 18 . 0. 0. 33. Ft 3. 3. 0. 21 . 0. 5. 32. fm 2. 1. 24. 0. 9. 12. 48. 0 3. 0. t. 1 . 0. 2. 7 . SS 1. 18. 0. 0. 0. 0. 19. K.J 9. 23. 3.. 48 . 9. 19. 138. TRANSITION PROBABILITY 0 S FI Fm 0 SS P ( I ) 0 0.0 0.0 O . U 0.S9 0 0 0.0 0.07 s 0.0 0.0 0.22 0.T8 0. .0 0.0 0. 16 FI 0.09 0.09 0.0 0.6G 0, .0 0. IS 0.22 F« 0.0« 0.02 0.50 0.0 0. . 19 0.23 0.33 0 0.43 0.0 0.14 0. U 0, .0 0.29 0.07 SS 0.03 0.93 0.0 0.0 o . 0 0.0 0. 14 Nunkt w o f I t e r a t i o n s • t E a t . . f o r Al • 1.80 Ea t . f o r BJ 0. 33 Ea t . , f o r At - 4.60 Ea t . f o r BJ 0. .96 E s t . f o r At * 6.40 Ea t . f o r BJ • 1, 46 Es t . , f o r At - 9.GO Ea t . f o r ej 2, .67 Cat. f o r At • 1.40 Ea t . f o r BJ 0. .34 E a t . . f o r At • 3.80 Ea t . f o r BJ o . 80 Number of I t e r a t i o n s • S E s t . , f o r At - 1.41 Ea t . f o r BJ 0, ,33 Ea t . , f o r At • 3.93 Ea t . f o r BJ '0. ,89 E s t . f o r A l • 3.98 E s t . f o r BJ 1 ,36 Ea t . , f o r At • 13.17 Ea t . f o r BJ 3 .07 E s t . f o r At • 1.10 Ea t . f o r aj 0. ,32 Eat , f o r At - 3.18 Ea t . f o r BJ 0, ,74 QUASI - INDEPENDENCE MATRIX (01) • at • B|) D s F I F M 0 SS XI . D 0.0 1 .23 1.92 4. .33 0.46 1 . .03 9.00 S 1 .30 0.0 3.37 12. . 13 1 .28 2. .93 23.01 f 1 1 .97 3.30 0.0 18. ,38 1 .94 4 , .44 32.03 F M 4.33 1 1 .66 17.90 0. .0 4 . 28 9. , 77 47 .94 0 0.36 0.97 1 .49 3. .37 0.0 0. 81 7.00 SS 1 .05 2.B2 4.33 9. .78 1 .03 0, .0 19.01 x-J 9.00 22.OO 31 .00 48. OO 9.00 19. 00 138.00 STANOAflOIZEO DIFFERENCES 0 S F 1 Fn 0 SS D 0.0 -1.12 -0. 66 1 . 76 -0. .68 -1.02 s •1 . 14 0.0 -0. 16 1 . 68 -1. . 13 -1.71 FI 0.T4 -1 . o o 0. 0 0. 61 -1, ,39 0.27 F« -1.12 -3. 12 1. 44 0. 0 2. .28 0.71 0 4.40 -0.99 -0. 40 -1 . 29 0. .0 1 .32 SS -O.03 9.04 -2. 08 -3. 13 - 1 , 02 0.0 GOOONESS OF FIT OF Q-I MQQCl CHI-SQUARE- tsa.iss WITH D.F.- 19 Pro©, o f e x c e e d i n g c h t s q u a r e • O.OOOOOO S t e p w i s e S e l e c t i o n of C,ella S t e o No.- 1 C e l t No.- 6. 2 A f t e r d e l e t i o n of c e l l , ne* model f i t i s : Chi s q u a r e - 76.40 Prot> -O.OOOOOO OOF -18 A f t e r s e l e c t i o n of 1 c e l l s : T o t a l t r a n s i t i o n c o u n t * 120.0 S t r u c t u r e ! z e r o e s * 7 Ste o No.* 2 C e l l No.- 5. i A f t e r d e l e t i o n o f c e l l , new model f i t i s : C n l a q u e r e - 43.39 ProO -0.000213 OOF -17 A f t e r s e l e c t i o n o f 2 c a l l s : T o t a l t r a n s i t i o n c o u n t - 117.0 S t r u c t u r a l z e r o e s * 8 Ste o No.- 3 C e l l No.- 6. t A f t e r d e l e t i o n of c e l l , new model f i t i s : C h i s a u a r e • 30.72 Proo -O.014394 OOF -16 A f t e r s e l e c t i o n of 3 c e l l s : T o t a l t r a n s i t i o n c o u n t - 116.0 S t r u c t u r a l z e r o e s * 9 S t e p No.* 4 C e l l No.- S. 4 A f t e r d e l e t i o n o f c a l l , new model f i t I S : C h i s a u a r e * 23.98 P r o o -0.063383 OOF -13 A f t e r s e l e c t i o n of 4 c e l l s ; T o t a l t r a n s i t i o n c o u n t * 114.0 S t r u c t u r a l z e r o e s * 1 0 S t e p No.* 3 C e l l No.* 3. 2 A f t e r d e l a t i o n o f c o l l , new modal f i t i s : C h l e q u a r e * 19.06 ProO *0.162642 OOF -14 A f t e r s e l e c t i o n of 3 c e l l s : T o t a l t r a n s i t i o n c o u n t * 111.0 S t r u c t u r a l z e r o e s * i 1 S t e o No.- 6 C e l l No.* 3. 1 A f t e r d e l e t i o n o f c e l l , new model f i t l a : C h i s o u e r e - 13.32 P r e * -0.423434 OOF -13 105 Table 4.4 Transition count matrix and expected transition count matrix for KM50.4. The Pearson X statistic indicates that the fit is poor, and a Markov dependence exists in the sequence. The stepwise selection procedure extracted four dominant transitions prior to termination. OBSERVED TRANSITIONS • S FI Fm 0 XI . 0 0. S. 5. 0. 0. 10. s 2. 0. 0. 1. 4 . 7. FI 0. 1 . 0. 0. 3. 4 . Fm 1 . 1 . 0. 0. 0. 2. 0 7. 0. 0. 0. 0. 7. X.] 10. 7. 5. 1. 7. 30. TRANSITION PROBABILITY 0 s F1 Fm 0 P(I) D 0.0 0 .50 0.50 0.0 0. ,0 0.33 S 0.29 0. .0 0.0 0. 14 0. .57 0.23 FI 0.0 0. .25 0.0 0.0 0. ,75 0. 17 Fro O.SO 0 .50 0.0 0.0 0 .0 0.03 0 1 .OO 0 .0 O.O 0.0 0 .0 0.23 Number of I t e r a t i o n s • 1 Est. f o r Al • 2.50 Est . f o r BJ 2. 00 Est. f o r Al - 1 .75 E s t . f o r BJ 1 . 22 Est. f o r Al - 1 .00 E s t . f o r BJ 0. 77 Est. f o r Al • 0.50 Eat. f o r BJ 0. 14 Eat. . f o r Al • 1.75 E s t . f o r BJ • 1. 22 Number of I t e r a t l o n a • 4 Eat. . f o r Al • 3. 12 Eat. f o r BJ 2. .20 Est. f o r Al • 1 .66 Eat. f o r BJ 1, . 16 Est. . f o r Al - 0.86 Eat. f o r BJ 0. .73 Est. . f o r Al « 0.38 Est . f o r BJ 0. . 14 Est. f o r Al • 1 .66 Eat. f o r BJ 1. . 16 QUASI - INDEPENDENCE MATRIX ( O i l - a l • b l ) 0 S F 1 Fm 0 XI 0 0.0 3.63 2.29 0. .43 3 .63 9.98 s 3.64 0.0 1 .22 0. .23 1 . .93 7.01 FI 1 .89 1 .OO O.O 0. . 12 1 .00 4.00 Fm 0.84 0.44 0.28 0. ,0 0 .44 2.00 0 3.64 1 .93 1 .22 0. .23 0 .0 7.01 X.J 10.OO 7 .00 5.00 1. .00 7 .OO 30.OO STANOAROIZEO DIFFERENCES D S FI Fn a 0 D O.O O. 72 1 . 79 -O. .63 - 1 . 91 S -0.86 0.0 -1 . IO 1 . .62 1 . 49 FI -1 .37 0.00 0. 0 -0. .34 2. 00 Fm 0.18 0.84 -0. 53 0. .0 -0. 67 0 1 .76 -1.39 -1. 10 -0. .48 0. 0 GOODNESS OF FIT OF Q-I MODEL CHI-SQUARE- 28.535 WITH D.F.- 11 Prob. of exceeding c h i square • 0.002679 Stepwise S e l e c t i o n of C e l l s Step No.- 1 C e l l No.- 3. 3 A f t e r d e l e t i o n of c e l l , new model f i t i s : Chi square - 24.33 Prop -0.006312 DOF -10 A f t e r s e l e c t i o n of 1 c e l l s : T o t a l t r a n s i t i o n count- 27.0 S t r u c t u r a l zeroes- 6 Step No.- 2 C e l l No.- 2. 3 A f t e r d e l e t i o n of c e l l , new model f i t 1s: Chi square - 19.79 Prop -0.019251 OOF - 9 A f t e r s e l e c t i o n of 2 c e l l s : T o t a l t r a n s i t i o n count- 23.0 S t r u c t u r a l zeroes- 7 Step No." 3 C e l l No.- 2, 4 A f t e r d e l e t i o n of c e l l , new model f i t Is: Chi square • 11.27 Prob -0.187104 OOF - 8 A f t e r s e l e c t i o n of 3 c e l l s : T o t a l t r a n s i t i o n count- 22.O S t r u c t u r a l zeroes- 8 Step No.- 4 C e l l No.- 3. 2 A f t e r d e l e t i o n o f c e l l , new model f i t i s : Chi square • 6.07 Prob -0.331199 OOF - 7 END OF SUCCESSFUL RUN OF MARKOV.S PROGRAM 106 the transition count matrix with a Pearson X 2 statistic of 28.5 with 19 degrees of freedom (prob. = 0.003). Four extreme cells could be selected before exceeding the significance level of 0.30. These dominant transitions are: Fm->SS, F1->SS, Fl->0, and Fm->F1 (Table 4.4). The facies relationship diagram constructed from these transitions involves four of the five facies, with the sand (S) facies randomly dispersed through the sequence (Fig. 4.2d). The sequence exhibits significant ordering, but again is not cyclic. This depositional pattern is most similar to the pattern observed at KM13.4C, but only marginally. The most striking similarity is that the erosional surface (SS) overlies both of the mud facies (FI and Fm). Results from KM53.0 Only two facies could be recognized in the cores from this site. No useful Markov analysis could therefore be undertaken because one facies would always alternate with the other (Figure 3.5g). Results from KM69.0 Although the sequence from this site is almost 5 m long, it contains only three facies and seven facies transitions. This sequence is unsuitable for Markov chain analysis (Figure 3.5h). 4.4 Summary Sequence for Alsek Valley Sediments To amalgamate the information gained through the preceding analysis, a summary facies sequence is constructed (Walker 1984). The significant sequences are tested for homogeneity and the statistically similar sequences are combined into a single transition count matrix. This matrix is analyzed in the usual manner and a summary facies relationship diagram is derived. The summary facies sequence is based on this diagram. Only matrices of equal rank can be tested for homogeneity, so the KM4.4, KM13.3C, and KM13.4D matrices (N=6) were tested, resulting in a Pearson X 2 = 146.78 with 48 degrees of freedom (prob. = 0.00). The large value indicates that the 107 hypothesis of similarity must be rejected and these sequences cannot be combined. Similarly large values resulted when the three pairs of sequences were tested against one another. This indicates that the facies sequences from different sites are spatially inhomogenous and cannot be combined. This result is perhaps not surprising in light of the difficulties encountered in stratigraphic correlations from site to site (section 4.5). A summary sequence which can be used as both a summary of the present observations and as a guide for future observations is nonetheless desirable. For these purposes, a composite sequence from KM13.4C is the best choice because (a) it has the largest number of observations and (b) facies relationships established for this site are most consistent with the facies interpretations of Chapter 3 (Table 4.2). This sequence includes glaciolacustrine (Fm, FI, and D), organic lacustrine (O), and fluvial (S) sediments, as well as an erosional unconformity (SS). 4.5 Correlation and Spatial Variability The stratigraphic principles used throughout much of this and the preceding chapter have provided the means for interpreting the temporal variability of Lake Alsek sediment cores. A fuller paleoenvironmental reconstruction of Lake Alsek can be gained by correlating the stratigraphic evidence from one core to that of other cores. This provides the means for interpreting the spatial variability of Lake Alsek sediments. The principle means of correlation are: paleomagnetism, tephra layers, paleosols, rhythmite counts, distinctive stratigraphic layers, chemical properties, microfossils, and radiometric dating (Lowe and Walker 1984, p.289). In this study, correlations between cores are made only on the basis of distinctive stratigraphic layers that can be traced from core to core. 4.5.1 Stratigraphic Correlations Cores from the same pond or lake can be confidently correlated on the basis of lithostratigraphy alone (Figs. 3.1a - 3.1h). These correlations are strongly supported by 108 the results of facies sequence analysis which showed that within-site similarities between sequences are statistically significant. Correlation between lakes by lithostratigraphy is possible to a very limited degree because of the areal continuity of Lake Alsek bottom sediments, but without supporting correlative evidence remain mostly speculative at this point. Difficulties are encountered primarily because indeterminate amounts of sediment have been removed at erosional unconformities. These erosional gaps in the sedimentary record result in time-transgressive stratigraphic gaps. Without a very large body of stratigraphic information, the time-transgressive nature of boundaries in the Lake Alsek sedimentary record may be virtually unresolvable (Lowe and Walker 1984, p.286). Furthermore, the repetitive nature of Lake Alsek events means that between any two sampling sites, similar depositional sequences may be preserved that are of markedly different ages. For these reasons, the difficulties of postulating meaningful stratigraphic correlations between the sites investigated in this study are considerable. A schematic diagram illustrates some of the complications involved in correlating Lake Alsek sediment sequences (Fig. 4.3). Similar to a Lowe and Walker (1984 p.287) diagram depicting the extent of glaciation over time, this represents the extent of Lake Alsek pondings in distance from Lowell Glacier, and through time (as proposed by Clague and Rampton 1982, fig. 11). It can be seen that the site closest to Lowell Glacier (site 1) is affected by fillings and drainings of Lake Alsek for almost every lake phase, while the most distal site (site 3) is affected much less. At sites 1 and 2 the environmental history is recorded in sequences of eutrophic lake muds, glaciolacustrine sediments, and coarser jokulhlaup-related sediments, whereas at site 3 only the first two of these are recorded. The sedimentary record at site 1, which underwent many inundations by Lake Alsek, may have many gaps because of susceptibility to erosion by draining lake water. By contrast, a thicker sequence of sediments may have accumulated at site 2 where erosional processes are less active and possibly subordinate to depositional processes. At site 3, where no drainage erosion occurs at all, an older sequence has accumulated which is potentially 109 Figure 4.3 Schematic representation of the extent of Lake Alsek inundations over time. Four hypothetical lake phases are shown, each with several filling and draining events. Hypothetical sediment sequences at three sites of increasing distances from the glacier dam are shown. For explanation, see section 4.5.1. D i s t a n c e f r o m L o w e l l G l a c i e r i c e dam 110 much thicker than sequences from the other sites when the entire depositional record of many lake phases is considered. This illustrates a major difficulty in correlating and interpreting Lake Alsek sediments because the thickest and most complete sequence of sediments may be preserved at a site that was not inundated by Lake Alsek for the greatest length of time. Moreover, the stratigraphic successions preserved at these sites may be very similar in many characteristics, but very different in age, a possibility very difficult to detect without other evidence. The depositional and erosional history of Neoglacial events in the Lake Alsek basin is thought to be complex (Clague and Rampton 1982), thus introducing great uncertainty into stratigraphic correlations between any but the most unambiguous sequences. The unreliability of correlation in the Lake Alsek sediments may be unresolvable without a detailed chronostratigraphic study of the sequences using some of the methods discussed above and in Chapter 5. 4.5.2 Spatial Variability Without meaningful correlations between depositional sequences preserved at different localities, discussion of the spatial variability of these deposits is based on premises that are less than perfect. The problems associated with missing time-stratigraphic correlations among the Lake Alsek sediments have also been found in the more extensive study of Lake Missoula sediments. The sediments associated with a single Lake Alsek phase undoubtedly vary throughout the basin, but because of time-transgressive erosional boundaries and a lack of chronostratigraphic data, correlations are speculative and the nature of spatial variability is unknown. 4.6 Discussion of Facies Sequences Application of Markov chain analysis was possible on four of the eight sedimentary sequences examined. In none of these is it possible to demonstrate cyclicity among the I l l facies as defined. Nevertheless, these results contain some valuable information which can greatly assist in environmental interpretation: 1) In two sequences (KM4.4 and KM13.4C) the diamict (D) facies were randomly dispersed. This result supports an earlier interpretation of this facies as an iceberg-rafted sediment that occurs at random in the sediment record. In the other sequence (KM13.4D) the diamict (D) facies were found to non-randomly overlie the carbonaceous mud (O) and laminated muds (FI) facies, as well as the erosional unconformities (SS), a result explained in part by the difficulty of defining the lower boundary of the diamict (D) facies. 2) In three sequences, the erosional unconformities (SS) were non-randomly overlain by the sands (S) facies; a result consistent with an interpretation of the S facies as a tractive current deposit. 3) The link between the glaciolacustrine (Fm and FI) facies and the sands (S) facies, (either directly or through the SS state), is apparent in all sequences. Interpretation of this link is examined in the following set of hypotheses: Hj: The sands (S) / fines (FI and Fm) unit is one annual composite varve in a sequence of several composite varves which correspond to a single lake filling. H2: The sands (S) / fines (FI and Fm) unit is one drain-and-fill event that corresponds to one lake filling. The sands (S) are a tractive current deposit while the fines (FI and Fm) are laminated lacustrine deposits and possibly varves. Hg: the sands (S) / fines (FI and Fm) unit is a single density current deposit initiated by inflowing river water into a filling or full Lake Alsek. H4/. a sands (S) / fines (FI and Fm) unit is a single density current deposit initiated by a valley side subaqueous slump in a filling or full Lake Alsek. Of these possibilities, is unlikely because of the coarseness of the sands (S) facies and the thickness of the couplets. H2 is difficult to test without an independent estimate of the hydraulics of draining lake water. The draining floodwaters may or may not have the tractive force necessary to transport and deposit sediment. Hg seems 112 unlikely since inflowing river water only carries suspended loads of a few hundred mg per liter, when at least 2 g per liter are necessary to initiate a density current (Gilbert and Desloges 1986). also seems unlikely because of the sorting observed in the discrete facies and because of the regularity of these sequences. 4) The non-random occurrence of the carbonaceous mud (O) facies in the sequences always overlies the massive and laminated muds (Fm and FI) facies and except for one transition of 0->D, underlies erosional surfaces (SS). This result is consistent with an interpretation of this facies as a carbonaceous mud produced by autochthonous vegetation in the ponds during non-Lake Alsek phases (as suggested above), but may have other interpretations. For example, floating forest litter stranded as lake water fell during a drainage event is another possibility. The observed paucity of subjacent rootlets and the sharp lower boundaries of this facies is consistent with this interpretation. 5) The erosional surfaces (SS) is non-randomly underlain by one of the glaciolacustrine mud (FI and Fm) facies. This transition suggests that glaciolacustrine muds deposited during a Lake Alsek filling are more resistant to erosion by draining lake waters than the other three facies. 113 Chapter 5 CHRONOLOGY OF NEOGLACIAL LAKE ALSEK EVENTS Given the stratigraphic record of Lake Alsek, it would seem relatively straightforward to establish age estimates for a number of stratigraphic layers, and using principles of age-equivalence and superposition, to reconstruct at least a partial chronological history of Lake Alsek. Dating a stratigraphic record of Neoglacial age is possible by a number of direct and indirect means (Table 5.1) several of which lend themselves to the Lake Alsek sediments. 5.1 Available Holocene Dating Techniques A number of absolute and relative dating methods are available to establish a realistic dating framework for the Lake Alsek stratigraphy. Each method is applicable only to a limited range of materials, and only to a specific segment of time. Available dating techniques that operate over the Holocene time range can be grouped into three broad categories (Lowe and Walker 1984, p.221): 1) Absolute age estimates provided by radiometric methods or by incremental methods, 2) Absolute age equivalences based on recognizable deposits or occurences of known age, and 3) Relative age estimates based on principles of superposition or based on the degree of degradation or chemical alteration. In addition, another more obvious technique may be added to these, 4) Absolute age estimates based on historical information. The principal techniques, the associated temporal range of applicability, and the materials necessary for application of each technique are compiled in Table 5.1. References are provided so a detailed discussion of the techniques themselves other than applications to Lake Alsek will not be given here. As can be seen, each technique is applicable to a limited and usually different range of materials. 114 Table 5.1 Available Methods for Dating Holocene Events Technique Range (a BP) Primary Material 1) Incremental Dendrochronology Varves Lichenometry 2) Radiometric Fission track TL 3 H 210 p b /206 p b 137 Cs 4 0 A r / 3 9 A r 14 C /14 N AMS 1 4 C Uranium series 3) Isochronic DRM Tephrochronology 4) Chemical Hydration Rinds Amino-acid racemization Fl/U/N content Weathering/ Pedogenesis 0 - 7x10° 0 - 2xl0 4 0 - 5xl0c 2xl0 2 - 109 102 - 108 0 - 1.6x10" 0 - 2x10" 0 - 28 103 - 105 3xl0X - 4xl04 3xl0 2 - 4xl0 4 (up to 8xl04) 2xl03 - 5xl05 0 - 7xl0l 0 - 10c 103 - 106 102 - 106 0 - 10c 103 - 106 5) Historical (Alsek Valley) Written 0-97 live or fallen trees with growth rings annually laminated lake sediments subaerial rocks tephra, basalt loess, lake silt recent sediment recent sediment, ice cores recent sediment tephra, extrusive igneous rocks organic material (i.e. carbon) organic material (i.e. carbon) travertine, marl, marine sediment lake/marine sediment tephra obsidian organic material bones rock surfaces, soils Verbal 0 - 136 Glave (1892) deLaguna(1972) a) Sources include: Bradley (1985), Goudie (1981), Lowe and Walker (1984), Mahaney (1984), and Wintle and Huntley (1982). 115 What is not shown in this table is an indication of potential precision and accuracy associated with dates obtained by each one of these methods. The precision of a date is the degree of uncertainty attached to the measurement used to define the date. In the context of Lake Alsek stratigraphy, which is of high resolution, relatively small imprecision of dates may render many techniques invalid (i.e. TL dating). Accuracy, on the other hand, is the closeness of agreement between the measured age and the true age. Sample contamination or mistaken provenance of materials used for dating are potential sources of inaccuracy in dating the Lake Alsek stratigraphy. Establishing many dates throughout the stratigraphy is a means of checking accuracy. 5.2 Relative Age Estimates The Lake Alsek stratigraphy alone offers some indication of the order (but not number, due to erosional unconformities) of events by obvious application of the principle of superposition. Stratigraphic position is used to establish relative age estimates, and these can be given absolute dates by using other estimates of age for chronologic control. Sedimentological interpretation of the stratigraphy remains ambiguous, so associated with each of these interpretations is a hypothetical set of mutually exclusive relative chronologies of events in the Alsek Valley. Only one absolute date was established in the Lake Alsek stratigraphy (section 5.3) but this single date becomes much more meaningful when considered in its stratigraphic context. To establish a time-scaled stratigraphy, sediment sequences are divided into units which correspond to intervals of time. Without a more detailed study of depositional rates, this technique is severely hampered. Establishing realistic depositional rates in geomorphological studies is a complex undertaking; for the purposes of chronostratigraphy however, order of magnitude estimates may suffice. Major filling events in the Alsek Valley may have lasted no more than about 8 years, and many smaller fillings may have lasted only a year or two (Table 2.3). On this basis, the glaciolacustrine facies (Fm and FI 116 facies) may be assigned a depositional rate per centimetre of about 1 month to 1 year. The iceberg-derived diamicton (D facies) may be assigned a rate per centimetre of about 1 minute to 1 hour. Catastrophic flood events may have lasted no more than a week or two for the largest fillings, so the sand facies (S) may be assigned a depositional rate per centimetre of about 1 day to 1 week. The intervening non-inundated phases may last 75 years (as in the present interval) or more, so the carbonaceous mud (O) facies may be assigned a depositional rate per centimetre of about 50 years. Restructuring the stratigraphies from Lake Alsek using a linear time scale and the estimates of depositional rates offers what would appear to be a useful means of correlation. However, the different geometries of sampling sites means there is no guarantee that a particular lake phase at one site corresponds to that at another site (section 4.5.1). Because of this, it cannot be claimed that chronostratigraphic boundaries are time-parallel. 5.3 Radiocarbon Result While the prospect of obtaining an excellent set of radiocarbon dates from the carbonaceous mud (O facies) layers seems remarkably good, most of these layers are too thin and the cores too small in diameter for enough datable material to be recovered. Furthermore, loss-on-ignition tests indicate only about 5% organic matter in these relatively organic-rich layers (Table 3.2). Only one aggregate sample of organic material corresponding stratigraphically with KM13.4C(77-117) 104 cm depth was obtained and successfully dated. The sample was mainly of fine-grained organic detritus mixed with silt and clay-sized mineral particles. The carbon content was analyzed to be approximately 0.2 g, therefore the sample was given an extended counting time of 4 times normal to reduce the statistical error to a reasonable amount. The sample was dated at 2840 ± 190 •*"4C years BP (Beta-15530). This date is unlikely to be biased by younger carbonaceous material since the sample was buried sufficiently deep to avoid contamination by modern 117 material. However, the date may be biased in the other direction if older detritus was transported to the site and deposited. If the radiocarbon date is accurate, as many as six (or more, depending on removal of sediment at erosional unconformities), or as few as four Lake Alsek phases have been recorded in the sediments since this time (Fig. 3.5c). The presence of Lake Alsek sediments under the dated layer indicate that lakes existed prior to 2840 ^ 4 C years BP. In fact, the Lake Alsek sediments are suspected to be at least 2.5 m deep at this site, based on exploratory probings with a soil auger (section 3.4.2). More than 1.46 m of Lake Alsek sediment must therefore underlie the dated material. If the date is accurate, then this result has important implications for the Holocene climatic chronology of the area. An advance of the Lowell Glacier terminus, sufficient to impound Lake Alsek must have occurred at least once prior to 2840 years BP, and probably many times if the sediments are similar in resolution to the upper 1.2 m sediments. The currently accepted date for the onset of Neoglaciation at approximately 2.8 ka BP (Rampton 1981) conflicts with this finding, therefore further evidence is needed. 5.4 White River Ash The White River Ash is bi-lobate tephra covering much of the southern Yukon Territory. The northern lobe has been dated at 1.980 ka BP, and the eastern lobe dated at 1.230 ka BP (Denton and Karlen 1977). The eastern lobe encompasses part of the Lake Alsek basin and can be found exposed in road cuts as a thin (1.0-10.0 mm) tephra layer 10-50 cm below the surface in areas of low slope. Of the sites sampled in this study, only KM69.0 lies within the boundaries of this ashfall. No other Neoglacial tephra can be found in this area, thus identification is straightforward. The tephra can be observed in soil pits (in the Lake Alsek basin) interbedded with fine-grained sediments that have been assigned a lacustrine origin, (Appendix VI), but it is not known if the tephra was deposited during an inundation of Lake Alsek. No tephra was detected in the sediment cores from 118 KM69.0, even though the 4.91 m long core probably spans the entire Neoglacial. This missing tephra may be explained in two ways; (a) the tephra deposit may be thin or non-existent in the lake sediments due to ice cover at the time of eruption or else to floating of the tephra out of the lake basin, (b) it has been found elsewhere that tephra deposits in lacustrine environments are spatially discontinuous due to differential settling rates (Anderson et al. 1984), so a single core from KM69.0 may be inadequate for detecting a discontinuous tephra layer. 5.5 Historical Information The most recent ponding phase of Lake Alsek occurred around 1909 A.D. (between 1882 and 1917 A.D.), filling to the 561 m asl level and reaching about 40 km upstream of the Lowell Glacier ice dam (Clague and Rampton 1982, p. 112). Evidence for this is somewhat equivocal, consisting of imprecise historical evidence, and a radiocarbon dated driftwood paddle of "modern" age. In spite of the high probability that the area was visited during this time, no historical reference to a lake has so far been found. The lake may have easily filled and emptied without being noticed since it could have done so in as few as 72 days (Table 2.3). During the summer of 1984, a broken wooden plank (about 50x20x2 cm in size) with traces of paint on the unexposed underside was found in the driftwood windrow at 561 m asl elevation near the confluence of Lava Creek and Alsek River (Fig. 1.1). This plank was probably made and painted by non-natives, whose presence in the area was extremely limited before about 1890. Thus, it is unlikely that the plank was deposited by a Lake Alsek ponding as old as the 1852 phase. This provides additional evidence for a separate, more recent ponding phase. In an account by Glave (1892) of a packhorse trip with Jack Dalton from Pyramid Harbour, Alaska to Kluane Lake, reference is made to "valleys of rocks" and "fertile grasslands" in the Dezadeash Valley that are possibly coarse gravel/cobble (jokulhlaup-119 related?) flood bars and areas of lake bottom sediments in the early stages of vegetation colonization. Two photographs (reproduced as drawings) included in this account are of recognizable locations on the Dezadeash River just upstream of the Kaskawulsh River confluence. These show that low coniferous vegetation in 1891 has become taller but maintained the same approximate distribution to the present day. Unfortunately, this account does not discuss the portion of the valley which would have been inundated by a 561 m asl Lake Alsek filling. 5.6 Further Chronological Work It should be possible to establish good chronological control on the Lake Alsek stratigraphy. Two levels of resolution are desirable, the first is to resolve the number of Lake Alsek phases (associated with an advanced terminus position of Lowell Glacier) during the Neoglacial, and the second is to resolve the filling and draining events of each major lake phase. The carbonaceous layers can yield more ^^C dates if enough material can be recovered from more cores, or better yet, from pits dug into the sediments. The lacustrine muds could be dated by TL if it can be shown that the TL clocks were effectively zeroed by sunlight during the residence time in Lake Alsek or during the fluvial passage before entering the lake. Interstratified "^C dates may provide independent control on dates obtained through TL methods and be used as an accuracy check in this way. More complete coring of KM69.0 should turn up the White River Ash deposit providing excellent chronostratigraphic control for the relatively complete sedimentary record found here. Finally, an exhaustive search of the historical literature from this area covering the first decade of this century may turn up some reference to a Lake Alsek filling, or at least to the recently emptied lake basin from a filling to the 561 m asl driftwood strandline. 120 Chapter 6 SUMMARY AND CONCLUDING REMARKS 6.1 Summary of results The major findings of this study can be summarized as follows: 1) The Neoglacial history of Lake Alsek events is recorded in varying degrees of completeness in the sediments of (a) small gravel-impounded and kettle-type ponds in the Alsek Valley (KM1.2, KM4.4, KM13.4C, KM13.4D, KM28.5, and KM50.4) and, (b) a >20 m deep lake in the Dezadeash Valley (KM69.0). 2) The contemporary sediments of these sites are carbonaceous muds in the shallow ponds and poorly sorted fine silts in the deeper lake. 3) The sediments associated with Lake Alsek are (a) thinly laminated muds (poorly sorted fine silts) with occasional dropstones, (b) massive muds (poorly sorted fine silts), and (c) occasional deposits of structureless diamicton (extremely poorly sorted medium sands). 4) The sediments associated with episodic drainage of the lake are silty sands and sandy silts (poorly sorted very fine sands) with occasional rip-up clasts of cohesive mud. These deposits are often underlain by erosional unconformities. 5) Erosional unconformities indicate relatively high-energy environments, and may represent major gaps in the stratigraphic record. These were found in the sediments from all sites investigated in the Kluane Ranges portion of Alsek Valley. At these sites, the combination of narrow valley cross-sections and high flux outburst floods may produce these high-energy environments. Sediments deposited at KM69.0 (in Dezadeash Valley) are not sensitive to any but the most extensive Lake Alsek- phases, and no erosional unconformities were observed here. 6) A summary sequence of Lake Alsek sediments cannot be produced without further coring and facies sequence analysis. As a guide for future studies, the sequence from KM13.4C is the best choice. This includes glaciolacustrine (Fm, FI, and D), eutrophic lacustrine (O), and flood (S) sediments, as well as an erosional unconformity. 121 7) At site KM13.4C, 1.04 m of Lake Alsek-related sediment overlies a carbonaceous mud layer dated at 2840 1 4 C years BP. This is apparently underlain by at least another 1.5 m of additional Lake Alsek sediment. 8) Dating the Lake Alsek stratigraphy was not accomplished satisfactorily, but datable carbonaceous mud layers are intercalated with the Lake Alsek sediments. The high-magnitude, low-frequency geomorphological events associated with pondings and drainings of Lake Alsek give rise to a depositional environment dominated by "long periods of boredom and brief periods of terror" (Ager 1981, p. 106). The sedimentary sequence records the Lake Alsek events to a higher resolution than the intervening non-Lake Alsek events. Details of Lake Alsek are represented by more core per unit time. The Lake Alsek depositional events are represented by more sediment, but in terms of a time-scaled stratigraphy, are less significant. 6.4 Concluding Remarks As stated in Chapter 1, the fundamental objective of this study was to investigate the characteristics and chronology of Neoglacial Lake Alsek by undertaking a stratigraphically based investigation of the Lake Alsek sediments. The portion of this objective concerned with the study of sediments and sediment sequences from an ice-dammed lake has met with some success. However, that part of the objective concerned with basin-wide correlation of sedimentary records and construction of a chronology of filling and draining events has been less successful. The difficulties encountered in this regard can be summarized as follows: 1) Sampling of the Neoglacial stratigraphic record is incomplete due to limited penetration of sediments by the coring techniques used. 2) Interpretation of observed sediments is severly restricted by the essentially one-dimensional view of the sequences afforded by a narrow sediment core. 122 3) Erosional unconformities that indicate sediment removal and may make much of the Lake Alsek stratigraphic record "a lot of holes tied together with sediment" (Ager 1981, p.35). 4) Incomplete analysis of the spatial variability in hydrodynamic conditions during Lake Alsek draining events. 5) Correlating stratigraphic records from different sites is problematic: this is because of both depositional variability and stratigraphic gaps at erosional unconformities. 6) Sufficient dating of the stratigraphic record (for both correlation and for construction of a Lake Alsek chronology) was not accomplished. Clearly, many more questions concerning the paleoenvironment of ice-dammed Neoglacial Lake Alsek need to be answered. Now that the sedimentary record has been located and identified, the work described here may lead to a more refined and comprehensive study of the spatial and temporal variability of these deposits. Further work will ultimately, lead to a more complete chronology of Neoglacial Lake Alsek. 123 REFERENCES CITED Ager, D.V. 1981. The nature of the stratigraphical record (2nd ed.). John Wiley and Sons, New York, 122 p. Anderson, R.Y., Nunfer, E.B. and Dean, W.E. 1984. Sinking of volcanic ash in uncompacted sediment in Williams Lake, Washington. Science, 2 2 5 , pp. 505-508. Ashley, G.M. 1975. Rhythmic sedimentation in Glacial Lake Hitchcock, Massachusetts-Connecticut. In Glaciofluvial and glaciolacustrine sedimentation. Edited by A.V. Jopling and B.C. McDonald. Society of Economic Paleontologists and Mineralogists, Special Publication 2 3 , pp. 304-320. ASTM, 1972. Standard method for particle-size analysis of soils, test D422-72. American Society for Testing and Materials, Philadelphia, PA, pp. 113-123. Atwater, B.F. 1984. Number, frequency, and relative magnitudes of last-glacial floods from Pleistocene Lake Missoula, Montana - Evidence from the Sanpoil River valley, northeastern Washington. EOS, 6 5 , p. 893. Baker, V.R. 1973. Paleohydrology and sedimentology of Lake Missoula flooding in eastern Washington. Geological Society of America, Special Paper 144, 79 p. Baker, V.R. and Bunker, R.C. 1985. Cataclysmic late Pleistocene flooding from Glacial Lake Missoula: a review. Quaternary Science Reviews, 4, pp. 1-41. Beget, J.E. 1986. Comment on "Outburst floods from Glacial Lake Missoula" by G.K.C. Clarke, W.H. Mathews, and R.T. Pack. Quaternary Research, 2 5 , pp. 136-138. Bostock, H.S. 1952. Geology of northwest Shakwak Valley, Yukon Territory. Geological Survey of Canada, Memoir 267, 54 p. Bostock, H.S. 1969. Kluane Lake, Yukon Territory, its drainage and allied problems. Geological Survey of Canada, Paper 69-28, 19 p. Bradley, R.S. 1985. Quaternary paleoclimatology: methods of paleoclimate reconstruction. Allen and Unwin, Boston, 472 p. Brooks, A.H. 1900. A reconnaissance from Pyramid Harbour to Eagle City, Alaska, including a description of the copper deposits of the upper White and Tanana Rivers. United States Geological Survey, Annual Report 21 (1899-1900), Part 2, pp. 331-391. Brown, M.B. 1983. Frequency tables. In BMDP: Biomedical Computer Programs, P Series. Edited by W.J. Dixon. University of California Press, Los Angeles, 734 p. Burrows, C.J. and Gellatly, A.F. 1982. Holocene glacier activity in New Zealand. In Holocene Glaciers. Edited by W. Karlen. Striae, 18, pp. 41-47. Campbell, R.B., and Dodds, C.J. 1975. Operation Saint Elias, Yukon Territory. In Report of Activities, Part A, Geological Survey of Canada, Paper 75-IA, pp. 51-53. 124 Campbell, R.B., and Rampton, V.N. 1980. The story of the rocks and glaciers. In Kluane, Pinnacle of the Yukon. Edited by J.B. Theberge. Doubleday Canada Ltd., Toronto, Ont., pp. 3-17. Carr, T.R. 1982. Log-linear models, Markov chains and cyclic sedimentation. Journal of Sedimentary Petrology, 52, pp. 905-912. Chambers, R.L. 1971. Sedimentation in Glacial Lake Missoula. Unpublished M.S. Thesis, University of Montana, 100 p. Church, M. 1980. Records of recent geomorphological events. In Timescales in Geomorphology. Edited by R.A. Cullingford, D.A. Davidson, and J. Lewin. Wiley and Sons Ltd., Chichester, pp. 13-29. Church, M., and Ryder, J.M. 1972. Paraglacial sedimentation: a consideration of fluvial processes conditioned by glaciation. Geological Society America Bulletin, 83, pp. 3059-3072. Church, M., and Gilbert, R. 1975. Proglacial fluvial and lacustrine environments. In Glaciofluvial and glaciolacustrine sedimentation. Edited by A.V. Jopling and B.C. McDonald. Society of Economic Paleontologists and Mineralogists, Special Publication 23, pp. 22-100. Clague, J.J. 1979. An assessment of some possible flood hazards in Shakwak Valley, Yukon Territory. In Current research, part B. Geological Survey of Canada, Paper 79-IB, pp. 63-70. Clague, J.J., and Mathews, W.H. 1973. The magnitude of jokulhlaups. Journal of Glaciology, 12, no.66, pp. 501-504. Clague, J.J., and Rampton, V.N. 1982. Neoglacial Lake Alsek. Canadian Journal of Earth Sciences, 19, pp. 94-117. Clague, J., Jozsa, L.A., and Parker, M.L. 1982. Dendrochronological dating of glacier-dammed lakes: an example from Yukon Territory, Canada. Arctic and Alpine Research, 14, no.4, pp. 301-310. Clarke, G.K.C. 1982. Glacier outburst floods from 'Hazard Lake', Yukon Territory, and the problem of flood magnitude prediction. Journal of Glaciology, 28, pp. 3-22. Clarke, G.K.C. 1986. Professor Mathews, outburst floods, and other glaciological disasters. Canadian Journal of Earth Sciences, 23, pp. 859-868. Clarke, G.K.C., and Mathews, W.H. 1981. Estimates of the magnitude of glacier outburst floods from Lake Donjek, Yukon Territory, Canada. Canadian Journal of Earth Sciences, 18, pp. 1452-1463. Clarke, G.K.C., Mathews, W.H., and Pack, R.T. 1984. Outburst floods from Glacial Lake Missoula. Quaternary Research, 22, pp. 289-299. Clarke, G.K.C., and Waldron, D.A. 1984. Simulation of the August 1979 sudden discharge of glacier-dammed Flood Lake, British Columbia. Canadian Journal of Earth Sciences, 21, pp. 502-504. 125 Cooper, M.A.R. 1974. Fundamentals of survey measurement and analysis. Crosby Lockwood Staples, London, 107 p. Day, J.H. 1962. Reconnaissance soil survey of the Takhini and Dezadeash Valleys in the Yukon Territory. Canada Department of Agriculture, Research Branch, 78 p. de Laguna, F. 1972. Under Mount Saint Elias: the history and culture of the Yakutat Tlingit. Smithsonian Conributions to Anthropology, no. 7 (3 pts), 1395 p. Denton, G.H. 1974. Quaternary glaciations of the White River Valley, Alaska, with a regional synthesis for the northern St. Elias Mountains, Alaska and the Yukon Territory. Geological Society of America Bulletin, 85, pp. 871-892. Denton, G.H., and Stuiver, M. 1967. Late Pleistocene glacial stratigraphy and chronology, northeastern St. Elias Mountains, Yukon Territory, Canada. Geological Society of America Bulletin, 78, pp. 485-510. Denton, G.H., and Karlen, W. 1973. Holocene climatic variations - their pattern and possible cause. Quaternary Research, 3, pp. 155-205. Denton, G.H., and Karlen, W. 1977. Holocene glacial and tree-line variations in the White River Valley and Skolai Pass, Alaska and Yukon Territory. Quaternary Research, 7, pp. 63-111. Dott, R.H. Jr. 1983. Episodic sedimentation - how normal is average? how rare is rare? does it matter? Journal of Sedimentary Petrology, 53, pp. 5-23. Douglas, G.W. 1974. Montane zone vegetation of the Alsek River region, southwestern Yukon. Canadian Journal of Botany, 52, pp. 2505-2532. Everitt, B.S. 1977. The analysis of contingency tables. Chapman and Hall, London, 128 p. Eyles, N., Eyles, C.H., and Miall, A.D. 1983. Lithofacies types and vertical profile models; an alternative approach to the description and environmental interpretation of glacial diamict and diamictite sequences. Sedimentology, 30 , pp. 393-410. Fienberg, S.E. 1980. The analysis of cross-classified categorical data, 2nd ed. MIT Press, Cambridge, Massachusetts, 198 p. Folk, R.L. 1964. A review of grain-size parameters. Sedimentology, 6, pp. 73-93. Folk, R.L. 1980. Petrology of sedimentary rocks. Hemphill, Austin, Texas, 182 p. Gilbert, R. 1972. Drainings of ice-dammed Summit Lake, British Columbia. Inland Waters Directorate, Water Resources Branch, Ottawa, Canada. Scientific Series, 20, 17 p. Gilbert, R., and Shaw, J. 1981. Sedimentation in proglacial Sunwapta Lake, Alberta. Canadian Journal of Earth Sciences, 18, pp. 81-93. Gilbert, R., and Desloges, J.R. 1986. Sediments of ice-dammed, self-draining Ape Lake, British Columbia. Unpublished manuscript, 37 p. Glave, E.J. 1892. Pioneer packhorses in Alaska. Century Magazine, 44, pp. 671-682, 869-881. 126 Goudie, A.S. editor. 1981. Geomorphological techniques. British Geomorphological Research Group. London, George Allen and Unwin, 395 p. Gretener, P.E. 1967. Significance of the rare event in geology. American Association of Petroleum Geologists Bulletin, 51, no. 11, pp. 2197-2206. Gustavson, T.C. 1975. Sedimentation and physical limnology in proglacial Malaspina Lake, southeastern Alaska. In Glaciofluvial and glaciolacustrine sedimentation. Edited by A.V. Jopling and B.C. McDonald. Society of Economic Paleontologists and Mineralogists, Special Publication 23, pp. 249-263. Hakanson, L. 1981. A manual of lake morphometry. Springer-Verlag, Berlin, 78 p. Hakanson, L., and Jansson, M. 1983. Principles of lake sedimentology. Springer-Verlag, Berlin, 316 p. Harbaugh, J.W., and Bonham-Carter, G.F. 1970. Computer simulation in geology. Wiley Interscience, New York, 575 p. Harper, C.W.,Jr. 1984a. Facies models revisited: an examination of quantitative methods. Geoscience Canada, 11, no.4, pp. 203-207. Harper, C.W.,Jr. 1984b. Improved Methods of Facies Sequence Analysis. In Facies Models. Edited by R.G. Walker. Geoscience Canada Reprint Series, no. 1, 2nd ed., pp. 11-13. Hilton-Johnson, W. 1982. Interrelationships among geomorphic interpretations of the stratigraphic record, process geomorphology and geomorphic models. In Space and Time in Geomorphology. Edited by C.E. Thorn. Allen and Unwin, London, pp. 219-243. Hughes, O.L., Rampton, V.N., and Rutter, N.W. 1972. Quaternary geology and geomorphology, southern and central Yukon (northern Canada). Guidebook, 24th International Geological Congress (Montreal, Quebec), Field Excursion All, 59 p. Johnson, F., and Raup, H.M. 1964. Investigations in southwest Yukon: geobotanical and archaeological reconnaissance. Papers of the Robert S. Peabody Foundation for Archaeology, 6, pp. 3-198. Johnson, P.G. 1986. Holocene paleohydrology of the St. Elias Mountains, British Columbia and Yukon. Geographie physique et Quaternaire, 40, pp. 47-53. Jopling, A.V. and Walker, R.G. 1968. Morphology and origin of ripple-drift cross-lamination, with examples from the Pleistocene of Massachusetts. Journal of Sedimentary Petrology, 38, pp. 971-984. Kindle, E.D. 1952. Dezadeash map-area, Yukon Territory. Geological Survey of Canada, Memoir 268, 68 p. Krumbein, W.C. and Dacey, M.F. 1969. Markov chains and embedded chains in geology. Mathematical Geology, 1, pp. 79-96. 127 Leeder, M.R. 1982. Sedimentology: process and product. George Allen and Unwin Ltd., London, 344 p. Leonard, E.M. 1986. Varve studies at Hector Lake, Alberta, Canada and the relationship between glacial activity and sedimentation. Quaternary Research, 25, pp. 199-214. Lerbekmo, J.F., and Campbell, J.A. 1969. Distribution, composition, and source of the White River Ash, Yukon Territory. Canadian Journal of Earth Sciences, 6, pp. 109-116. Lerbekmo, J.F., Westgate, J.A., Smith, D.G.W., and Denton, G.H. 1975. New data on the character and history of the White River volcanic eruption, Alaska. In Quaternary Studies. Edited by R.P. Suggate and M.M. Cresswell. The Royal Society of New Zealand, Wellington, pp. 203-209. Lewin, J. 1980. Available and appropriate timescales in geomorphology. In Timescales in Geomorphology. Edited by R.A. Cullingford, D.A. Davidson, and J. Lewin. Wiley and Sons Ltd., Chichester, pp. 3-10. Lowe, J.J., and Walker, M.J.C. 1984. Reconstructing Quaternary environments. Longman Group Ltd, New York, 389 p. MacDonald, G.M. 1982. Late Quaternary paleoenvironments of the Morley Flats and Kananaskis Valley of southwestern Alberta. Canadian Journal of Earth Sciences, 19, pp. 23-35. Mahaney, W.C. editor. 1984. Quaternary dating methods. Developments in paleontology and stratigraphy, 7. Elsevier, New York, 431 p. Mair, S.G. 1982. Use of the digitizer. University of British Columbia Computing Centre documentation, l i p . McLaren, P. 1981. An interpretation of trends in grain size measures. Journal of Sedimentary Petrology, 51, pp. 611-624. Miall, A.D. 1973. Markov chain analysis applied to an ancient alluvial plain succession. Sedimentology, 20, pp. 347-364. Miall, A.D. 1978. Lithofacies types and vertical profile models in braided rivers: a summary. In Fluvial Sedimentology. Edited by A.D. Miall. Memoirs of the Canadian Society of Petroleum Geologists, 5, pp. 597-604. Nye, J.F. 1976. Water flow in glaciers: jokulhlaups, tunnels and veins. Journal of Glaciology, 17, pp. 181-207. Passega, R. 1964. Grain size representation by CM patterns as a geologic tool. Journal of Sedimentary Petrology, 34, pp. 830-847. Pickrill, R.A., and Irwin, J. 1983. Sedimentation in a deep glacier-fed lake-Lake Tekapo, New Zealand. Sedimentology, 30, pp. 63-75. Porter, S.C, and Orombelli, G. 1985. Glacier contraction during the middle Holocene in the western Italian Alps: evidence and implications. Geology, 13, pp. 296-298. 128 Post, A., and Mayo, L.R. 1971. Glacier dammed lakes and outburst floods in Alaska. U.S.G.S., Dept. of Interior, Hydrologic Investigations Atlas HA-455, 3 sheets (1:1,000,000), 10 p. Powers, D.W., and Easterling, R.G. 1982. Improved methodology for using embedded markov chains to describe cyclical sediments. Journal of Sedimentary Petrology, 52, pp. 913-923. Rampton, V.N. 1970. Neoglacial fluctuations of the Natazhat and Klutlan Glaciers, Yukon Territory, Canada. Canadian Journal of Earth Sciences, 7, pp. 1236-1263. Rampton, V.N. 1981. Surficial materials and landforms of Kluane National Park, Yukon Territory. Geological Survey of Canada Paper, 79-24, 37 p. Read, W.A. 1969. Analysis and simulation of Namurian Sediments in Central Scotland using a Markov process model. Mathematical Geology, 1, pp. 199-219. Reading, H.G. 1978. Facies. In Sedimentary environments and facies. Edited by H.G. Reading. Blackwell, Oxford, pp. 4-14. Romesburg, H.C, Marshall, K., and Mauk, T.P. 1981. Fitest: A computer program for "Exact Chi-Square" goodness-of-fit significance tests. Computers and Geosciences, 7, pp. 47-58. Ryder, J.M., and Thomson, B. 1986. Neoglaciation in the southern Coast Mountains of British Columbia: chronology prior to the late Neoglacial maximum. Canadian Journal of Earth Sciences, 23, no.3, pp. 273-287. Schwarzacher, W. 1975. Sedimentation models and quantitative stratigraphy. Elsevier, Amsterdam, 382 p. Simpson, G.G. 1963. Historical science. In The fabric of geology. Edited by CC. Albritton, Jr.. Freeman, Cooper and Company. Stanford, California, pp. 24-48. Sly, P.G., Thomas, R.L. and Pelletier, B.R. 1983. Interpretation of moment measures derived from water-lain sediments. Sedimentology, 30, pp. 219-233. Sticht, J.H.H. 1951. Geomorphology and glacial geology along the Alaska Highway in Yukon Territory and Alaska. Ph.D. thesis, Harvard University, 176 p. Tempelman-Kluit, D. 1980. Evolution of physiography and drainage in southern Yukon. Canadian Journal of Earth Sciences, 17, pp. 1189-1203. Turk, G. 1979. Transition analysis of structural sequences, discussion. Geological Society of America Bulletin, 90, pp. 989-991. Waitt, R.B. 1984. Periodic jokulhlaups from Pleistocene glacial Lake Missoula - new evidence from varved sediment in Northern Idaho and Washington. Quaternary Research, 22, pp. 46-58. Walker, R.G. 1979. Facies and facies models, In Facies models. Edited by R.G. Walker. Geological Association of Canada Reprint Series 1, pp. 1-8. 129 Walker, R.G. 1984. General introduction: Facies, facies sequences and facies models, In Facies models. Edited by R.G. Walker. Geological Association of Canada, Reprint Series 1, 2nd ed., pp. 1-9. Williams, G.P. 1984. Paleohydrologic equations for rivers. In Developments and applications of geomorphology. Edited by J.E. Costa and P.J. Fleisher. Springer-Verlag, Berlin, pp. 343-367. Wintle, A.G., and Huntley, D.J. 1982. Thermoluminescence dating of sediments. Quaternary Science Reviews, 1, pp. 31-53. Wrigley, N. 1985. Categorical data analysis for geographers and environmental scientists. Longman Group Ltd, New York, 392 p. Young, G.J. 1980. Monitoring glacier outburst floods. Nordic Hydrology, 11, pp. 285-300. Zhao, X.F., and Hsu, K.J. 1984. Grain size analyses of Zubo sediments, In Quaternary geology of Lake Zurich: An interdisciplinary investigation by deep-lake drilling. Edited by K.J. Hsu and K.R. Kelts. Contributions to Sedimentology 13, E. Schweizerbart'sche Verlagsbuchhandlung (Nagele u. Obermiller) Stuttgart, pp. 75-86. Zhao, X.F., Hsu, K.J., and Kelts, K.R. 1984. Varves and other laminated sediments of Zubo, In Quaternary geology of Lake Zurich: An interdisciplinary investigation by deep-lake drilling. Edited by K.J. Hsu and K.R. Kelts. Contributions to Sedimentology 13, E. Schweizerbart'sche Verlagsbuchhandlung (Nagele u. Obermiller) Stuttgart, pp. 161-176. 130 APPENDIX I LEVELLING SURVEY RESULTS Table 1.1 contains the levelling survey results for the one directional traverse conducted from the Alaska Highway (USCGS Benchmark R8 at 628.834 m asl) to Lowell Glacier. This survey was conducted under the direction of S.G. Collins and G.K.C. Clarke over a period of several years. Discussion of the methods employed can be found in section 2.3. 131 F i l e JEFF .D4A A l s e k R i v e r p r o f i l e from 1984 su r v e y (1985-04-01)0 * * 1. T h i s d a t a f i l e c o n t a i n s the r e s u l t s of a l e v e l l i n g l i n e from * the USCGS Benchmark R8 (G28.834 m a s l ) near M a c i n t o s h Lodge, * and e n d i n g a t the I n t e r i m a l t i t u d e Marker PITON, c o n s i s t i n g * of a d i s c a r d e d i r o n c l i m b i n g p l t o n d r i v e n f i r m l y i n t o an * o u t c r o p of metamorphlc "greenstone" ( c h l o r i t e s c h i s t ) near the * unimproved t r a i l a t i t s e n t r a n c e t o Kluane Park. * 2. A l s o i n c l u d e d a r e a l t i t u d e s d e t e r m i n e d by computations * completed 6 Feb 1985 e x t e n d i n g the t r a v e r s e from i n t e r i m * marker "PITON" on downstream r i g h t s i d e o f Dezadeash R i v e r * t o marker "BC", i n s t a l l e d i n an o u t c r o p of bedrock t h a t * r i s e s d i r e c t l y from the r i v e r a c r o s s from Beachview Creek * f a n . * 3. A l t i t u d e s d e t e r m i n e d by computations completed e a r l i e r . * on 19 Jan 185, and s u b s e q u e n t l y m o d i f i e d t o c o n s t i t u t e * independent l i n e s c o n t i n u o u s w i t h those f u r t h e r upstream, * t h i s f i n a l l i n e b e g i n n i n g a t marker "BC" and c o n t i n u i n g down * the A l s e k R i v e r and e n d i n g near the terminus of Lowell G l a c i e r , * a t Marker "WB", a l a r g e w h i t e b o u l d e r near a v e r y prominent * b l a c k b o u l d e r on the edge of the r i p p l e - m a r k e d a r e a at the * h y d r o l o g i c r i g h t s i d e of A l s e k R i v e r , about one k i l o m e t r e * upstream w i t h i t s c o n f l u e n c e w i t h Lowell G l a c i e r . A l l * computations r e c h e c k e d f o r c o n s i s t e n c y and sequence 3 Mar 85; * l a b e l s f o r r i v e r - l e v e l s h o t s , s i d e s h o t s , and permanent markers * added 10 Mar 1985. [Notes from SGC computer o u t p u t ] Ref . Type E1evat i o n 0 i s t a n c e (m a s l ) (m) ooo F B 0 O O BM 6 2 8 . 8 3 4 OOOO OOO 0 0 1 F B # 2 3 IP 6 3 0 . 5 6 2 1 4 0 4 3 1 0 0 2 FB/C23 TP 6 7 1 .404 3 8 7 7 4 3 0 0 3 F B # 2 2 IP 6 3 1 .691 6 2 3 4 4 5 0 0 4 F B # 2 2 TP 6 3 1 . 1 2 2 5 0 2 5 7 0 0 5 F B # 2 1 IP 6 3 1 . 2 8 9 1 2 9 7 9 2 0 0 6 FB/C21 TP 6 3 0 .601 8 9 1 6 0 0 0 7 F B 0 2 O IP 6 2 8 6 1 1 2 5 4 4 5 2 0 0 8 F B # 2 0 TP 6 2 6 . 6 0 9 1 9 3 3 7 2 0 0 9 F B 0 1 9 IP 6 2 3 .644 181 2 7 4 0 1 0 F B # 1 9 TP 6 2 0 . 3 0 7 2 3 5 4 3 5 0 1 1 F B # 1 8 IP 6 1 8 3 3 8 108 5 2 1 0 1 2 F B / M 8 TP 6 1 4 . 1 7 8 1 8 9 0 4 0 0 1 3 F B # 1 7 IP 6 1 2 6 8 9 1 0 8 3 6 1 0 1 4 FB/SM7 TP 6 0 3 1 6 6 3 4 2 8 6 5 0 1 5 F B / M 6 IP 5 9 9 9 1 9 1 5 4 4 0 0 0 1 6 F B # 1 6 TP 5 9 8 0 5 7 5 7 2 1 8 0 1 7 FB/f 1 5 IP 5 9 6 4 2 6 6 8 0 4 3 0 1 8 F B # 1 5 TP 5 9 5 7 0 6 7 7 1 2 3 0 1 9 FB/f 14 IP 5 9 4 9 2 2 1 13 8 6 0 0 2 0 F B # 1 4 TP 6 0 7 5 5 7 1 2 6 148 0 2 1 F B # 1 3 IP 5 9 8 6 0 6 1 8 0 6 7 1 0 2 2 F B # 1 3 TP 6 1 6 0 4 2 131 8 5 7 0 2 3 F B # 1 2 IP 6 2 3 5 9 0 4 3 8 5 5 3 0 2 4 F B # 1 2 TP 6 3 8 2 2 9 1 1 4 3 7 3 8 0 2 5 FB/C1 1 IP 6 2 8 2 9 4 6 6 2 7 6 0 0 2 6 F B 0 1 1 TP 6 1 7 8 2 7 2 4 8 0 9 2 0 2 7 FB/C05 IP 6 2 4 5 0 4 2 8 5 198 0 2 8 FB/C05 TP 6 0 1 4 14 7 6 9 7 5 5 0 2 9 F B # 0 4 IP 5 9 9 8 4 2 5 5 0 7 8 1 0 3 0 FBJC04 TP 5 9 8 187 6 8 1 3 5 6 0 3 1 F B # 0 3 IP 5 9 2 2 6 9 5 1 1 2 1 1 0 3 2 F B # 0 3 TP 5 8 1 9 2 4 3 5 5 6 0 7 0 3 3 F B # 0 1 IP 5 9 5 5 1 0 5 3 6 8 7 5 0 3 4 F B # 0 1 S1 5 7 3 7 8 0 1 1 1 5 2 4 0 0 3 5 F B # 0 1 S2 5 7 3 8 1 4 4 7 4 5 2 2 0 3 6 F B # 0 1 S 3 5 8 3 4 5 2 6 7 2 4 5 0 0 3 7 F B # 0 1 S 4 5 7 3 3 2 4 6 8 3 . 5 4 4 0 3 8 F B # 0 1 TP 5 9 7 4 0 2 174 . 148 0 3 9 F B # 0 2 IP 5 9 9 6 4 5 2 5 2 . 0 0 9 Notes USCGS BM nr M a c i n t o s h Lodge Marker PITON nr Dezadeash R bend (GKCC) Rvr l v l a t ?? (poor d e t e r m i n a t n ) Rvr l v l a t S i t e "17.5" ( c o r r e c t e d f o r HT) Lake 17.5 f l o o d i n l e t ( c o r r e c t e d f o r HT) Lake 17.5 s f c l v l 132 040 FB#02 S5 574 .029 810 639 Lake 18 sfc Ivl (KM50.4) 041 FB#02 S6 573 .537 695 773 Rvr lvl at 18 (KM50.4) 042 FB#02 TP 607 .919 511 203 043 FB/C06 IP 598 .976 809 370 044 FB#06 TP 575 253 593 122 045 FB0O7 IP 579 026 500 751 046 FB/C07 TP 586 231 520 679 047 FB#08 IP 574 938 406 556 048 FB#08 S7 573 121 58 763 Rvr lvl 049 FB#08 TP 597 .947 669 583 050 FB#09 IP 582 . 782 532 696 051 FB0O9 TP 580 169 590 827 052 FB/SMO IP 578 019 259 173 053 FB/MO S8 572 620 145 014 Rvr lvl (GKCC was rod man I think) 054 FB#10 TP 592 358 891 750 Marker BC (rck bif f acr fm Beachvlew 055 FB#01 IP 572 581 648 336 056 FB/fOI TP 575 757 515 1 1 1 057 FB/C02 IP 572 980 727 302 058 FB#02 TP 572 057 541 724 Rvr lvl 059 FB#03 IP 572 397 432 388 060 FB#03 TP 572 083 624 654 061 FB#04 IP 572 570 1399 725 062 FB#04 TP 572 791 667 265 1.5 m above rvr lvl 063 FB#05 IP 571 429 514 426 064 FB/C05 TP 570 125 623 032 Rvr lvl 065 FB0O6 IP 569 146 499 433 066 FB#06 TP 569 719 557 227 067 FB0O7 IP 567 785 370 849 068 FB#07 TP 566 729 391 301 Rvr lvl 069 FB#08 IP 566 461 418 643 070 FB^ S TP 564 466 728 867 Rvr lvl 071 FB/C09 IP 563 523 445 106 072 FB#09 TP 561 797 597 335 Rvr lvl 073 FB/C10 IP 560 682 643 589 074 FB*10 TP 558 209 776 359 075 FB/M 1 IP 557 509 525 378 076 FB#1 1 TP 556 434 570 237 077 FB#12 IP 554 609 889 607 078 FB#12 TP 552 564 749 279 Rvr lvl 079 FB#13 IP 55 1 817 904 978 080 FB/SM3 TP 575 862 699 857 Marker KFE (Kask fan end) 081 FB#14 IP 551 540 1056 735 082 FB(f 14 TP 576 552 143 951 083 FB#15 IP 548 560 868 226 084 FB/M5 TP 546 966 731 930 1.0 m above rvr lvl 085 FB/C16 IP 547 820 1034 247 086 FB/M6 TP 544 388 525 040 Rvr lvl 087 FB#17 IP 544 226 754 938 088 FB#17 TP 543 373 508 782 089 FB#18 IP 542 174 745 854 090 FB/M8 TP 541 065 616 487 Rvr lvl 091 FB#19 IP 540 354 653 234 092 FB#19 TP 539 743 617 628 093 FB(C20 IP 539 222 400 721 094 FB#20 TP 537 636 733 856 095 FB#21 IP 537 056 513 513 096 FB#21 TP 536 247 496 183 Rvr lvl 097 FB022 IP 536 065 655 504 098 FB/C22 TP 534 941 506 917 Rvr lvl 099 FB#23 IP 534 310 725 203 100 FB#23 TP 533 082 526 369 "t" only at rvr lvl (t=toward TP) 101 FB#24 IP 534 048 767 882 102 FB#24 TP 532 055 665 300 103 FB#25 IP 532 662 517 326 104 FB#25 TP 529 780 703 785 105 FB#26 IP 529 232 631 352 106 FB#26 TP 527 720 669 01 1 Rvr lvl 107 FB/l'27 IP 527 014 799 618 108 FB/C27 TP 525 762 1006 692 0.5 m above rvr lvl 109 FB#28 IP 525 433 779 510 1 10 FB#28 TP 523 291 683 738 Rvr lvl at KM13.4 1 1 1 FB/i'29 IP 523 879 467 524 133 112 FB029 TP 522 .264 422 .588 " t " and "ro" at rvr l v l (m=midline) 1 13 FB#30 IP 524 .028 597, .439 114 FB#30 TP 521 .455 552, . 735 1.5 m above rvr l v l 1 15 FB#31 IP 520 .545 668 .764 1 16 FB#31 TP 520 .599 20, .653 117 FB#32 IP 521 .216 720, .948 Marker P i c P i c [at shrp bnd Alsk R 2-3 m abv] 1 18 FB#32 TP 517. .993 794. .986 " t " only at rvr l v l 1 19 FB#33 IP 517 . 559 664 . 160 120 FB/C33 TP 516. . 122 685. .204 121 FB#34 IP 527 .318 770. .312 122 FB#34 TP 514. .623 556. .251 Rvr l v l 123 FB035 IP 514, .971 543. .797 124 FB#35 TP 513. .266 716. , 129 1.0 ro above rvr l v l 125 FB/«36 IP 512. .626 727. 253 126 FB#36 TP 511. . 108 663. ,851 Rvr l v l 127 FB#37 IP 510. .220 682. ,304 128 FB#37 TP 509. 425 480. 535 Rvr l v l 129 FB038 IP 509. . 150 338. 311 130 FB#38 TP 506. .352 672. 885 " t " only at rvr l v l 131 FB#39 IP 509. .443 582. , 198 IA ca 3 m above rvr l v l 132 FB#39 TP 505. ,532 547. 969 133 FB#40 IP 505. .261 765. 082 134 FB#40 TP 504. .563 675. 705 Large f l a t phaneritc rock on righ t bank. 135 F B # 4 ^ IP 527 . . 158 295. 328 136 FB041 . TP 529. 673 60. 799 Marker WB (White Boulder) 1 km fm Lowell G 134 APPENDIX II GRAIN SIZE DISTRIBUTION RESULTS The figures presented in this appendix represent graphs of grain size distributions of selected sediment samples from the Lake Alsek basin. The horizontal axis represents the diameter of the sediment in each sample from coarsest (left) to finest (right) in phi units at the graph bottom and in millimetres at the graph top. The left vertical axis represents the mass fraction retained on the sieves corresponding to the histogram bins. The right vertical axis represents the "fraction finer than" by weight given by the graphed curve. All samples were sieved at 1/2 phi intervals, and hydrometer results are presented at this same interval. The histogram bin at the fine end of the distribution contains all sediment finer than its size. 135 F r a c t i o n Flnor Than by Valght Fraction Flnor Than by Watght F r a c t i o n Flnor Than by Woight Froctlon Finer Than by Weight Fraction Flnor Than by Weight Fraction Finer Than by Weight Fractional Retained by Weight P P P P P p P Fractional Retained by Weight P P P P P P P P P P P P P w « A. a 9 s| Fraction Finer Than by Weight Fractional Retained by Weight P P P P P P P P P P P P P P P M tfl w a >i a Fraction Finer Than by Weight :1 | So™, 1 x. • " X u> n : S ; • / S i l t sj 1 •si . : Cloy P p p p P p M w *- u a sj Fraction Finer Than by Weight F r a c t i o n Finer Than by Weight Fraction Flnor Than by Weight by Weight Fractional Retained by Waight P P P P r p p p p p p p ' a 3 F r a c t i o n a l Retained by Waight p p p p p p p p o p r o — M U f c U a si •> •) D 1 Cronwlo* — X a j Sand S IO M >J o "1 / S U t 7 cioy P P P P P P P P P P r F r a c t i o n F l n o r Than by Waight F r a c t i o n a l Retained by Waight P P P P P P P P P P r ) Q •• M Ut a- U) B NJ a o o P P P P P P P P P P r o — N u ». u a NJ » « Q F r a c t i o n Flnor Than by Weight Frocttonal Retained by Woight P P P P P P P P P P Fraction Finer Than by Woight Fractional Retained by Weight P P P P P P P P P P r , O ju «J * ui a M • o o o — M >• >v u a •«* »j (o o -Froetlon Finer Than by Weight 140 •S.O -4.0 - i . 0 -4.0 -7.0 -B. 0 -4LO -10, O Si ova S l z a Cphi> 141 APPENDIX III A REVIEW OF MARKOV CHAIN ANALYSIS III. 1 Introduction 142 111.2 Structuring of Test Sequences 143 111.3 Initial Analytic Method 145 III. 4 Shortcomings and Improvements in Method 148 111.4.1 Inappropriate Chi-square Tests 148 111.4.2 Modelling of Expected Values 149 III. 4.3 Significance of Important Transitions 151 111.4.4 Recognition of Outliers 152 111.4.5 Homogeneity and Stationarity 153 111.4.6 Limitations of First Order Markov Chain 154 III. 5 Choice of Significance Level 157 III. 6 Chi-square Test Assumptions 157 111.7 Loglinear Models 159 111.8 Discussion and Conclusions 159 142 III. 1 Introduction Markov chain analysis is a statistical technique that attempts to determine if there is non-random ordering in a sequence and what this ordering may be. Assuming that the sequence of depositional environments and associated sedimentary processes have a definable statistical character, and that this character is recorded in a sedimentary sequence, analysis can proceed. By analyzing sedimentary records in a vertical sequence, it is possible to determine statistically what temporal relationships in the sequence might exist, and by comparing sequences, what spatial relationships might exist (Schwarzacher 1975, p.91). For a Markov process, there is a probablistic dependence on preceding states in a sequence. The fundamental principle of Markov chain analysis is that discrete parts of a sequence, (such as a lithofacies type), determine, in a probabilistic way, the overlying discrete facies type. When this principle is used to analyze sequences of sedimentary data, the technique is known as Markov chain analysis (see Harbaugh and Bonham-Carter 1970). A Markov process can be defined as one that "describes a sequence of states, or events, for which the occurrence of one state may exhibit a dependence on a previous state or states" (Powers and Easterling 1982, p.913). If a sequence is adequately described by a Markov process, the sequence can be called a Markov chain, and the chain can be said to exhibit the Markov property (Harbaugh and Bonham-Carter 1970, p.99). The use of Markov chains to analyze stratigraphic sequences was first introduced by Vistelius (1949, reported in Powers and Easterling, 1982), and has since been widely advocated as an important analytic tool in stratigraphy and sedimentology (see Walker 1984, p.3). The real value of the analysis has not been so much in modelling and prediction of sedimentary sequences, but in the subsequent evaluation and geologic interpretation of the statistical results. With this in mind, Markov chain analysis is seen as a special technique in the broader context of facies sequence analysis. The purpose of the analysis is to extract 143 information concerning the temporal evolution of past depositional environments and sedimentary processes. Markov chain analysis has undergone many changes since it was first used; in particular, the statistical foundations for analysis of embedded Markov chains has been the subject of a number of papers. Explanation and useful application of the initial analytic method is found in Krumbein and Dacey (1969), Miall (1973), and Walker (1979, 1984). Objections to the initial method have been based mainly on statistical grounds, and improvements have been suggested by Carr (1982), Powers and Easterling (1982) and Harper (1984a). All of the above papers are geologically oriented, but a substantial body of work concerned with this technique is found also in the statistical literature, where Markov chain analysis is treated as a special case of contingency table or frequency table analysis (Fienberg 1980, Brown 1983, Wrigley 1985). Details of the analytic method will be discussed in the following sections. Structuring a vertical sequence of sediments into an analyzable form is common to all methodologies, and will thus be discussed first. A summary of the initial analytic method follows this discussion, then the shortcomings of and improvements on this method will be discussed. III.2 Structuring of Test Sequences The first requirement of Markov chain analysis is that the stratigraphic column be divided into a sequence of discrete states such as biofacies or lithofacies. Using observable characteristics to define such units requires a consistent and objective typology with a minimum of three states (such as lithofacies types). The maximum number of states is limited only by the total number of facies transitions observed. Vertical, upward transitions from one facies type to another are tallied in a square matrix, in which facies types occur as both row and column labels. The matrix has rank N, where N is the 144 number of facies types. Transition from one facies type to another is coded by assigning the row label to the lower facies type, and the column label to the overlying facies type. Deciding what constitutes a transition can be accomplished in two different ways. The first method uses some regular interval (e.g.: 10 cm) and notes facies transitions at this interval. These are tallied in a square matrix where the underlying facies is the row (i subscript) and the overlying facies is the column (j subscript). The completed matrix records all transition pairs at the observational interval chosen and can be summed to check on this (Krumbein and Dacey 1969). In this method, the choice of observational interval is quite important. If the observational interval is too large, then thicker facies will be represented by reasonable values on the principal diagonal, but the thinner facies may not show up in the matrix at all. Choosing an interval small enough to include even the thinnest facies may result in overrepresentation of the thicker facies as excessively large values on the principal diagonal. Despite its apparent importance, there does not seem to be a means of determining objectively the optimal observational interval. Because of this, unacceptable bias may be introduced (Miall 1973). The second method of structuring disregards facies thicknesses entirely. Instead, upward transitions are tallied only at facies boundaries. In this method, the matrix elements record the total number of transitions from the underlying facies type (of row i) to the overlying facies type (of column j). The resulting transition count matrix cannot have any values on the principal diagonal because a transition from one facies type into itself is not observable. A transition count matrix with this property is called an embedded matrix, and the facies sequence may be called an embedded Markov chain (Krumbein and Dacey 1969). This method of structuring does not suffer from observational bias, therefore it will be used exclusively in the ensuing discussion and application of this technique. Zero entries in these matrices have a significant effect on the statistical analysis, so are classified into two types (Carr 1982). Structural (a priori) zeroes are those that 145 occur because certain transitions cannot be observed: for example, upward transitions from one facies type into itself. In most analyses, it is difficult to define structural zeroes other than this sort; therefore structural zeroes are found exclusively on the principal diagonal, which must always contain only structural zeroes. Sampling zeroes, on the other hand, occur when a particular transition is possible but is not observed; this can happen when there is a small total number of transitions. Sampling zeroes can be reduced by two means: (a) the sequence is lengthened so that the total number of transitions increases, or (b) the number of cells in the matrix is reduced by combining facies types (Fienberg 1980, p. 140). IXI.3 Initial Analytic Method The methods described in this section are compiled from those of Krumbein and Dacey (1969), Miall (1973) and Walker (1979), but the notation has been modified for consistency. The conventions used here are as follows: a matrix is referred to by an upper case letter (such as M), while the matrix elements are referred to by a subscripted lower case letter (such as m.y). Summation across subscripts is indicated by + (such as N m + j , which is equivalent to £my) . Computed test statistics are designated by the appropriate Arabic letter (such as X or G ), whereas a probability distribution is designated by the appropriate Greek letter or by name (such as chi-square distribution) Once the observed upward transition count matrix X has been compiled (an embedded transition count matrix if structural zeroes are on the principal diagonal), an upward transition probability matrix P is computed as follows, Pij = *ij / x i + where pjj is the probability of transition from the facies type of row i to the facies type of column j, X|j is the observed number of transitions from row i to column j, and X| + is the total for row i, summed over the j columns. The P matrix is a means of standardizing the 146 observed data, and as such it may yield some geologic insight into the character of the sedimentary sequence. This can be misleading however, because the total number of transitions in each row (xj + ) is possibly quite different and this difference is not taken into consideration when calculating the Pjj's. A row (i.e. facies type) with a small number of transitions has a probability sum of 1.0, as does a row with a very much larger number of transitions, so the disproportionate numbers of facies types in the sequence is not adequately represented. The P matrix is then compared statistically with an independent-trials probability matrix (or random matrix). The latter contains expected values (thus is designated the E matrix) calculated so that given row and column totals, the grand total, and the constraint that rows must sum to 1.0, the transitions occur randomly in the matrix: eij = + 1 ( x++" x +j } ey = 0 i=j where x + j is the total for column j, x + + is the grand total, and ey is the random probability of transition from facies i to facies j, (Read 1969, p.201). In this way, the modelled elements e^ are constrained so that the elements e + j will match the elements p+j. The structural zeroes on the principal diagonal (i = j) have simply been excluded in these calculations. To determine if the observed P matrix is a Markov chain or a random sequence, it is tested for goodness-of-fit against the random E matrix. The fit of the two matrices is o 9 tested using either a Pearson chi-square statistic (X ) or a likelihood ratio statistic (G ). The general forms of the X and G goodness-of-fit test statistics commonly used are: 9 N N 9 147 G 2 = 2!j where p— is the observed transition probability from i to j, and ejj is the expected transition probability as given by the random model above. These test statistics are then compared to a chi-square distribution with (N-l)^ - N degrees of freedom (Powers and Easterling 1982, p.914). The null hypothesis being tested is that the P matrix and the E matrix are statistically good fits at some chosen level of significance. If the null hypothesis is rejected then it is concluded that a random model does not fit the observed transition data and that there must be non-random ordering in the facies sequence. Furthermore, because the sedimentary sequence was initially structured into one-step transitions, it is concluded that the observed facies ordering is the result of a first order Markov process. To explore the dominant pattern of transition ordering, comparison of the two matrices is then undertaken by constructing a difference matrix D as follows, dy = Pij - % Positive entries in the D matrix should indicate which transitions have occurred with greater than random frequency, while negative or zero entries indicate transitions that occur randomly. These positive values are interpreted as the "preferred" or "dominant" transitions in the sequence. The geologic significance of this analysis is then investigated by constructing a facies relationship diagram (Walker 1979) that traces the positive values in the D matrix and sets down the Markov sequence. It is these dominant transitions that are largely responsible for the poor fit of the observed data to the random data, hence are presumed to be the most important transitions in an ideal sedimentary cycle. 148 Introduction and use of this technique has been hailed as a significant advance in stratigraphy and sedimentology (Dott 1983, p.22). The technique is nonetheless in need of improvement, for faults in the technique have been identified. III.4 Shortcomings and Improvements in Method The literature concerning critical discussion of the analytic method outlined above can be broadly grouped as follows: 1) The effect of structural zeroes in the observed transition matrix on the goodness-of-fit tests (Powers and Easterling 1982; Schwarzacher 1975). 2) Inappropriate method of modelling the expected (independent) transition frequencies (Powers and Easterling 1982, p.915; Turk 1979; Carr 1982; Harper 1984a). 3) Finding a rigorous means of determining the statistical significance of departures from the random (independent) model for the individual matrix elements, and of selecting "preferred" transitions (Powers and Easterling 1982; Carr 1982). 4) Definition and identification of data outliers (Harper 1984a). 5) Problem of non-stationarity or non-homogeneity in the sedimentary successions (Schwarzacher 1975, p.Ill; Powers and Easterling 1982, p.919). 6) Analytic limitations of an analysis that only considers a Markov chain of the first order (Powers and Easterling 1982; Schwarzacher 1975, p. 105). III.4.1 Inappropriate Goodness-of-Fit Test Statistics A goodness-of-fit test against the chi-square distribution is valid within the limitations of assumptions of a minimum sample size and of minimum expected cell values. Potential violation of these assumptions is critical to the value of the method; therefore discussion of these general assumptions is left to a separate section (III. 2 4). The presence of structural zeroes on the principal diagonal in an embedded matrix disrupts the desired property of statistical independence (randomness) in the modelled 149 expected values (E matrix). That is, transition probabilities are constrained to take a value of zero for these elements instead of participating in the equal allocation of the observed row marginal total. The result is that if the E matrix is calculated using the above method, marginal totals for the rows are maintained but column totals are distorted. There is an equal division of the column totals among the N elements in that column instead of the N - 1 elements. By not constraining the observed and expected marginal totals to match, the calculated goodness-of-fit statistics do not approximate the chi-square distribution (Powers and Easterling 1982). The solution to this problem is to modify the model of independence to account for the presence of structural zeroes by setting e^ = 0 and adjusting the rest of the E matrix so that marginal totals are maintained. This new model is called quasi-independence and is like independence as it applies to the non-empty cells of a matrix (Fienberg 1980, p. 144). Using a model of quasi-independence, appropriate values can be determined for the E matrix, and the problem of a valid chi-square test disappears. Estimation of the expected cell values must now be done iteratively to comply with this new constraint. This method is discussed in the next section. III.4.2 Modelling of Expected Values Under the model of quasi-independence, preservation of row totals while simultaneously preserving column totals requires an iterative proportional fitting procedure to estimate each e^ value. This constraint on the row and column totals of the expected transition matrix is necessary so that the presence of structural zeroes does not disrupt the desired quality of randomness. Using a probabilistic argument, Powers and Easterling, (1982, p.915), show that this constraint is necessary to both the assumption of randomness and to counteract the actual loss of information that results from structural zeroes. For embedded matrices the model of quasi-independence holds if the estimates of 150 e-j are: eij = *ihy 1 ^ J where the parameters a- and bj are estimated using an iterative proportional fitting procedure as follows; First Iteration: ajd) = x i + / (N - 1), i = 1, 2.....N bj(l) = x+j/.Said), j = 1,2,....N The initial value for aj given here is simply the N-l average of the i ^ row entries, but elsewhere other choices for initial values (such as unity) have been used (Fienberg 1980, p. 144). Ith Iteration: a:(I) = x i + / 2b:(I-l), i = 1,2 N 1 I T (#>) J b:(I) = x + -/-2ai(I), j = 1,2 N J + J (tfj) 1 A FORTRAN subroutine was written to do these iterative calculations until a convergence criterion of 0.5% was satisfied. This criterion was chosen so that the ajbj products would have a maximum error of 1%. This means that the row and column totals in the quasi-independence matrix should have the same whole-number values as those found in the observed transition matrix. Since the random transition matrix is thus constrained, the test for goodness-of-fit under a model of independence becomes a test under a model of quasi-independence. 151 Provided the other assumptions are met, using the e^ 's computed from the above iterative procedure in the general expression for X or for G provides a statistic that is 9 approximately chi-squared distributed with (N - 1) - N degrees of freedom, (Powers and Easterling 1982, p.916). III.4.3 Significance of Important Transitions When the D matrix is calculated, large positive differences are considered to be "preferred" transitions, but no attempt is made to determine statistically if, and by how much, these transitions are significant. This is an important stage in the analysis because it bridges the results of statistical analysis with geologic interpretation. This bridge is precarious, though, in that definition of a preferred or significant transition is not solidly based on either statistical or geologic grounds. Instead, the transitions with the largest d-j's were taken ad hoc from the D matrix as needed and used to construct "facies relationship diagrams" and "facies models" (Walker 1979). Regardless of the value of conclusions based on this method, a better definition of preferred transitions can be made on statistical grounds. Three ways of doing this are mentioned here; the last of which is most compatible with the rest of the analysis, and widely used due to its origins in conventional contingency table analysis (Brown 1980). 1) Harper (1984b) proposed a procedure that enables a significance test against the binomial distribution to be undertaken. This method uses the non-standardized difference matrix of early transition analyses and tests each of these elements for significance. This method is unnecessary in light of the following stepwise selection procedure. 2) Powers and Easterling (1982) and Turk (1979), suggest that the differences can be normalized and compared as a matrix, looking for large values as the preferred transitions. A normalized difference matrix can be used to compare transitions, but a statistically sound comparision is still required. 152 3) Carr, (1982) draws on conventional contingency table analysis and suggests that significant transitions can be chosen one by one in a stepwise procedure until only random transitions are left. Using the normalized difference matrix, the matrix element containing the largest value is identified as the first preferred transition. Then this cell is treated as a structural zero and the quasi-independent model is fit to the remaining data, with the loss of one degree of freedom. If this fit is again significantly different, then the stepwise procedure will select the cell that will reduce the matrix X value the most when it is deleted by actually evaluating the X statistic for each possibility. At each step a cell is selected and eliminated and the remaining cells refit under the model of quasi-9 independence until the probability of the X statistic exceeds some threshold level of significance. The stepwise selection of cells is also stopped if the degrees of freedom is reduced to 1. Carr (1982) as well as the BMDP program P4F (Brown 1983) both advocate a short-cut method to this procedure apparently in an attempt to reduce computing time. However this short-cut method may adversely affect the analysis, especially when the observed data sets are small (as they often are in facies sequence analysis) and Harper (1984a, p.207) recommends that the short-cut method not be used. III.4.4 Recognition of Outliers Wrigley (1985, p.293) defines "outliers" or "rogue cells" as those isolated large anomalies which can be identified on the basis of large residual values. In this sense, an outlier and a preferred transition are one and the same. However, Harper (1984a) claims that the procedure used to identify preferred transitions is susceptible to two types of errors. First, "masking" occurs if some preferred transitions are not recognized due to other highly significant transitions. Secondly, "swamping" occurs if more transitions are identified than should be. To identify outliers, Harper (1984a) proposes the use of median tetrads and half-normal plots. This procedure is unnecessary however, if the stepwise 153 selection procedure explained in the previous section is used to identify preferred transitions (outliers). Masking or swamping cannot happen if the most significant transitions are identified and their influence removed from the remaining data. For this reason, no separate technique is required to identify outliers. III.4.5 Homogeneity and Stationarity If more than one facies sequence is investigated, then these can be tested for similarity. Similarity between different sections of the same sequence implies stationarity (temporal similarity), while similarity between different sequences implies homogeneity (spatial similarity). Stationarity applies if sedimentary processes have not changed significantly through the length of time represented by the sedimentary record. This means that the temporal variability of sedimentation processes is similar throughout the sequence and over the interval of time that this represents. Interestingly, Schwarzacher (1975, p.223) points out that true stationarity is hardly ever seen in the stratigraphic record because of long-term fluctuations in the processes of deposition at any one locale. Homogeneity means that transition probabilities are constant laterally, and that spatial variability of the sequence probabilities is not significant. If homogeneity is proven then there is great advantage to merging these data into a composite matrix in order to increase the total number of observations and reinforce strong tendencies. To test two or more sedimentary sequences for homogeneity or a single sequence for stationarity, another goodness-of-fit test is undertaken (Powers and Easterling 1982, p.920). Let k be the number of embedded matrices to be compared, x^ be the number of transitions from state i to state j in the k matrix and p ^ be the corresponding transition probability. If all k matrices are homogeneous then the probabilities p ^ and py should be equivalent and the expected frequencies defined as: eijk = ^ij + ^ + k) / x i + + 154 To test this hypothesis, a Pearson X statistic and a likelihood ratio statistic G are computed as follows, x 2 = H f ( X i ik- e i ik) 2 / e i jk G 2 = 2 22Zx i j k ln (x i j k / e i j k ) These statistics are approximately chi-square distributed with (k - 1)(N - 2)N degrees of freedom (Powers and Easterling 1982, p.919). These test statistics are subject to the same chi-square test assumptions discussed in section III.6. If the sequences are found to be dissimilar, then the same techniques used to identify preferred transitions can be used here to identify the most strongly dissimilar transitions. III.4.6 Limitations of First Order Markov Chain By definition, an embedded Markov chain has a single-step first-order dependence, meaning that the state at time t depends only on the state at time t - 1. However, if a first order Markov dependence is detected then it is quite possible, even probable, that a second order or even higher order Markov dependence exists (Schwarzacher 1975, p. 105). Once a sequence has been found to behave as a first order Markov process, another chi-square test for second or higher order dependence should logically follow. In a similar way, a facies state dependence on a complex combination of preceding facies states at any number of preceding intervals may be possible. Since the complexity of possible combinations of dependence, order and step length quickly grows, a classification of some hypothetical combinations is given in Fig. III. 1 (Harbaugh and Bonham-Carter 1970, p. 128). Testing for a single dependence of any order is easily done by counting the upward transitions using the desired step size; that is, for a second order single dependence case 155 Figure ELI. 1 Classification of Markov chains according to various combinations of dependence, order, and step length. Upper row chains are single dependence, chain (a) is first order because state j at time t is dependent upon immediately preceding state i at time t-1. Chains (b) and (c) are higher order because they are dependent on earlier states, even though only a single dependence is involved; u denotes step length. Middle row shows examples of double dependence (d, e, f) where t denotes step length for earlier of the two steps involved in the dependence relationship. Triple dependence is illustrated in the last row, where p denotes length of initial step in dependency relationship. Distance between each rung in ladders is of unit length. (After Harbaugh and Bonham-Carter, 1970) a C ot T3 G 01 a. oi a 60 a t - 1 (b) t - 2 (c) F i r s t Order Second Order t - 5 F i f t h Order oi 01 •a a 01 a. oi a 01 2 (d) 3 o a t - 1 t - 3 t - 1 t - 6 Second Order Third Order Sixth Order oi o a 01 T3 a oi a oi a t - u t - 3V (b) t - 3 t - 5 0) x - 2 -1 - 4 -' , t - 5 Third Order F i f t h Order F i f t h Order 156 the transitions from state t - 2 to state t are tallied. The computational procedures and chi-square test assumptions are identical to those of a first order single dependence case. For double (or greater) dependence relationships, data sets must be very large, otherwise many zeroes appear in the expected transition matrices. Each additional preceding state used in the analysis means another N matrices; for example, in the double dependence case there are N matrices, each N x N. While the number of matrix elements goes up, the total number of transitions does not, so many more sampling zeroes must result. A corresponding increase in very low values in the expected values matrix would now severely limit the worth of conclusions based on the goodness-of-fit statistics (Fienberg 1980, p.172). Observed two-step transition frequencies are counted and tabulated in N matrices that are each N x N, with the Xy k 's representing the transition count from state i to j to k. Then using marginal totals, the expected values ejjk are computed by equal allocation of these totals under the assumption of a first-order chain: eijk = (Xij + ) ( x - r j k ) / x + j + Structural zeroes are not a problem in computing expected values since independence is not a desired property in this test, but they are subtracted from the total degrees of freedom. A goodness-of-fit test, using either a Pearson chi-square statistic (X2) or a likelihood ratio statistic (G ) is undertaken, using the following forms: x 2 = 2 2 :S(x i j k -e i j k )2 /e i j k G 2 = 2 3 ? 2 x i j k ln(xijk / eijk) which are approximately chi-square distributed with N(N - 2) degrees of freedom (Powers and Easterling 1982, p.921). If these statistics indicate that the double-dependence model 157 is significantly different from the single-dependence model, then further testing is desirable but only possible if data sets are very large. III.5 Choice of Significance Level In Markov chain analysis a level of significance (Q!) must be chosen when testing the fit of a transition matrix against the quasi-independence model, and when identifying the preferred transitions using the stepwise procedure. In the first case, the significance levels may need to be set quite high to minimize the probability of a Type II error, where a false null hypothesis is incorrectly accepted. This kind of error is undesirable because analysis would be stopped at that point and further insight into the facies sequence would unnecessarily be lost. For the purposes of geologic insight, the danger of rejecting the null hypothesis when it is actually true (Type I error) is not as hazardous as the danger of making a Type II error. Conventional significance levels (Oi= 0.005, 0.01, 0.05) are too low to avoid possible type II errors, so significance levels of (X= 0.10, 0.25, or even 0.40 can be used. In the test for homogeneity among sequences, the probability of a Type I error, where the true null hypothesis of similarity is incorrectly rejected, must be minimized. A significance level of 0.10 is used for this test, but this could be set even lower. III. 6 Chi-square Test Assumptions That the chi-square distribution is an approximation of the distribution of the X or G statistics is derived under an assumption that the expected values are not "too small" (Everitt 1977, p.40). How small is too small seems to be a matter of some debate. For example, Fienberg (1980, p. 172) states that there is a widely accepted rule of thumb that "a chi-square test is valid only if none of the expected cell frequencies are less than five". Cochran (1954, given in Fienberg 1980) states that if one or more of the cell expectations is less than one, or twenty percent or more are less than five, then the chi-square testmay 158 be in error. Fienberg (1980, p. 172) says that several expected cell frequencies may actually be zero without endangering the validity of the test. Lewontin and Felsenstein (1965, quoted in Everitt 1977, p.40) assert that minimum expected values of 0.5 are adequate and that even this rule is conservative. Facies sequence analysis is often restricted to a small total number of transitions relative to the number of possible transitions in the transition count matrix. For example, most Lake Alsek sedimentary sequences have less than 55 facies transitions when subdivided into 5 facies types. The transition count matrices have 20 cells (52 - 5) at this level of subdivision, which results in more than a few sampling zeroes and in many low expected cell values. Depending on whose rule for minimum expected cell frequencies is used, this may endanger the value of the goodness-of-fit tests. Collapsing two or more facies types into a single type in order to reduce the matrix size and increase the expected frequencies is one way to overcome the problem, but is objectionable because of the associated loss of information (Fienberg 1980, p. 172). Alternatively, several transition matrices may be combined and analyzed as a single matrix if it can be shown that the sequences are statistically homogeneous by using a test for similarity. Combining several sequences into a larger single sequence on this basis is frequently done (Powers and Easterling 1982). Elimination of sampling zeroes by addition of a small constant, generally 0.5, to each cell in the matrix is a procedure advocated in the BMDP program P4F (Brown 1983). Wrigley (1985, p.280) notes however, that the justification for the procedure and the specific choice of 0.5 as the constant has little basis beyond convenience. An alternative goodness-of-fit test known as the "Exact Chi-Square" or ECT test has less restrictive assumptions regarding small sample sizes, but is not widely used because of computational difficulties (Romesburg et al. 1981). 1 5 9 III. 7 Log-linear Models The analysis of contingency tables is improved by using log-linear instead of linear models (as used in Markov chain analysis) to estimate expected values. The basic advantage of log-linear models is that "they provide a systematic approach to the analysis of complex multidimensional tables," and that they provide a hierarchical means of assessing the relationships among variables (Everitt 1977, p.81). The general log-linear model is easily derived from the general linear model of expected frequencies, making it additive instead of multiplicative, with two benefits; 1) easier computation, and 2) a form analogous to the analysis of variance model (Fienberg 1980; Wrigley 1985). In Markov chain analysis, log-linear models do not need to be invoked because only two-dimensional tables are analyzed, and only the hypothesis of quasi-independence is of interest. If a log-linear approach were used, no parameters could be eliminated and estimates of expected values would be the same as those estimated by the linear model. III.8 Discussion and Conclusions In order to test the validity of the improved procedures outlined above, a FORTRAN program, MARKOV.S, was written to perform the necessary computations (Appendix VI). In addition, the BMDP program P4F (Frequency Tables) was used as a check on the basic algorithms used in the program MARKOV.S. The program was used to model expected frequencies from the observed transition frequencies, to test the model for goodness-of-fit, to identify preferred transitions, and to test sequences for similarity. MARKOV.S was run using data sets analyzed by Powers and Easterling (1982) and an artificial data set in order to test the validity of the methods. First, the data sets reported in Powers and Easterling (1982) were run and MARKOV.S succeeded in correctly determining expected frequencies, the goodness-of-fit statistics, and the set of preferred transitions. Secondly, a random data set was generated by the procedure used to compute the E matrix (quasi-independence matrix) and used in 1 6 0 turn as observational data for input into MARKOV.S. The hypothesis of no significant difference between this artificial data set and the expected matrix was clearly accepted. These two procedural tests of the methods involved in Markov chain analysis indicate that the method performs as expected, even with an artificial extreme data set. Limitations imposed by the small sample sizes affects interpretation of goodness-of-fit statistics and must be accounted for. In this context, combining data that has been shown to be statistically similar is a useful undertaking. Excessively large sample sizes may also pose problems. Schwarzacher (1975, p.121) claims that if sample sizes are large enough to represent time scales that are long enough, most geologic sequences would exhibit the Markov property. Using Markov chain analysis to test for a Markov dependency in a sedimentary sequence that has been truncated by erosion at some point may introduce additional difficulties. Erosional unconformities may correspond to a considerable loss of the deposited sedimentary record, therefore a sequence analysis may be inappropriate, unless the eroded surface itself is considered as part of the record. By including a "scoured surface" (SS) in the transition count matrix, Walker (1984) was able to generate a complete sedimentary cycle. Nonetheless, the loss of information due to erosion of the sediment column poses a serious problem since it is apparent that most sedimentary records are inherently discontinuous (Ager 1973, Dott 1983, Hilton-Johnson 1982). In the study of Lake Alsek sedimentary records, facies sequence analysis is an important means of deciphering a complex and apparently repetitive record of environmental and sedimentary conditions. That there may be sound evidence of sequential ordering and possibly a cyclic signal in the sediment column greatly assists in the interpretation of past sedimentary conditions in the Lake Alsek basin. 1 6 1 APPENDIX IV MARKOV.S: A FACIES S E Q U E N C E ANALYSIS P R O G R A M This appendix contains the computer code (in FORTRAN IV) for analysis of sedimentary sequences by Markov chain analysis. Explanation of the program algorithms are given as comment statements throughout the program. This code is available from the author for unlimited use. 162 1 C * * • • « • « * « » * * « * « * PROGRAM: MARKOV.S * » • » • « • • « . * * * » • • * « * * * * « • « * • « » » 2 C P I L E NAME: MARKOV.S 3 C LANGUAGE: FORTRAN IV 4 C OATA F I L E : MKV.01, MKV.02. ETC. 5 C PURPOSE: To ana 1yso s t r a t i g r a p h i c s u c c e s s ! o n s on t h e b a s i s of 6 C a f a c i a s t r a n s i t i o n c o u n t m a t r i x . To t o s t t he hypot n a s t s of 7 C r a n d o m n e s s i n t h e s t r a t i g r a p h i c s e q u e n c e s - To a i d i n t h e 8 C c o n s t r u c t i o n of f a c i a s m o d e l s . 9 C AUTHOR: J e f f r e y Schmok ( J a n . 19BS, Mar. 19861, b a s e d on p a p e r s lO C by Powers a n d E a s t e r l i n g ( 1 9 8 2 ) , and H a r p e r ( 1 9 8 4 1 . t i C « * • * * « * « « * « « • * • • « • • • • « « « • « • « * * « • • • * * • • * • * • * * « • « • « « * * « * « * * * • • * * * * • • 12 C DATA F I L E SETUP: 13 C CAR0 1. go c h i r i e t t r a l p h a m e r t c t i t l e 14 C CARO 2. N * r a n k of s q u a r e m a t r i x , (GS.O) 15 C FMT1• f o r m a t of m a t r i x rows. (4A4) 16 C CARO 3,4... X( I ,J) t r a n s i t i o n c o u n t mat r i x 17 C . . . -N w i t h N c a r d s ( r o w s ) 18 C CAROS 1 • N i n r « p « i t « d u n t i l a l l mat r i c e s e r e e n t e r e d . 19 C * » • » * • • • « * • « « » • * • • « « « . « « . . . « . . , > . . . « - « « • « • - « « - • • - « - « « • • • - . - . - - . « . . 20 C PREAMBLE: A Markov p r o c e s s d e s c r i b e s a s e q u e n c e of s t a t e s , or 21 C e v e n t s , f o r w h i c h t h e o c c u r r e n c e of one s t a t e e x h i b i t s a 22 C p r o b a b l i s t i c d e p e n d e n c e on a p r e v i o u s s t a t e or s t a t e s . 23 C A s e q u e n c e of f a c i e s , or l t t n o l o g i e s . c a n be t e s t e d 24 C f o r t h i s p r o p e r t y a g a i n s t a n h y p o t h e s i s of p u r e l y 25 C i n d e p e n d e n t o r d e r i n g . T h i s p r o g r a m c o n d u c t s t h i s t e s t w i t h a 2G C c n i - s q u a r e s t a t t s t i c , and out put s t h e mat r t x of s t a n d a r d 1 x e d 27 C d i f f e r e n c e s . To a i d t n t h e i n t e r p r e t a t i o n of a s e d i m e n t a r y 28 C s e q u e n c e , t h e c e l l s w h i c h c o n t r i b u t e most t o t h e Markov 29 C p r o p e r t y a r e i d e n t i f i e d w i t h a s t e p w i s e p r o c e d u r e . F i n a l l y , 30 C t h e s e t o f mat r t o e s i n t h e i nput f i l e a r e t e s t e d f o r 31 C homogene i t y w i t h a c h i * s q u a r e s t a t i s t i c . For more 32 C b a c k g r o u n d s e e Wa1ker( 1978) , H a r p e r ( 1 984 ) , C a r r ( l 9 6 2 ) , 3 3 C and e s p e c t a l l y P o wers and E a s t e r 1 i n g l l 9 8 2 ) , on w h i c h some of 34 C t h e b a s i c a l g o r i t h m s u s e d i n t h i s p r o g r a m were b a s e d . 35 £ . « . . . . . . * . . « « . . * . « • . . . « « * . . . . « . . « . • « . « . . . • . . • . . . » • - < - * . . * . » • > . . - . • > -36 C 37 c**«* S t a r t ma i n r o u t i n e . . . . 38 C 39 REAL*8 X ( 2 0 , 2 0 ) , P I 2 0 . 2 0 ) 40 REAL*8 0 ( 2 0 , 2 0 ) , 20 I F F ( 2 0 , 2 0 ) . SI 20) 4 1 01 MENS 1 ON HT0T(SO) , VT0TI5O] . LET{20 ) 42 DIMENSION T I T L ( 2 0 ) . FMT1(7) 43 C 44 c.**** You must put t h e f a c i e s names h e r e In t h e same f o r m a t as t h e 45 C f o l l o w i n g names, f r o m row 1 down 46 C 47 DATA LET /'D', ' S ' , ' F T , 'Fm', '0', 'SS'/ 48 C 49 C NT c o u n t s t h e number o f m a t r i c e s i n t h e i n p u t f i l e . . . 50 NT s O 5 1 C 52 C **** t h i s i s s t a r t of pr i m a r y l o o p , one c y c l e p e r i n p u t mat r i x 53 DO 40 K • 1, 20 54 C 55 C * * * * * * * * Read a n d w r i t e t h e t i t l e and c o n t r o l c a r d s . . . . 56 C 57 READ (S. 60,END-BO ) (T I TL( I ) , I * 1 . 20) 5 8 READ ( S , 7 0 , e N 0 » S O ) N, (FMT1 ( I ) , I = 1 . 4 ) 5 9 c •* * * i n c r e m e n t t h e m a t r 1 x c o u n t e r . . . . GO c G1 NT • NT • 1 62 C 63 WRITE ( 6 . l O O ) 64 WRITE (6. 6 0 ) ( T I T L ( I ) , I • I , 2 0 ) 65 WRITE ( 6,90) N, (FMT t ( I ) , 1«1 ,4 ) 86 C 87 c Read t r a n s i t i o n c o u n t m a t r i x ( t h e X m a t r i x ) 88 C 69 DO lO I • 1 , N 70 READ (5,FMT1,END-20) ( X ( I , J ) , J " I . N ) 71 to CONTINUE 72 20 CONTINUE 73 C 74 c **** T h i s i s a n I/O t o g g l e f o r w r i t i n g o u t p u t ; 75 c 1 • y e s a n d O • no ( a l w a y s 1 e x c e p t f o r xtrerne and compar) 76 IWRIT e i 77 C 78 C » - • » • » • » « • • • « - . • « « - - - - • - • « - . « . - - - - « . . - - - . . - . - - - . « » « . - - - « • < - . - « . . • 79 C «»** T h i s b l o c k of s u b r out i n e s i s c a l l e d f o r e a c h mat r i x t o c *••* 81 C **•* The XMAT r o u t i n e t a l l i e s row, c o l u m n and g r a n d t o t a l s . . . 82 C 83 CAL L XMAT(X, TOT. HTOT, VTOT, N, LET, I WRIT) 64 C 65 C •««« ThQ COMPAR r o u t i n e g e n e r a t e s a 3*0 m a t r i x a t t h i s p o i n t f o r 86 C l a t e r u s e i n t h e compar i s i o n f o r s i m i l a r i t y bewt een m a t r i c e s . 87 C 88 CALL C0MPAR(X, N. NT, IWR IT) 69 C 9 0 c • * • • PMAT comput es probatai l i t y o f t r a n s i t i o n s FROM any row t i t l e . . . 91 C (N.B. t h i s Is not p a r t of t h e raarkov c h a i n a n a l y s i s ) 92 C 93 CALL PMAT(P, X, S, HTOT. VTOT. TOT, N. LET) 94 C 95 C • • • • XPECTD c o m p u t e s t h e Q u a s i - I n d e p e n d e n c e m a t r i x u s i n g an 96 C i t e r a t i v e pr opor t i ona1 f i t t i n g pr o c e d u r e w h i c h w i l l p r e s e r v e 9 7 C b o t h t h e row and c o l u m n t o t a l s . The e s t i m a t i o n s f o r a i and 98 C b j a r e out put f o r t h e f i r s t and l a s t i t e r a t i o n , a l o n g w i t h 99 C t h e f i n a l m a t r i x . A c o n v e r g e n c e c r i t e r i o n of 1 * i s u s e d 100 C but t h i s c a n be m o d i f i e d i f d e s i r e d 10 1 c 102 CALL XPECTD(0. X, HTOT. VTOT. TOT, N, LET. IWRIT) 103 C 104 c * * « * NDIFF comput e s t h e s t a n d a r d i z e d d i f f e r e n c e s between t h e 105 C o b s e r v e d t r a n s i t i o n c o u n t a n d t h e e x p e c t e d t r a n s i t i o n 10S c c o u n t u n d e r t h e model of q u a s i I n d e p e n d e n c e . 107 C The f o r m u l a u s e d i s Z l j «* ( O i j • E l j ) / S Q R T ( E i j ) 108 C 109 CALL NDIFPIZOIPP, X. Q, N. LET, IWRIT) 1 to C 111 C *•*«» NSTRUC i s no. of s t r u c t u r a l z e r o e s i n t r a n s i t i o n m a t r i x . . . 112 NSTRUC • N 113 C 114 C •* * * CHISOR comput es t he c h i s q u a r e s t a t i s t i c , OOF, a n d p r o b e b f l i t y MS C o f e x c e e d i n g t h e c o m p u t e d c h i s q u a r e . The DOF a r e c a l c u l a t e d 116 C a s f o l l o w s , DOF c (N - 1) ** 2 - NSTRUC. where t h e number 1 6 3 117 C of s t r u c t u r a l l e r o o s Is MSTRUC 118 C 119 CALL C H I S O R ( Z D I F F , M. CHI SOU, PROB, NSTRUC, NOOF) 120 C 121 C •>*« x TREME US«S a s t e p w i s e p r o c e d u r e t o s e l e c t t h e c e l l s w h i c h 122 C c o n t r i b u t e most s t r o n g l y t o t h e Markov p r o p e r t y . The 123 C r o u t i n e w i l l s e l e c t c e l l s u n t i l an a l p h a l e v e l i s p a s s e d 124 c or u n t i l s o many c e l l s a r e d e l e t e d t h a t t h e OOF becomes 125 C l e s s t h a n one 128 C 127 ALPHA = 0.30 128 IF (PROB . GE . ALPHA) WRITE f 6,30) PROB 129 I F (PROB GE ALPHA) CO TO 40 130 30 FORMAT (//4X, ' « " » « N 0 CELL SELECTION DONE: PROB » * , FS.4) 13 1 C 132 CALL XT REME(X, 0. ZDIFF, N) 1 33 C 134 C «•»» T h i s i s t h e end of p r i m a r y l o o p 135 40 CONTINUE 136 C 137 C « • • • • c o n t r o l t r a n s f e r r e d t o h e r e a f t e r l a s t X m a t r i x f s r e a d . . . . 138 SO CONTINUE 139 C 140 c • •«« I WP. I T now t e l l s COMPAR t o comput e and w r i t e r e s u l t s . . . . 141 [WRIT c 2 1 42 C 143 I F ( N T .LE. 1) GO TO 61 144 c •»»« COMPAR t e s t s f o r nomogene11 y b e t w e e n t h e i n p u t mat r i c e s w i t h a 145 C c h i s q u a r e s t a t i s t i c . T h i s r o u t i n e u s e s 3 d t mans f ona1 146 C m a t r i c e s a n d f o l l o w s t h e a l g o r i t h m s u g g e s t e d In A p p e n d i x 11, 147 CALL C0MPAR(X, N, NT, IWRIT) 148 C 149 61 CONTINUE 1 50 1 5 1 152 60 FORMAT I20A4) 1 S3 70 FORMAT (C 3 . 0 . IX. 4A4 1 1 54 SO FORMAT (I X . ' D a t a T i t l e : ' , 20A4) 155 IO FORMAT (' Rank of I n p u t M a t r i M B ' , 13. /' I n p u t F o r m a t * 4A4 } 1S6 100 FORMAT t ' 1 ' ' « • • • • P RO CRAM: MARKOV.S VERSION: 11 1 57 c 15« WRITE I 6. 1 IO] 159 1 10 FORMAT Ml. • ENO OF SUCCESSFUL RUN OF MARKOV.S PROGRAM ' ' ) 1 60 STOP 1 S 1 END 1 B2 C 163 c « » * « « « « » « « * « « * * « « « « " « « " " * « « « « » « « « « " « e e « « « " » « * « » « " e « e « * « e » « * e « « « » * » » « 164 c***« T h i s r o u t i n e t a l l i e s row M T O T ( X i - ) a n d c o l u m n V T O T ( X . j ) t o t a l s 165 C and t h e g r a n d t o t a l T O T ( X . . ) and w r i t e s t h e X m a t r i x . Note 166 C t h a t H a r p e r ( 1 9 0 4 ] c i t e s M M l « r ( 1 9 8 3 ) a n d F e i n b a r g l 19 80 J 167 C as s t a t i n g t h a t t r a n s i t i o n c o u n t s may be l e s s t h a n one i n 16 8 C some of t he mat r i x e l anient s 1 69 C 170 SUBROUTINE XMATfX, T O T . HTOT, VTOT. N, LET, IWRIT) 17 1 C 172 REAL * 8 X(20, 20] 173 0 IMENS I ON HTOT(SO\ . VTOT(SO) . L E T \ 2 0 ) 174 DATA HSUM / ' X i . ' / , VSUM / ' X . j ' / n s c 176 C * * * * Do s uraraations a c r o s s r ows a n d c o l u m n s 177 TOT « O . O 176 DO IO I » 1, N 179 V T O T ( I J « 1.OE-1O ISO HTOT ( I 1 B 1 .OE- IO 181 I O CONTINUE 182 00 30 I = 1 . N 1 S3 00 20 J = 1 , N 184 H T O T ( I ) = HTOTI I ) • X ( I , J | 185 V T O T U ) a V T O T ( J ) • X ( | , J | 186 20 CONTINUE 1*7 T O T « TOT • H T O T ( I ) 188 30 CONTINUE 189 C ISO C • • • « • • • • • • • * • « « « « PRINT X MATRIX AND SUMS 19 1 C 192 I F ( I W R I T . E Q . O ) RETURN 193 W R IT E ( 6 , S O ) 194 WRITE ( 6. SO 1 | L E T ( J ] , J » 1 ,N| , HSUM 195 DO 40 I a 1, N 196 WRITE ( 6 . 7 0 ) L E T ( I ) , ( X ( I , J ) , J « 1 , H ) , H T O T ( I ) 197 40 C O H T l H U e 198 WRITE ( 6,80 ) VSUM, ( V T O T ( J ) ,J>1,N) , TO T 199 C 200 SO F O R M A T (/, 6X. 2 0 ( 3 X , A 3 ) ) 201 60 FORMAT (/5X , 'OBSERVEO TRANSITIONS', /. , 4X, 2 0 ( ' *)1 202 TO FORMAT ( I X , A3, I X , 2 0 ( 1 X , F 5 . 0 ) ) ~ 203 S O FORMAT (/, I X , A3. I X , 20(1X,FS.O1) 204 C 205 RETURN 206 END 207 CMll>lt*>aMll«lftMIMMM*ll>*t..*MMMtMtMI«iM«llt*.fMII 206 C * * * • • « » * « * * • * • COMPUTE PROBABILITY MATRIX 209 c * * « * comput e s p r o b a b i 1 i t y of t r e n s i t i o n s FROM any r ow t i t l e . . . . 210 C ( t h i s i s not p a r t of t h e mar Kov c h a i n a na l y s i s ) 2 11 C 212 SUBROUTINE PMAT[P. X, S. HTOT, VTOT, TOT, N, LET, IWRIT) 2 13 C 2 14 REAL'S X ( 2 0 , 2 0 ) , P ( 2 0 , 2 0 ) , S ( 2 0 ) 215 DIMENSION HTOT(SO), V T O T ( S O ) , L E T ( 2 0 ) 216 OATA AD /'P( I ) ' / 2 17 C 2 18 10 DO 20 I * 1 , N 2 19 D O 20 J s 1 , N 220 S | J ) « V T O T U I / TOT 221 IF (1 E O . J I P ( I , J I B O.O 222 IF (I .EQ. J ) C O T O 20 223 I F (HTOT( I ) E O . O . ) C O TO 20 224 P ( l . J ) B X I I . J ) / HTOT(I) 225 20 CONTINUE 226 C 227 WRITE 16,60) 228 WRITE (6, 5 0 ) ( L E T ( J ) , J " = 1 , N ) , AO 229 DO 30 I e 1, N 230 WRITE ( 6 , 4 0 ) L E T ( I ) , I P ( I , J ) , J » 1 , N ) , S ( I ) 231 30 CONTINUE 232 C 164 233 40 FORMAT ( I X , A3, IX, 2 0 I 2 X . P 4 . 2 I ) 234 50 FORMAT (/. 5X, 2 0 I 2 X . A 4 ) ) 235 60 FORMAT (/5X, 'TRANSITION PROBABILITY •, /. '•', 4X. 2 2 ( ' _ ' ) > 236 C 237 RETURN 236 BHD 239 2*0 C * B • • t h i s c o m p u t e s t h e O u e s i - I n d e p e n d e n c e m e t r l x u s i n g en 24 1 c i t e r a t i v e p r o p o r t i o n c l f i t t i n g p r o c e d u r e w h i c h w i l l p r e s e r v e 242 c b o t h t h e row e n d c o l u m n t o t a l s - The e s t i m a t i o n s f o r e l end 243 c b j a r e o u t p u t f o r t h e f i r s t e n d l a s t i t e r a t i o n , a l o n g w i t h 244 c t h e f i n a l m a t r i x . A c o n v e r g e n c e c r i t e r i o n o f 1% i s u s e d 245 c but t h i s c a n be m o d i f i e d i f d e s i r e d 246 c 247 SUBROUTINE XPECTOtO, X, HTOT. VTOT. TOT, N. LET, IWRIT) 248 c 249 REAL'S X I 2 0 . 2 0 ) , 0 ( 2 0 . 2 0 ) 250 REAL * 6 A ( 2 0 ) . B ( 2 0 l , SUMAI20), SUMBI20), SUMAAI20I 25 1 REAL'S A A I 2 0 ) , A A P R E V I 2 0 ) . BPREVI20) 2S2 0 1 M e N S I ON N T 0 T I 5 O I , VTOT(SO), L E T I 2 0 ) 253 OATA HSUM .•/, VSUM / • X . j • / 254 c 255 c • B * » i n i t i a l i z e v a r i a b l e s a n d comput e f i r s t e s t i mat es of A t . . . . 256 DO 10 1 a 1, N 257 A l l ) a HTOTI I) / (N - 1 I 256 A A ( I ) a o . O 259 B 1 I ) a O . O 2BO 5UMA( I ) a O . O 261 i o CONTINUE 262 c 263 c B B • « comput e SUMA. t h e sum of t h e fit's.... 264 00 30 I a 1, N 26S DO 20 J a 1. N 266 IF (I . EO . J ) GO TO 20 267 SUMA(I) s SUMA( I I + A ( J ) 266 20 CONT1NUE 269 30 CONTINUE 270 c 27 1 c > * * * c ompute t h e f i r s t e s t i m a t e s f o r B j 272 00 40 J a 1 , N 273 IF (SUMA(J) .EO. O.O) GO TO 40 274 6 ( J ) s V T O T I J I / SUMA(J) 275 40 CONTINUE 276 c 277 c F i r s t I t e r a t i o n i s c o m p l e t e , t h e r e f o r e HIT • 1 276 c 279 N I T s 1 280 c 26 1 c • « » • W r i t e out r e s u l t s o f f i r s t e s t i m a t e s o f A i and B j 262 IF (IWRIT .EO. O) GO TO SO 263 WRITE 16.2601 NIT 264 00 50 I a 1. N 265 WRITE (6.2SO) A l l ) . 6 ( 1 ) 286 SO CONTINUE 287 c 286 c« a * * T h i s i s s t a r t of t h e main i t e r e t i v e l o o p 289 60 CONTINUE 290 c 29 1 c * * * * i n i t i a l i z e v a r i a b l e s 292 00 70 I « 1. N 293 SUMS( I 1 n O.O 294 SUMAAlI) - O.O 295 70 CONTINUE 296 c 297 c • * * * s a v e r e s u l t s o f t h e p r e v i o u s i t e r a t i o n f o r c o n v e r g e n c e t e s t . . 298 DO 60 I > 1 . N 299 AAPREV ( I ) s AA ( I ) 300 BPREV( 1 ) s B 1 I ) 30 1 SO CONTINUE 302 c 3 0 3 c • as* compute sum of t h e B j ' s 304 00 l o o I s 1 , N 305 00 BO J a 1, N 306 IF (I . EO . J ) GO TO 90 307 SUMB(I) B SUMB(I) • B I J ) 308 90 CONTINUE 309 lOO CONTINUE 3 1 0 c 3 1 1 c • * « • c o m p ute A i ' s . c a l l e d A A i ' s . s i n c e A i ' s i n f i r s t i t e r a t i o n . . . 3 1 2 DO 1 1 0 I • 1 . N 3 1 3 IF ( S U M B ( l ) .EO. O.O) CO TO IIO 3 1 4 A A ( I ) B H T O T I I I / SUMB(I) 315 1 t o CONTINUE 3 1 6 c 3 1 7 compute sums o f A A i ' s 318 DO 130 I B 1 . N 3 1 9 DO 120 J B 1 , N 320 IF (I .EO. J l GO TO 120 32 1 SUMAA(I) B SUMAAlI) * A A I J I 322 1 20 CONTINUE 323 1 30 CONTINUE 324 c 325 c * * * a 326 00 140 I B 1 , N 327 IF (SUMAAII) .EO• O.O) GO TO 140 328 8 ( 1 ) B V T O T I I ) / SUMAAII) 329 1 40 CONTINUE 330 c 33 1 332 NIT B NIT • 1 333 c 334 c« • « « i f y o u want t o g e t v a l u e s o f A i and B j f o r o t h e r i t e r a t i o n s 335 IF (IWRIT .EO. 01 GO TO 160 336 IF (NIT .CT. 1) GO TO 160 337 WRITE (6.2BOI NIT 338 DO 150 I B 1, N 339 WRITE [ 6,2501 AAI I ) , B ( I ) 340 1 50 CONTINUE 34 1 1 60 CONTINUE 342 c 343 c * s s a '« T e s t f o r c o n v e r g e n c e t o l e s s t h a n 1 p e r c e n t c h e n g e 344 c 345 00 1 TO I B 1 , N 346 ABSA B D A B S ( A A ( I I - A A P R E V I I I I 347 ABS8 • 0 A B S I 8 I I I - B P R E V ( I I ) 348 ACRIT B AA( I I " O.01 1 6 5 349 BCP.IT a B i l l * O.OI 350 IF IABSA .GE. ACRIT) CO TO 60 351 IF (ABS B .GE. BCRIT) GO TO SO 352 1TO CONTINUE 353 C 354 c**«* Write, out f i n a l I t e r a t i o n c o u n t a n d e s t i met e s . . . . 355 IF (IWRIT .EO. O) CO TO 1 BO 35S WRITE (6.2SO) NIT 357 00 ISO I s 1, N 35S WRITE ( 6 , 2 5 0 ) AAI I ) , B I I ) 359 ISO CONTINUE 3BO C 361 c **•• C a l c u l a ' t e v a l u e s i n q u a s i - i n d e p e n d e n c e m a t r i x 362 DO 200 I = 1, N 363 DO 190 J a 1 , N 364 IF (I .EO. J l 0 ( 1 , J ) s O.O 365 IF (I EO. J ) GO TO 1BO 366 Ol I .J) = AA( I I • B I J ) 367 190 CONTINUE 36S 2O0 CONTINUE 3B9 C 370 C *•*• Do s u m m a t i o n s a c r o s s rows a n d c o l u m n s 371 TOT s O.O 372 DO 210 I a 1, N 373 VTOTI .OE- 10 374 H T O T ( I | « t.O E - 1 0 375 210 CONTINUE 376 DO 230 I e 1. N 377 00 220 J = 1 . N 376 H T O T I I ) • HTOTII) * 0 ( 1 , J ) 379 V T O T I J ) « V T O T I J ) • 0 ( 1 . J l 350 220 CONTINUE 351 TOT a TOT • HTOTII) 352 230 CONTINUE 353 C 354 c * * s s * w r i t e o ut v a l u e s i n q u a s i - i n d e p e n d e n c e m a t r i x 385 IF (IWRIT EO. Ol RETURN 386 WRITE 16,2901 387 WRITE ( 6 . 2 8 0 ) ( L E T ( J ) , J s 1,N), HSUM 358 DO 240 I = 1. N 389 WRITE 16,270) L E T ( I ) , ( 0 ( I . J ) . J • 1 . N ) . H TOTII) 390 240 CONTINUE 391 WRITE 1 6 , 3 0 0 ) VSUM, ( V T O T I J ) . J s I , N I . TOT 392 C 3B3 250 FORMAT I' E s t . f o r A i « *, F6.2, ' E s t . f o r Bj s F6.2) 394 260 FORMAT (/' Number of I t e r e t i o n s a 14, /] 396 270 FORMAT ( I X , A3, IX, 2 0 ( 1 X , F 7 . 2 ) ) 396 280 FORMAT (/, 6X. 2 0 I 4 X . A 4 ) ] 397 290 FORMAT I/5X. 'OUASI - INDEPENDENCE MATRIX ( O i j » a i • B j ) ' . /. 398 1 '•' . 4X, 43( '_' I I 399 300 FORMAT (/. IX, A3, IX, 2 0 ( 1 X , F 7 . 2 I I 400 C 401 RETURN 402 END 403 c * » * * * « * « * * * s « . * . * * s * s * « * « * * * * * * . . « * « . * * s * « * . • • • • • » » • • • • « • « • « « « • • * * 404 c » « « « « « * " « * « » * « Compute s t a n d a r d i z e d d i f f e r e n c e s . . . . 405 C • *•« t h e s t a n d a r d i z e d d i f f e r e n c e s b e t w e e n t h e o b s e r v e d t r a n s i t i o n 406 C c o u n t a n d t h e e x p e c t e d t r a n s i t i o n c o u n t under t h e model of 407 C q u a s i - i n d e p e n d e n c e i s co m p u t e d h e r e . The f o r m u l a u s e d 406 C i s Z l j a ( O i j - E i j l / S O R T I E i j ) 40B C 410 SUBROUTINE NO 1FF(ZD I F F , X. 9, N. LET. IWRIT) 4 11 C 412 REAL'S 0 ( 2 0 , 2 0 ) , X ( 2 0 , 2 0 ) , ZD I F F I 20,20 I 413 DIMENSION L E T I 2 0 ) 4 14 C 415 C ***• compute 2D I F F , t h e s t a n d a r d i z e d d i f f e r e n c e s 4 16 00 lO I a 1, N 417 DO 10 J a 1 . N 418 IF [I EO. J ) I D I F F I I . J l » O.O 419 IF (I .EO. J l GO TO lO 420 IF ( 0 ( 1 . J ) .EO. O.O) Z O I F F I I . J l a O.O 421 IF ( 0 ( 1 . J l .EO. 0.01 GO TO 10 422 Z O I F F I I . J ) a ( X I I , J ) - 0 ( 1 . J l ) / OSORT(0( I . J I I 423 10 CONTINUE 424 C 425 C a s * * W r i t e gut t n e r e s u l t s 42S IF (IWRIT .EO. Ol RETURN 427 WRITE (6,SO) 428 WRITE 16.401 I L E T ( J ) , J a I . N ] 429 C 430 00 20 I s t, N 431 WRITE ( 6 . 3 0 I L E T I I I . (20 IF F I I . J | .Ja1 ,NI 432 20 CONTINUE 433 C 434 30 FORMAT ( I X , A3, IX. 2 0 ( 2 X , F 6 . 2 ) | 435 40 FORMAT (/. 4X. 20 I 4 X . A 3 I ) 436 SO FORMAT I/5X. 'STANDAROIZEO DIFFERENCES •. /. 4X. 2 4 ( ' _ ' l l 437 RETURN 438 END 439 c « * * a a * « * * * * * * * * * a s a a a * a a a * * * « * * a s * * a a * * * * a a » * s s a a a « * . « • • « • • • • « • • 440 c **** Compute C h t - s q u a r e s t a t i s t i c t o t e s t e a c h m a t r i x 441 c f o r q u a s i I n d e p e n d e n c e . . . . 442 C «'** t h i s c o m p u t e s t h e c h i s q u e r e s t e t t s t t c , DOF, end p r o b e b i l l t y 443 C of e x c e e d i n g t h e computed c h i s q u a r e . The DOF e r e c a l c u l a t e d 444 C es f o l l o w s , DOF a (N - 1) *« 2 - NSTRUC. where t h e number 445 c of s t r u c t u r a l z e r o e s i s NSTRUC 446 C 447 SUBROUTINE CHISORIZOIFF, N, CHISOU, PROS, NSTRUC, N00FI 448 C 449 REAL * 6 « A L I 2 0 , 2 0 ) , Z D I F F ( 2 0 , 2 0 ) 450 C ***** Compute CHISOU, t h e s t a t i s t i c 45 1 C 452 CHISOU a O.O 453 DO 20 1 a 1 , N 454 DO lO J a t. N 455 IF II .EO. J ) GO TO lO 456 V A L 1 I . J I a Z D I F F ( I . J ) 2 457 CHISOU a CHISOU * V A L ( I , J I 466 10 CONTINUE 459 20 CONTINUE 460 C 461 NDOF s (H - 1) •* 2 - NSTRUC 462 C 463 C E v a l u a t e p r o b a b i l i t y f o r e x c e e d i n g c h i s q u e r e 464 C N o t e : t h e f u n c t i o n PCH1 I s s u p p o r t e d by t h e UBC comp. c e n t e r o n l y . . 1 6 6 465 C 486 PROB « PCH I (CHI SOU,HOOF) 467 C 468 IP (N .NE. NSTRUC) CO TO 70 469 C 470 WRITE (6, 3 0 ) 47 1 WRITE (6,40 ) CH I SOU 472 WRITE (6.SO) NOOF 473 WRITE (8, 6 0 ) PROB 474 C 4 75 30 FORMAT (//' GOODNESS OF F I T OF 0-1 MODEL '/'*', IX, 28 ( ' ) ) 476 40 FORMAT (/7X. * CHI'SQUAREc ', F9.3) ~ 477 50 FORMAT (7X, ' WITH O.F.= *. IS) 478 SO FORMAT (7X, ' P r o b . of e x c e e d i n g C h i s q u a r e = F 8 . 6 ) 479 70 CONTINUE 480 RETURN 481 ENO 482 ^ • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • « • * * • * « • • • • • • * • • • • « « • « • • • « • • • 483 : « • « » • • * • « « FUNCTION PCHISO ( a d a p t e d f r o m B e v l n g t o n , P.R. ( 1 9 6 9 ) ) 484 C T h i s f u n c t i o n e v a l u a t e s p r o b a b i l i t y f o r e x c e e d i n g c h i s q u a r e 4*5 C Usage: 486 C r e s u l t « PCHISO(CHISQR. NOOF) 467 C N . B . : 486 C c a l c u l a t i o n i s a p p r o x i m a t e f o r NOOF odd a n d c h i s q u a r e 489 C g r e a t e r t h a n SO. 490 C 491 FUNCTION PCHI SO FREE / 2.0 500 NEVEN s 2 * (NDOF/2) 50 1 IF (NOQF - HEVENt 40, « 0 , 1OO 502 C 503 C ««»•« NOOF IS EVEN 504 40 IMAX = NOOF / 2 505 TERM » 1 . O 506 SUM ti O.O 507 SO DO BO I s 1. IMAX SOS FI a I SOS SUM « SUM • TERM SIO TERM a TERM * Z / FI 5 11 60 CONTINUE S12 70 PCHISO " SUM * OEXP(-Z) 5 13 C 514 W R I T E ( 6 , 9 0 ) 515 WRITE (6.SO) CHI SOU, F I . NDOF, TERM, PCHISO 5 15 SO FORMAT (/, 4X, ' CHISOU"'. F8.4, /* F I * ' , F8.4, * N O O F a ' , 16, /. 5 17 1 ' TERMa', 08.4, /' PCHISO*'. F8.4) 516 SO FORMAT (/5X, ' D e b u g g i n g PCHISO S t a r t /, *•', 4X, 2 2 ( ' — ' ) ) 51 9 C 520 GO TO 190 52 1 522 C*« » « • NOOF IS 000 523 lOO IF (2 - 26) 120, 120, 1IO 524 1 IO Z a CHI SOU * 1 FREE - l.O) / Z.O 525 CO TO 40 526 120 PWR * FREE / 2.0 527 TERM - l . O 528 SUM • TERM / PWR 529 130 00 ISO I a l , 1OOO 530 FI • I 531 TERM a •TERM * Z / F I 532 SUM - SUM + TERM / (PWR + F I ] 533 140 IF (DABS(TERM/SUM) - .OOOO1) ISO, ISO, ISO 534 ISO CONTINUE 535 ISO PCHISO • 1.0 - ( Z * « P W R ) * SUM / SAMMA(PWR) 536 C 537 WRITE ( 6 , I S O ) 536 WRITE 16, 170) CHISOU. PWR, SUM. PCH ISO 539 170 FORMAT (/, 4X. * CHISOU"*, F8.4, /' PWR-', F8.4, /* SUM**, 08.4. / 540 1 , ' PCHISO"'. F8.4) 541 1 SO FORMAT (/EX. ' D e b u g g i n g PCHISO e n d . . '. /. 4X , 22( ' ' ) ) 542 C 543 190 CONTINUE 544 RETURN 545 END 546 C«MIIMMI«tl.«IMMIMIU«t.«l»l >I*M tUltlttlMIIUII 547 C *aaa T h i s r o u t i n e w i l l s e l e c t c a l l s u n t i l an a l p h a l e v e l I s p a s s e d 548 C or u n t 1 1 so many c e l l s a r e d e l e t e d t h a t t h e OOF becomes 549 C l e s s t h a n one 550 C 551 SUBROUTINE X TREME(X, 0, ZOIPF, N) SS 2 C 55 3 REAL"8 0 ( 2 0 , 2 0 ) , ZD I P F ( 2 0 , 2 0 ) , X ( 2 0 , 2 0 ) 554 DIMENSION HTOT(SO), VT0T15O), L E T ( 2 0 ) 555 DIMENSION IROW(SO), IC 0 L ( 5 O ) 556 ALPHA • O.30 557 KO UN T e O 558 IWRIT '<*< O 559 C 560 C •**» s t a r t o f t h e main l o o p . . . . 56 1 C 562 IO CONTINUE 563 (COUNT * KOUNT * 1 564 C • « « • d e t e r m i n e no. of s t r u c t u r a l z e r o e s now i n m a t r i x 566 C 566 NSTRUC a KOUNT • H 567 NDOF a (N - 1) *• 2 * NSTRUC 568 C 569 C «a«a C a l l t h e s t e p w i s e p r o c e d u r e t o s e l e c t t h e c e l l w i t h 570 C t h e maximum s t a n d a r d i s e d d e v i a t e i n t h e c u r r e n t m a t r i x . . . . 57 1 C 572 CALL S E L E C T ! ZDIFF, N, PROBV. I ROW, ICOL. KOUNT, NSTRUC) 573 C 574 C •»*» i f t h e l a s t c e l l d e l e t e d i n c r e a s e s t h e PROB t o g r e a t e r 575 C t h a n t h e s p e c i f i e d ALPHA l e v e l t h e n r e t u r n t o main p r o g r a m . . . 57B C 577 IF (PROBV .GT. ALPHA) RETURN 578 C S78 C ••»« . . . . O t h e r w i s e , s e t p r e v i o u s l y s e l e c t e d c e l l s t o x e r o and 5SO C c o n t I n u e t h e c e l l s e l e c t i o n p r o c e d u r e by comp1et e1y 167 581 C 682 C 6 83 c 564 585 515 c 5«T 568 c ssa 580 c 59 1 592 c S93 S94 595 595 597 599 c 599 c 600 601 c 602 603 604 c 60S 60S 607 c 60S c 609 c 610 c 6 1 1 c 612 c 613 c 6 1 4 615 c 6 1 6 11 7 6 1 8 6 T 9 c 620 62 1 622 c 623 624 625 626 627 c 628 c 629 630 631 632 633 634 c 635 c 636 c 637 638 c 639 c 640 64 1 642 c 643 644 c 645 646 647 648 c 649 c 6SO 96 1 852 6S3 6S4 65S c 666 c 8S7 c 656 669 c 660 c 66 1 662 c 663 c 664 c 666 c 666 c 667 c 668 c 669 c 670 c 67 1 c 672 c 673 c 674 c 676 676 677 678 676 660 68 1 c 662 683 664 c 665 c 686 c 687 686 669 690 69 1 692 693 694 c 695 c 666 c r « c o a p u t I n g a l l o b s e r v e d t o t a l s , i l l e x p e c t e d v a l u e s , and t h e m a t r i x o f s t a n d a r d i s e d d i f f e r e n c e s « I 1 R 0 W C K O U N T ) . 1 C D L | K O U N T ) ) • O.O 0(IROW(KOUNT).ICOL(KOUNT)) B O.O CALL XMAT(X, TOT. HTOT. VTOT, N, LET, IWRIT) CALL XPECTOlO, X. HTOT, VTOT, TOT. N. LET. IWRIT) CALL ND I F F ( ZOI FF . X, 0, N, LET, IWRIT) WRITE [ 6 . 2 0 I KOUNT WRITE (6, 3 0 ) TOT. NSTRUC 20 FORMAT (/' A f t e r s e l e c t i o n o f '. 12, * c e l l s : ') 3 0 FORMAT (' T o t a l t r a n s i t i o n c o u n t * ' , F 6 . 1, /' S t r u c t u r a l z e r o e s 1 121 • « • ' I f t h e OOF I s l e s s t h a n 1, end l o o p h e r e and w r i t e m essage.. IF (NOOF .GT. 1) GO TO IO WRITE (6. 4 0 ) 40 FORMAT (/, ' C e l l s e l e c t i o n p r o c e d u r e e n d e d ; OOF < 1 '] RETURN END • aaa T h i s u s e s a s t e p w i s e p r o c e d u r e t o s e l e c t t h e c e l l s w h i c h c o n t r i b u t e mast s t r o n g l y t o t h e Markov p r o p e r t y . The r o u t i n e w i l l s e l e c t t h e c e l l w h i c h w i l l r e d u c e t h e m a t r i x c h i s q u a r e v a l u e t h e most when It i s d e l e t e d by a c t u a l l y e v a l u a t i n g t h e c h i s q u a r e f o r e a c h p o s s i b i l i t y SUBROUTINE S E L E C T ( Z O I F F , N, PROBV. IROW. ICOL. KOUNT, NSTRUC) REAL * 6 2 0 1 F F ( 2 0 , 2 0 ) , Z 0 I F ( 2 O , 2 O I , Z0UM(2O,2OI OIMENSION IR0WI5O), ICOL(SO) CHI LOW B 10000.0 a a • « IROW(KOUNT) • 1 ICOL(KOUNT) = 1 IF (N .NE. NSTRUC - 1) GO TO 20 WRITE ( 6 . I O ) IO FORMAT I / ' S t e p w i s e S e l e c t i o n o f C e l l s ' . /. '•', IX, 2 7 ( ' _ ' J ) 20 CONTINUE •a*a s e t t h e m a t r i x t o be t e s t e d ( t h i s i s c l u m s y . . ) DO 40 K s I, N DO 30 L s I . N ZD I F 1 K . L ) » Z O I F F ( K . l ) 30 CONTINUE 40 CONTINUE aaaa t h i s i s main l o o p t o f i n d c e l l w h i c h w i l l r e d u c e t h e m a t r i x c h i s q u a r e v a l u e t h e most when i t i s d e l e t e d SO CONTINUE aaaa s t a r t l o o p t o compute t h e l o w e s t c h i s q u a r e v a l u e 00 IOO I • 1. N DO 90 J • 1. N IF II .EO. J ) GO TO SO aaaa s k i p o v e r n e w l y e m p t i e d c e l l s 00 60 K B 1 , KOUNT IF IIROW(K) .EO. I .AND. ICOL(K) .EO. J ) CO TO 90 60 CONTINUE aaaa now. r e s e t t h e m a t r i x ( t i l l s i s c l u m s y . . ) DO SO K • 1. N DO 70 L • 1 , N 2 DUM(K.L) • Z O I P t K . L I 70 CONTINUE » 0 CONTINUE aaaa s e t c e l l l . j t o x e r o . . . ZDUMII.J) » O.O aaaa e v a l u e t a new c h i s q u a r e s t a t i s t i c . . . . CALL CHISORIZDUM, N, CHISOU. PROB, NSTRUC. NDOF) WRITE ( 8 . 1 SO) WRITE ( 6 , 1 SO) DO 140 II B 1, N WRITE ( 6 . 1 7 0 ) (ZOUMI I I , J J ) ,JJ»1 ,N) 140 CONTINUE ISO FORMATI/SX.' S t e p w i s e d e l e t i o n o f c e l l s ' , / ' • * . 4X, 2 2 ( ' ') 160 FORMAT (/I 170 FORMAT I IX, 2 0 I 2 X . F 5 . 2 ) ) •aaa s e l e c t t h i s c e l l If f t has t h e l o w e s t c h i s q u a r e r e s u l t . . . IF (CHI SOU GE. CHILOW) GO TO 90 CH1L0W o CHISOU PROBV B PROB NOEGF B NDOF IROWIKOUNT) • 1 ICOL(KOUNT) • J a s a a 90 CONTINUE IOO CONTINUE •aaa w r i t e o ut t h e r e s u l t s . WRITE ( 6 , 1 1 0 ) KOUNT, IROW(KOUNT), ICOL(KOUNT) WRITE ( 6 , 1 2 0 ) WRITE ( 6 . 1 3 0 ) CHILOW, PROBV, NDECF 110 FORMAT I/2X. • S t e p N O . B ' , 12, ' C e l l No.a'. 12. 121 120 FORMAT (2X. ' A f t e r d e l e t i o n o f c e l l , new model f i t i s : ' ) ISO FORMAT I4X, ' C h t S q u a r e ' ' . F8.2, / < « , ' P r o b a ' , F8.I 1 /4X, ' OOF B - . 121 aaaa now, c h e c k t o s e e i f t h e l a t e s t r e d u c t i o n i n c h i s q u a r e pL t h e p r o b a b i l i t y a t more t h a n a l p h a a o . 3 0 1 6 8 697 C 696 C IP (PROBY . L T . 0.30 .AND. NDEGF .GT. 1) GO TO SO 699 C WRITE (6.240) CHI LOW. NDEGF, PROBY TOO C 701 140 FORMAT [/4X, * F i n a l C h i - S q . B * , F8.4, /. 4X, ' N D D F B ' , 13. /4X, 702 I * P P O d a b n i t y - * , PB.4) 703 RETURN 704 E NO 705 c*•••••**•••*••*••••••••**••*•••*•••••**•••***••••**•*••• 706 C •«->* t h i s t e s t s f o r h o m o g e n e i t y oetwoon t h e i n p u t m a t r i c e s w i t h a 707 C c h i s q u a r e s t a t i s t i c . 70S c 709 SUBROUTINE COMPAR(X, N, N T , IWR IT) 710 C 711 REAL * ft X(20,20) , [KPL 712 REAL * ft C(2 0 . 2 0 , 2 0 ) , XX(20,20.20) 7 13 C 7 14 C 7 15 IF (IWRIT .GE. 2) GO TO 30 718 C •»•« G e n e r a t e 3-D m a t r i x 717 K - NT 718 DO 20 I a t , H 7 19 DO lO J » 1, N 720 X X ( I . J . K ) o X ( I , J ) 721 lO CONTINUE 722 20 CONTINUE 723 C 724 C »* * * Oon't do a camper i s1 on u n l e s s a l l mat r i c e s n a v e b e e n r e a d . . . 725 C 726 IF (IWRIT .NE. 2) RETURN 727 IF ( N T . LE . 1) RETURN 728 C 729 30 CONTINUE 730 C G e n e r a t e e x p e c t e d v a l u e s m a t r i x 73 1 DO SO K = I, NT 732 DO 70 I B t, N 733 DO SO J a 1. N 734 P L I J « O 735 IKPL = O 736 P L I P L o O 737 IF [I . E O . J ) GO TO SO 738 DO 40 MM « 1 , NT 739 P L I J * P L I J * > X X ( I . J . MM) 740 40 CONTINUE 74 1 DO SO NN » 1 . N 742 IKPL « IKPL • XX(I,NN.K) 743 50 CONTINUE 744 P L I P L • P L I J • IKPL 745 IF ( P L I P L .EO. O ) GO TO 60 746 C ( I , J , K ) = ( (PL I J ) * ( IKPL) ) / ( P L I P L I 747 C 748 60 CONTINUE 749 70 CONTINUE 750 SO CONTINUE 75 1 C 752 CHISOU = O.O 753 VAL » O.O 754 C • ••*•* Sum f o r s t a t i s t i c . . . 755 DO 1 10 K «* 1 , NT 756 00 lOO I B T, N 757 DO 90 J • 1, N 7S6 IF (I .EO. J ) GO TO 90 7S9 IF ( C ( I . J . K ) .EO. O.O) GO TO 90 7BO VAL • ( X X l I . J . K ) • C l l . J . K I ) 2 / C ( 1 , J , K ) 761 CHISOU • CHISOU * VAL 762 90 CONTINUE 763 lOO CONTINUE 764 1lO CONTINUE 765 C 786 C * * * * D e t e r m i n e t h e d e g r e e s o f f r e e d o m now 767 C 788 HOOF • |HT - I) « (N • 2) • N 769 C « » • * E v a l u a t e p r o b a b i l i t y f o r e x c e e d i n g c h i s q u a r e 770 C 77 1 PROS B PCHI(CHISOU.NOOF) 772 C 773 C *»*» w r i t e out r e s u l t s 774 WRITE (6,160) 775 WRITE (6.120) NT 776 WRITE (6,130) CHISOU 777 WRITE ( 6 . 1 4 0 ) NOOF 778 WRITE ( 8 , I S O ) PROB • 779 C 780 120 FORMAT (/7X. ' T o t a l number o f m a t r i c e s * * ' . 12) 78 1 130 FORMAT (7X, ' T o t a l c h i - s q u a r e * ' , P6.3) 782 140 FORMAT (7X, ' w i t h DOF-'. 13) 783 ISO FORMAT (7X. ' P r o b . o f e x c e e d i n g c h i s q u a r e " ' , P8.61 784 160 FORMAT I//2X, ' H o m o g e n e i t y of Input M a t r i c e s ' , /, *•*. 2K. 785 1 29( ' ' ) ) 786 C ~ 787 RETURN 788 END 169 APPENDIX V STRATIGRAPHY OF SEDIMENTS IN SOIL PITS APPROXIMATELY 8 KM WEST OF HAINES JUNCTION Soil pit profiles of Lake Alsek sediments have been published by various authors. This appendix contains reproductions of profiles from approximately the same location in Dezadeash Valley. The most recent work on the soil stratigraphy is by C. A. S. Smith of the Yukon Territorial Government. However, this remains unpublished and is reproduced here in preliminary form. Discussion of the profiles is in sections 1.1.5, 1.4, and 3.1.2. The profiles are reproduced here to illustrate both the evolving understanding of these sediments, and the variability of stratigraphy between sites in relatively close proximity in the Lake Alsek basin. The first profile is from Johnson and Raup (1964), the second from Rampton (1981), the third from Clague and Rampton (1982), and the fourth from Smith (1985, personal communication). Note that the last of these includes a detailed soils description on a separate page. 170 0 T Lake I S 40 Lake 14 E o ^ ^ t _ ^ Dark brown silt filled with humas and roots of present prairie plants. Cray silt, partially leached of carbonates beneath the existing turf. ^Wi^JJ; ^ Organic layer (ca 1/2"). Gray silt unleached. 80- Lake13 4-1 a. 2-3 cm then the corer tube edge has to either push the clast out of the way or else it is stopped. In stiff sediments, clasts as small as 1-2 cm have stopped coring. 3. Very deep water. Lake water depth is probably limited by the number of rod extensions you have. Lake muds have been successfully sampled in water depths up to about 20 m in the Lake Alsek study. 4. Very stiff sediment. In most applications, the force applied to the coring tube is limited to the brute force of two people pushing on the rods. In stiff sediments, this factor is what will determine the depth of sediment sampled. 5. A mud-water interface greater than about 0.5 m. This means that overlap between the square-rod samples and the plexiglass tube or crust-freezer samples will not be sufficient. Comox Lake, for example, is reported to have a 'soupy' mud-water interface of >2 m thickness. III. MODIFIED DESIGN (UBC 1985) SQUARE-ROD CORER This design is basically a modified Wright (1967) square-rod piston sampler for lake sediments. The most important changes are in the materials used to build the corer, making it a much stronger and more reliable tool while still retaining the light weight and simplicity of the original design. Mr. D. Schreiber, technician in the Dept. of Geophysics and Astronomy, built this particular corer. 177 The optimal configuration for a corer is one with maximum strength but with a minimum wall thickness of core tube. The limit on corer penetration into stiffer sediments is the cross-sectional area of the tube wall because this determines how much sediment has to be pushed out of the way for the core tube to penetrate. This area Atube is, Atube =*"(rout2 - r m 2 ) A rule of thumb recommended by Hvorslev (1949, reported in Hakanson and Jansson 1983, p.32) is that Atube should be less than 10% of the core cross-sectional area. In addition, the core tube should be smooth on the inside and have a sharpened cutting edge. This corer uses two 2.375 inch OD, (2.25 inch ID, 1/16 inch wall), stainless steel tubes, 1.20 m long, one with a tempered steel serated cutting edge, and one with a tempered steel beveled cutting edge. With these coring tubes, Atube is 0.454 inch2, and the sample area is 3.98 inch2, therefore the Hvorslev requirement is very nearly met. The square rod is a 0.750 inch2 piece of stock stainless steel, and all its fittings match it in strength and material. The head pieces are machined stainless steel and interchangeable. The pistons are brass with two rubber seals which can be adjusted by a washer and nut assembly. Another similar corer tube made of clear plexiglass, 2.25 inch ID (2.50 OD, 1/8 inch wall) and 1.20 m long, can be used in place of the stainless tubes has similar fittings, and are completely interchangeable. To protect and transport the corer, ABS pipes were fitted with custom inserts to hold the corers firmly inside. Drive rods are five foot lengths of 1-5/16 inch diameter zirconium-magnesium joined with coarsely threaded stainless couplings. These are standard rotary drill rods (XRT zirc-mag) chosen for rigidity and light weight. A T-extension was made to fit these. IV. DRILLING PLATFORM A stable drilling platform is essential for this corer to work properly. The best platform is one designed exclusively for use as a drilling platform, but a multi-use platform, such as a 12 ft Zodiac, can be used quite successfully. An ice-covered surface is probably the optimal drilling platform if the associated logistical difficulties can be overcome. Also, a mechanical drill driver can be mounted on the ice to increase the drilling force. One example of a well designed and proven drilling platform is the Cwynar design (L. Cwynar, personal communication). This is relatively inexpensive, is helicopter transportable, and stable enough for 3 people to work on. Basically, one 4x8 sheet of plywood separable into 3 pieces, reinforced by two 2x4:s running lengthwise along each edge, is lashed to'! two 6-8 ft rubber rafts. A drilling hole is cut in the center and cleats, blocks, sills, etc., are attached to the upper surface. Spare equipment, such as core boxes and rods, are kept in the rafts themselves, while the sampler and people stay on the platform. A small tender with oars is neccessary with this setup to place the anchors. A very similar design uses large styrofoam floats instead of rubber rafts (P. Jones, personal communication). This has been proven to work well and is somewhat less expensive to put together than a rubber raft design. A 12 ft Zodiac can be used unmodified by drilling off of the transom. The boat is stable enough in practise so that two people can push or pull on the rods at the same time. Anchoring of the drilling platform is extremely important. For very small lakes and ponds, two long ropes strung perpendicularly from shore to shore so that they cross above the drill site is best. Cleats and/or prussiks are very 178 useful for attaching the platform securely to these ropes. In larger lakes anchors (such as bags of rocks) have to be used. A minimum of three anchors must be used, but four anchors, one at each corner, usually works best. These are rowed out from the platform and dropped so that when the anchor lines are pulled taut, the anchors are equispaced. Inadequate anchoring will have two possible consequences: 1. the platform moves sideways and the drilling force is wasted and core quality is reduced, or 2. the platform moves sideways and the rods break, or even worse, the square rod becomes disastrously bent. Casing the corer helps to reduce these problems, but careful anchoring is important and may even require more time than the coring itself. In many applications, a drill hole casing is needed. The two main reasons for a casing are, 1. to relocate the drill hole so that successively deeper cores can be obtained in 1 m increments from the same borehole, and 2. to give the drill rod system more vertical rigidity so that bending of the rods is reduced. Casing made from 4 inch ABS tubing cut into 2.5, 5, and 10 foot lengths and fitted with standard threaded ABS couplings is satisfactory. V. CORING PROCEDURES Before coring begins, the stable drilling platform should be established and all equipment, cleaned and operative, placed on the platform. The water depth beneath the platform is then determined by sounding to within a few cm of a stable datum on the platform. A. Water-sediment interface samples To sample the water-mud interface, use the clear plexiglass tube. Determine the water depth from some known datum on the platform to within 10 cm. The clear plexiglass tube is taped securely to a rod section and the spare piston is positioned at the lower end with the clothesline wire attached. Tape the top of the tube to prevent pulling the piston out. Lower the tube to 20-30 cm above the bottom and secure the piston wire. The drive should be slow and careful and should go for the entire length of the plexiglass tube. The unit is brought to the surface where it is kept vertical while a second piston is plugged into the lower end of the tube while still underwater. The top piston has to be moved upwards so that the sediment-water column is not compressed. The tube can now be removed from the water but must be kept vertical. B. Continuous samples with square-rod corer Samples are removed in up to one meter (where possible) segments from a continuous borehole. This can be accomplished using the following step by step guide: 1. Attach the clothesline wire to the piston wire loop, using plier tightened knots or crimps. Tape the free end to the wire and keep the wire free of kinks. 2. Wet the barrel and piston. Adjust the middle nuts on the piston in 1/8 turn increments until the two piston rubbers fit snugly but can still run smoothly. If the piston is too loose then vacuum will be lost and the sample will be contaminated or lost. Too tight and the coring drive becomes too difficult and the whole system is overly strained. This is an important step. 3. With the piston at the bottom of the tube and the square rod on the 179 piston, lower the sampler to the coring depth with the second person adding rods as you go. Note that this must be done by keeping the tension on the wire, not the rods, so that the square rod does not slip out. If it is necessary to go through sediment already cored, the rods will have to be pushed while some tension is maintained on the wire. 4. Once the sampler is at the desired coring depth, the piston wire is securely fixed on the platform. Vice-grips or a cleat are useful, and the wire cannot move. A reel with a ratchet, fixed to the platform, is a good solution to tangled and bothersome wires. At this point, the rod itself should be marked at some fixed datum with a piece of tape, so that the actual length of the drive can be measured. Suitable baseline datums are waterline, or the top of the platform. 5. Now the square rod is cocked by lifting up on the rods while jiggling them slightly until the bottom nut hits the headpiece. Then turn the rods clockwise 1/4 turn so that the metal flanges are abutting the stops on the headpiece. The rods are now resting on the top of the headpiece and the sampler is ready to drive. Note that the taped mark on the rods is now exactly 1 m above baseline datum so that a full 1 m drive will push the mark back down to datum. 6. The drive is smooth if possible, and any hard pushing must be done with the sampler kept vertical. The t-extension allows two people to exert a more even force on the rods. The drive ends either when the mark is reached or when the sampler cannot be pushed any further. If the mark is not reached then the difference should be measured and the actual drive distance noted. 7. To remove the sample, release the piston wire, and while firmly holding it against the rods, lift the rods and wire out together. The initial pull can be difficult but the wire and rods must be synchronized at all times. Bring the sampler to the surface by having the second person remove rods as they surface. Turn the sampler horizontal quickly as it comes out of the water so that sediment cannot fall out. 8. To extrude the core make sure that there is plenty of room and a tray is clean. First uncock the square rod and slip it back into the headpiece until it hits the top of the piston. Then the square rod is held steady while the tube is pulled back and the core is deposited in a tray held by an assistant. If the sediment is somewhat stiff, the square rod can be held against an immoveable block while both people pull carefully on the tube. If the sediment is quite stiff, or if there is coarse sand, (which can bind the piston), the core may not be extrudable by hand and a mechanical advantage using a pulley or something must be obtained. Sediment which is too liquid to hold its shape is extruded by holding the corer upside down, stretching a rubber skirt around the barrel, and pushing down on the barrel in 5 or 10 cm increments while funneling the sediment into a ziploc bag. 9. Once the core has been extruded into a tray or onto the platform, it is carefully wrapped in plastic wrap, then in aluminum foil. Mark the inside and outside of the foil wrap with the 'up' direction, the measured length of the core (ie: 220-310 cm), and the core location. Store the core in a core box or encased in two core trays. 10. Before the next drive, clean the sampler in lake water so that the piston is not bunged up with sediment and runs smoothly. If you replace the rods in the same order as the previous drive, the tape mark is moved up the rod the same distance as the previous drive, and this new mark will indicate the start of this drive. 180 11. The casing is usually not put into place for the first drive. Before the second drive, enough casing is lowered and assembled to be pushed firmly into the bottom sediment to within 20-30 cm of the first drive. The casing top is secured to the raft and steps 2) through 10) are repeated making sure to overlap the first cased drive with the uncased drive. Differences between drive length and core recovered are often encountered and are usually due to; 1. compression of sediment during the coring procedure, 2. sediment lost from the tube bottom, 3. bore hole infilling between drives. Driving lengths are independent of core recovered for drilling purposes and should be recorded as the actual length of sediment sampled. VI. STORAGE AND TRANSPORTATION OF EXTRUDED CORES The drilling logbook should record the following basic information: 1) Lake name or number 2) Location of drill site on lake 3) Water depth at drill site Columns in the drilling logbook should have the following headings: 1) Core number 2) Drive number 3) Drive distance 4) Recovery length 5) A comments field Cores which can hold their shape are wrapped carefully in plastic wrap, then aluminum wrap, as described in section 3. Core boxes should not be placed on end and should be marked thus if they are to be transported commercially otherwise the cores might compress. Styrofoam blocks placed in gaps between cores are useful as a check on accidental compression. If the material. is too unconsolidated to hold its shape, then extrusion in 5-10 cm sections into whirlpak bags is the usual alternative. VII. SPLITTING AND PHOTOGRAPHY OF CORES For core splitting it can usually be expected that if the mud was possible to sample with a piston corer in the first place, it can be split with the piano-wire technique. This is uncomplicated. The whole core is placed in a split tube tray and a piece of taught piano wire .is pulled along the top of the tray to slice the core in half. Tilting the tray separates the top portion from the bottom. How easy this actually is depends on the nature of the specific core. Homogeneous muds may require splitting with an electro-osmotic knife which is based on the electro-osmotic behaviour of clays, i.e., if an electric field is applied to a clay, the interstitial water will flow to the negative electrode (Hakanson and Jansson 1983, p.48). The exposed surface has to be carefully cleaned with a knife or glass slide edge used to scrape the surface laterally. Laminations are scraped along their axis. Sedimentary structures are easier to see if the core surface is allowed to air-dry for several days or even weeks. Thin laminations are most visible if allowed to dry, then lightly sprayed with an atomizer spray bottle just prior to photography. Most studies photograph the cleaned and scraped surfaces of split cores. Black and white or infrared photos are the easiest to enlarge and manipulate if a darkroom is available. Use balanced light sources which match the film being used, and use a tripod to ensure that the film plane and core surface plane are 181 parallel to avoid distortion. Lake sediments are studied for their physical, biological, and chemical properties so that an historical record of the hydroecological conditions of deposition may be derived. Stratigraphic relations and reference horizons in undisturbed lake sediments make a relative chronology of recent environments relatively easy to obtain. For reviews of lake sediment studies see Lowe and Walker (1984), O'Sullivan (1983), and Renberg and Segerstrom (1981). VHI. CRUST-FREEZER SAMPLER This device requires two people to work it once the platform is stabilized. About 2-10 kg of dry ice and about 2-10 liters of methyl hydrate, or some other anti-freeze liquid, are needed for each core. Dry ice can be purchased commercially or made in the field with a dry ice maker, (such as Frigimat; Technilab model 378), and cylinders of liquid C0 2. The insulated box used to hold the dry ice is also used to store the frozen sediment. The coring procedure is as follows: 1) Using gloves, fill the corer box with chunks of dry ice. Then, being careful to avoid the splashing and bubbling, pour enough anti-freeze into the box so that the required length of sample is less than the level of anti-freeze. Do not attempt to pack the chunks of dry ice down into the box bottom as this will probably freeze your fingers and the chunks will fall down on their own as they sublimate anyway. 2) Attach the cover plate and rod section being careful to compress the rubber gasket and seal the box. Attach the gas relief valve to the rod top and the cable to the rod ring. 3) Lower the corer through the lake water as quickly as possible, but stop well before the mud-water interface is reached and lower the corer smoothly into the mud. When the sampler is felt to stop penetrating the mud easily, secure the cable at that point and allow the mud to freeze onto the box for 5-10 minutes. The box has to be brought back to the surface before all the dry ice sublimates, so watch that bubbles of C0 2 keep rising to the surface. Leaving the box too long might result in the frozen crust slipping off of the box. 4) After pulling the box onto the platform, remove the cover plate and pour the anti-freeze and any remaining dry ice into a bucket for reuse. Then, fill the box with lake water so that the inner layer of frozen crust melts and the crust can be removed from the box as planar chunks. These chunks are then quickly scraped of unfrozen mud, wrapped in plastic and aluminum foil, marked with the 'up' direction and core location, and placed in a freezer box for storage. Do not allow the sample to thaw. IX. PARTS SUPPLIERS XRT drill rods: J.K.S. Boyles International Inc. 182 X. REFERENCES Bouma, A.H. 1969. Methods for the study of sedimentary structures. Wiley, New York, 458 p. Cushing, E . and Wright, H.E. 1965. Hand operated piston corers for lake sediments. Ecology, 46, pp.380-384. Cwynar, L.C. 1982. A late-Quaternary vegetation history from Hanging Lake, Northern Yukon. Ecological Monographs, 52 pp. 1-24. Emery, K.O., and Hulsemann, J . 1964. Shortening of sediment cores collected in open barrel gravity corers. Sedimentology 3, pp. 144-154. Hakanson, L . , and Jansson, M. 1983. Principles of lake sedimentology. Springer-Verlag, Berlin, 316 p. Huttunen, P., and Merilainen, J . 1978. New freezing device providing large, unmixed sediment samples from lakes. Annales Botanici Fennici 15 pp. 128-130. Lowe, J.J. and Walker, M.J.C. 1984. Reconstructing Quaternary environments. Longman Group Ltd, New York, 389 p. Mackareth, F.J .H. 1969. A short core sampler for lake deposits. Limnology and Oceanography, 3 pp. 181-191. O'Sullivan, P.E. 1983. Annually-laminated lake sediments and the study of Quaternary environmental changes~a review. Quaternary Science Reviews 1 pp. 245-313. Renberg, I. 1981. Improved methods for sampling, photographing, and varve-counting of varved lake sediments. Boreas, 10 pp. 255-258. Renberg, I. and Segerstrom, U. 1981. Application of varved lake sediments in paleoenvironmental studies. Wahlenbergia 7 pp. 125-133. Wright, H.E. Jr. 1967. A square-rod piston sampler for lake sediments. Journal of Sedimentary Petrology 37 pp. 975-976. Wright, H.E. Jr. 1980. Cores of soft lake sediments. Boreas, 9 pp. 107-114.