Glaciotectonic Shear Zones: Surface Sample Bias and Clast Fabric Interpretation. Elliot Charles Klein

Transcription

1 Glaciotectonic Shear Zones: Surface Sample Bias and Clast Fabric Interpretation A Thesis Presented by Elliot Charles Klein to The Graduate School in Partial Fulfillment of the Requirements for the Degree of Master of Science in Department of Earth and Space Sciences State University of New York at Stony Brook May, 2002

2 State University of New York at Stony Brook The Graduate School Elliot Charles Klein We, the thesis committee for the above candidate for the Master of Science degree, hereby recommend acceptance of this thesis. Daniel M. Davis, Advisor, Professor Department of Earth and Space Sciences E. Troy Rasbury, Assistant Professor Department of Earth and Space Sciences William E. Holt, Professor Department of Earth and Space Sciences This thesis is accepted by the Graduate School Dean of the Graduate School ii

3 Abstract of the Thesis Glaciotectonic Shear Zones: Surface Sample Bias and Clast Fabric Interpretation by Elliot Charles Klein Master of Science in Department of Earth and Space Sciences State University of New York at Stony Brook 2002 Long Island surface geology is diverse in glacial settings and glaciotectonic landforms. I present two models that elucidate the generation of glaciotectonic push moraines with examples from eastern Long Island. One model, with prolonged glaciotectonic pushfrom-behind, contracts glacial sediment and strata in a manner analogous to larger scale processes in thin-skinned small-scale fold-and-thrust belts. In a prolonged glacial advance, proglacial material shortens at the glacier margin into the form of a critical taper, a wedge shaped packet of material containing the deformed structures in cross section. An alternative model, with repeated glaciotectonic push-from-behind, deforms less proglacial material since the ice, which is doing the pushing, melts back before the deforming sediment can form a critical taper. Structures produced by repeated glaciotectonic push-from-behind are iii

4 generated by seasonal or annual readvance of the glacier margin during a period of overall glacial retreat. Two field areas located within the Ronkonkoma Moraine of Long Island are documented to provide clear examples of glaciotectonic push-from-behind. At the Ranco Quarry site, measured sections suggest emplacement by prolonged glaciotectonic pushfrom-behind. Ground penetrating radar (GPR), seismic studies, topographic analysis, and measured sections at the Hither Hills site indicate emplacement by repeated glaciotectonic push-from-behind. Quantitative clast fabric analysis, despite its limitations, is a worthwhile analytical tool in glacial diamict studies. More robust than graphical methods, clast fabric analysis allows quantification of otherwise descriptive three-dimensional fabrics. In conjunction with the orientation of the long-axes, short-axes preferred direction could further establish the nature of shear, emplacement, and deposition in glacigenic settings. Field measurement of clast orientation, however, produces a systematic sampling bias in favor of clasts normal to outcrop surface. This surface sampling bias is a function of the orientation of outcrop surface to the fabric and can affect the inferred fabric strength (eigenvalues) enough to influence interpretation. I use simple calculations and computer generated random clast populations to quantify this bias and I find that it is greatest for those clasts best suited to fabric analysis (those that are rod-like in shape). True fabric strengths and orientations (eigenvectors) are misrepresented due to the surface sampling bias. Fortunately, the sampling bias effect upon strong fabric orientation is generally small. iv

5 Table of Contents List of Figures vi List of Tables....viii Acknowledgments......ix I. Push Moraine Glaciotectonics: Examples from Eastern Long Island 1 Introduction...1 Setting...6 Rationale...9 Glaciotectonic Deformation on Pleistocene Long Island...10 Glacigenic Deposits of Long Island Glaciotectonic Deformation Observed Within Eastern Long Island Moraines..22 References II. Surface Sample Bias and Clast Fabric Interpretation 42 Abstract Clast Fabric Analysis in Glacial Sediment Surface Sample Bias...57 Quantifying the bias in limiting cases. 61 Implications for field studies Conclusions. 70 References...71 Appendix A. 74 v

6 List of Figures I. Push Moraine Glaciotectonics: Examples from Eastern Long Island Figure 1. Map of Long Island Figure 2. (S 1, S 3 ) Eigenvalue Plot Figure 3. Last Glacial Maximum of the Laurentide Ice Sheet Figure 4. Digital Elevation Model of Long Island Figure 5. Critical Wedge Built by Prolonged Glaciotectonic Push-From-Behind..12 Figure 6. Moraine Formation by Seasonal Glaciotectonic Push-From-Behind...13 Figure 7. Mechanical and Glaciotectonic Parameters of Push Moraines Figure 8A. Applied Glacial Stress: Push-From-The-Rear Figure 8B. Applied Glacial Stress: Gravity-Spreading Figure 8C. Applied Glacial Stress: Compression-From-Within Figure 8D. Applied Glacial Stress: Gravity-Sliding Figure 9A. Schematic Illustration of the Critical Taper Figure 9B. Interpretive Cross-Section of the Western Taiwan Fold & Thrust Belt..19 Figure 10. Exposure at Ranco Quarry Figure 11. Montage of Aerial Photos of Hither Hills Figure 12. Hither Hills Ridge Orientation Packets Figure 13. Hither Hills Ridge Elevation Transect Figure 14A. Glaciotectonic Folds Exposed Along the Shoreline at Hither Hills...27 Figure 14B. Seismic Reflection Section of an Anticline-Cored Hill in Hither Hills..27 Figure 15. Map of the Power Line Cut in Hither Hills Figure MHz Radargram of the Power Line Cut Figure 17. Map of Rocky Point in Hither Hills Figure MHz Radargram of the Dominant Hill Structure at Rocky Point...32 Figure 19. Three Measured Sections at Rocky Point II. Surface Sample Bias and Clast Fabric Interpretation Figure 1. Map of Long Island Figure 2. A South-Facing Sea Cliff at Ditch Plains Figure 3A. Rose Diagram of 150 long-axis Clast Orientations at Ditch Plains...46 Figure 3B. Equal-Area Stereonet of the same 150 long-axis Clast Orientations...46 Figure 3C. Contoured Equal-Area Stereonet of the long-axis Clast Orientations...46 Figure 4A. Equal-Area Stereonet of short-axis Clast Orientations at Ditch Plains...49 Figure 4B. Contoured Equal-Area Stereonet of the short-axis Clast Orientations...49 Figure 5A. (S 1, S 3 ) Eigenvalue Plot Figure 5B. Isotropy-Elongation Ternary Diagram Figure 6A. (S 1, S 3 ) Eigenvalue Plot with Isotropy-Elongation Diagram labels...53 Figure 6B. Isotropy-Elongation Ternary Diagram with (S 1, S 3 ) Eigenvalues...53 Figure 7A. Isotropy-Elongation Ternary Diagram divided into Four Equal Areas...54 vi

7 Figure 7B. Four Equal Areas Skewed on a (S 1, S 3 ) Eigenvalue Plot...54 Figure 8. (S 1, S 3 ) Eigenvalue Plot of Glacigenic Fabric Domains Figure 9. Cross-Section of an Ellipsoidal Clast Being Exposed With Time.. 58 Figure 10A. Moraine Eroding in a Time Order Series: Erosion Begins Figure 10B. Moraine Eroding in a Time Order Series: After time Figure 10C. Moraine Eroding in a Time Order Series: Erosion Continues Figure 11. (S 1, S 3 ) Eigenvalue Plot of Computer Generated Clast Fabrics...63 Figure 12. (S 1, S 3 ) Eigenvalue Plot of Generated Clast Fabrics with a.r. = Figure 13. Observed Eigenvectors for Strong Fabrics...65 Figure 14. Observed Eigenvectors for Moderate Fabrics...66 Figure 15. (S 1, S 3 ) Eigenvalue Plot: Domains and Generated Clast Fabrics...69 vii

8 List of Tables Table 1. Eigenanalysis Results for Clast Orientations Recorded at Ditch Plains...52 Table 2. Computer Generated Random Clast Fabric Data Sets viii

9 Acknowledgments This thesis would not be possible without the support of Dan Davis, my thesis committee, and the Department of Geosciences. I could not thank my advisor, Dan Davis often enough for everything he has done for me at Stony Brook. Besides making unintuitive concepts understandable, Dan guided me with great care through the biggest transitions of my life. My background as a visual artist never impaired Dan s view of my capabilities and he worked with me despite my deficiencies in mathematical and scientific concepts. His lab and ideas have always been fully open and shared with me and for this I am truly grateful. When I first approached Dan for some help on understanding the orientation tensor and its relationship to pebble orientations, I remember Dan saying with a smile that he previously thought so little of pebbles that he would, without much thought, gently toss them over his shoulder. Since that day Dan Cosine is My Friend has continually enriched my life as well as my math and geophysics knowledge. I appreciate the time that Dan Davis, Bill Holt, and Troy Rasbury spent as my committee members reviewing my thesis. Their scientific insight and discussions were quite helpful in preparation for my defense. I am grateful to Troy Rasbury and Bill Holt for helping me begin my graduate career in geosciences. They have been great examples to me because of their success as scientists. Neither Troy nor Bill can stop questioning anything about geosciences or science in general. Plus they throw fun parties. Bill Meyers allowed me to pursue undergraduate research in glacial diamict and clast fabric analysis when he knew that I lacked formal training in sedimentology. To this day I often wonder why Bill gave me this research opportunity. I am truly indebted to him. Bill taught me how ask scientific questions with intelligence and purpose. I may never forget some of his responses to my written words (e.g., huh!). Bill always mentioned that good science requires a good write up. I hope this thesis lives up to Bill s expectations. I give special thanks to Don Lindsley for supporting me in his experimental petrology lab during my second summer at Stony Brook where he continually encouraged me to find a way to conduct research in geosciences. I thank Gil Hanson for sharing his general knowledge of glaciology and Long Island geology with me. I am also grateful to the Long Island Geologists because this organization (ran by Gil) provided me opportunities to submit papers and give oral presentation on glaciotectonics and clast orientation. Lianxing Wen is thanked for graciously supplying me with desk, workstation, and inspiration to carry out this work. ix

10 Saad Haq Do you have time for a game (of wiffleball) has been a superb lab mate and friend. He has been generous with his advice and time. Saad has literally saved me from embarrassing technical disasters as I prepared graduate circus talks and other presentations. I have been very fortunate to be surrounded by great characters and maturing geoscience graduate students. In particular Lucy Flesch and Andy Winslow have always made me laugh even if I was not in the mood. I thank Wen-Che Yu, Brian Hahn, and Yi Wang for bringing life and culture into our geophysics group. My friends Rob Finkenthal and Ed Keegan each deserve a hearty thank you since these guys always stuck by me. They have supported me through thick and thin ever since I was a young boy. Gail Schaefer has lifted my confidence on many occasions and I am indebted to her. Yi-Ju Chen deserves mention for constantly encouraging me. Of course my parents George and Marcy, brothers David and Louis, sister Diane, brother-in-law Rich, grandmother Sue, grandmother Adele, great aunts Olga and Charolette, aunt Rosy, uncle Lawernce, and great uncle Paul all merit special thanks for their belief in me and for giving me unconditional support throughout my life. x

11 I. Push Moraine Glaciotectonics: Examples from Eastern Long Island Introduction Late Pleistocene Long Island landforms are a product of glacial sediment availability and glaciotectonic deformation when sea level was substantially lower (by 10 s of meters) than at present. Data obtained through geologic and geophysical studies, developed and refined into landform evolution models, can help to elucidate the evolutionary path of Long Island glacial landforms, and Pleistocene Long Island landforms can be compared with evolutionary path models for each distinct Quaternary glacial landform type. There are few studies of glacial deposits and structural features that present Long Island landform evolution models, but it is clear that there is substantial local diversity in glacial settings and glaciotectonic landforms (e.g., Bernard, 1998; Meyers et al, 1998) (Fig. 1). I have concentrated my effort in the interpretation of eastern Long Island glacigenic sediment and structure, as a first step in amassing the data necessary to develop landform evolution models. We consider the landforms at Hither Hills (Fig. 1) as a late Pleistocene ice marginal push moraine based on the results of interpreted (migrated and topography corrected) ground penetrating radar radargrams, earlier topographic and seismic survey analyses (e.g., Bernard, 1998), correlation of ridges (spacing and amplitude) and glaciotectonic structural styles to modern push moraine analogs, and from detailed field analysis of lateral variation in stratigraphy. There is also the evidence of syntectonic deposition found in sediments exposed along the shoreline in sea cliffs. Future work in this setting will, it is hoped, measure glaciotectonically-shortened features and report on their change in rate of contraction. The fieldwork effort that figures most prominently in this thesis included the measurement macro-clast orientations within the stratified diamict at Ditch Plains, Long Island (Fig. 1). Clast fabric analysis of the measured clasts indicates a preferred subhorizontal, slightly west of north, long-axis orientation consistent with subglacial shear due to ice advance from that direction. Existence of potential sampling biases in clast orientation measurements collected from outcrop surfaces led me to design numerical and analytic models testing the degree to which clast axis ratios, outcrop surface orientations, and clast residence times in an eroding outcrop influence resulting clast fabric analyses. The models incorporate uniform erosion of an outcrop from the outcrop exposure surface normal direction leading to preferentially over-sampled and under-sampled clast orientations. Favorably oriented clasts remain embedded in an outcrop as it erodes, producing over-sampled orientations. Meanwhile, other clasts roll out of the outcrop after a relatively short time and can not be measured producing under-sampled orientations, often dramatically distorting the results of clast fabric analyses. Surface sampling bias modeling predicts that weakly anisotropic axial clast fabrics 1

12 2 Figure 1. Map of Long Island. Shaded in green are the Harbor Hill and Ronkonkoma moraines. Ditch Plains field study location is indicated by the small red dot on the south fork.

13 measured and quantified from surface sampled clast orientations produce large observation errors in eigenvalue strength and eigenvector direction compared with clast fabrics calculated from the entire volume of clast orientations. Volumetric clast fabric analysis would include clast orientations within the outcrop, instead of being limited to the clasts exposed at an outcrop surface. A highly anisotropic clast fabric represents the statistical distribution of clasts with a very strong preference for axial orientation in close alignment with one dominant direction. I find that the error in preferred fabric direction for sediments with a moderate or strong fabric due to surface observation bias is small, indeed smaller than the error anticipated for field measured clast orientations. Unfortunately, my results also show that clast fabric eigenanalysis of surface clasts can lead to uncertainty in eigenvalues that is much larger than the uncertainty in eigenvalue distributions associated with other sources in field measurement of clast orientations, and compares to typical differences between sediments from different glacial environments. Application of eigenanalysis to quantifying glacial sediment fabrics and other geophyical problems (e.g., Watson, 1966; Anderson and Stephens, 1972; Mark, 1973; Woodcock 1977; Woodcock and Naylor, 1983) developed into a diagnostic tool intended to deduce the genetic origin of glacial deposits (e.g., Mark, 1974, Lawson 1979; Dowdeswell et al., 1985; Rappol, 1985; Dowdeswell and Sharp, 1986; Benn, 1994, Ham and Mickelson, 1994; Hicock et al., 1996; Larsen et al., 1999; Kjaer et al., 2001) (Fig. 2) and to infer relative strain within glacigenic sediments (e.g. Hicock, 1992; Hart, 1994; Benn, 1995; Benn and Evans, 1996; Rijsdijk, 2001). Large-scale glacial landforms are interpreted by a variety of means which often includes the quantitative clast fabric analyses of glacial sediments which are then related to emplacement by sedimentary (e.g., Johnson and Gillam, 1995; Mattsson, 1997; Karlstrom, 2000; Munro-Stasiuk, 2000; Ward 2000; Hambrey et al., 2001; Kjaer and Krüger, 2001) or deformational processes (e.g., Hicock and Dreimanis, 1992; Hambrey and Huddart, 1995; Zelcs ˇ and Dreimanis, 1996; Hart, 1997, 1998; Hart and Smith, 1997; Bennett et al, 1999a; Dreimanis, 1999; Hicock and Lian, 1999; Johnson and Hansel, 1999; Blake, 2000; Evans, 2000; Cofaigh and Evans, 2001; Hart and Rose, 2001; Henriksen et al., 2001). Due to the growing importance of quantitative clast fabric analysis in landform and ice sheet reconstructions, fundamental questions continue to be raised about the statistical reliability and accuracy of the eigenanalysis method (e.g., Ringrose and Benn, 1997; Kjaer and Krüger, 1998; Krüger and Kjaer, 1999; Bennett et al, 1999b; Millar and Nelson, 2001a, 2001b; Benn and Ringrose, 2001). I find that the surface sampling bias in clast fabric analysis does not affect inferences regarding ice-flow or shear direction for strongly oriented fabrics, but it severely limits the usefulness of the technique as an indicator of glacial sediment genesis. Future research aimed at producing Quaternary glacial landform evolution models can still integrate directional data resulting from the eigenanalysis of clast orientations, but should do so with caution. Detailed geologic and geophysical analyses of glacial landforms place vital constraints on spatial, temporal and climatic reconstructions of Pleistocene ice sheets and their settings. Recent studies of contemporary glaciers worldwide has led to a more accurate understanding of the kinematics, physics, and mechanics involved in the evolution of glacial landforms by modern icesheet environments. Major advances in classifying sedimentary, structural, and geomorphological variations at active glaciers corroborate the notion that landform assemblages are shaped by a variety of complexly related depositional and structural processes that are dependent on the conditions that exist during landform generation (e.g., Lawson, 1979; Bluemle and Clayton, 1984; Boulton, 1986; Hart and Boulton, 1991; Benn and Evans, 1996; Bennett et al., 2000; Boulton et al, 2001a; Bennett, 2001; Houmark-Nielsen et al., 2001). The incorporation into models of what has been learned 3

14 4 Figure 2. S1 vs S3 eigenvalue plot of Dowdeswell et al., (1985) comparing the standard deviation and mean eigenvalues of four modern glacigenic fabric domains with debris-rich basal ice. Numbered data sets represent Svalbard tillite samples.

15 about landform evolution at an active glacier is a primary way to delineate processes responsible for landform generation in the Pleistocene epoch (e.g., Hambrey and Huddart, 1995; Boulton et al., 1996a, 1996b; Hart and Watts, 1997; Dreimanis, 1999; Evans, 2000; Bennett, 2001; Boulton et al, 2001b; Hart and Rose, 2001; Russell et al, 2001). A complementary approach to field identification of the kinematics and mechanical conditions responsible glacial landform evolution should include laboratory scale analog modeling comparable to that commonly used in structural geology and tectonics (e.g., Davis et al., 1983; Dahlen et al., 1984; Davis and Engelder, 1985), and finite difference and/or element modeling (e.g., Hsui et al, 1990; Wang, 1996) to compile a set of modeling predictions for strain patterns as a function of the governing conditions, including the shape and velocity of the advancing mass and the yield/flow criteria and thickness of sediments. In this way a link between the recent classifications of modern glacial landform assemblages and the corresponding predicted deformation patterns should then make it possible for field observations of structures and fabrics to be used more effectively to draw conclusions about the geologic and climatic conditions at the time of the deformation. Deformation imprinted on glacial landforms covers a broad range of length scales, from tiny grain size particles (e.g., van Der Meer, 1993; van Der Meer, 1997b; Menzies et al, 1997) to thousands of square kilometers (e.g., Zelcs ˇ and Dreimanis, 1997; Hart and Smith, 1997; Boulton et al., 2001b). For example, strain can be observed in the microscopic fracturing and folding of glacial sediments (e.g., Lachniet et al., 1999; Van der Wateren, 1999), in cm-to-meter scale glaciotectonic structures within the mass of deformed materials (e.g., Croot, 1987; Hart, 1990; Benn, 1995; Aber and Ruszczynska-Szenajch, 1997; Klein and Davis, 1999; Schlücther et al., 1999; Boulton et al., 2001a) as well in the macroscopic form of entire landform assemblages (e.g., Bennett et al., 1999b; Evans et al., 1999; Benn and Clapperton, 2000; Bennett et al., 2000). Landforms resulting from Pleistocene glaciations continue to be examined by traditional geological analyses (i.e., sedimentary, stratigraphic, structural, aerial photographic, topographic, clast fabric, and micro-morphologic) and by geophysical field surveying methods (i.e., seismic refraction and reflection, ground penetrating radar, satellite imagery, downhole geophysical logging, magnetometer, and resistivity-meter measurements), which yield data that can be incorporated into landform models. Comparisons made between relative strain accumulation and strain pattern in microstructures, and macroscopic deformation features within the same glacial landform can aid in constraining a particular landform evolution model. Pleistocene glacial landform models can be compared with models of landform evolution by active glaciers with similar glaciological setting (e.g., Van der Wateren, 1985; Evans et al, 1999; Bennett, 2001, Khatwa and Tulaczyk, 2001; Piotrowski et al., 2001) and with appropriate numerical and analog models in order to synthesize a collection of consistent and reliable Quaternary glacial landform models that will aid in the reconstruction of former ice sheets. 5

16 Setting North America was repeatedly covered with continental glaciers in the Pleistocene Epoch. During the last glacial maximum, one of these massive ice sheets had an extended southward advance from Artic North America to the southern shore of modern Long Island (e.g., Dyke, 2002) (Fig. 3). Other earlier Pleistocene ice sheets may also have extended as far south as Long Island. Ice sheet lobes distributed glacigenic sediments and created landforms on Long Island before finally retreating. The erosion of bedrock in the Long Island Sound basin, New York State, and in the southern New England region, during ice sheet advances, produced and transported source materials ranging widely in particle size and lithology that ultimately became surface and near surface sediment deposits on the Long Island platform (e.g., Lewis and Stone, 1991) (Fig. 4). The relatively warm climate of the Atlantic coastal plain, which included the landlocked Long Island platform in the late Pleistocene, slowed the southward advance of the spreading ice sheet. The warmer coastal climate gradually weakened the basal coupling of the warm-based ice sheet and melted large volumes of ice. The local climate reduced the total glaciotectonic stress needed to overcome the basal shearing resistance, as the glacier continued its push forward onto the Long Island platform. The temperature became warm enough for the ablation rate at the ice margin to be nearly equal in magnitude to the ice advance rate causing the ice sheets to nearly stall. Once the ablation rate overcame the forward advance rate of the glacier, the ice sheet melted back to the north off the Long Island platform. Global ice melting allowed the sea level to rise slowly to its present day level. As a consequence, Long Island glacial sediments were exposed to direct ice contact at the glacier margin only for a limited amount of time (perhaps centuries), which strongly influenced the strain histories, deformation styles, and geometric shapes of Long Island glacial landforms. Geological dating of the Pleistocene glaciation or glaciations is not well established for Long Island but deposition and deformation of the surface deposits are genetically related to one or possibly more than one glacial cycle (e.g., Lewis and Stone, 1991). Small to moderate size temperature fluctuations at the ice sheet terminus contributed to lateral variations in stratigraphy and to complexities in glaciotectonic structure observed in Long Island moraine environments. Although Long Island was landlocked in the late Pleistocene, it was still near the relatively warm waters of the Atlantic Ocean. The temperature contrast between the glacial ice and the ocean water caused a thermal gradient that locally altered ice flow and glaciation dynamics. The ice sheet terminus was influenced by the magnitude of the local temperature gradient so that prolonged periods of extreme cold and glacier advance must have been difficult to maintain and probably occurred infrequently. Therefore, the size, shape, location, and distribution of glacial lobes as well as the amount of sediment, ice, and water that these lobes carried, deposited, and deformed was primarily function of local temperature variation with respect to time. The geomorphology of Long Island was, in part, shaped by sudden surging of ice sheet advance during a period of overall glacial retreat and by relative motions between glacial lobes during a relatively short time span. Such relatively sudden changes in ice flow dynamics probably occurred even on annual-to-decade-to-century time scales and were in direct response to change in magnitude of the local temperature gradient. 6

17 Figure 3. Last Glacial Maximum of the Laurentide Ice Sheet as defined by Dykes et al. (2002). Ice sheet margins are shaded white. Ice surface contours are based predominately on direct mapping of elevations along the Last Glacial Maximum ice margin and topographic high points that were overridden by ice. 7

18 New York Connecticut LONG ISLAND SOUND BASIN Harbor Hill Moraine New Jersey ATLANTIC OCEAN BASIN Ronkonkoma Moraine LONG ISLAND scale (km) ' 73 30' 73 00' 72 30' 72 00' elevation (m) Figure 4. Digital elevation model of Long Island. Based on data from Sterner (1994) ' 41 00' 8

19 The elevated Rokonkoma and Harbor Hill moraines are the dominant topographic features on Long Island (Fig. 4). The glacial margins of the Laurentide Wisconsinan ice sheet of the late Pleistocene generated the Ronkonkoma moraine of central Long Island (Fig. 1). This thin moraine trends roughly WSW-ENE for nearly the entire length of the island as a succession of numerous interconnected kilometer-scale lobate shaped ridges. The easternmost portion of the moraine meets the Atlantic Ocean, at Montauk Point on the south fork of Long Island. As the Laurentide Wisconsinan ice sheet retreated from Long Island it likely stalled temporarily to create the WSW-ENE trending Harbor Hill moraine. This moraine is another narrow elevated landform made up of many interlinked lobate shaped ridges, on the scale of a few kilometers, that traverse the entire length of Long Island, mainly along the northern shore, reaching Orient Point on the north fork. The moraine systems each contain segments formed by the Hudson, Connecticut and Connecticut-Rhode Island lobes. Besides the two distinct moraines, the glaciers left Long Island with outwash plains, glaciotectonic hill-hole pairs, tunnel valleys, and deltaic sequences, each with distinct structural and depositional features. Glaciotectonically deformed strata, in exposed sections of Long Island moraines, contain contracted strata shortened by folding and faulting processes (e.g., Merrill, 1986; Nieter et al., 1975; Fullerton et al., 1992). Little about the subsurface geometry of Long Island glaciotectonized folds or faults is known, so the mechanism of their emplacement or formation remains unclear. Only modest effort has thus far gone into glaciotectonic analog or numerical models or to comparing them with contemporary and Pleistocene landforms. This research is needed in order to differentiate ice sheet dynamic influences on the generation of Long Island landforms. Rationale Long Island glacial sediments are well suited for geological and geophysical field investigations. In addition to being geologically and economically important, the spatial arrangement of Long Island glacial sediments and associated glaciotectonic structures plays an important role in controlling hydrologic fluid flow paths. Improved analysis of landform geomorphology and near-surface hydrology, through the investigation of three-dimensional heterogeneities in glacial strata, will likely influence the next generation of Long Island groundwater flow models. Existing groundwater flow models do not adequately account for lateral variability in glacial sediments and neglect the inclusion of identified regions of glaciotectonic folding or thrusting in Long Island deposits. Glaciotectonically altered strata and abrupt changes in the sedimentology of glacial strata often redirect groundwater flow (e.g. Sminchak, 1996; Beres et al., 1999; Boyce and Eyles, 2000; Gerber et al., 2001, Regli et al., 2002). Incorporating quantitative structural and sedimentological anisotropies into future Long Island groundwater and contaminant flow models is absolutely necessary given the large population (approximately 2.7 million) who directly depend on the groundwater pumped from fragile aquifer systems as their sole water supply. 9

20 Geologic studies characterizing sedimentary and structural field relationships in glacial diamict and associated sediments are fundamental in establishing glacigenic facies (e.g., Krüger and Kjaer, 1999). Measured sections at outcrops serve as ground truth for Long Island glacigenic facies which represent the depositional and emplacement associations for the wide spectrum of deformed and undeformed sediment, strata, and structure that are found within the deposits of glacigenic sediments (e.g., Meyers, 1998). Since not all of these glacigenic facies are documented, and others have not been correlated, Long Island glacial strata, diamict and associated sediments, as well as glaciotectonic features should continue to be sedimentologically, structurally, and stratigraphically characterized at field sites. All known glacigenic facies ought to be integrated into sedimentological, glaciotectonic, and landform evolution models. Application of geophysical instruments, imaging unconsolidated glacial sediments on Long Island, has proven to be extremely valuable in shallow surface surveying because these tools provide representation of otherwise inaccessible deposits or structures (e.g., Bernard, 1998; Davis et al., 2000). Geophysical survey methods such as seismic reflection and refraction, ground penetrating radar (GPR), and resistivity-measurement can estimate and differentiate physical property variations in glacial sediments at a wide range of depths and resolutions. This is especially useful in terrain that is not well exposed, providing two or three-dimensional images of the near subsurface (e.g., Hansen et al., 1997; Ramage et al, 1998; Beres et al., 1999; Penttinen et al., 1999; Gerber et al., 2001; Overgaard and Jakobsen, 2001; Williams et al., 2001). Combining geophysical investigations and geologic fieldwork studies, particularly at outcrop exposures or at excavated sites, strengthens correlation between ground penetrating radar, resistivity, seismic, and glacial facies (e.g., Harris et al., 1997; Davis et al., 2000; Eden and Eyles, 2001; Ékes and Hickin, 2001; Salem, 2001; Regli et al., 2002). Glaciotectonic Deformation on Pleistocene Long Island On Long Island, the Harbor Hill moraine ridge topography often exceeds 60 m and frequently the Ronkonkoma moraine ridges top 90 m, attaining maximum elevation at roughly 128 m above sea level (Fig. 4). Glacial erosion and transport of Long Island Sound basin bedrock material by ice sheets were the chief sedimentological processes contributing to the deposition of the enormous supply of sediments that evolved into glaciotectonically thickened and deformed Long Island moraine landforms. The emplacement mechanisms and landform evolution paths of moraines on Long Island, though still poorly understood, included subglacial and proglacial deposition and deformation processes. The range of glaciotectonic structures, evident at a variety of scales, provides insight into the development of Long Island landforms, as well as other Quaternary landforms built by similar warm based, weakly coupled glacier marginal systems. Therefore, documenting the distributions and complexities of glaciotectonic structures within the landforms of Long Island is necessary. 10

21 At present, I believe that two principal glaciotectonic mechanisms are responsible for generating much of the deformed proglacial structures on Pleistocene Long Island. These structures were generated glaciotectonically by either a prolonged push-from-behind, a seasonal push-from-behind, or a mix of these two mechanisms. This is consistent with many of the commonly observed deformation features found in Long Island proglacial sediments (e.g., Meyers, 1998; Bernard, 1998; Klein and Davis; 1999). Prolonged glaciotectonic push-from-behind thin-skinned deformation shortens glacial strata by producing fold-and-thrust structures that must be accommodated by a décollement, a weak accommodating layer at depth in which there develops a shear zone, typically with a strong shear-related fabric (Fig. 5). The glacial sediments and structures involved in prolonged glaciotectonic push-from-behind often contract into the form of a critical taper (e.g., Schlüchter et al., 1999; Williams et al., 2001) a wedge shaped packet of material in cross section containing the deformed structures. This has been documented on a larger scale in thin-skinned small-scale fold-and-thrust belts (e.g., Davis et al., 1983; Dahlen et al., 1984). Seasonal glaciotectonic push-from-behind thin-skinned contraction is not capable of folding and thrusting as much sediment since the ice, which is doing the pushing, melts back and retreats before the deforming sediment can form a critical taper (Fig. 6). The spatial and temporal patterns associated with push-from-behind glaciotectonic deformation events that contracted Long Island glacial sediments are still uncertain. This is particularly true in terms of discriminating deformation patterns involving newer glaciotectonic structures overriding previously deformed structures regardless of how any of the structures were glaciotectonically emplaced. The descriptive terms used to characterize proglacial glaciotectonic deformation are subdivided by size differentiations that are often arbitrarily defined (e.g., Aber et al., 1989; Hambrey and Huddart, 1995; Benn and Evans, 1998). Bennett (2001), clarifies some of the confusion in taxonomy by using the term push moraine to define the product of construction by the deformation of ice, sediment, and/or rock to produce a ridge, or ridges, oblique or transverse to the direction of ice flow at, in front of, or beneath and ice margin. The formation of a moraine by advance of the glacier margin is thus what defines push moraines, not whether or not the moraines were formed by seasonal or prolonged glaciotectonic push. In push moraine systems sediment displacements and dislocations can range from a few meters to several kilometers horizontally and up to 200 m vertically producing larger and thicker moraine sizes with more distinctive internal tectonic style as displacement of pushed sediment progresses (e.g., Boulton, 1986; Hart and Boulton, 1991; Lehmann, 1993; Boulton and Caban, 1995, Boulton et al., 1999). Included in the definitions of push moraine by Bennett (2001) are thrust moraines, thrust-block moraines, composite ridges and hill-hole pairs as long as they can be clearly linked to have occurred at, or close to an ice margin. My thesis adopts the push moraine definition of Bennett (2001) and recognizes that glaciotectonic push-from-behind, whether seasonal or prolonged, terrestrial or marine, generated by gravity spreading or glaciodynamic pushing forces, was the pushing mechanism which drove the deformation producing push moraine ridges, regardless of generated ridge amplitude, ridge spacing, or the state (whether lithified or frozen) of the pushed sediment. Small push moraines built by annual or seasonal push-from-behind glaciotectonic mechanisms at contemporary glacier margins develop into annual or seasonal push moraines 11

22 Figure 5. Illustration in time sequence of the growth of a critical wedge in a prolonged glaciotectonic push-from-behind setting. The youngest deformation is concentrated toward the distal end of the wedge. Note that the anticlines are typically cored by imbricate thrust faults. 12

23 Figure 6. Three schematic cross-sections (in time sequence) illustrating one possible model for the formation of a seasonal glaciotectonic push-from-behind moraine. Sediments can vary greatly over short distances. Note how sediments from an impounded lake between the glacier and previously formed ridges are emplaced in the next-formed ridge. In this model, ridges young toward the ice, in the opposite direction than in a critical wedge. 13

24 that usually form ridges 5 m in height (e.g., Boulton, 1986; Bennett, 2001). Large push moraines generated by prolonged or large-scale glaciotectonic push-from-behind at glacier margins produce ridges with heights 5 m in a sustained glacial advance often due to a change in glacier mass balance (e.g., Boulton, 1986; Bennett, 2001; Russell et al, 2001). The main distinction is not the size of the moraine, since the 5 m height is an arbitrary cutoff, but whether the moraine was produced by seasonal or annual readvance or by a more sustained advance at the glacier margin. When significant deformation has been transmitted horizontally beyond the glacier margin, multi-crested push moraines are generated with deformation style usually involving, multiple folds, fans of listric thrusts, fans of imbricate thrusts, or superimposed sub-horizontal nappes produced by overthrusting (e.g., Bennett, 2001). The glaciotectonic process responsible for the initiation, excavation and elevation of proglacial materials is similar for either size push moraine: the main difference between the two can characterized by the amount of deformed outwash fan sediment present at the glacier margin (e.g., Boutlon, 1986; Benn and Evans, 1998; Bennett, 2001; Russell et al., 2001). Small push moraines grow by a glaciotectonic push and deformation of the proximal outwash fan slopes of the glacier margin. Large push moraines, on the other hand, grow not only by glaciotectonically shoving the proximal outwash fan slopes, but also by pushing much or the entire outwash fan. Asymmetric ridges that have steep distal and shallow proximal flanks tend to be formed in small push moraines (e.g., Sharp, 1984). Additionally, the moraine ridges may push their own pretectonic and syntectonic outwash sediment as well as override subglacially lain tills, if there are any, during a readvance of the glacier margin (e.g., Boulton, 1986; Bennett, 2001). Ridge amplitude is predominately controlled by sedimentological factors such as sediment character and availability as well as the duration of a glacial advance. Small glacial advances commonly produce 1 to 2 m ridge height amplitudes such as those formed by the seasonal readvances of the Breidamerkurjökull glacier ice margins, in Iceland between 1965 and 1981 (e.g., Boulton, 1986). Small push moraines are associated with the formation flute and show variation in pattern of sedimentary activity along the ice margin (e.g., Boulton, 1986; van der Meer, 1997a; Bennett, 2001). In the glaciotectonic development of small push moraines, the ice advance is often annual and proceeds much like an oversized bulldozer blade plowing through loose, water saturated sediments that deform into a series of ridges. When a glacier ablates, water, ice and debris are sloughed off the snout to build up outwash fans and glaciofluvial streams. During readvances the glacier pushes and partially overrides the fans. The forward moving glacier oversteepens the distal slopes of the fans which receive a new layer of debris when the glacier retreats. Sediment that was overridden in an advance is incorporated into the subglacial environment where it is deformed into a thickening wedge of till beneath the margin. Both large and small push moraine systems preserve of at least 25% of the glaciotectonic structures involved in the push-from-behind deformation process and the syntectonic plus pretectonic proglacial materials make up over 25% of a moraine system unit area (e.g., Benn and Evans, 1998). The extent of proglacial deformation resulting from push-from-behind glaciotectonics is highly variable, but in general is a function of time, climate, and glacier margin environment. For example, if an advancing glacial margin begins to deform proximal materials but stalls after a short period (e.g. one season) then that glaciotectonic push will 14

25 not have deformed much material. On the other hand, if instead of stalling, the glacial margin continued for a prolonged advance (e.g. multiple seasons) then much more proglacial material will be glaciotectonically pushed and deformed. Some important parameters affecting the deformation front and the growth of a critical taper include seasonal temperature and moisture variation, porewater pressure gradient, coupling of basal ice with the subglacial bed, glacial lobe height and its basal area, and the rate of glacial advance and ablation. Other factors influencing the magnitude of the push-from-behind deformation are sediment size, state, and type, local topographic relief, friction on the décollement, and availability of standing water and outwash materials. Proglacial environment and climate conditions directly affecting physical characteristics of the glacial margin including local pore fluid pressure, evolution of drainage, sediment and glacier bed state, aspect ratio of foreland wedge, foreland rheology and strength, basal shear traction, depth and slope of the décollement (e.g., Bluemle and Clayton, 1984; van der Wateren, 1985, 1986; Boulton, 1986; Hart and Boulton, 1991; Boulton and Caban, 1995, Etzelmüller et al., 1996; Dell Isola and Hunter, 1998; Boulton et al, 1999; Schlüchter et al., 1999; Bennett et al, 2000, 2001, Boulton et al, 2001a). (Fig. 7) The growth of push moraines relies on the large-scale displacement of proglacial materials within shear zones due to stresses imposed by the gravity spreading of a glacier (Fig. 8A). The gravity spreading model demonstrates that the total glaciotectonic stress needed to push glacial material from behind, permitting proglacial sediment failure and glaciotectonic thrusting, is obtained by the translation of compressive stress due to the weight of a spreading ice mass (e.g., van der Wateren, 1985; Aber et al., 1989; Benn and Evans, 1998; Bennett, 2001). Important components of the total glaciotectonic stress field include the glaciodynamic stress (basal shear stress) and the horizontal cumulative compressive stress transferred from the normal stress (glaciostatic stress) generated by the static weight of the ice over a given area. Failure can take place on a plane when the total glaciotectonic stress exceeds or equals the shear resistance. Push-from-the-rear, gravity sliding, and compression-from-within models represent other mechanical ways to produce push moraines (e.g., Bennett, 2001) (Fig. 8B,C,D). Glaciotectonically deforming sediment blocks are folded and thrust into push moraine systems by the gravity sliding of surging glacial ice moving down a slope by the driving force of its own weight. The laterally compressive push-from-the-rear mechanism directly shoves, folds and thrusts sediment wedges by the forward motion of glacial ice into the foreland. Compression from within the terminal zone of the glacier occurs if there is deceleration of ice flow, strongly coupled subglacial and proglacial zones which behave as a single unit that is deformed by listric faults, and a décollement which lies below both the glacier and its foreland (e.g., Hart, 1990; Hambrey and Huddart, 1995; Bennett, 2001). Development of glaciotectonic clast fabrics within internal structures of push moraines has the possibility of shedding light on wedge propagation and paleo-ice flow directions as well as potentially distinguishing amongst glacial stress fields and seasonal or prolonged push deformation styles (e.g., Sharp, 1984). The main section of this thesis concentrates on the applicability of clast fabric analysis as an interpretative tool in glacial settings based on clast orientations measured from surfaces of outrcrop exposures. In the glaciotectonic evolution of large push moraines, the recently deformed sediments are in contact with and are actively shoving the more distal sediments as the system advances forward. The geomorphology of a moraine ridge generally reflects the 15

26 Schematic model showing how some push moraines may relate to selected variables used to define a broad matrix. Figure 7. Important mechanical and glaciotectonic parameters involved in the structural development of push moraines. Schematic model after Bennett (2001). ) 16

27 A B C D Figure 8. Models of push moraine Fig. 18. Models structural of applied glacial evolution stress. and morphology due to the effect of applied glacial stress as shown by Bennett (2001). A. Push-from-the-rear. B. Gravity-spreading. C. Compression-from-within. D. Gravity-sliding. 17

28 shape of the thrusting glacier margin, and individual ridge crests correspond to the crests of internal folds (e.g., Boulton, 1986; Aber et al., 1989; Benn and Evans, 1998). One pushfrom-behind, glaciotectonic model for building a push moraine is the glacial analog of the thin-skinned wedge model of tectonic deformation, which produces a critical taper in mountain belts (e.g., Davis et al., 1983; Dahlen et al., 1984) (Fig. 9A,B). Glaciotectonic push-from-behind compression that has evolved into a small-scale thin-skinned fold-andthrust belt often exhibits multiple thrust sequences with piggyback structures being the most common of the glaciotectonic structure produced (e.g., Van der Wateren, 1985; Schlüchter et al., 1999). In some push moraine systems the largest ridge is the most proximal to the glacier margin and was produced first. As the ice continued to push, a new ridge formed in front of the previously formed ridge so that ridges are youngest in the forward direction of the advancing ice. Newer ridges are more distal and are somewhat less elevated than those ridges formed earlier, so ridge amplitudes decay with distance away from the glacial margin (e.g., Croot, 1987; Hambrey and Huddart, 1995; Boulton, et al., 1999). The cross-sectional taper of a growing wedge-shaped mass of overthrust material is dependent upon the cohesive strength of the deforming material at the time of deformation, its thickness, and the strength of its coupling to the base (e.g., Davis et al., 1983; Schlüchter et al., 1999; Williams et al, 2001) (Fig. 5). If a glacier margin is pinned at the margin front but continues to advance from the rear, the shortening within the mass of sediments increases causing folds to be pushed or pinched-out. The most internally shortened structures and the shortest wavelengths between ridges are closest to the glacial margin since the cumulative shortening of the pushed sediments is the furthest distance from the distal extremity of the push moraine (e.g., Boulton et al., 1999). If on Long Island the glacier moved in a sustained advance by prolonged push then one should observe diminishing ridge heights in cross-sectional shape (a critical taper) as one moves away from the suspected glacial margin. One would also find a décollement, a gradient in strain magnitude (more intense in the hinterland to the north), a gradation in syntectonic deposition (finer, more distal facies to the south), little lateral variation in glacial stratigraphy, and a general northward sweeping in the dips of sediments and thrust faults. Proglacial glaciotectonic deformation through seasonal or annual meters-scale glacier margin surges or thrusts generates push moraines during a period of overall ice sheet retreat. The timing of the advance and the deformation it causes can be seasonal-to-annual, or decadal, but if the precise time intervals of the surges are unknown then the glaciotectonic process is simply referred to as annual or seasonal push but probably ought to be called repeated push. Repeated (seasonal or annual) push glaciotectonic deformation is suspected in the growth of push moraine structures at Hither Hills, Long Island and has likely contributed to the creation of other portions of Long Island push moraines (e.g., Klein and Davis, 1999). Local climate regime and temperature fluctuation at the glacier margin, over relatively short periods, can promote repeated push glaciotectonic deformation which influences push moraine ridge geometries and internal structures, as well as the hydrogeology of the push moraine foreland. Unlike prolonged glaciotectonic push-from-behind where contractional deformation is sustained over many seasons without retreat, repeated glaciotectonic push-from-behind involves push from the rear during almost every ice advance season (typically, winter) 18

29 A B 19 Figure 9. A. Schematic illustration of force balance calculation used in deriving the critical taper and the orientations of the principal stress axes throughout a wedge of material everywhere on the verge of failure as shown by Davis et al., (1983). An element of wedge is subject to stresses due to body forces from the side and at its base, as well as graviational stresses. B. Interpretive cross-section through the foothills of the western Taiwan fold & thrust belt (e.g., Davis et al., 1983). Note the overall wedge taper and stacking of thrust sheets over the décollement.

30 glaciotectonically deforming the glacial margin. The seasonal or annual ice marginal advance is immediately followed by an ice ablation season (typically, summer) where the glacial margin retreats so this coupled with ice marginal advance and ablation form a repeating pattern over multiple seasons (or years) deforming and elevating substantial push moraine topography. The seasonal ice advances and retreats resulting in push moraine ridges are thought to be due to thermally activated changes in ice flow dynamics that stimulate rapid surge forward of the glacial margin that later melts back close to its original position prior to the advance. Repeated glaciotectonic push develops new push moraine ridges because the glacier in overall retreat deposits sediments in front of the glacier during a warm period when ablation is most rapid, and then internally folds those sediments during a partial readvance stimulated by a colder period (e.g., Boulton, 1986; Hart and Watts, 1997; Bennett, 2001). The repeated push glaciotectonic deformation process, once initiated, shoves and contracts proximal subglacial and proglacial sediments by high-angle thrusting and folding that develop into a push moraine ridge during the ice advance (Fig. 6). After the first ice advance the glacier stalls and eventually retreats depositing new sediment loads. The next glacial advance imparts further compressive deformation to the previously formed ridge and generates a new ridge. Over time, repeated glaciotectonic push-from-behind is capable of developing extensive push moraine ridge systems (e.g., Hart and Watts, 1997). Repeated push or surge of the ice sheet provides the stress necessary to shorten or extend nearby landforms and allow the opportunity for substantial variation in depositional environments. With the seasonal ablation of the ice sheets, lowlands open up between the most recently generated ridge and retreating glacial margin trapping ice, sediment, and melt-water to form proglacial lakes and outwash fan systems. If the glacier margin did propagate by repeated glaciotectonic push-from-behind, then one would observe non-systematic variation of ridge amplitude in cross-section throughout the push moraine ridge system. In other words, there would be no through going décollement or obvious critical taper of the push moraine ridges. One would also expect to find asymmetric internal folding and substantial lateral variation in glacial stratigraphy and syntectonic deposition (e.g., Sharp, 1984; Boulton, 1986; Hart and Watts, 1997). Along with geologic dating techniques such as lichenometry used by Hart and Watts (1997), the evaluation of clast fabrics within the internal structures of push moraines can, in principle, be used to distinguish seasonal or prolonged glaciotectonic pushfrom-behind deformation styles. 20

31 Glacigenic Deposits of Long Island Classification of glacial sediments and the description of glacial facies are interpretive and frequently controversial due to the enormous varieties of glacial deposit types and the complex stratigraphic and structural relationships present in glacial environments (e.g., Dreimanis, 1989; Meyers et al., 1998, Krüger and Kjaer, 1999; Ruszczynska-Szenajch, 2001). Rapid changes in deposition and deformation rates at the glacier margin can syntectonically thicken glaciotectonic structures and drastically affect vertical and lateral stratigraphic sequences over short distances. The glacial facies system describes the products of glaciation by classifying and organizing glacigenic sediments, spatially and temporally, at a wide range of scales, with the purpose of genetically relating glacial deposition and deformation to glacier erosion, transport, and melt. Descriptions of glacial facies rely on process assemblages to reflect the variety of processes that were active in the arrangement of a particular glacial environment over a range of length scales and time spans (e.g., Benn and Evans, 1998). Sedimentary deposits of unknown genetic origin called diamicts exist in the Harbor Hill and Ronkonkoma moraines and because glaciation is thought to be responsible for almost all shallow surface sedimentation on Long Island these sediments are referred to as glacial diamicts. Diamict deposits (glacial or non-glacial) contain a broad mix of particles varying in shape and angularity that range in size from mud to boulder all incorporated into a poorly sorted matrix. Important Long Island glacial diamicts, genetically known as primary tills (or primary glacigenic deposits), can be produced by either deformation, lodgement, or melt-out processes, but most primary tills are made from a mixture of these till producing processes. Other genetic glacial sediment on Long Island, known as secondary tills, are sediments which have been remobilized by some form of non-glacial process that has reworked the primary tills (e.g., Lawson, 1982). Glacial diamicts and associated sediments produced either by subglacial, proglacial, or combined processes can not be identified by only one diagnostic criterion so observing a set of sedimentary characteristics may aid in differentiating the genesis of diamict depositions (e.g., Hicock, 1990; Krüger and Kjaer, 1999; Kjaer et al., 2001). Glacial diamict deposits include oriented clasts within stratified or massive stratigraphic units or within isolated lumps, lenses, or layers contained within a glacigenic sedimentary unit or bounded by one or more glacigenic units. Clast fabric analysis statistically represents the fabric shape, as a frequency distribution of oriented clasts by chosen axial direction (i.e., long-axis). Eigenanalysis (Chapter 2) performed on a group of clasts quantitatively describes clast fabric shapes by normalized eigenvalues, corresponding to at least one eigenvector, describing the likelihood that any clast axis (i.e., long-axis) from the group of clasts is potentially pointed in one of three mutually orthogonal minimum, intermediate, and most preferred eigenvector directions. Clast fabrics of glacigenic sediments may reveal the sense of motion in a shear zone and indicate the relative strain but fails to adequately delineate types of glacial diamict. 21

32 Glaciotectonic Deformation Observed Within Eastern Long Island Moraines Long Island glacigenic sedimentation processes were often complex so that interpreting the history of these deposits, especially for the Harbor Hill and Ronkonkoma moraines, has been problematic. Geophysical surveys and geological fieldwork studies of these moraines conform both the stratigraphic and structural complexities of these settings. Much of the sediment deposition and deformation occurred in front of and beneath glacier ice so that most Long Island glacial sediments have under gone some glaciotectonic pushing, shearing, or folding, and may have experienced syntectonic deposition or re-deposition. In the exposed sediments at Ranco Quarry (Fig. 1) within the Ronkonkoma moraine, coherent glaciotectonic thrust blocks have been mapped for several tens of meters above and hundreds of meters or more laterally from their source, with gravel-rich thrust zones as indirect evidence for the sediment having been permafrost (e.g., Meyers et al., 1998) (Fig. 10). Glaciotectonic deformation studies in the Ronkonkona moraine of eastern Long Island were conducted in Hither Hills State Park, (Fig. 1) revealing very different sediments and structures than found 75 km southwest at Ranco Quarry. Sedimentary, structural, seismic, and GPR surveying at Hither Hills show evidence of syntectonic deposition, folded strata, cm-scale faulting, and lateral variation of sedimentary layers. Aerial photographs reveal dozens of ridges with nearly parallel strike directions (fig. 11). Topographic analysis of the Hither Hills region grouped ridges of similar height, spacing, and orientation into packets with consistent azimuths (Fig. 12). Transects of typical ridge successions indicate either a slight increases in maximum ridge heights from north to south or no obvious ridge elevation pattern (e.g., Bernard, 1998) (Fig. 13). Photographs of sea cliff exposures and seismic data obtained from ridges in the state park by common mid-point seismic reflection, after stacking, revealed glaciotectonic structures in the moraine (Fig. 14A,B). After processing, seismograms depict a gently folded antiformal layer beneath a shorter and more tightly folded antiformal piggyback structure. These shallow surface folds are a few meters beneath the surface and have amplitudes in meters and wavelengths that are tens of meters long (e.g., Bernard, 1998) (Fig. 14B). In general, Long Island glacigenic sediments and glaciotectonic structures are poorly exposed and are often difficult to evaluate with more traditional geophysical techniques such as seismic reflection surveying. Ground penetrating radar, another geophysical tool appropriate for use in glacigenic sediment, was employed to acquire radar data at the power line cut in Hither Hills (Fig. 15). Radar data are processed and analyzed in a manner similar to seismic reflection data, allowing the establishment of radar facies, which, like seismic facies, describe distinct structural and stratigraphic changes in depositional sequences. GPR surveys at Hither Hills reproduced the shallow structures inferred from the seismic study of Benard (1998) but also reveal many more antiformal folds and piggyback structures at greater depths than achieved by seismic techniques (Fig. 16). The radargrams obtained with the 50, 100, and 200 MHz antennas provide complementary and far more complex images of the folded strata previously detected by the seismic reflection survey. The Hither Hills radargrams also provide higher resolution to a greater penetration depth (as deep as 45 m) than accomplished with the seismic reflection survey. In detail, the Hither Hills radargrams imaged many folded strata having an appearance resembling imbricate thrust systems. Some of these imbricate thrusts 22

33 23 Figure 10. Coherent thrust sheets exposed in a quarry in cenral-eastern Long Island, NY (e.g., Meyers et al., 1998). Unit 4 is a marine beach/barrier sand that can be demonstrated to have been transported, albeit folded and microfractured, over a long distance. Note the classic fold-thrust belt ramp-flat geometry, typical of critical wedge fold and thrust belts.

34 Rocky Point 24 0 km 1 km 2 km Power Line Cut Figure 11. Montage of aerial images of Hither Hills. The ridges strike roughly SW-NE. The ridges do not show the clear N-S size progression expected for critical wedges. There are two sites (indicated in red) that have been documented with GPR. Many ridges are exposed by beach erosion along the north shore.

35 25 Figure 12. Pleistocene glaciotectonic ridge systems like Hither Hills in eastern Long Island, NY (e.g., Klein and Davis, 1999) can cover several square km and contain large numbers of parallel ridges. Analysis of Hither Hills ridge heights, spacing and orientations shows that the ridges are grouped into 'packets' of similiar size with consistent mean azimuths (typically ±5 ).

36 S Elevation Along a Transect in Hither Hills N Elevation (ft) Position Along Transect (ft) Figure 13. Typical S-N elevation transect of a ridge succession at Hither Hills. Elevations were determined on the basis of 5 ft contour intervals provided by a topographic map of the region (e.g., Bernard, 1998). 26

37 A) B) CDP Location (m) Time (sec) Figure A). Glaciotectonic folds in the Ronkonkoma moraine, exposed along the shoreline at Hither Hills, NY. Each anticline corresponds to one of a series of subparallel ridges (figure 12), and is believed on the basis of mapping and geophysical data to contain a thrust fault below present-day sea level (e.g., Klein and Davis, 1999). B) Seismic reflection section from Bernard et al., (1998) of an anticline-cored hill in Hither Hills, showing a. shallow reflector (a folded bed). The seismic survey line is on a ridge flank that dips to the left (north) at 7, so the bed dips 18 to the horizontal at left (N) and 4 on the right (south) limb. There is little or no vertical exaggeration. This fold is similiar in wavelength, amplitude, and shape to that in (A). 27

38 power line cut 1000 feet power line cut Figure 15. GPR survey (red line) conducted with 50 MHz antennas along a power line cut in Hither Hills (figure 11). Seismic survey of Bernard et al., (1998) begins at the southern end of the GPR survey profile. Note topographic contours indicate the number of feet above sea level. 28

39 Distance (m) Figure 16. Topography corrected and migrated 50 MHz radargram of a 150 m N-S portion of the power line cut at Hither Hills (figure 15). Dashed purple lines indicate dip-domains in sediments, and heavy red dashed lines indicate possible faults. Depth (m) 29

40 core the ridges and exhibit an over-printed geometry consistent with a push-from-behind glaciotectonic mechanism being responsible for the ridge formations found at this location in the Ronkonkoma moraine. The GPR survey using the 50 MHz antennas overlapped the seismic line of Bernard et al., (1998) along the power line cut in Hither Hills and extended to the north for an additional 150 meters (Fig. 15). Interpreted radargrams of the Hither Hills shallow subsurface indicate substantial sequences of syntectonic depositions since radar reflectors (dip domains) thicken away from fold hinges and the cores of the imaged ridge structures are more tightly folded than is the overlaying topography (Fig. 16). Hither Hills Radargrams processed from radar data acquired near Rocky Point (Fig. 17), show the presence of complexly folded glacigenic strata of wavelength and amplitude similar to those seen in exposed sea cliffs along the north coast of the park (Fig. 18). This is consistent with my measured stratigraphic sections (e.g., Klein and Davis, 1999) along the north shore of Hither Hills that document lateral variation of stratigraphy on the scale of 10 s to 100 s of meters (Fig. 19). Seismic reflection and refraction studies, measured sections, topographic analyses, ridge transects, and GPR surveys have led to the conclusion that seasonal glaciotectonic push-from-behind (e.g., Boulten, 1986; Bennett, 2001) is the most viable explanation for the development of the push moraine features observed at Hither Hills. As described early in the next chapter of this thesis, clast fabric analysis along the shoreline at Ditch Plains clearly shows a fabric consistent with glacial advance from the NNW. In conjunction with future Long Island geophysical surveys, measured sections and clast orientations should be recorded. Careful geologic and geophysical studies will continue to elucidate a more complete picture of the dynamic strain histories and complex sedimentation processes associated with glaciotectonic deformation. Long Island is currently loaded with a wide variety of glacial diamict deposits generated by Pleistocene continental ice sheets justifying continued emphasis on quantitative methods for establishing local ice flow direction based on clast orientation. 30

41 Column C. - Small Anticline Column B. - Rocky Point Column A.. - Dominant Hill Structure 00m 0 10 Figure 17. Location of a GPR survey (red line) conducted with 200 MHz antennas across the dominant hill structure at Rocky Point, Hither Hills (figure 11). The positions of three measured stratigraphic sections at sea cliff exposures are indicated. Note topographic contours indicate the number of feet above sea level. 31

42 32 Figure 18. Topography corrected, migrated, and interpreted 200 MHz radargram. The radargram sampled 130 m of the dominant hill structure at Rocky Point, Hither Hills (figure 17). Changes in surveying direction are indicated in degrees. Interpreted radar facies of this complicated ridge structure are shown in color.

43 Column A.. - dominant hill structure Column B. - Rocky Point Column C. - small anticline 10 m clay sand diamict cover offset roots 8 m 6 m 4 m 2 m clay - boulder clay - boulder clay - boulder Figure 19. Three measured stratigraphic sections spaced about 100 m (about one ridge wavelength) apart in the sea cliffs exposed along the northeast coast of Hither Hills (figure16). Note the thin clay in column C. (small anticline) and the great variation in sedimentology and stratigraphy between the sections. 33

44 References Aber, J.S., D.G. Croot, and M.M. Fenton, Glaciotectonic Landforms and Structures, Kluwer Academic Publisher, Aber, J.S., and Ruszczynska-Szenajch H., Origin of Elblag Upland, northern Poland, and glaciotectonism in the southern Baltic region, Sedimentary Geology, 111, , Anderson, T.W., and Stephens M.A., Test for randomness of directions against equatorial and bimodal alternatives, Biometrika, 59, , Benn, D.I., Fabric shape and the interpretation of sedimentary fabric data, Journal of Sedimentary Petrology, A64, , Benn, D.I., Fabric signature of subglacial till deformation, Breidamerkurjökull, Iceland, Sedimentology, 42, , Benn, D.I., and Clapperton, C.M., Pleistocene glacitectonic landforms and sediments around central Magellan Strait, southernmost Chile: evidence for fast outlet glaciers with cold-based margins, Quaternary Science Reviews, 19, , Benn, D.I., and D.J. A. Evans, The interpretation and classification of subglacially-deformed materials, Quaternary Science Reviews, 15, 23-52, Benn, D.I., and Evans D.J.A., Glaciers & Glaciation, 734 p., Arnold Publishers, New York, Benn, D.I., and Ringrose, T.J., Random variation of fabric eigenvalues: implications for the use of A-axis fabric data to differentiate till facies, Earth Surface Processes and Landforms, 26, , Bennett, M. R., The morphology, structural evolution and significance of push moraines, Earth-Science Reviews, 53, , Bennett, M.R., Hambrey, M.J., Huddart, D., Glasser, N.F., and Crawford, K. The landform and sediment assemblage produced by a tidewater glacier surge in Kongsfjorden, Svalbard, Quaternary Science Reviews, 18, , 1999a. Bennett, M.R., Waller, R.I., Glasser, N.F., Hambrey, M.J., and Huddart, D., Glacigenic clast fabrics: genetic fingerprint or wishful thinking? Journal of Quaternary Science, 14, , 1999b. Bennett, M.R., Huddart, D., and McCormick, T., An integrated approach to the study of glaciolacustrine landforms and sediments: a case study Hagavatn, Iceland, Quaternary Science Reviews, 19, , Bernard, M., 1998, Shallow Seismic Studies in Glaciotectonic Hither Hills, Long Island [M.S. thesis]: State University of New York at Stony Brook, 81 p. Beres, M., Huggenberger, P., Green, A.G., and Horstmeyer, H., Using two- and threedimensional georadar methods to characterize glaciofluvial architecture, Sedimentary Geology, 129, 1-24, Blake, K.P., Common origin for De Geer moraines of variable composition in Raudvassdalen, northern Norway, Journal of Quaternary Science, 15, , Bluemle, J.P., and Clayton, L., Large-scale glacial thrusting and related processes in North Dakota, Boreas, 13, , Boulton, G.S., Push-moraines and glacier-contact fans in marine and terrestrial environments, Sedimentology, 33, ,

45 Boulton, G.S., The origin of till sequences by subglacial sediment deformation beneath mid-latitude ice sheets, Annals of Glaciology, 22, 75-84, 1996a. Boulton, G.S., Theory of glacial erosion, transport and deposition as a consequence of subglacial sediment deformation, Journal of Glaciology, 42, 43-62, 1996b. Boulton, G.S., and Caban P., Groundwater flow beneath ice sheets: part II its impact on glacier tectonic structures and moraine formation, Quaternary Science Reviews, 14, , Boulton, G.S., Van der Meer, J.J.M., Hart, J.K., Beets, D., Ruegg, G.H.J., Van der Wateren, D.F.M., and Jarvis, J., Till and moraine emplacement in a deforming bed surge an example from a marine environment, Quaternary Science Reviews, 15, , Boulton, G.S., Dobbie, K. E., and Zatsepin, S., Sediment deformation beneath glaciers and its coupling to the subglacial hydraulic system, Quaternary International, 86, 3-28, 2001a. Boulton, G.S., Dongelmans, P., Punkari, M., and Broadgate, M., Palaeoglaciology of an ice sheet through a glacial cycle:; the European ice sheet through the Weichselian, Quaternary Science Reviews, 20, , 2001b. Boulton, G.S., van der Meer, J.J.M., Beets, D.J., Hart, J.K., and Ruegg, G.H.J., The sedimentary and structural evolution of a recent push moraine complex: Holmstrømbreen, Spitsbergen, Quaternary Science Reviews, 18, , Boyce J.I. and Eyles N., Architectural element analysis applied to glacial deposits: internal geometry of a late Pleistocene till sheet, Ontario, Canada, Geological Society of America Bulletin, 112, , Cofaigh, C.Ó., and Evans D.J.A., Sedimentary evidence for deforming bed conditions associated with a grounded Irish Sea, glacier, southern Ireland, Journal of Quaternary Science, 16, , Croot, D.G., Glacio-tectonic structures: A mesoscale model of thin-skinned thrust sheets?, Journal of Structural Geology, 9, , Croot, D.G., Glaciotectonics: Forms and processes, 212 p., A.A. Balkema, Rotterdam, Dahlen, F. A., Suppe, J., and Davis, D., Mechanics of fold-and-thrust belts and accretionary wedges: Cohesive Coulomb theory, Journal of Geophysical Research, 89, 10,087-10,101, Davis D.M., Winslow III, F.D., Klein, E.C., and Holt, W.E., Contrasting radar signals in sand and till: an example from Grandifolia Sandhills, eastern LI, Conf. On Geology of Long Island and Metropolitan NY, Program_4_00.htm (April 2000). Davis, D., Suppe, J., and Dahlen, F.A., Mechanics of fold-and-thrust belts and accretionary wedges, Journal of Geophysical Research, 88, , Davis, D., and Engelder, R., The role of salt in fold & thrust belts, Tectonophysics, 119, 67-88, Dell Isola, F., Hutter, K., What are the dominant thermomechanical processes in the basal sediment layer of large ice sheets? The Royal Societyof London Proceedings: Mathematical, Physical and Engineering Sciences,454, , Dowdeswell, J.A., Hambrey, M.J., and Wu, R., A comparison of clast fabric and shape in late Precambrian and modern glacigenic sediments, Journal of Sedimentary Petrology, 55, ,

46 Dowdeswell, J.A., and Sharp, M.J. Characterization of pebble fabrics in modern terrestrial glacigenic sediments, Sedimentology, 33, , 1986 Dreimanis, A., Tills: their genetic terminology and classification. In: Goldthwait, R.P., Matsch, C.L. (Eds.), Genetic Classification of Glacigenic Deposits, Balkema, Rotterdam, 17-83, Dreimanis, A., Need for three-dimensional analysis of structural elements in glacial deposits for determination of direction of glacier movement, Geological Society of America, Special Paper 337, 59-67, Dyke, A.S., Andrews, J.T., Clark, P.U., England, J.H., Miller, G.H., Shaw, J., and Veillette, J.J., The Laurentide and Innuitian ice sheets during the Last Glacial Maximum, Quaternary Science Reviews, 21, 9-31, Eden, D.J., and Eyles, N., Description and numerical model of Pleistocene iceberg scours and ice-keel turbated facies at Toronto, Canada, Sedimentology, 48, , Ékes, C., and Hickin, E.J., Ground penetrating radar facies of the paraglacial Cheekye Fan, southwestern British Columbia, Canada, Sedimentary Geology, 143, , Etzelmüller, B., Hagen, J.O., Vatne, G., Ødegård, R.S., and Sollid, J.L., Glacier debris accumulation and sediment deformation influenced by permafrost: examples from Slavbard, Annals of Glaciology, 22, 53-62, Evans, D.J.A., Quaternary geology and geomorphology of the Dinosaur Provincial Park area and surrounding plains, Alberta, Canada: the identification of former glacial lobes, drainage diversions and meltwater flood tracks, Quaternary Science Reviews, 19, , Evans, D.J.A., Lemmen, D.S., and Rea, B.R., Glacial landsystems of the southwest Laurentide Ice Sheet: modern Icelandic analogues, Journal of Quaternary Science, 14, , Eybergen, F.A., 1986, Glacier snout dynamics and contemporary push moraine formation at the Turtmannglacier, Wallis, Switzerland. In: Van der Meer, J.J.M., (Ed.), Tills and Glaciotectonics. Balkema, Rotterdam, , Fullerton, D.S. and 9 others, Quaternary Map of the Hudson River 4 x 6 Quadrangle, United States and Canada, Map I-1420, U.S. Department of the interior, U.S.G.S., Gerber, R.E., Boyce, J.I., and Howard, K.W.F., Evaluation of heterogeneity and field-scale groundwater flow regime in a leaky till aquitard, Hydrogeology Journal, 9, 60-78, Ham, N.R., and Mickelson, D.M., Basal till fabric and deposition at Burroughs Glacier, Glacier Bay, Alaska, Geological Society of America Bulletin, 106, , Hambrey, M.J., Davies, J.R., Glasser N.F., Waters, R.A., Dowdeswell, J.A., Wilby, P.R., Wilson D., and Etienne, J.L., Devensian glacigenic sedimentation and landscape evolution in the Cardigan area of southeast Wales, Journal of Quaternary Science, 16, , Hambrey, M.J., and D. Huddart, Englacial and proglacial glaciotectonic processes at the snout of a thermally complex glacier in Svalbard, Journal of Quaternary Science, 10, , Hansen, L.A., Jensen, J.B., and Mads, T., Pleistocene sedimentary record of the Falsterselv area, Jameson Land, East Greenland, Geology of Greenland Survey Bulletin, 176, 84-36

47 88, Harris, C., Williams, G., Brabham, P., Eaton, G., and McCarroll, D., Glaciotectonized quaternary sediments at Dinas Dinlle, Gwynedd, North Wales, and their bearing on the style of deglaciation in the Eastern Irish sea, Quaternary Science Reviews, 16, , Hart, J.K., Proglacial glaciotectonic deformation and the origin of the Cromer Ridge push moraine complex, North Norfolk, England, Boreas, 19, , Hart, J.K., Till fabric associated with deformable beds, Earth Surface Processes and Landforms, 19, 15-32, Hart, J.K., The relationship between drumlins and other forms of subglacial glaciotectonic deformation, Quaternary Science Reviews, 16, , Hart, J.K., The deforming bed/debris-rich basal ice continuum and its implications for the formation of glacial landforms (flutes) and sediments (melt-out till), Quaternary Science Reviews, 17, , Hart, J.K., and G.S. Boulton, The interrelation of glaciotectonic and glaciodepositional processes within the glacial environment, Quaternary Science Reviews, 10, , Hart, J, and Rose, J., Approaches to the study of glacier bed deformation, Quaternary International, 86, 45-58, Hart, J.K., and Smith, B., Subglacial deformation associated with fast ice flow, from the Columbia Glacier, Alaska, Sedimentary Geology, 111, , Hart, J.K., and Watts, Robert J., A comparison of the styles of deformation associated with two recent push moraines, south Van Keulenfjorden, Svalbard, Earth Surface Processes and Landforms, 22, , Henriksen, M., Mangerud, Jan., Maslenikova, O., Matiouchkov, A., and Tveranger, J., Weichselian stratigraphy and glaciotectonic deformation along the lower Pechora River, Arctic Russia, Global and Planetary Change, 31, , Hicock, S.R., Genetic till prism, Geology, 18, , Hicock, S.R., Lobal interactions and rheologic superposition in subglacial till near Bradtville, Ontario, Canada, Boreas, 73-88, Hicock, S.R., and Dreimanis, A., Deformation till in the Great-Lakes region - implications for rapid flow along the south-central margin of the Laurentide Ice Sheet, Canadian Journal of Earth Sciences, 29, , Hicock, S.R., Goff, J.R., Lian, O.B., and Little, E.C., On the interpretation of subglacial till fabric, Journal of Sedimentary Research, 66, , Hicock, S.R., and Lian, O.B., Cordilleran Ice Sheet lobal interactions and glaciotectonic superposition through stadial maxima along a mountain front in southwestern British Columbia, Canada, Boreas, 28, , Hsui, A.T., Wilkerson, S.M., and Marshak, S., Topographically driven crustal flow and its implication to the development of pinned oroclines, Geophysical Research Letters, 17, , Houmark-Nielsen, M., Demidov, I., Funder, S., Grøsfjeld, K., Kjær, K.H., Larsen, E., Lavrova, N., Lyså, A., and Nielsen J. K., Early and Middle Valdaian glaciations, icedammed lakes and periglacial interstadials in northwest Russia: new evidence from the Pyoza River area, Global and Planetary Change, 31, ,

48 Johnson, M.D., and Gillam, M.L., Composition and construction of late Pleistocene end moraines, Durango, Colorado, Geological Society of America Bulletin, 107, , Johnson, W.H., and Hansel, A.K., Wisconsin episode glacial landscape of central Illinois: a product of subglacial deformation processes?, Geological Society of America, Special Paper 337, , Karlstrom, E.T., Fabric and origin of multiple diamictons within the pre-illinoian Kennedy Drift east of Waterton-Glacier International Peace Park, Alberta Canada, and Montana, USA, Geological Society of America Bulletin, 112, , Khatwa, A., and Tulaczyk, S., Microstructural interpretations of modern and Pleistocene subglacially deformed sediments: the relative role of parent material and subglacial processes, Journal of Quaternary Science, 16, , Kjær, K.H., Demidov, I., Houmark-Nielsen, M., and Larsen, E., Distinguishing between tills from Valdaian ice sheets in the Arkhangelsk region, Northwest Russia, Global and Planetary Change, 31, , Kjær, K.H., and Krüger, J., Does clast size influence fabric strength?, Journal of Sedimentary Petrology, 68, , Kjær, K.H., and Krüger, J., The final phase of dead-ice moraine development: processes and sediment architecture, Kötlujökull, Iceland, Sedimentology, 48, , Klein, E. C., and Davis, Dan M., Glaciotectonic process and glacigenic sediments on eastern Long Island, in Geology of Long Island and Metropolitan New York, SUNY Stony Brook, p , Krüger, J., and Kjær, K.H., A data chart for field description and genetic interpretation of glacial diamicts and associated sediments with examples from Greenland, Iceland, and Denmark, Boreas, 28, , Lachniet, M.S., Larson, G.J., Strasser, J.C., Lawson, D.E., Evenson, E.B., and Alley, R.B., Geological Society of America, Special Paper 337, 45-57, Larsen, E., Lyså, A., Demidov, I., Funder, S., Houmark-Nielsen, M., Kjær, K.H., and Murray, A.S., Age and extent of the Scandinavian ice sheet in northwest Russia, Boreas, 28, , Lawson, D.E., A comparison of the pebble orientations in and deposits of the Matanuska Glacier, Alaska, Journal of Geology, 87, , Lawson, D.E., Mobilization, movement and deposition of active subaerial sediment flows, Matanuska Glacier, Alaska, Journal of Geology, 90, , Lehmann, R., The significance of permafrost in the formation and appearance of push moraines, Cheng Guodong (chairperson), Permafrost; Sixth international conference proceedings, International Conference on Permafrost, Proceedings, 6, Vol. 1, p , 1993., Beijing, China, July 5-9, Lewis, R.S., and Stone, J.R., Late Quaternary stratigraphy and depositional history of the Long Island Sound Basin: Connecticut and New York, Journal of Coastal Research, 1-23, Mark, D.A., Analysis of axial orientation data, including till fabrics, Geological Society of America Bulletin, 84, , Mark, D.A., On the interpretation of till fabrics, Geology, 2, , Mattsson, Å., Glacial striae, glacigenous sediments and Weichselian ice movements in 38

49 southernmost Sweden, Sedimentary Geology, 111, , Menzies, J., Zaniewski, K., and Dreger, D., Evidence, from microstructures, of deformable bed conditions within drumlins, Chimney Bluffs, New York State, Sedimentary Geology, 111, , Merrill, F.J.H., On the geology of Long Island, Annals New York Academy of Sciences, 3, , Meyers, W.J., Boguslavsky, S., Dunne, S., Keller, J., Lewitt, D., Lani, M., McVicker, A., and Cascione, Matituck Cliffs and Ranco Quarry: Models for Origin of Roanoke Point and Ronkonkoma Moraines?, Conf. On Geology of Long Island and Metropolitan NY, 83-90, April Millar, S.W.S., and Nelson, F.E., Sampling-surface orientation and clast macrofabric in periglacial colluvium, Earth Surface Processes and Landforms, 26, , 2001a. Millar, S.W.S., and Nelson, F.E., Clast fabric in relict periglacial colluvium, Salamanca reentrant, southwestern New York, USA, Geografiska Annaler, 83A, , 2001b. Munro-Stasiuk, M.J., Rhythmic till sedimentation: evidence for repeated hydraulic lifting of a stagnant ice mass, Journal of Sedimentary Research, 70, , Nieter, W., B. Nemickas, E.J. Koszalka, and W.S. Newman, The Late Quaternary Geology of the Montauk Peninsula, Montauk Point to Southampton, Long Island, New York, p , in New York State Geological Association Guidebook, 47th Annual Meeting, Overgaard, T., and Jakobsen, P.R., Mapping of glaciotectonic deformation in an ice marginal environment with ground penetrating radar, Journal of Applied Geophysics, 47, , Penttinen, S., Sutinen, R, and Hänninen, P., Determination of Anisotropy of tills by means of azimuthal resistivity and conductivity measurements, Nordic Hydrology, 30, , Piotrowski, J.A., Mickelson, D.M., Tulaczyk, S., Krzyszkowski, D., and Junge, F.W., Were deforming subglacial beds beneath past ice sheets really widespread?, Quaternary International, 86, , Ramage, J.M., Gardner, T.W., and Sasowsky I.D., Early Pleistocene Glacial Lake Lesley, West Branch Susquehanna River valley, central Pennsylvania, Geomorphology, 22, 19-37, Rappol, M., Clast-fabric strength in tills and debris flows compared for different environments, Geologie en Mijnbouw, 64, , Regli, C., Huggenberger, P., and Rauber, M., Interpretation of drill core and georadar data of coarse gravel deposits, Journal of Hydrology, 255, , Rijsdijk, K.F., Density-driven deformation structures in glacigenic consolidated diamicts: examples from Traeth Y Mwnt, Cardiganshire, Wales, U.K., Journal of Sedimentary Research, 71, , Ringrose, T.J., and Benn, D.I., Confidence regions for fabric shape diagrams, Journal of Structural Geology, 12, , Russell, A. J., Knight, P. G., and Van Dijk, T. A. G. P., Glacier surging as a control on the development of proglacial, fluvial landforms and deposits, Skei arársandur, Iceland, Global and Planetary Change, 28, , Ruszczynska-Szenajch, Lodgement till and deformation till, Quaternary Science 39

50 Reviews, 20, , Salem H.S., Modeling of lithology and hydraulic conductivity of shallow sediments from resistivity measurements using schlumberger vertical electrical soundings, Energy Sources, 23, , 2001 Schlüchter, C., Gander, P., Lowell, T.V., and Denton, G.H., Glacially folded outwash near Lago Llanquihue, Southern Lake District, Chile, Geografiska Annaler, 81A, , Sharp, M., Annual moraine ridges at Skálafellsjökull, south-east Iceland, Journal of Glaciology, 30, 82-93, Sminchak, J.R., Dominic, D.F., Ritzi, R.W.Jr., Indicator geostatiscal analysis of sand interconnections within a till, Ground Water, 34, , Sterner, R., Hudson Canyon, hudson_canyon.html (1994). Thomas, G.S.P., The origin of the glacio-dynamic structure of the Bride Moraine, Isle of Man, Boreas, 13, , van der Meer, J.J.M., Microscopic evidence of subglacial deformation, Quaternary Science Reviews, 12, , van der Meer, J.J.M., Short-lived streamlined bedforms (annual small flutes) formed under clean ice, Turtmann Glacier, Switzerland, Sedimentary Geology, 111, , van der Meer, J.J.M., Particle and aggregate mobility in till: microscopic evidence of subglacial processes, Quaternary Science Reviews, 16, , 1997b. Van der Wateren, D.F.M., A model of glacial tectonics, applied to the ice-pushed ridges in the Central Netherlands, Bulletin of the Geological Society of Denmark, 34, 55-74, Van der Wateren, D.F.M., 1986, Structural geology and sedimentology of the Dammer Berge push moraine, FRG, INQUA symposium on the Genesis and Lithology of Glacial Deposits, Amsterdam, In: Tills and Glaciotectonics, A.A. Balkema, Rotterdam, , Van der Wateren, D.F.M., Structural geology and sedimentology of push moraines: processes of soft sediment deformation in a glacial environment and the distribution of glaciotectonic styles, Mededelingen Rijks Geologische, 54, 1 168, Van der Wateren, D.F.M., Structural geology and sedimentology of the Heiligenhafen till section, Northern Germany, Quaternary Science Reviews, 18, , Wang, W.-H. and Davis, D.M., Sandbox model simulation of forearc evolution and noncritical wedges, Journal of Geophysical Research,101, , Ward, B., and Rutter, N., Deglacial valley fill sedimentation, Pelly river, Yukon territory, Canada, Quaternary International, 68-71, , Watson, G.S., The statistics of orientation data, Journal of Geology, 74, , Williams, G.D., Brabham, P.J., Eaton, G.P., and Harris, C., Late Devensian glaciotectonic deformation at St. Bees, Cumbria: a critical wedge model, Journal of the Geological Society of London, 158 part 1, , Woodcock, N.H., Specification of fabric shapes using eigenvalue method, Geological Society of America Bulletin, 88, , Woodcock, N.H., and Naylor, M.A., Randomness testing in three-dimensional orientation data, Journal of Structural Geology, 5, ,

51 Zelcs, ˇ V., and Dreimanis, A., Morphology, internal structure and genesis of the Burtnieks drumlin field, Northern Vidzeme, Latvia, Sedimentary Geology, 111, 73-90,

52 II. Surface Sample Bias and Clast Fabric Interpretation Abstract Quantitative clast fabric analysis, despite its limitations, is a useful analytical tool in glacigenic sediment studies. More powerful than graphical methods, eigenanalysis allows quantification of otherwise descriptive three-dimensional fabrics. In conjunction with the orientation of the long-axes, short-axes preferred direction could further establish the nature of shear, emplacement, and deposition in glacigenic settings. Field measurement, however, produces a systematic sampling bias in favor of clasts normal to outcrop surface. This sampling bias is a function of the orientation of outcrop surface to the fabric and can affect the perceived fabric strength (eigenvalues) enough to influence interpretation. I use simple calculations and numerically generated random clast populations to quantify this bias and I find that it is greatest for those clasts best suited to fabric analysis (those that are rod-like in shape). The effect of this surface bias can be mitigated with careful sampling and interpretation. Fortunately, its effect upon strong fabric orientation (eigenvectors) is generally small. Clast Fabric Analysis in Glacial Sediment Glacial deposits typically incorporate a very broad range of particle sizes that reflect the source and quantity of sediment supplied to the glacier. One way glacial diamict can evolve is when glacially entrained rock, bedrock fragments, and sediments are crushed, ground, deformed and mixed while being carried in the ice and particle rich basal zone of a glacier, producing deposit types which include a heterogeneous (clay to boulder) mixture of poorly sorted particles (e.g., Alley et al., 1999). Prolonged glacial transport modifies the textural maturity of glacial sediments, so clast roundness and sphericity are functions of the distance the clast has been transported (e.g. Boulton, 1978; Boulton, 1996). Moraine landforms are often composed of relatively large volumes of glacial diamict deposits. In these settings, diamict units can be massive, stratified, laminated, imbricated, reworked, resedimented, sheared, and they can include clast pavements. Diamicts include either isolated or graded clasts within a matrix of finer material or an interstitial matrix of finer material within supporting clasts (e.g. Benn and Evans, 1998; Krüger and Kjaer, 1999). A clast fabric defines a distribution of clasts by the degree to which clast axes orientations are clustered. Such fabrics range from very weak (nearly isotropic) to very strong (highly 42

53 directional). The manner in which glacigenic clasts are arranged provides outcrop-scale information that may help to constrain the physical dimensions of stratigraphic units and also provides clues to the nature and sequence of glaciotectonic and depositional processes (e.g., Hicock, 1990). A quantitative description of a clast fabric, defining the degree of preferred orientation observed in exposed diamict settings, is obtained by a combination of field measurement and vector analysis (e.g., Mark, 1974). The statistical description of a glacigenic sediment clast fabric reflects the effect of both depositional conditions and the subsequent strain, in principle enabling the differentiation of the structural contrasts recorded in glaciotectonic shear zones. Published research (e.g., Lawson, 1979; Dowdeswell et al., 1985; Dowdeswell and Sharp, 1986) on a variety of glacial diamict deposits has shown a broad range of preferred clast orientations, producing fabrics of strengths that differ with depositional setting. The glacigenic settings which produce these differing results range from weakly organized clast orientations (e.g., water-lain glacigenic sediment) preserved during outwash deposition to highly uniform lineations (e.g., subglacial melt-out till) generated by well-developed shear zones that are later preserved during subglacial melt-out deposition from motionless glacial ice. The magnitude of clast orientation preference is a measure of the degree to which all preserved and observed clasts measured are aligned subparallel to a unique direction. Strongly organized, highly anisotropic clast orientations derived from subglacial-meltout or lodgment processes are typically subparallel with local ice flow direction (e.g., Lawson, 1979; Dreimanis, 1999). Fabric strength generally decreases as water content in the sediment-ice mixture increases. Clast fabric is not uniquely indicative of sediment genesis, but can be an indicator of relative strain within a stratigraphic unit (e.g., Bennett, et al. 1999; Karlstom, 2000). I have studied glacigenic sediments at a site within the Ronkonkoma Moraine of Long Island, New York. My field area, Ditch Plains, roughly 5 kilometers west of Montauk Point on the south fork of eastern Long Island (Fig. 1), is spaced along several kilometers of the shoreline. Bluffs and vertical exposures outcrop at heights as great as 10 to 15 meters, as the Ronkonkoma Moraine resists erosion caused by ocean surf and other weathering processes. I interpret the general stratigraphy to include two distinctive glacial diamict deposits. A massive upper diamict unit (Dmm) overlies a stratified diamict unit (Dms) (Fig. 2), (e.g., Klein, et. al., 1998, 2001). The preferred direction of the long axes for the rodshaped clasts (subhorizontal N-S direction) is evident within the stratified diamict unit. The north-south long axis orientation of most clasts in this unit is consistent with shear due to glacial advance from the north. One way to distinguish differences in fabric strength is with a rose diagram. This graphical method plots frequency distribution versus bearing. The number of clasts plotted for a given range of bearings corresponds simply to the number of clasts with a particular axis (e.g., long) pointed within a narrow range of bearing (commonly ±5º) of that direction. A major drawback in using rose diagrams for fabric descriptions is their inability to differentiate between shallowly and steeply plunging clast axes, which limits the usefulness of physical interpretation based on rose diagrams alone. Plotted on a rose diagram are the long axis bearings of 150 clast orientations collected from the stratified diamict unit at Ditch Plains (Fig. 3A). The rose petal in figure 3A, pointing to the northeast between 340º to 350º defines the bearing direction for 23% of the measured long axes. 43

54 44 Figure 1. Map of Long Island. Shaded in green are the Harbor Hill and Ronkonkoma moraines. Ditch Plains field study location is indicated by the small red dot on the south fork.

55 1 m Dmm massive diamict (flow till?) Dms stratified diamict (melt-out till?) Covered Figure 2. The south-facing sea cliff at Ditch Plains (Figure 1). Most of the 10 m exposure consists of a stratified diamict, tentatively interperted as a melt-out till. The top part of the exposure is a massive diamict which is probably a flow till. Insert is a typical 1 m square clast orientation sampling area (red square). 45

56 A N B C Figure 3. A. A rose diagram showing the 150 long axis clast orientations recorded in the stratified diamict unit (Dms) at Ditch Plains. The rose diagram perimeter corresponds to 23%. B. Equal-area stereonet projection of the same 150 clast orientations C. Equal-area stereonet projection of the same 150 clast orientations contoured at 2% intervals for 1% area. The contour plot of these clast orientations unambiguously suggests a subhorizontal preferred direction just west of north. Note, north is located at the top of each of the diagrams in the figure. 46

57 Equal-area stereonet projection is a more commonly used graphical method that expresses three-dimensional clustering in direction of a specified clast axis. In this projection, each point represents the bearing and plunge of one clast. In figures 3B and 3C, I present the same clast data from the lower stratified diamict unit projected and contoured on equalarea stereonets. The 150 long axes cluster at shallow plunge near a bearing of 345. Unlike rose diagrams, stereonets are capable of graphically establishing a three-dimensional maximum preferred orientation for a single axis (e.g. long-axis) distribution where one might also be able to infer the least and intermediate preferred directions. Eigenanalysis, a more formal quantitative technique, can statistically establish three mutually orthogonal (most, intermediate, and least) preferred directions for a single axial data set, but its advantage is greater resolution and more precision than stereonet projection. In clast fabric analysis, each individual clast may be approximated by a triaxial ellipsoid with three mutually orthogonal principal axes. It is convenient to determine such principal axes for each clast by first measuring the length and orientation of the longest axis through the clast center. The lengths and orientations of two mutually perpendicular and orthogonal axes to the established long axis orientation can then be established and recorded. I record the orientation (bearing and plunge) of any two of these principal axes, from which I can calculate the third. The bulk fabric information gleaned from in-situ measurement of axial orientations for a given number of clasts (vectors) in a clast distribution is transformable into a three-dimensional orientation tensor (e.g., Mark, 1973; Woodcock, 1977). The elements of the orientation tensor, defined in a 3 x 3 symmetric matrix, indicate the degree to which a particular clast axis (this technique is usually applied to the long axis) tends to align in a given direction. Clast shape can play a significant role in the determination of a clast orientation. Rod-like clast shapes, however, have an obvious long axis orientation, allowing them to contribute to a more robust orientation tensor. The analytical solution obtained by eigenanalysis determines three normalized eigenvalues (a maximum, an intermediate, and a minimum) and assigns one eigenvalue to each one of three mutually perpendicular preferred eigenvector directions. These eigenvectors can be thought of as the most, intermediate, and least preferred clast-vector directions (all mutually perpendicular). The corresponding normalized eigenvalues describe the degree of preference for each of these directions. For example, a normalized eigenvalue equal to one would mean that all axes point exactly that way, while an eigenvalue of zero would indicate that they are all 90 from that direction. Three-dimensional eigenvalue analysis in glacially-derived diamicts has been used by a number of researchers studying clast fabrics as a diagnostic tool intended to deduce the genetic origin of glacial deposits (e.g., Mark, 1974, Lawson 1979; Dowdeswell et al., 1985; Rappol, 1985; Dowdeswell and Sharp, 1986; Benn, 1994, Ham and Mickelson, 1994; Hicock et al., 1996; Larsen et al., 1999; Kjaer et al., 2001) and to infer relative strain within glacigenic sediments (e.g., Hicock, 1992; Hart, 1994; Benn, 1995; Benn and Evans, 1996; Rijsdijk, 2001). Clast shapes affect the nature and strength of the clast orientation fabric. If a is the long axis, b the intermediate axis, and c is the short axis for a particular clast, then the clast shape is defined as a rod if a>b c, a disc if a b>c, a blade if a>b>c, or a spheroid if a b c (e.g., Zingg, 1935; Sneed and Folk, 1958). Spheres have infinite combinations of three mutually perpendicular axes so spheroidal clasts make a very weak fabric, have no preferred directions, and indicate little about emplacement or strain. Glacigenic deposits composed exclusively of sediment of extreme 47

58 textural maturity are therefore the poorest targets for clast fabric analysis. Rod-like clasts are often the easiest shapes in the field to measure, so they are the most ideal for the determining long axis orientation. A more complete description of clast fabric includes the clast short-axis directions since it is equally possible to study the distribution of long or short axes, although the literature emphasizes long axis orientations, particularly for blunt rods (aspect ratio a/b of at least 3/2) (e.g., Bennett et al., 1999; Kjaer and Krüger, 1998; Krüger and Kjaer, 1999; Millar and Nelson, 2001a,b). To this date there is limited reporting of measured short axis orientation distributions in glacigenic sediment studies. Combining long axis and short axis orientations of the same set of clasts may permit clearer discernment of the nature of shear responsible for emplacement of the clasts. A clast fabric describing short axis orientations for a set of blade-like (triaxial) clasts (a>b>c) may help in identifying pure shear uniaxial shortening. For example, if a randomly oriented clast set is subjected to pure uniaxial compression, then the clasts long axes will tend to girdle in orientations radially normal to the compression direction and the clasts short axes orientations will be apt to cluster subparallel to the compression direction. For simple shear, long axes tend to cluster in the shear direction (e.g., Dreimanis, 1999), although some studies describe quasiperoidic behavior (e.g., Lindsay, 1968). The short axis direction of the clasts measured at Ditch Plains have a great propensity for being aligned nearly vertical. This, like the long axis orientation, is consistent with emplacement and subsequent shear from the north due to glacial advance. The short axes of 141 of the 150 Ditch Plains clasts cluster on a stereonet in a near-vertical orientation more tightly aligned than the long axes (Fig. 4). The distribution of the long and short axial orientations of the Ditch Plains clasts is suggestive of a well-developed fabric, enhanced by considerable shear. The orientation of the long axis of a single clast is usually described in terms of a vector with a given bearing and plunge. It is also possible, however, to describe that vector in terms ofthe angle it makes with respect to each of the three coordinate axes (x, y, z) or (E, N, vertical). The three direction cosines (m, l, n) are simply the cosines of the angles from that pebble axis vector to the positive x, y, and z-axis directions, respectively. They can also be thought of as the x, y, and z components of a unit vector pointing in the same direction as the pebble axis, with values of +1 or -1 (if pointing along a given axis) and 0 if perpendicular to that axis. Following the example of other workers, (e.g., Mark, 1973; Woodcook, 1977), I use the direction cosines (l, m, n) of measured pebble axis vectors to produce a symmetric matrix that describes the direction and intensity of the overall pebble fabric. This matrix, called the orientation tensor, is the sum over all N pebbles of sets of products of the direction cosines. It is written formally as A = 1 N [ ] Σl i 2 Σm i l i Σn i l i Σl i m i 2 Σl i n i Σm i Σm i n i Σn i m i Σn 2 i (1) The contribution of a single clast to the xx-component (a xx ) is simply the first 48

59 A B Figure 4. A. Equal-area stereonet projection of 141 short axis clast orientations recorded in the stratified diamict unit at Ditch Plains. B. Equal-area stereonet projection of the same 141 clast orientations contoured at 2% intervals for 1% area. Note, north is located at the top of each of the steronet diagrams in the figure. 49

60 direction cosine of the axis (usually the long axis being considered), l, multiplied by itself. Similarly, the other components along the diagonal (a yy and a zz ) are squares of those direction cosines. The off-diagonal terms are simply products of the corresponding pairs of direction cosines. For example, a xy and a yx are each the products of the x- and y-axis direction cosines, l, and m. For this reason, the matrix is symmetric (a xy = a yx, a xz = a zx, and a yz = a zy ). Such a matrix can be written for an individual clast. By summing the values of each matrix component over all N measured clasts (the subscript i is a counter for individual clasts) and then normalizing by that value N, one arrives at the normalized orientation matrix A (e.g., Mark, 1973; Woodcock, 1977). This matrix contains information about the statistics of clast axes the frequency to which they point in any given direction. The normalized orientation matrix A is written with respect to the x-, y-, and z-axes. The mutually orthogonal set of axes corresponding to the fabric (the directions in which clasts most and least strongly point) can be oriented at any arbitrary angles to that (x, y, z) coordinate frame. Ideally, I would like to rewrite the normalized orientation matrix with respect to a reference frame that has a physical meaning for the clast fabric, rather than the arbitrary (x, y, z) axis reference frame. The diagonalization of the matrix A does just that. The eigenvectors of the matrix are simply vectors indicating the mutually orthogonal axes that define that natural reference frame. One of these axes, written as vector V 1, is the most preferred direction for clast axes. The degree of that preference is given by a scalar value the greatest eigenvalue S 1. The least preferred direction, V 3, is an eigenvector that is normal to the V 1 axis, and its corresponding eigenvalue is called S 3. A third direction, V 2, is perpendicular to both the V 1 and V 3 directions and has an eigenvalue S 2 that indicates the relative propensity of clasts to align in that direction. If, as in my calculations, the eigenvalues are normalized to one, I have S 1 +S 2 +S 3 =1, with S 1 S 2 S 3. When the matrix has been rewritten in diagonalized form, in which it is defined with respect to the fabric axes (eigenvectors), the off-diagonal terms are all zero. The three remaining non-zero terms, along the diagonal, then correspond to the three eigenvalues. Fabric strengths derived from eigenvalue analysis in the literature have been depicted graphically in two distinct ways. Both (S 1,S 3 ) eigenvalue plots (Fig. 5A) and isotropyelongation ternary plots (Fig. 5B) encompass the entire range of possible sets of eigenvalues, defined by the relations S 1 +S 2 +S 3 =0 and 0 S 3 S 2 S 1 1 (e.g., Benn, 1994). The two types of plots can be 1:1 mapped onto each other. In an (S 1,S 3 ) eigenvalue plot (Fig. 5A), the horizontal axis corresponds to the magnitude of the largest eigenvalue (S 1 ) and the vertical axis indicates the smallest eigenvalue (S 3 ). Each point on the plot corresponds to a unique (S 1,S 2,S 3 ) set, since the third eigenvalue S 2 =1-S 1 -S 3 is uniquely determined in terms of the other two. All possible (S 1,S 2,S 3 ) combinations are contained within a skewed triangular region, bounded on the bottom by the S 3 =0 line, at left by the S 2 =S 1 line and at top by the S 2 =S 3 line. In an isotropy-elongation ternary diagram (Fig. 5B) the set of all fabric eigenvalues is plotted within an equilateral triangular region, as in any other ternary diagram. The vertical (isotropy) axis measures the ratio (S 3 /S 1 ). When this ratio equals one, all three eigenvalues are equal and the fabric is purely isotropic. When the isotropy (S 3 /S 1 ) equals zero, 0=S 3 S 2 S 1, and the fabric lies somewhere between a girdle (S 3 =0 ; S 2 =S 1 =.5) and a cluster (S 3 =S 2 =0 ; S 1 =1.0). The elongation axis, 60 clockwise of the isotropy axis, measures the value 1-(S 2 /S 1 ). The third axis value in a ternary diagram, in this case (S 2 -S 3 )/S 1, is 50

61 A isotropi pic S 3.20 S 2 = S 3 = S 1 girdle cluster B 1.0 isotropic 0 Isotropy (S /S ) Elongation 1-(S /S ) girdle 1.0 cluster Figure 5. A. (S 1, S 3 ) eigenvalue plot. Note that possible eigenvalue combinations can fall only within the triangular area of the plot bounded by the lines S 1 =S 2, S 2 =S 3, and S 3 = 0.0 with S 1 eigenvalues ranging from 0.50 to 1.0 respectively. B. Isotropy-elongation ternary diagram. Ideal fabric shapes (isotropic, girdle, cluster) indicated at the apices of the useable triangle in the (S 1, S 3 ) eigenvalue plot and in the isotropy-elongation ternary diagram. 51

62 redundant because the three coordinate values sum to one and are thus not mutually independent. Although the mapping between the two diagrams is 1:1, their relative scaling is not uniform. The eigenvalue plot stretches the lower right-hand (cluster) area of the plot (Fig. 6A,B), so that the extreme corner of the plot (where S 1 1) is scaled up in area by a factor of 4.5:1 compared to the ternary plot. Likewise, the top of the ternary diagram exaggerates the near-isotropic region (where S 3 S 1 ) by a factor of 6:1 compared to the eigenvalue plot. Thus, equal-area regions near the isotropic and cluster extremes of one of these diagrams will appear on the other diagram to encompass corresponding regions of greatly different size, differing by a factor of up to 27 at the far corners of the diagrams. Although this difference in scaling is less extreme for larger regions that extend away from the corners, it remains significant. For example, the ternary diagram can be divided into four equal areas, corresponding to sets of eigenvalues that tend toward being roughly isotropic, clustered, or girdle-like, plus those that fall in between (Fig. 7A). Plotted on an eigenvalue diagram, however, those four regions are skewed and far from equal in area (Fig. 7B). The nearly isotropic 20% of the ternary diagram (shaded dark in Fig. 7A) occupies only 1/12 th of the total area on the eigenvalue diagram (5 times smaller than the nearly-clustered region, which also covers 20% of the ternary diagram). Thus it may not be surprising that relatively few published observations fall in the near-isotropic region. Also, existence of naturally occurring truly isotropic fabrics is thought to be extremely rare due to the influence of depositional or mechanical boundary conditions such as surface and stress field orientations that often promote particle alignment (e.g., Benn and Ringrose, 2001). Particular care must be exercised in evaluating data points that plot close to the boundaries in eigenvalue diagrams, where at least two eigenvalues are nearly equal in strength. Such a condition can lead to misinterpretation of the preferred orientation of a clast fabric. Random sampling can easily cause nearly equal eigenvalues to reverse their magnitude order, leading to a dramatic and spurious change (one axis for another) in eigenvector direction (e.g., Ringrose and Benn, 1997; Benn and Ringrose, 2001). In fact, as I will show, there is an additional reason. There is anobservational bias that tends to produce spurious measurements away from this region even when the true clast fabric is indeed quite isotropic. Various authors have attempted to associate fabric domains in (S 1, S 3 ) space with the mode of genesis of glacial till in modern glacigenic sediments, although in relatively few cases are there such data where sedimentary processes are unambiguously observed (e.g., Lawson, 1979; Dowdeswell et al., 1985). The Ditch Plains sediment orientation data give well-determined eigenvalues and eigenvectors (Table 1) that can be placed on an S 3 versus S 1 eigenvalue plot. long-axis eigenvalues long-axis eigenvectors short-axis eigenvalues short-axis eigenvectors S 1 =.792 V 1 = (342, 06 ) S 1 =.911 V 1 = (003, 73 ) S 2 =.150 V 2 = (072, 01 ) S 2 =.051 V 2 = (260, 04 ) S 3 =.058 V 3 = (178, 84 ) S 3 =.038 V 3 = (169, 16 ) Table 1. Eigenanalysis results for long and short axis clast orientations recorded at Ditch Plains. 52

63 A isotropic Isotropy (S /S ) Elongation 1-(S /S ) girdle 1.0 cluster B isotropic S S 1.60 girdle cluster Figure 6. Illustration of 1:1 mapping and non uniform scaling between the (S 1, S 3 ) eigenvalue plot and the isotropy-elongation ternary diagram. A. The useable triangular domain of the (S 1, S 3 ) eigenvalue plot mapped to and labeled with isotropy-elongation ternary diagram eigenvalue ratio combinations. B. Isotropy-elongation ternary diagram mapped to and labeled with (S 1, S 3 ) eigenvalue plot eigenvalue ratio combinations. Ideal fabric shapes (isotropic, girdle, cluster) indicated at the apices of both diagrams. 53

64 A isotropic Isotropy girdle 1.0 B isotropi pic S 3.20 S 2 = S S 1 girdle cluster Figure 7. A. Isotropy-elongation ternary diagram divided into four equal area regions that correspond to general fabric shape domains. B. The equal area general fabric shape domains of the isotropy-elongation ternary diagram are skewed substantially in area on the (S 1, S 3 ) eigenvalue plot. 54

65 The Ditch Plains long axis stratified diamict clast fabric falls into the lower righthand region of the plot, near where Lawson (1979) and Dowdeswell et al. (1985) plot sediments with strong fabrics, such as lodgement and subglacial melt-out tills (Fig. 8). The result of calculated by eigenanalysis is consistent with the rose diagram (Fig. 3A) and the long-axis stereonet plots diagram (Fig. 3B,C) depicting the majority of the measured pebbles aligned subhorizontally with north-south direction. Each of the fabric domains described by an ellipse on an eigenvalue plot (S 3 versus S 1 ) denotes a genetic type of glacial sediment recorded in-situ (Fig. 8), but glacial facies or the degree of glaciomechanical influence on fabric development should not be interpreted based on fabric Although glacial ice might be responsible for clast fabric generation and the subsequent deposition of lodgement, deformation, and subglacial-meltout tills from its basal zone, direct evidence for regional ice movement can not easily be delineated by fabric geometry exclusively. For instance, ice-marginal moraines formed by many locally small and structurally complex ice tongues may introduce so much genetic (modal) and directional variability that it is often not possible to establish a single local ice movement direction (e.g., Dreimanis, 1999). Inferring strain accumulation during the emplacement of glacial diamicts is a formidable challenge. Many kilometers of ice may, before melting, shear past a point very near the front of the moraine, leaving no further trace. An indeterminate amount of that shear may occur in the ice, as opposed to the sediments. For these reasons, absolute strain levels may not be recorded clearly in sediments. Therefore, clast fabric analysis is best suited as a relative (as opposed to absolute) strain indicator within a single set of glacial sediment deposits (e.g., Bennett et al., 1999). The strength of the clast fabric and the magnitude of bulk strain do not necessarily have a one-to-one relationship: quite different deformation ellipsoids (and eigenvalues) can result, depending upon whether the clasts undergo passive (March) rotation with the matrix, or make a more independent (Jefferey) rotation in response to force couples resolved across them (e.g., Jeffrey, 1922; Benn and Evans, 1996; Hooyer and Iverson, 2000). In some sediments, fabric strength may even be cyclical with strain (e.g., Lindsay, 1968). Many different processes can produce similar clast fabrics (e.g., Bennett, et al., 1999; Krüger and Kjaer, 1999; Karlstrom, 2000; Benn and Ringrose, 2001; Kjaer, 2001), so one must also use other sediment properties to classify till. Reconstruction of shear as recorded in quantitative clast fabric analysis opens up the possibility of mapping patterns of glaciomechanical strain. It is possible with eigenanalysis to map the directional changes associated with ice movements and to relate sediment emplacements to depositional processes that are evident in more traditional examination of the outcrop. Researchers (e.g., Hart, 1998) have sought to correlate quantitative measures of clast fabrics with shear zones. In most subglacial tills, preferred direction is parallel with the movement of ice and the clasts preferentially plunge upglacier (e.g., Krüger, 1970; Dreimanis, 1999). The preferred long axis eigenvector (342, 06 ) obtained by eigenanalysis (Table 1.) for the stratified diamict unit at Ditch Plains suggests that the pebble orientation in that unit may be subparallel with the Pleistocene ice flow direction believed to be from the NNW (Klein and Davis, 2001). The short axes of these clasts predominately plunge steeply north (upglacier) which suggests that the long and short axes directions are not orthogonal. I infer that the stratified diamict unit originated by subglacial melt-out deposition in the local direction of glacier 55

66 isotropic Waterlain Glacigenic Sediment Subglacial Flow Till Melt-out Till Lodgement Till Debris-rich Basal Ice S3 0.1 girdle cluster S 1 Figure 8. Modified S1 vs S3 eigenvalue plot of Dowdeswell et al., (1985) comparing the standard deviation and mean eigenvalues of four modern glacigenic fabric domains with debris-rich basal ice. Note that possible S1 vs S3 eigenvalue combinations can only fall within the white triangular area of the plot. 56

67 movement based on the strongly preferred orientations of the long and short axes. A clast fabric is a bulk volume property requiring careful sampling acquisition and statistical interpretation. Randomness of the sample is assumed when performing eigenanalysis and the resultant eigenvectors obtained from the analysis are constrained to be orthogonal although the most and least preferred directions do not need to be normal to each other. Furthermore, because nearly equal eigenvalues yield a girdle of unresolved vectors making impossible a distinction of orthogonal directional preference, the eigenvalue S 3 found for long axes eigenanalysis is not necessarily the preferred direction (S 1 ) for the short axes. Surface Sample Bias Since clast fabrics are a bulk volume property, one ideally measures all clasts in a defined volume of an outcrop. This, however, is usually not possible. More commonly, clast orientations recorded from, at, and near an outcrop surface plane are assumed to represent statistically the clast fabric of an associated unit volume (e.g., Benn and Ringrose, 2001). Clast orientations more likely will be preserved in a cohesive matrix containing a high proportion of clay and silt-size particles with relatively tightly packed pore spaces. The best outcrops for measuring clast orientation are quite hard, making them resistant to late slumping and facilitating accurate determination of orientations. Unfortunately, hardness also makes deep excavation difficult. Therefore, measurements are often constrained to at and near exposed surfaces. Investigators commonly sample as many clast orientations as possible in order to maximize data sets and minimize random measurement errors, but do so with limited excavation. Such surficial (rather than volumetric) sampling can lead to a systematic sampling bias (e.g., Millar and Nelson, 2001a,b). Furthermore, the surface sampling bias can skew by a large degree (dependent on true eigenvalue) the eigenvalues reported in eigenvalue studies of fabric shape. Eigenvectors determined by eigenanalysis are much less susceptible to this bias than are eigenvalues especially if S 1 >>S 2. The likelihood of sampling an individual clast depends on the angle that its long-axis makes with the plane of an exposed outcrop surface. As a diamict outcrop erodes, more and more clasts are gradually exposed (Fig. 9). The long axis of a rod-like clast aligned parallel to the strike direction of an eroding outcrop surface will fall out of an outcrop significantly sooner after first exposure than would the equivalent clast with its long-axis direction oriented normal to the exposed surface. The determination of clast fabric strength from clast orientations recorded at an outcrop surface therefore contains an intrinsic bias, because elongated clasts aligned parallel with an eroding outcrop surface will be undersampled and elongated clasts oriented normal to the eroding surface will be oversampled (Fig. 10). Clast orientation sampling can be termed volumetric if outcrop material is excavated to a distance much greater than the mean particle size of the clasts being measured. Absent such excavation, the bias inherent in surface sampling will 57

68 eroding outcrop surface time Y N φ n c D φ P (x p,y p ) X a Figure 9. A cross-section of an ellipsoidal clast in the (X, Y) plane being exposed with time by an eroding outcrop surface. The clast is centered at the origin of the (X, Y) plane and is positioned with its long-axis, a, parallel to the X-axis and its short-axis, c, parallel to the Y-axis. The eroding surface, a plane containing the Z-axis oriented normal to the (X, Y) plane, appears as a trace in the cross-section at some angle, φ, to the Y-axis. As the outcrop surface begins to erode a point, P (x p,y p ) is exposed on the ellipsoidal clast. The vector normal, N, and the unit vector, n, orignate at the point P (x p,y p ). The length, D, is the shortest distance from the center of the ellipsoidal clast to the eroding surface. The magnitude of D, thus indicates how much erosion and time are required before the clast is half exposed. This is used an approximate indicator of the clasts exposed lifetime. 58

69 Surface erosion begins Exposed surface after time moraine air A B C Clasts oriented parallel to eroding surface Clasts oriented normal to eroding surface Figure 10. Three cross sections of a moraine eroding from right to left in a time ordered series. As the outcrop surface gradually erodes clasts oriented normal to the exposed surface are over-sampled, while clasts oriented parallel to the surface are under-sampled. A. Erosion begins at the vertical surface of an outcrop slowly exposing the clasts of a weakly oriented fabric that is composed of equivalent shaped clasts. B. After time, an individual clast with a long-axis orientation closely parallel with the eroding surface falls out of the outcrop while the other clasts remain embedded in the moraine C. As erosion continues, another clast falls out of the outcrop exposure. Three clasts oriented approximately normal to the eroding surface remain cantilevering out from the moraine, and would be over-sampled if recorded. The two clasts orientations that had been oriented nearly parallel to the outcrop surface would be under-sampled since these orientations could no longer be recorded. 59

70 inevitably appear in the data. The magnitude of the surface sampling bias and its effect on the apparent strength of a long-axis clast fabric is not only a function of the angle made between the preferred volumetric long-axis direction of the recorded clast orientations and the orientation of the eroding outcrop surface but it is also a function of the average clast aspect ratio measured therein. Sphere-like clasts have no bias, but they also convey no directional information. Elongate rods with aspect ratio approaching infinity (c b<<a) are the ideal clasts to indicate lineation fabric (e.g. Millar and Nelson, 2001b). Because of erosion, rods parallel to outcrop surface erode away almost as soon as they appear, but those with long axes normal to the outcrop surface are preferentially exposed. Waterlain glacigenic sediments and flow tills are deposits that are particularly vulnerable to the affect of sampling bias. These settings produce weak fabrics (e.g., Lawson, 1979) which could be misidentified as moderately well-developed fabrics if the sample of elongated clasts are limited to those visible at or near the outcrop surface. The probability of a clast being observed in situ depends upon the period of time over which it is exposed at the surface. Clasts with their shortest axes normal to the eroding outcrop surface will be removed by erosion quite soon after they are first exposed. Relatively few such clasts will be observed at any given time. Conversely, pebbles with their long axes normal to the eroding outcrop surface will be exposed over an extended period time before they are removed. They will be relatively over-represented in field studies. I will assume that each clast is ellipsoidal and that it remains embedded in the outcrop until some fraction of its volume (here, I assume 1/ 2 ) is exposed (Fig. 9). Results seem to depend only weakly upon this assumption. The degree to which clasts of a given orientation are over- or under-represented in a surficial sample can then be calculated as a function of the orientation of its axes with respect to the surface. If the erosion rate is either constant or random over time, I can calculate this degree of preference simply by determining the projection of the pebble normal to the outcrop surface (the distance D in Figure 9). The surface of an ellipsoidal clast is given by the equation. I can define a function [ x2 / a 2 + y2 / b 2 + z2 / c 2 ] = 1 (2) F(x,y,z) = ( 1 / a 2) x2 + ( 1 / b 2) y 2 + ( 1 / c 2) z 2-1 (3) such that the ellipsoid is the locus of points F(x,y,z)=0. Thus, N is a vector normal to the ellipsoid at location (x,y,z). N = (N x, N y, N z ) = 2 [ x / a 2, y / b 2, z / c 2 ] = K n = K( n x, n y, n z ) (4) In general, N is not the unit normal n, but a normal vector of some length k. So n = ( n x, n y, n z ) = 2/K [ x / a 2, y / b 2, z / c 2 ] (5) The vector P to a point on an ellipsoid of this family is 60

71 P(x,y,z) = (Ka 2 n x / 2, Kb 2 n y / 2, Kc 2 n z / 2 ) (6) Because the point falls on the ellipsoid, this must satisfy eqn. 2. Therefore, and [ (Ka 2 n x / 2 ) 2 / a 2 + (Kb 2 n y / 2 ) 2 / b 2 + ( Kc 2 n z / 2 ) 2 / 2 ) 2 / c 2 ] = 1 (7) so 4/K 2 = a 2 n x 2 + b 2 n y 2 + c 2 n z 2 (8) K = 2 [ a 2 n x 2 + b 2 n y 2 + c 2 n z 2 ] -1 /2 (9) The component of P in the n direction is simply the projection D of the clast in the direction of the outcrop surface, so D = n P = (Ka 2 n x 2 / 2 ) + (Kb 2 n y 2 / 2 ) + (Kc 2 n z 2 / 2 ) D = (K/ 2 ) [ (a 2 n x 2 ) + (b 2 n y 2 ) + (c 2 n z 2 ) ] Or D = [ a 2 n x 2 + b 2 n y 2 + c 2 n z 2 ] 1 /2 (10) Quantifying the bias in limiting cases I use numerical methods and Monte Carlo calculations to quantify the sample bias due to recording clast orientations from an outcrop surface. For simplicity, it is assumed that clasts are sufficiently widely spaced so that their interactions can be ignored. Each clast is treated as an ellipsoid of rotation, (either b=a or b=c), in which case D (eqn. 10) gives the relative exposure duration of a clast. Larger clasts are obviously exposed longest but for any given size clast, shape and orientation also matter. For more nearly spheroidal clasts (a c), D is less strongly a function of axial orientation, but for geologically significant aspect ratios a>>c, D is a rather strong function of axial orientation leading to a significant sample bias. In Monte Carlo calculations, I randomly create thousands of rod-like clasts per numerical experiment. For each experiment I assume a uniform clast size and aspect ratio (a/c), as well as a prescribed true anisotropy in clast orientation introduced by modifying the random statistics to produce a preferred fabric. I then define an outcrop surface orientation anywhere from normal to parallel to the fabric axis. I calculate the true eigenvalues and eigenvectors of this synthetic clast data set, as if observed by deep excavation. Simultaneously, I carry out the same calculations for the population of pebbles exposed at an outcrop surface, using D (eqn. 10) for each pebble to determine its relative likelihood of 61

72 appearing on the surface. For each experiment, I calculate for a population of 5000 pebbles and compare eigenvalues: the true S 1, S 2, S 3 vs. the observed S 1 ', S 2 ', S 3 '. As a limiting case, I performed a set of numerical experiments with a true eigenvalue isotropic distribution (S 1 = S 2 = S 3 = 1/3) by numerically generating clasts of aspect ratio a/c approaching infinity (thin rods). This produced an observed fabric of S 1 = 0.5 and S 2 = S 3 = Thus, even a totally isotropic distribution composed of elongated clasts can look anisotropic. The effect of the bias on the fabric depends on the angular separation (θ) between the most preferred eigenvector direction of the true fabric with the orientation of the sampling surface. In addition, the effect is smaller with smaller a:c aspect ratio (Fig. 11) but still produces a significant sampling bias on clasts with realistic aspect ratios of a:c 2.5 (Fig. 12). The results of surface sampling experiments show that the sampling bias has very little impact on observed eigenvectors of strong fabrics (Fig. 13), which remain fairly stable even though the observed eigenvalues are distributed widely. Moderate and weakly preferred surface sampled fabrics can have extreme observational errors in calculated eigenvector direction (Fig. 14). These large errors are reproducible and cast serious doubt on the accuracy of interpreted eigenvector directions that result from clast orientation data sets measured at poorly excavated field sites. True eigenvalues of clast fabrics can not be reproduced from surface sampled clast orientations (Fig. 13,14, Table 2, Appendix A) but in some instances the eigenvalue ratios of both true and observed eigenvalues are serendipitously about equal, due to trade-offs among sets of eigenvalue combinations. Implications for field studies The sampling bias associated with clast axes measured in situ has very little impact on reliable eigenvector determination for strong fabrics. This bias causes deviations from true fabric directions (by a few degrees) that are smaller than the typical uncertainties due to various inevitable observational errors. The effect of the bias on direction is more pronounced for increasingly weak fabrics but is not large (tens of degrees) except for those fabrics which very nearly isotropic. Extremely weak fabrics are inherently variable, giving little preferred orientation information, so they would not typically be used as directional indicators when reconstructing former ice sheet movement even if they were not effected by such a bias. The sampling bias effect always skews the true eigenvalue distribution. Nearly isotropic distributions can be falsely perceived as only a moderate lineated fabric if observed at a surface that is nearly parallel to the preferred clast long axis. Likewise, moderately clustered lineations can be misidentified either as nearly isotropic fabrics or as fabrics of greater than their true strength. Variability in flow strength and direction is undoubtedly responsible for the large eigenvalue range found for the flow till domain (Fig. 15). The surface sampling bias is by itself, capable of producing spurious results that vary by amounts larger than typically cited sizes of genetic till domain ranges (e.g., Dowdeswell et al., 1985). Figure 15, illustrates the effective eigenvalue range produced by the bias on several data 62

73 63 Figure 11. (S 1, S 3 ) eigenvalue plot of computer generated random clast fabric eigenvalue data sets for volume measured fabrics (single data points) along with surface sample bias related eigenvalue observation errors. These eigenvalue observation errors are a function of clast aspect ratio (a.r = a/c) and angular separation, θ, between the most preferred eigenvector direction of the volume measured fabric and the orientation of the sampling surface. The bias related observation errors plot as individual curves linking eigenvalue data points calculated with values of θ = 90, 60, 45, 30, and 0 (small dots from left to right on each curve). The aspect ratio for each curve link data points that are indicated by small black dots for a.r. = 5.0 and small colored dots for a.r. = 2.5. In addition, we plot computer generated clast fabric data in order to test the robustness of field data obtained by surface sampling the stratified diamict unit at Ditch Plains.

74 64 Figure 12. (S 1, S 3 ) eigenvalue plot presenting the a.r. = 2.5 computer generated random clast fabric eigenvalue data sets of figure 11, for volume measured fabrics (single data points) along with surface sample bias related eigenvalue observation errors. The bias related observation errors plot as individual curves linking eigenvalue data points calculated with values of θ = 90, 60, 45, 30, and 0 (small dots from left to right on each curve).

75 Figure 13. Detail of (S 1,S 3 ) eigenvalue plot shown in figure 12, along with eigenvector observation errors. In strong fabrics, surface sampled eigenvectors are approximately equal to the true population (volume measured) eigenvectors. Thus, ice flow direction can be inferred from eigenvector directions of strong clast fabrics. 65

76 S S 3.20 = S 2 3 S 2 =S 1.10 Outcrop orientation Observation error Outcrop orientation Observation error ± ± ±1 41 ±3 28 ±2 19 ±2 Generated data (volume measured fabric) 60 9 ± ± ±1 0 0 ± S 1 Figure 14. Detail of (S 1, S 3 ) eigenvalue plot shown in figure 12, along with eigenvector observation errors. In fabrics of modest strength, surface sampled eigenvectors are significantly displaced from the true population (volume measured) eigenvector. 66

77 Table 2. Computer generated random clast fabric data sets for volume measured fabrics (eigenvalues and eigenvectors) along with observational errors and standard deviations due to the surface sampling bias. Surface measured eigenvalues and observation errors are based on clast aspect ratio, a/c, (a.r. = 2.5 and a.r. = 5.0) and the angular separation, θ, (0, 30, 45, 60, and 90 ) between the preferred eigenvector direction of the volume measured fabric and the orientation of the outcrop surface responsible for creating the sample bias. volume surface obs. error (a.r. = 2.5) obs. error (a.r. = 5.0) s1 s3 s1 s3 θ ( ) AVG ( ) ST DEV ( ) AVG ( ) ST DEV ( )

78 volume surface obs. error (a.r. = 2.5) obs. error (a.r. = 5.0) s1 s3 s1 s3 θ ( ) AVG ( ) ST DEV ( ) AVG ( ) ST DEV ( )

79 69 Figure 15. (S 1, S 3 ) eigenvalue plot of glacigenic diamict fabric domains by standard deviation ellipses (e.g., Dowdeswell et al., 1985). In addition, the plot contains the a.r. = 2.5 computer generated random clast fabric eigenvalue data sets shown in figure 12, for volume measured fabrics (single data points) along with surface sample bias related eigenvalue observation errors. The surface sample bias observation errors plot as individual curves with range in S 1 and S 3 magnitudes produced by the sample bias that often exceeds the standard cited sizes of glacigenic diamict fabric domains.