EVALUATING THE RELATIONSHIP OF BLUFF RECESSION TO MORPHOLOGIC AND HYDRODYNAMIC VARIABLES AT SEVENTEEN BLUFF SITES ALONG THE MICHIGAN COAST OF LAKE MIGHICAN By Elizabeth M. Spitzer A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Geography – Master of Science 2021 ABSTRACT EVALUATING THE RELATIONSHIP OF BLUFF RECESSION TO MORPHOLOGIC AND HYDRODYNAMIC VARIABLES AT SEVENTEEN BLUFF SITES ALONG THE MICHIGAN COAST OF LAKE MICHIGAN By Elizabeth M. Spitzer Changes in Lake Michigan bluff recession rates over time were evaluated by comparing rates derived from a study done in the 1970s (Buckler 1981) to modern rates (since the 1970s) generated by data collection for this thesis. A subset of the 118 original sites were mapped, totaling 17 bluff sites along the Michigan coast of Lake Michigan. Modern rates were calculated through field RTK-GPS surveys and digitizing bluff crests from historical aerial imagery. Recession rates were correlated to various site morphologic and hydrodynamic variables to assess their role in facilitating change over space and time. Major conclusions of this study are: (1) no tested variables correlated well with morphologic or hydrodynamic variables, and thus cannot explain the large spatiotemporal variability in recession rates along the MI coast, (2) more focus should be paid to quantifying and understanding the coupled changes at the bluff toe and crest, to better constrain the 3-D geomorphic recession patterns of Great Lakes bluffs, and as processes driving recession at the bluff toe could not be meaningfully correlated to recession rates, (3) utilizing digitized bluff crest locations from georeferenced aerial imagery to derive bluff recession rates introduces errors associated with human interpretation that can be overcome now with geospatial technologies including RTK-GPS, LIDAR, and sUAS (i.e. drones), and (4) as small-scale processes appear to play an important role in driving recession, more attention should be given toward documenting small-scale changes in recession and processes at individual sites, and to avoid over-generalization of the dynamic geomorphic environment. ACKNOWLEDGEMENTS I would like to thank my academic advisor, Dr. Ethan Theuerkauf, for being endlessly supportive and kind, as well as for his guidance and insight though every step of this process. I would also like to thank Dr. Alan Arbogast and Dr. David Lusch for serving on my committee and for their invaluable encouragement, advice, and knowledge. Further, I would like to thank the MSU Geography Department faculty, staff, and students for creating a welcoming and inclusive environment in which to learn and do research. Additionally, thank you to Mr. Tomago Collins for the generous donation that funded this project and made it possible. I would also like to thank the Joint Airborne Lidar Bathymetry Technical Center of Expertise (JALBTCX) for kindly supplying data crucial to the completion of this study. Lastly, I would like to give a huge thanks to everyone who has supported me along this journey and on the journeys to come. iii TABLE OF CONTENTS LIST OF TABLES ........................................................................................................................ vi LIST OF FIGURES ..................................................................................................................... viii Introduction ......................................................................................................................................1 Background ..........................................................................................................................1 The Bluff Retreat System.....................................................................................................3 Drivers of Bluff Recession...................................................................................................4 Wave Climate...........................................................................................................4 Groundwater Seepage ..............................................................................................5 Mediators of Bluff Recession ..............................................................................................6 Water Level ..............................................................................................................6 Human Disturbance ...............................................................................................12 Lithologic Characteristics ......................................................................................14 Geomorphic Characteristics ...................................................................................15 Nearshore Bathymetry ...........................................................................................16 Beach Morphology.................................................................................................17 Interconnection of the Drivers and Mechanisms of Bluff Recession ................................17 Mechanisms of Bluff Recession ........................................................................................18 Statement of Problem.........................................................................................................21 Methods..........................................................................................................................................23 Summary of Buckler 1981 Methods ..................................................................................23 Methods of Subsampling Buckler 1981 Bluff Sites ..........................................................25 Evolution of Methodology Efficacy over Time ................................................................26 Thesis Methods ..................................................................................................................28 Property Owner Determination and Property Access Requesting .........................28 Site Determination and Confirmation ....................................................................29 RTK-GPS Field Surveys ........................................................................................32 Aerial Image Acquisition .......................................................................................33 Aerial Image Georeferencing .................................................................................35 Digital Shoreline Analysis System ........................................................................36 Bluff Crest Extraction from Topobathy LiDAR ....................................................40 Lithology ................................................................................................................41 Hardened Shoreline Analysis.................................................................................41 Nearshore Bathymetry and Beach Topography Extraction ...................................42 Beach, Bluff, and Nearshore Morphodynamic Extraction.....................................45 Results ............................................................................................................................................47 Bluff Recession Rates ........................................................................................................47 Lithology ............................................................................................................................54 Hardened Shoreline Analysis.............................................................................................68 iv Nearshore Bathymetry .......................................................................................................72 Beach, Bluff, and Nearshore Morphodynamics.................................................................74 Relationships Between Beach, Bluff, and Nearshore Morphometrics and bluff Recession Rates...................................................................................................................................75 Fetch, Shoreline Azimuth, and Wave Power .....................................................................80 Relationship Between Fetch, Shoreline Azimuth, Wave Power, and Bluff Recession Rates...................................................................................................................................83 Discussion ......................................................................................................................................87 Long-term Rates of Bluff Recession..................................................................................88 Comparison of Historic Recession Rates to Modern Rates ..................................88 Sites with Generally Increasing Recession Rates over Time.................................90 Sites with Generally Decreasing or Constant Recession Rates over Time ...........97 Correlation to Site Morphometrics and Hydrodynamics .................................................103 Commentary on Methodology .........................................................................................104 Future Work .....................................................................................................................107 Conclusion ...................................................................................................................................110 BIBLIOGRAPHY ........................................................................................................................113 v LIST OF TABLES Table 1: Summary table of rate-of-change at site M42 in Manistee, Michigan. Transect 5 includes the GLO and Buckler (1981) points. To the right is the table denoting the average of all transects per timestep and the associated standard deviation. For the 1839-1977 timestep, the total change calculated by Buckler is normalized by decade to follow the near-decadal pattern of the more recent time steps. For the 1839 - 1977 and 2018 - 2020 time steps, the standard deviation is zero since there is only one data point for each time step ..........................................39 Table 2: Results from DSAS showing bluff crest position change over time in meters. The first documented bluff crest position (mid-1800s) is represented as zero. Negative number indicates landward movement of bluff crest .................................................................................................48 Table 3: Magnitude normalized by time of all sites from GLO survey to Dr. Buckler’s survey (mid-1800s-1977) compared to Dr. Buckler’s survey to present (1977-2020), as well as the difference between the rates and if they are increasing, decreasing, or remaining constant over time based on the threshold of +/- 0.25m of change ......................................................................49 Table 4: The key and unit descriptions associated with the stratigraphic analysis of sites as redrawn from Buckler (1981 Table B2)........................................................................................ 55 Table 5: Bluff stratigraphy of the seventeen sites, redrawn and modified from Buckler (1981 Table B2) ...................................................................................................................................... 56 Table 6: Lithology of each site at the bluff toe, bluff face, and crest, as interpreted from Buckler (1981 Table B2) ............................................................................................................................ 57 Table 7: The interpreted lithology of the sites from Farrand and Bell (1982) .............................. 59 Table 8: The distance in kilometers (south of or north of) from a shore-perpendicular structure to the local bluff site, and the structure type ......................................................................................70 Table 9: The interpretation of plotting the bluff sites on the hardened shoreline classification layer (coast.noaa.gov/digitalcoast/data/hardened-shorelines.html) showing approximate time of construction, primary and secondary structure type, and relative condition .................................72 Table 10: Morphological characteristics of bathymetry at each site in terms of presence of inner bar, outer bar, and berm, relative to the monthly mean water level at the time of data collection. N/A values represent missing data. *M2 utilizes topobathymetric data from 2008 instead of 2012 due to inadequate data ....................................................................................................................74 Table 11: The 2012-2020 bluff recession magnitude normalized by time in meters per year and the associated nearshore slope, bluff face slope, bluff crest elevation, bluff toe elevation, beach width, and beach slope parameters from 2012 at each site ............................................................75 vi Table 12: Summarization of linear regression analysis results of relationships between beach, bluff, and nearshore morphometrics and bluff recession rates including the R-squared value and P-value ...........................................................................................................................................76 Table 13: The 2012-2020 bluff recession magnitude normalized by time in meters per year and the azimuth of the shoreline in degrees measured clockwise from 270 ........................................81 Table 14: The 2012-2020 bluff recession magnitude normalized by time in meters per year and the maximum perpendicular fetch distance (km) relative to each site’s shoreline azimuth, and the azimuth of the perpendicular fetch orientation ..............................................................................82 Table 15: The 2012-2020 bluff recession magnitude normalized by time in meters per year and the daily mean wave energy from 2012-2020 in square meters ....................................................83 Table 16: Summary of linear regression analysis results of relationships between maximum perpendicular fetch, mean daily wave energy, bluff azimuth, and maximum perpendicular fetch azimuth and bluff recession rates including the R-squared value and P-value..............................84 vii LIST OF FIGURES Figure 1: The monthly mean water level (blue) and long-term annual average water level (red) of the Lake Michigan-Huron basin from 1918 to present. Figure is from the USACE Detroit District Website ............................................................................................................................................7 Figure 2: Output of the USACE multi-method model simulating future Lake Michigan-Huron water levels. Figure is from the USACE Detroit District Website ..................................................9 Figure 3: Aerial image of home collapsing into Lake Michigan following bluff failure in late 2019. Image is from Fox2 Detroit website ....................................................................................11 Figure 4: Location of sites from Buckler (1981) plotted on Michigan coastal dune locations. Green points represent sites that were kept for further analysis. Red points are sites that were eliminated from further analysis. The “X” symbols denote sites that had already been field evaluated prior to digital site evaluation. The coastal dunes polygon layer is an open-source data set from the Michigan Environmental Council website ................................................................31 Figure 5: Map showing the location of the seventeen bluff sites that were examined in this study. Map generated in ArcGIS 10.6 ......................................................................................................32 Figure 6: Results of DSAS analysis on site M42 generated in ArcGIS 10.6 showing digitized shorelines (blue), a baseline (black), GLO, Buckler (1981) and 2020 survey points, and section corner (green points, left to right respectively), transects (red), and transect intersection points (red point), superimposed over 2018 NAIP imagery .....................................................................39 Figure 7: Locations of the profiles for site M43 shown on the 2008 topobathy LiDAR (JALBTCX) .................................................................................................................................. 44 Figure 8: Nearshore bathymetry map generated from single-beam sonar data collected by the MSU Coastal Lab in October of 2020. The dotted line shows the track of the survey vessel…...44 Figure 9: Magnitude normalized by time of all sites from GLO survey to Dr. Buckler’s survey (mid-1800s-1977) compared to Dr. Buckler’s survey to present (1977-2020) ............................ 49 Figure 10: Magnitude of change normalized by time in meters per year for all seventeen bluff sites. M44-projected represents projected rate of change for modern time step due to shoreline modification at site.........................................................................................................................50 Figure 11: Bluff sites with recession rates in meters per year that are generally increasing over time ................................................................................................................................................51 Figure 12: Bluff sites with recession rates in meters per year that are generally increasing that had a peak in rate of recession in the 1970s-1990s ........................................................................51 Figure 13: Bluff sites with recession rates in meters per year that are generally increasing that did not have a peak in rate of recession in the 1970s-1990s ................................................................52 viii Figure 14: Bluff sites with recession rates in meters per year that are generally decreasing or near-constant over time. M44-projected represents the projected modern recession rate due to shoreline modification at the site ...................................................................................................53 Figure 15: Bluff sites with recession rates in meters per year that are generally decreasing or remaining constant that had a peak in rate of recession in the 1970s-1990s .................................53 Figure 16: Bluff sites with recession rates in meters per year that are relatively constant that did not have a peak in rate of recession in the 1970s-1990s ................................................................54 Figure 17: Bluff sites plotted on the Quaternary Geology of Southern Michigan map (Farrand and Bell 1982). Inset map site names from south to north: bottom left: sites M1, M2, M3, M4, M5, M6, bottom right: sites M9, M10, M14, top left: sites M37 and M38, top middle: sites M42, M43, M44, and M45, top right: sites M54 and M55 .....................................................................58 Figure 18: Oblique imagery of site M1. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................59 Figure 19: Oblique imagery of site M2. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................60 Figure 20: Oblique imagery of site M3. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................60 Figure 21: Oblique imagery of site M4. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................61 Figure 22: Oblique imagery of site M5. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................61 Figure 23: Oblique imagery of site M6. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................62 Figure 24: Oblique imagery of site M9. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................62 Figure 25: Oblique imagery of site M10. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................63 Figure 26: Oblique imagery of site M14. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................63 Figure 27: Oblique imagery of site M37. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................64 Figure 28: Oblique imagery of site M38. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................64 Figure 29: Oblique imagery of site M42. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................65 ix Figure 30: Oblique imagery of site M43. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................65 Figure 31: Oblique imagery of site M44. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................66 Figure 32: Oblique imagery of site M45. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................66 Figure 33: Oblique imagery of site M54. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................67 Figure 34: Oblique imagery of site M55. Retrieved from: https://toolkit.climate.gov/tool/greatlakes-shoreviewer. Site is represented as red star ................67 Figure 35: Side by side change of site M44 in Manistee Michigan. The left was taken in 2005, prior to modification, and the right is in 2018 after the bluff was graded out, bluff stabilization measure installed, and concrete-lined channel established ............................................................69 Figure 36: Bluff site locations relative to the hardened shoreline classification layer generated by NOAA (coast.noaa.gov/digitalcoast/data/hardened-shorelines.html)........................................... 71 Figure 37: The 2012-2020 bluff recession magnitude normalized by time of the seventeen sites plotted against the average nearshore slope ...................................................................................76 Figure 38: The 2012-2020 bluff recession magnitude normalized by time of the seventeen sites plotted against the average bluff face slope ...................................................................................77 Figure 39: The 2012-2020 bluff recession magnitude normalized by time by time of the seventeen sites plotted against the average bluff crest elevation ...................................................78 Figure 40: The 2012-2020 bluff recession magnitude normalized by time of the seventeen sites plotted against the average bluff toe elevation ..............................................................................79 Figure 41: The 2012-2020 bluff recession magnitude normalized by time of the seventeen sites plotted against the average beach width ........................................................................................79 Figure 42: The 2012-2020 bluff recession magnitude normalized by time the seventeen sites plotted against the average beach slope .........................................................................................80 Figure 43: The 2012-2020 bluff recession magnitude normalized by time for the seventeen sites plotted against the maximum perpendicular fetch distance ...........................................................84 Figure 44: The 2012-2020 bluff recession magnitude normalized by time for the seventeen sites plotted against wave energy in meters squared..............................................................................85 Figure 45: The 2012-2020 bluff recession magnitude normalized by time for the seventeen sites plotted against bluff azimuth .........................................................................................................85 x Figure 46: The 2012-2020 bluff recession magnitude normalized by time for the seventeen sites plotted against the perpendicular fetch azimuth ............................................................................86 Figure 47: Satellite image of M54, M55, and surrounding geography. Image taken from Google Earth Pro ......................................................................................................................................101 xi Introduction Background Coastal cliffs are dynamic landscapes that evolve across a variety of temporal and spatial scales in response to a wide range of physical drivers. These coastal landforms support tourism, housing, infrastructure, and numerous species of flora and fauna. Furthermore, these environments are important for sustaining coastal geomorphic systems as they provide sediment to the adjacent beaches and nearshore systems (Carter et al. 1990; Young and Ashford 2006; Johannessen 2010). It is estimated that coastal cliffs make up approximately 52% of the global shoreline, making them the most prevalent coastal feature (Young and Carilli 2019). However, the dynamic nature of these landscapes presents a variety of management challenges centered around the fact that coastal cliffs erode and fail over time, leaving any infrastructure and habitat on or near the edge vulnerable. A strong scientific understanding of the processes that lead to failure, as well as the causes of spatiotemporal variability in bluff retreat, is needed to predict the future of coastal cliffs and develop policies that help preserve and protect these landscapes and their associated natural and human-built ecosystems. The term ‘coastal cliff’ refers to a steeply sloping surface where elevated terrain converges with the shoreline (Hampton 2004). Coastal cliffs can form along marine, estuarine, and freshwater bodies. However, these landforms differ in many aspects, including lithology and environment in which they were created. Many geomorphologists have studied the various types of coastal cliffs: Hustra et al. (2016) studied retreat rates of chalk cliffs on the south coast of Great Britain; Eposito et al. (2018) investigated retreat rates of pyroclastic cliffs in southern Italy, and Pierre (2006) analyzed retreat rates of clay and sandstone sea cliffs in northern France, along with many other examples. 1 Among the many synonyms for coastal cliffs is the term coastal bluff, which lacks a consistent definition, but most often is thought of as a steeply, sloping surface formed in unconsolidated sediment, such as clay, sand and gravel, whose crest plateaus above the shoreline (Collins and Sitar 2004; Maine Geological Survey 2011; Stanchev et al. 2013; Krueger et al. 2020; among others). There exist regional discrepancies in the exact characteristics of a coastal bluff, primarily resulting from differences in depositional material and landscape history. For example, Collins and Sitar (2008) studied coastal bluffs in California that formed during periods of tectonic-forced uplifting and fluctuating sea levels, whereas Shipman (2008) examined bluffs in the state of Washington that were derived from glacial activity. In the Great Lakes region of North America, the local definition of a coastal bluff is more precise. Researchers have uniformly defined the coastal bluffs along the Great Lakes as steeply sloping surfaces with an upland crest adjacent to the shoreline, comprised of unconsolidated Quaternary material sediments, that were deposited during the Pleistocene Epoch 2.6 MYA to 11.7 KYA (e.g., Vallejo and Degroot 1988; Jibson et al 1994; Mickelson et al. 2004; Castedo et al. 2013; Zoet and Rawling 2017). The Laurentide Ice Sheet (LIS), the largest continental glacier that affected North America, is thought to have formed and advanced 15 to 20 times during the Pleistocene Epoch. In the Great Lakes Region, with each advance and retreat, the LIS carved the lake basins deeper and wider until they reached their current size beneath the last major ice advance during the Wisconsinan glaciation between 25,000 and 10,000 years ago (Larson and Schaetzl 2001; Mickelson et al. 2004). The LIS carried eroded rock and unconsolidated sediments along with it as it flowed, often depositing this debris in front as till as the glacier retreated. Glacial till is an unsorted and unstratified mixture of clay, sand, gravel, and boulders deposited directly by glacial 2 ice (Mickelson et al. 2004). With each readvance, sediments of slightly different composition were deposited, and these layers are now exposed in the coastal bluffs along the shoreline (Mickelson et al. 2004). In between these till layers are lenses of sand and gravel that were deposited in water as beaches and deltas ahead of the retreating glacier; between advances, silt and clay were deposited in proglacial lakes formed near the margin of the ice sheet (Mickelson et al. 2004). As these proglacial lakes drained, the modern Great Lakes emerged, and through wave action, eroded these glacial deposits into the bluffs that are seen along the Great Lakes shores today (Mickelson et al. 2004). Bluffs along the Great Lakes shorelines provide not only a breathtaking landscape for recreational and ecological benefit, but also supports critical built infrastructure. However, due to a combination of driving factors, bluffs consistently erode and fail over time, moving the bluff crest position further landward as they recede. With bluff failure and recession comes destruction and loss of homes, habitat, and infrastructure that were positioned on or near the crest. Understanding the mechanisms and reasons for spatiotemporal variability of bluff retreat are crucial for predicting rates of future retreat, as well as for making sound policy and management decisions to effectively manage these complex landscapes. The Bluff Retreat System From a systems approach, the rate of bluff retreat is a function of the relative role of assailing drivers mediated by environmental/anthropogenic resisting factors (Sunamura 1983). Drivers are natural forces that directly or indirectly cause an environmental change. Mediators are environmental or anthropogenic factors that modulate the impacts of drivers. On Great Lakes bluffs, the assailing drivers include wave action, groundwater seepage, surface runoff, and spring thaw (Buckler 1981; Buckler and Winters 1983; Amin and Davidson-Arnot 1995; Brown 2005; 3 Castedo et al. 2013; Zoet and Rawling 2017; Volpano et al. 2020; among others), as well as variations in wave climate (Vallejo and Degroot, 1988; Swenson et al., 2006; Castedo et al., 2013; among others), and/or the presence of shore ice (Davis 1973; Barnes et al. 1944; Zumberge and Wilson 2000). Resisting factors, or mediators, include the compressive strength of the bluff material (Sunamura 1983; Kamphius 1987), bluff height, the slope of the bluff face, and bluff azimuth, water levels, nearshore bathymetry, and bluff lithology and stratigraphy (Alden 1918; Gray 1975; Edil and Vellejo, 1977; Mickelson et al., 1977; Buckler 1981; Carter and Guy, 1988; Rovey and Borucki, 1993; Johnson and Johnston, 1995; among others). Other mediating factors include vegetation presence (Gray 1975; Dai et al., 1977; Hall and Ludwig 2011; among others), and shoreline structures and armoring (Ball 1938; Omohundro 1973; Wood 1988; Shabica et al. 2010; among others). Some local factors can influence the strength of either the assailing drivers, the resisting mediators, or both. For example, while the presence of shoreline armoring, such as seawalls, may inhibit incoming waves from reaching the bluff toe, the presence of a jetty or groin may disrupt the natural longshore sediment transport process, thus reducing the replenishment of protective beaches in front of the bluff toe. Overall, erosion can occur when the assailing forces overcome the resisting (or mediating) factors (Sunamura 1983; Kamphius 1987). Drivers of Bluff Recession Wave Climate Wave climate is defined as the distribution of wave characteristics averaged over a period of time and for a particular location (Bramante and Woods Hole Oceanographic Institution 2017). Wave conditions can be analyzed using different metrics such as wave height, wave period, and wave direction, as well as derived metrics such as wave power and wave energy. 4 Overall, it is well documented that wave action generates removal of bluff material at the toe, and subsequent slope instability and bluff crest recession (Amin and Davidson-Arnot 1995; Quigley et al. 2017; Edil and Vallejo 1977; Browen et al. 2005; Castedo et al. 2013; among others). Amin and Davidson-Arnot (1995) concluded that deep water wave energy, strong winds, and storm surge were the most important drivers of bluff toe erosion. Keillor and DeGroot (1978) showed that bluff recession, in part, can be attributed to greater exposure of the bluff toe to wave action, however this is interdependent on the relative position of water level and the adjacent beach width. Regardless of the role that individual wave climate parameters may play in driving bluff recession, it is well supported in the literature that wave action plays a primary role in driving bluff recession. Groundwater Seepage Groundwater conditions, including the elevation of the groundwater table, can be an important driver of bluff retreat (Wyllie and Mah, 2017). Higher groundwater table elevations may result in more bluff material being saturated along the slope, resulting in a weakening of the slope stability. Castedo et al., (2012) found that recession rates were higher in saturated materials than in dry. Water table elevations are products of the recharge regime where heavy precipitation coupled with high infiltration rates raises the local groundwater table level (Thorsen 1987). The influence of groundwater on bluff recession is mediated by the bluff lithology. In a scenario where outwash overlies proglacial lakebed clays in the bluff, groundwater can move down through the porous outwash and then laterally along the outwash-clay boundary to form a groundwater seep along the bluff face (Hampton and Griggs 2004). Overall, the presence of groundwater seeps in the bluff face diminishes the resistance of the bluff sediments to slope failure. 5 Mediators of Bluff Recession Water Level Bluff recession is mediated by many factors including fluctuating water levels, human disturbance, and the lithologic and geomorphic characteristics of a bluff (Buckler and Winter 1983), but water level is one of the most important. Davis et al. (1973) suggest that water level plays a passive role and sets a foundation by providing the correct environment for bluff recession to occur. Greater storm frequency and intensity increase the impact of wave action during storm events, but during periods of high lake levels, the threshold wind and wave conditions needed to cause wave erosion of the bluff toe is much lower than during low lake level periods (Amin and Davidson-Arnot 1995). Ultimately, however, the removal of bluff material and subsequent crest retreat is driven by powerful waves generated by intense storms. Wave action will always be occurring regardless of water level fluctuations, however the amount of impact that wave action can have on bluff toe erosion is reliant on water levels being high enough to surpass the adjacent beach, if one is present. Water levels in the Great Lakes fluctuate on a range of time scales from seasons to millennia (Colman et al., 1994; Thompson and Baedke 1995; 1997). On decadal time scales, water levels vary from periods of above average lake levels, such as those observed in the 1980s and from 2012 to 2021, to periods of below average water levels, such as the period that occurred in the early 2000s (US Army Corps of Engineers [USACE], Great Lakes Water Level Data). Great Lakes water level fluctuations reflect the precipitation and temperature patterns of the region on both short and long-term scales (Gronewald et al. 2013). Figure 1 shows the long- term (1918 to present) record of mean monthly water levels superimposed on the long-term average annual water level of the Lake Michigan-Huron basin. In the beginning of the 21st 6 century, water levels remained at or below the long-term average for approximately 15 years, followed by a continuous rise in lake level which has remained at or above the long-term average since approximately 2015. Figure 1: The monthly mean water level (blue) and long-term annual average water level (red) of the Lake Michigan-Huron basin from 1918 to present. Figure is from the USACE Detroit District Website. Researchers use a water-balance approach and quantify total water volume by calculating net basin supply (NBS), which is the net amount of water entering each Great Lake, not including the input of water from the upstream lakes (Neff and Nicholas 2005). NBS is derived in a mass balance equation where change in volumetric storage of the lake is calculated by the addition of over-lake precipitation, runoff inflow, groundwater inflow, and the subsequent subtraction of surface evaporation, connecting channel outflow, consumptive use, diversion, and change in local storage due to thermal expansion or contraction (Lewis 2005). The US Army Corps of Engineers utilize historical NBS data and model-derived NBS outputs to generate a range of potential water levels that could occur in the future (NOAA, GLERL, GLSHyFS) (US Army Corps of Engineers, Experimental 5-year Outlook of Great Lakes Water Levels). Figure 2 shows the model output for the long-term water level outlook for the Lake Michigan-Huron basin, which suggests that in the future, Lake Michigan-Huron water levels may stay at or above the long-term average of approximately 176.5m. The left panel of Figure 2 depicts predicted monthly mean water levels using a historical supplies and simulated supplies approach. The yellow shaded area indicates the range of variability in monthly mean water levels 7 that result when an ensemble of historical NBS sequence is used to predict water levels and the yellow line shows the median outlook from this approach. The blue shaded area indicates the range of variability in the monthly mean water levels that result from a physically based modeling system that incorporates initial conditions as well as seasonal climate outlook, and the blue line represents the median outlook from this approach. The right-hand panels of Figure 2 show the range of variability of monthly mean water levels within the subsequent two to four 12- month periods of outlook horizon (shaded, vertical lines), as well as the annual mean water level projected for those 12-month period (large, dark points). The red horizontal line depicts the long- term average annual water level, the black bars depict long-term average monthly mean water levels, and gray bars depict long-term maximum and minimum mean water levels. 8 Figure 2: Output of the USACE multi-method model simulating future Lake Michigan-Huron water levels. Figure is from the USACE Detroit District Website. Studies suggest that an increase in anthropogenically emitted greenhouse gases have caused an increase in rainfall intensity in the northern hemisphere and Great Lakes region (Vavrus and Dorn 2010; Allan 2011), which is likely driving changes in Great Lakes basin hydrology, which could lead to higher lake levels. Countering this, warming conditions result in reduced winter ice cover, which can increase evaporation rates, possibly resulting in lower lake levels (Blanken et al. 2011). Though it is extremely difficult to predict future patterns of precipitation and evaporation for the Great Lakes basin and thus lake level, bluff recession is an ever-present problem regardless of whether lake level is above or below average. During low 9 lake level periods, bluff recession can result from surface water runoff or groundwater movement (Wyllie and Mah 2017). Given that bluffs formed during past glacial conditions, once they erode that material cannot be replenished and that portion of the bluff is lost permanently. This necessitates studies aimed at gathering the knowledge necessary to accurately predict future bluff retreat rates in order to protect infrastructure and habitat associated with coastal bluffs. Due to the combination of present high lake levels and the predicted future high lake levels it is important to document coastal impacts related to high lake levels. However, even if lake levels drop from their present highs, bluff recession will continue and may in fact increase in the years immediately following high lake levels (Roland et al. 2021). My thesis study documents rates of bluff recession across a range of lake level fluctuations in order to better understand what drives bluff change. These data encompass periods of both slowly fluctuating and rapidly fluctuating lake levels, which account for the range of conditions that may be experienced in the future. Rapid fluctuations between low and high lake level periods, such as what was documented between 2013 to present, may be the result of climate change in the Great Lakes region (Zhong et al. 2016) and will likely result in changes to bluff recession patterns and processes. Furthermore, Roland et al. (2021) concludes that seasonality in lake level drives bluff recession; seasonal high lake level periods in the summer are associated with peaks in recession during that time. 10 Figure 3: Aerial image of home collapsing into Lake Michigan following bluff failure in late 2019. Image is from Fox2 Detroit website. However, similar to predicting future lake levels, it is also difficult to predict future storm frequency and intensity (Mortsch et al. 2013). Studies suggest it is likely that there will be an increase in future frequency of storm events in the Great Lakes region due to anthropogenically driven climate change impacts (Bechle et al., 2016; Catto et al., 2019). Warmer temperatures, caused by increased greenhouse gas (GHG) emissions, result in more evaporation, leading to an increase in atmospheric moisture and more intense precipitation (Winkler et al., 2012). The range of variability in increased intensity of precipitation events depends on the GHG emission scenarios used (Winkler et al., 2012). Model projections suggest higher emission scenarios produce a larger percentage of more intense precipitation events (Vavrus and Dorn, 2010; Mackey 2012). With the potential for an increase in storm event frequency and intensity superimposed upon rapidly fluctuating lake levels, it is important to understand how Great Lakes 11 bluffs respond to different storm event characteristics under varying lake level conditions to avoid future loss of property, infrastructure, and habitat (Figure 3). Human Disturbance Another factor that mediates the rate and extent of bluff recession is the degree of human disturbance. In response to erosion from high lake levels and storms, people have constructed numerous and often extensive coastal defense structures. Property owners and local, state, and federal agencies often construct groins and seawalls composed of wood, concrete, or steel, rock revetments, or conduct beach nourishment for coastal stabilization (Wood 1988). Coastal areas with such structures are termed “armored shorelines” and structures can be described as shore- parallel (sea walls and revetments) or shore-perpendicular (groins and jetties). Although armored shorelines can stabilize coastlines and may capture sediment, there are often unintended negative impacts that are not immediately apparent since they are gradual and/or are occurring beneath the lake surface (Dethier et al., 2016). The primary negative impacts of shoreline armoring include (1) trapping of sand on the updrift side of structures thus preventing sand accumulation on downdrift shorelines; (2) enhancing erosion downdrift of shoreline armoring due to sand starvation and/or scouring; and (3) eroding the lakebed nearshore of the structure, resulting in enhanced wave attack on the coast. Wood (1988) concluded that shore-perpendicular structures on Great Lakes shorelines have an impact on sediment transport in that beaches downdrift of the sea wall are starved of sediment as the structure will trap sand updrift of it. He also concluded that the outer sand bar on armored shorelines failed to reestablish itself post high lake level. Beletsky and Schwab (2008) created maps of climatological circulation in Lake Michigan and concluded that on annual timescales, currents on the eastern shoreline of Lake Michigan are dominantly oriented north to 12 south, resulting in sediment deposition on the north side of shore-perpendicular structures along this extent of the lake. It is important to note thought that there are both spatial and temporal variations in this general trend related to the geometry of the shore as well as seasonal and event hydrodynamics. Starvation of sediment downdrift of shore-perpendicular structures reduces the width of the beach fronting bluffs, resulting in more wave attack on the bluff and increased recession rates (Lin and Wu 2014). Shore-parallel structures, mainly seawalls and revetments, reduce the amount of sediment that is added to the littoral system from erosion, which deprives adjacent beaches of sand to replenish themselves (Griggs 2005; Lin and Wu 2014). Over time, the shoreline migrates landward to and potentially beyond a shoreline structure constructed on an eroding coastline (Griggs 2005). The result of this migration will be the gradual loss of beach lakeward of the structure as the water deepens and the shoreface moves landward. This process involves wave action actively eroding the lakebed and scouring out material, which prevents beach recovery after storms or when lake level falls (Ruggiero 2009; Morang et al. 2011; Lin and Wu 2014). Shore-parallel structures also result in scouring of adjacent, non-protected shoreline due to refraction as wave energy impacts the structure. This also leads to the gradual loss of beach and eventual possible failure of the structure (Griggs 2005). In response to these erosive processes, beaches narrow over time, which enhances bluff toe vulnerability (Lin and Wu 2014). Another form of human disturbance is the alteration of bluff landscapes for various reasons and uses that are beyond shore protection structures. Examples of this include roads running both adjacent and perpendicular to the bluff crest and grading out of the bluff for a beach entry road. Further, bluffs may be altered in an attempt to stabilize them where the material is graded to a lower slope angle, slope stabilization mechanisms established, and concrete-lined 13 channels installed to redirect the flow of surface runoff. To date, there are no studies that investigate the role such structures have on bluff retreat rates, however, it is reasonable to assume that such modifications act as a ‘stabilizing’ agent and reduce retreat rates. Great Lakes coastlines are increasingly being populated and urbanized, and it is predicted that this region will be a large center for population growth into the future (Amen et al. 2009). With increased population growth and urbanization, demand for armored shorelines will likely increase, which highlight the importance of understanding how armored and modified shorelines impact coastal erosion. Lithologic Characteristics Local lithologic characteristics of a bluff can influence the recession rate (Johnson and Johnston 1995). Great Lakes bluffs are composed mostly of unconsolidated glacial sediments that exhibit spatial variability in lithology (Jibson and Odum 1994). Under the same physical forces, bluff materials recede at different rates due to differences in their relative physical strengths. Previous studies of Great Lakes bluffs have reported varying conclusions on the influence of lithology on recession rates. Carter (1976) concluded that along the southern shore of Lake Erie, glaciolacustrine clay eroded fastest, bedrock eroded the slowest, and till recession rates were intermediate. Brennan and Calkin (1984) found that on Lake Ontario, bluffs composed of glaciolacustrine clay eroded slower than bluffs with sand or sand and gravel units. Johnson and Johnston (1995) reported that on western Lake Superior, recession rates were highest for coastal areas comprised of clay and water-laid sand and gravel, intermediate for sandy till, and lowest for bedrock. Swenson et al. (2006) found a strong relationship between recession rates of bluffs with comparable lithologies, i.e., bluffs made of similar material 14 exhibited similar recession rates. In contrast to this, Buckler and Winters (1983) reported no conclusive relationship between lithology and recession rates. The range of conclusions on the role of bluff lithology on recession rates is likely due to the difficulty in fully accounting for the three-dimensional spatial variability of lithology within unconsolidated glacial till bluffs, which often contain sandy and clayey units that are mixed, interspersed, and discontinuous across the bluff face. Variability in the susceptibility of a bluff sediment to erode may also be linked to lake level. Seibel (1972) contends that clay till bluffs retreat at higher rates than sandier bluffs during downward trends in lake level. Despite varying conclusions on the role of site-specific bluff lithology on recession rates, it is accepted that lithology does exert a control on recession, however the exact nature of this relationship is not yet well-constrained. Geomorphic Characteristics Geomorphic characteristics of a bluff, such as bluff height, bluff face slope, and bluff azimuth, also impact the rate of bluff retreat. Edil and Vallejo (1980) found that bluffs with a high bluff face slope will become unstable faster than one with a gentler slope. Conversely, Buckler and Winters (1983) reported no distinct relationship between bluff height and recession rates but suggest that steeper bluffs may exhibit higher magnitude episodes of erosion during which large amounts of material are removed due to oversteepening of the bluff by erosion at the toe. The most apparent relationship between recession rates and bluff height and face slope is that taller and steeper bluffs require more time to reach equilibrium as the crest must recede a larger amount than that of a shorter, more gently sloped bluff (Buckler and Winters, 1983). Adjacent shoreline azimuth of the bluff influences recession rate by affecting the amount of wave action that will reach the toe. The azimuth is also important in terms of the fetch 15 lakeward of the bluff. The larger the fetch, the longer wind can work against the lake surface and build large waves that increase erosion rates at the toe (Mickelson et al. 2004). Buckler and Winters (1983) found that bluff sites with higher recession rates are direct exposed to high- energy storm winds. Siebel (1972) concluded that shoreline azimuth, fetch, and associated large waves are crucial factors influencing bluff recession in that azimuth and fetch dictate the amount of wave exposure and energy at the bluff toe, respectively. In all, azimuth of the shoreline and adjacent bluffs coupled with fetch dictate the size and strength of waves at the bluff toe; azimuth determines the extent of exposure of the bluff toe to incoming wave action while fetch controls the distance and amount of time in which waves can gain power before reaching the bluff toe. Nearshore Bathymetry Beach and nearshore geomorphic characteristics also exert an influence on bluff recession rates. Davis et al. (1973) suggests that bluff recession along the eastern Lake Michigan shore is most attributable to the presence or absence of nearshore sandbars and human-built structures. Nearshore sandbars act as a barrier to dissipate incoming wave energy; without a sandbar to scatter incoming wave energy, there is a larger potential for wave erosion at the bluff toe. Additionally, nearshore lakebed downcutting, the product of lakebed erosion due to wave action, also contributes to bluff instability (Brown et al. 2005). Kamphius (1987) showed that nearshore wave processes may affect long-term rates of bluff recession in coastal areas where downcutting has created a deeper water condition as this allows greater wave energy to impinge upon the bluff toe. In contrast, Carter and Guy (1988) concluded that nearshore slope controls the frequency and intensity of waves before breaking on the beach; a steeper-sloped nearshore zone may dissipate incoming wave energy more than a gentler sloped nearshore zone. 16 Beach Morphology Morphologic characteristics of the beach fronting a bluff may influence recession rates since a beach acts as a protective barrier for the bluff toe by absorbing wave energy. With narrow beaches, resulting from beach erosion or simply from inundation during high lake level, waves can break directly on the bluff toe, resulting in a greater proportion of wave energy reaching the bluff toe and higher erosion potential (Carter and Guy 1988; Sorensen 1997). On wider beaches, wave energy is dissipated across the beach face resulting in either no or minimal wave energy reaching the bluff toe. Beach slope also mediates erosion at the bluff toe in that a more steeply sloping beach may prevent waves from reaching the bluff toe (Castedo et al. 2013). Overall, beach slope is an important control on the frequency and intensity of wave run-up impacting the bluff toe (Castedo et al. 2013). Interconnection of the Drivers and Mediators of Bluff Recession Bluffs along the Great Lakes shoreline recede at different rates in response to many multiple drivers and mediators, although a few are likely to be dominant at a given location. The variability in impact of erosional drivers is due to the variability in resisting factors (mediators), such as bluff lithology and beach and nearshore geomorphic characteristics. Many researchers have attempted to isolate the impacts of these drivers and mediators on bluff recession. However, this presents a challenge, as many of the processes are interconnected and superimpose upon each other, which can both attenuate and enhance the impacts. Many of the processes and characteristics that influence bluff recession rates are temporally dynamic, for example beach width, which changes seasonally. Such temporal variability makes it very difficult to assess the relationships between bluff recession rates and the various environmental mediators. Despite discrepancies in the literature surrounding the role of driver/mediator relationships in bluff 17 recession, it is accepted that to some degree, these factors all have an impact on bluff recession rates. With this, it is important to attempt to identify all variables associated with bluff recession and understand how each individually drive recession and how they superimpose upon each other. Mechanisms of Bluff Recession Once erosional forces initiate bluff recession, the style of recession or failure can happen in several ways. A fundamental mechanism for bluff failure in the Great Lakes is undercutting at the bluff toe (Carter and Guy 1988). Erosion and undercutting at the bluff toe are driven by incoming wave energy being expended on the toe, resulting in the removal of material, creating an unstable slope, leading to eventual failure (Carter and Guy 1988). Erosion and removal of bluff material by wave action occurs by abrasion, the gradual wearing away of bluff material by sediments entrained in the waves, and by quarrying, the removal of discrete pieces of bluff material; combined these two processes result in undercutting (Carter and Guy 1988). As undercutting occurs, the bluff slope becomes more vertical and less stable, resulting in eventual failure (Castedo et al. 2013). Displaced material from slumping accumulates downslope and acts as a temporary protective barrier against wave attack on the bluff toe; eventually, this dislodged material is removed by wave action and added into the nearshore system as littoral sediment (Edil and Vallejo 1980). Another mechanism of bluff recession is through creep, which is the slow removal and downslope movement of discrete pieces of material at the bluff crest (Zoet and Rawling 2017). This process occurs in locations with saturated or nearly saturated sediments on slopes that are steep enough that the force of gravity overcomes the frictional resistance of the material; this results in slow, internal deformational flow and the associated downslope movement of the 18 material. This process can occur at the crest or downslope of the crest where groundwater seepage along the bluff face removes material and moves it further downslope. Rates of surface runoff are controlled primarily by land cover, land use, and precipitation amounts (Maxwell and Miller 2004; Wang et al. 2020). Groundwater seepage is controlled by precipitation amount, infiltration capacity of the soil and the permeability of the bluff sediments (Montgomery 1998; Maxwell and Miller 2004; Kornelsen and Coulibaly 2014). Similar to toe undercutting, any bluff material dislodged by creep eventually can accumulate at the bluff toe or adjacent foreshore zone, where it is then entrained by wave action and added to the nearshore system as littoral sediment. Creep is one of several types of mass movement involving the transportation of material by gravity down a slope face (Dufrense and Davies 2009). Other than the slow progression of creep, most other mass movement processes are sudden, intermittent events that results in significant amounts of dislodged bluff material. Mass movement of bluff material usually occurs in the form of block falls, rotational and translational slumps, and debris flows (Carter and Guy 1988). A rotational slump (or landslide) occurs when the failure plane is rounded concavely upward, and the slumping movement is roughly rotational about an axis that is parallel to the ground surface and spans across the slump. A translational slump (or landslide) occurs when the slumping mass moves along a nearly planar surface with little to no rotation or backward tilting. A block fall (or slide), which is a translational slump, occurs when the slumping mass consists of a single unit or few associated units that move downslope as a congruent mass. Mass movement events can be triggered by spring ground thaws, as well as surface runoff and groundwater seepage (Zoet and Rawling 2017; Volpano et al. 2020; Roland et al. 2021). Further, mass 19 movement can be triggered by wave action undercutting the bluff toe which leads to eventual failure of the bluff face (Brown et al. 2005; Castedo et al. 2013). Carter and Guy (1988) concluded that bluffs made of Quaternary deposits exhibit a typical annual slope cycle in which spring thaws trigger mass wasting and accumulation of material at the bluff toe, which is then removed by wave action in the summer, resulting in undercutting at the toe. The undercutting increases slope instability leading to block falls in late summer and fall. This process results in a steeper, smoother slope that persists through the winter as shorefast ice protects the toe from wave action and freezing temperatures maintain the strength of the cohesive material (Carter and Guy, 1988). However, recent studies suggest that there is a delay or lag in bluff response to erosional driving forces (Castedo et al., 2013; Volpano et al., 2020; Roland et al., 2021). In a study by Castedo et al. (2013) where effects of wave erosion on bluff stability were examined by modeling erosion, they concluded that high lake levels promote bluff recession but at an inter-annual scale. The relationship is more complex at the seasonal scale where most bluff recession occurred in the summer, rather than the spring when water level was higher during the duration of the study. In a study by Roland et al. (2021) examining the seasonality of cold coast bluff recession processes, they concluded that the peak crest recession rates did not occur in summer when lake level was generally highest, but rather in late winter and spring. Volpano et al. (2020) reported that the largest mass wasting events occurred in spring during freeze-thaw events and that long-term above average lake levels are necessary to instigate large episodes of bluff recession. These various conclusions demonstrate that bluff recession is complex at seasonal and interannual timescales and that erosional processes should not be assumed to be constant nor linear. Constraining the relative role of individual driving forces and mediating factors on bluff recession rates, as well as identifying 20 reasons for spatiotemporal variabilities in the recession rates will help landowners and decision makers proactively plan for future bluff failures. Statement of Problem My thesis study investigated the complex linkages between bluff recession and fluctuating lake levels based on the hypothesis that coastal erosional processes are being exacerbated by recent above-average lake levels. The relationships between bluff recession rates and various site characteristics (i.e., shoreline protection structures, lithologic and geomorphic characteristics of the bluff, nearshore bathymetry, and beach morphology, etc.) were analyzed with respect to how driving forces and mediating factors interact and impact recession rates. This study sought to advance our understanding of bluff hazards by specifically looking at the processes driving and resisting change through space and time. Such enhanced understanding may contribute to more accurate predictions of future bluff recession rates. Short-term and long-term temporal variations in bluff recession rates were accounted for in order to better understand the complexity of recession and enhance our knowledge of how dynamic coastal processes irregularly change bluffs over time, rather than assuming that bluffs recede at a constant rate, which is a common practice due to inconsistencies in historical data. Large-scale spatial patterns in bluff recession were assessed by utilizing bluff sites spanning the Lake Michigan coast of Michigan. This research also aimed to contribute fundamental knowledge of the temporal component of processes associated with cliff failure. By better understanding past patterns of recession, future estimations of bluff recession can be improved. A broader goal of this study was to improve knowledge and management of these landscapes beyond the Great Lakes Region, since bluffs composed of unconsolidated sediments are not singular to this region. This study was designed to provide insight into how various factors, such 21 as rapid water level fluctuations and wave climate, as well as human disturbance, affect coastal bluffs. The findings from this study also contribute to methodological improvements related to data collection and analysis of bluff crest positions both in the Great Lakes Region as well as in areas with unconsolidated cliffs along marine coasts. Overall, the problem I examined was how do varying geomorphic and hydrodynamic variables, such as nearshore slope and wave energy, affect coastal bluffs in the presence of changing anthropogenic variables, such as shoreline protection structures. In total, seventeen sites along the eastern shore of Lake Michigan were analyzed (Figure 5). I hypothesized that higher energy wave climate and the geomorphic setting, such as nearshore slope and bluff face slope, at each bluff site will dictate bluff recession rates in that: (1) a steeper nearshore slope will increase recession rates, (2) a steeper bluff face slope will increase recession rates, (3) a higher bluff crest elevation will increase recession rates, (4) a higher toe elevation will decrease recession rates, (5) a wider beach width will decrease recession rates, (5) a steeper beach slope will decrease recession rates, (6) a shoreline azimuth that is subject to more frequent wave action will increase recession rates, and (7) a bluff and adjacent shoreline that has a higher maximum perpendicular fetch will have higher recession rates. Additionally, it is hypothesized that bluff sites with shoreline armoring will have decreased recession rates relative to those without. Furthermore, presence of nearshore geomorphic features will dictate recession rates in that if sand bars are present, less wave energy will reach the bluff toe as the energy is being expended earlier on the sand bar. Also, since lake levels are projected to remain high or increase in the near term, it is likely that bluff recession rates may accelerate due to the lagged response of bluffs to recessional driving forces. This scenario has significant implications for coastal planning and management at the state and local level. 22 Methods This research builds on a previous study conducted at Michigan State University by Dr. William R. Buckler during the 1970s (Buckler 1981). Within this study, Buckler resurveyed and analyzed 118 bluff sites along the Lake Michigan coastline in Wisconsin (62) and Michigan (56) and compared 1970s bluff crest positions to those denoted in General Office Land (GLO) surveys conducted in the early to mid-1800s. The objectives of Buckler’s study were to (1) delineate the extent of long-term bluff erosion at numerous Lake Michigan sites, (2) analyze bluff recession magnitude in relation to various bluff and lakeshore characteristics, and (3) identify spatial relationships of recession rates with respect to opposing shorelines and the hypothesis that long-term recession rates are greater on the eastern than the western shore (Buckler and Winters 1983). My study focused on the cohesive bluff sites along the Michigan Lake Michigan coast (17 of the 56 sites) from Buckler’s work. Summary of Buckler 1981 Methods Buckler focused on study sites along two segments of Lake Michigan Shoreline, extending northward from both the Illinois-Wisconsin border and Indiana-Michigan border, totaling in 118 sites. Study sites were selected by locations where a bluff crest intersected with U.S. Public Land Survey section lines. At this intersection, Buckler determined long-term bluff crest position by comparing the 1970s bluff crest location to those denoted by the General Land Office (GLO) surveys conducted in the early to mid-1800s. In this study, Buckler only utilized sites where bluffs were thought to exist; he defined a bluff as a “lakeward-facing steep bank or sharp slope composed of unconsolidated material landward of the shoreline”. Buckler argues that the crest of a bluff offers a reliable standardized line at which measurements can be made as the position is discernible and does not fluctuate to the same magnitude as water lines. 23 Measurements of bluff change were characterized as the landward displacement or lakeward accretion of the bluff crest. Bluff crest positions were measured using standard surveying techniques during the field seasons of 1976 and 1977; these crest positions were compared to those in the original GLO surveys, and total recession amount and average annual recession rates were calculated. The General Land Office surveys of Michigan and Wisconsin offer the earliest quantitative records of lake Michigan bluff crest position. In the surveys, distances from section and quarter section corners within one mile of Lake Michigan to the “meander line” are recorded. Powers (1958, pp. 98-90) argues that the GLO surveys do not specifically define what the “meander line” is, but it is highly unlikely that it was ever identified with the water line and that the measurements were clearly made at or near a bluff edge. For the sites in Buckler’s study, it was assumed that the meander line represented the lakeward bluff crest and all resurveys followed this rule. Possible site locations were eliminated by Buckler due to the uncertain relationship between the meander line defined by the GLO surveys and the bluff crest location in the 1970s. Resurveying and comparing these section line distances with the GLO measurements, long-term and average annual bluff crest changes were ascertained at places where the section line intersects the lakeshore bluff. Field measurements were obtained by measuring the distance along the section line using a 100-foot engineers steel tape and/or by stadia method in which a Philadelphia or stadia rod, and transit were utilized. Buckler followed standard surveying procedures as set by previous researchers (Davis, Foote, and Kelley, 1966; Brinker, 1969; Breed, Hosmer, and Bone, 1970) and some distances were acquired from earlier performed surveys by registered land surveyors (R.L.S.). Buckler reported the probable error in measurement being ranged from one foot in 24 5,000 feet for the R.L.S. and an error of approximately 0.25% or less for stadia method; with this, all measurements were to the crest of the lakeshore bluff. At sites where pedestrian or vehicular activity had made dips in the bluff crest so that a sharp change in slope was not discernable, Buckler carried the resurvey to an imaginary line connecting the bluff edge on either side of the site line. Furthermore, at sites where the bluff crest was rounded, a slightly arbitrary edge position was established, resulting in an estimated error of less than three feet, as reported by Buckler. On site selection, all section lines intersecting Lake Michigan within the study areas were investigated and 118 lines were resurveyed; bluff crest position changes were calculated by comparing to measurements from the GLO surveys. At each location, the following site observations were recorded: bluff composition, bluff height, shoreline orientation, groundwater and artificial drainage, beach width, presence of shoreline structures, bluff stability, and each site documented by photos. Methods for Subsampling Buckler 1981 Bluff Sites Buckler (1981 p. 61) grouped bluff site profiles into four general sedimentary categories: dune sand, water-laid sand, clay, and till. In his study and following publications, non-clay dominated sites were categorized as bluffs composed of dune sand or “dunal bluffs” being deposited by eolian processes (i.e., Buckler 1981, Abstract; Buckler and Winters 1983, p. 94), although this terminology does not exist elsewhere in literature on Great Lakes bluffs. Arnott and Guelph (2016) use the term “cohesive shoreline” to describe cliff coasts where the bluff profile and nearshore environment is composed of high silt and clay content sediments and describe that such shorelines account for about 40% of the lower Great Lakes shoreline. Zoet and Rawling (2017) describe bluffs along the Wisconsin coast as being composed of unconsolidated glacial 25 deposits. Jibson and Odum (1994) describe the bluffs used in their study as being deposited by late-Wisconsin glacial activity. Multiple publications describe Great Lakes bluffs as being of composed of high silt and clay content sediments deposited by late-Quaternary glacial activity is prevalent in the literature (Mickelson et al., 2004; Castedo et al., 2012; Volpano et al., 2020; among others). With this, any sites that are formed of predominantly non-cohesive material, such as those Buckler termed dunal bluffs, were removed from my thesis. Evolution of Methodology Efficiency over Time Throughout the progression of these surveys, from the original GLO surveys in the mid- 1800s, Dr. Buckler’s surveys in the late 1970s, and the modern surveys in 2019-2021, the general accuracy, precision, and efficiency of the surveys has improved. In the GLO surveys, distances from section and quarter section corners to where they intersected with a “meander line” is reported. The Land Ordinance of 1785 mandated that land be surveyed using the Rectangular Survey System and divided into Townships and Ranges; this method is still in use today (University of Oregon Libraries Website). These GLO surveys were conducted using iron chains for distance measurements and compasses for azimuth measurements (Osher Map Library) and the system of metes and bounds was used to categorize the land (University of Oregon Libraries Website). The standard chain measured 66 feet in length (~20.2m), equivalent to 4 rods of 16½ feet each (~5.03m), with each chain divided into 100 links and 1 link representing 0.66 inches (~0.017m) (citation). The GLO surveys were not originally conducted with the idea of bluff recession and coastal geomorphology in mind, however, were used to classify the land and offers the earliest quantitative records of Lake Michigan bluff line position. When resurveying these sites, Dr. Buckler used a transit and stadia rod and/or tape to measure distance from section corner to bluff crest location. The stadia is a surveying method of 26 measuring distances with a telescope and a graduated rod where the intersection of the stadia hairs in the eyepiece with the height on the distanced rod is used to measure distance (Breed and Hosmer 1977). The modern surveying methods utilize an RTK-GPS unit (Trimble R10-2 with GNSS Antenna; geospatial.trimble.com), a base station, and a rover receiver. This methodology is discussed in further detail in the next section below. Over the progression of these surveys denoting bluff crest location, the approaches have improved in terms of efficiency, physical effort needed, accuracy, and precision. In terms of inaccuracy, the use of iron chains by the GLO surveys was most inaccurate as the chains stretched with use where an error of one link (about 8 inches/20.32 cm) in 3 to 5 chains was considered normal (Virtual Museum of Surveying Website). To reiterate from a previous section, Dr. Buckler denoted the probable error in measurement being ranged from one foot in 5,000 feet for the R.L.S. (or 1m in 1524m) and an error of approximately 0.25% or less for stadia method (1 foot in 3250, or 1m in 990.6m) (Buckler 1981). Today, an approach called Real-time Kinematic GPS can be used, which is more accurate than these previous methods as it achieves centimeter- level positioning based off of the location of at least two GPS receivers, a base station, and a rover receiver. Even though accuracy of measuring bluff crest location has improved throughout time using these new methods, accuracy differences between new and old data must be considered when analyzing long-term patterns of bluff recession. Future studies on bluff recession will improve this richness of this dataset through using higher accuracy methods on older records of bluff crest location (i.e., aerial photographs) as well as through field surveys using high accuracy instruments (i.e., UAVs and RTK-GPS). Recent technological improvements will also allow for more field surveys and points of data collection to be completed as the amount of time and effort needed by the user to complete the work is greatly 27 reduced. This concern of historical data accuracy is the case with other disciplines, such as climate modeling and projecting (Winkler 2004), however these issues will resolve over time with technological improvements and continued effort in understanding bluff recession. Thesis Methods Property Owner Determination and Property Access Requesting In order to conduct field surveys of bluff crest location and other data collections (i.e., UAV flights, single-beam sonar boat surveys), property access permission needed to be granted by the property owners of each bluff site. First, the section corners used to measure bluff crest change at each site were identified using Dr. Buckler’s dissertation (Buckler 1981). These locations are represented in Table C2 on page 172 of Dr. Bucklers location as section line location and point of survey origin for the section corner (e.g. for M1: South Line/ Sec 4/ T5s, R19W and SE corner) These descriptions were then converted based on a dataset generated after the Survey and Remonumentation Act of 1990 (Michigan Legislature Website) to fit current PLSS surveying standards (e.g. for M1: 05S19WG03). A shapefile containing all the remonumented section corners available for the state of Michigan projected to Michigan GeoRef was received through personal communication with Dr. David Lusch at MSU (2019). Using the converted section corner locations, individual section corners were located using the Select by Attribute function and associated latitude and longitude positions (decimal degrees) were noted. The decimal degree coordinates of the selected set were converted to Universal Transverse Mercator (UTM), Zone 16 North. From these section corners, the E-W section line, which is in accordance with the northing component of the UTM coordinate, were prolongated in ArcGIS to locate where each survey line intersected the coastline. The coordinates of this intersection were 28 utilized to determine land ownership of the sites by consulting tax accessor websites from each county to determine the name of the landowner. Open-source websites were then used to find landowner phone numbers and the resident was contacted to ask for land access permission. These websites are county-specific tax accessor websites where, depending on the portal, latitude and longitude or address can be used to search for land parcels and associated owners (e.g., for M1: https://beacon.schneidercorp.com/Application.aspx?AppID=346&LayerID=4427&PageTypeID= 1&PageID=261). If the input of latitude and longitude was not possible for that specific county tax accessor website, the address was inputted and found by searching the coordinates in Google Maps. Property owner information was recorded, and phone numbers were located using the website www.Whitepages.com which is a pay-to-use online phonebook. Additionally, an official letter from the MSU Coastal Lab was drafted and sent to all 56 sites (this was prior to selected site refinement as discussed below). Prior to each field survey expedition, the appropriate landowner was contacted to confirm land access permission. Site Determination and Confirmation The fundamental methodology of this study is founded on the procedures carried out by Buckler (1981), however this study utilized higher accuracy approaches. Unlike Buckler’s (1981) research, this study only examined bluffs along the eastern shoreline of Lake Michigan and focused on a subset of 17 cohesive bluffs sites out of the original 56 sites that Buckler (1981) studied along this shore. For general geomorphic setting confirmation and validation (cohesive vs non-cohesive), all sites were plotted in ArcGIS on a 2016 map of Michigan dunes (Figure 4) obtained from the Michigan Environmental Council (MEC) website 29 (www.environmentalcouncil.org/valuing_michigans_coastal_dunes). Starting in 2016, this project was an amalgamation of earlier work from Dr. Arbogast, Dr. Nicholls, and Dr. Richardson (Michigan State University), community partners Elaine Sterrett Isely (West Michigan Environmental Action Council), Jonathan Jarosz (Heart of the Lakes) and Alek Kreiger (Ducks Unlimited), and a group of dune stakeholders from other organizations and the general public. The results of plotting the bluff sites on the 2016 Michigan dune layer showed that 32 sites fell within a dune area, indicating that they were not cohesive bluffs, the focus of this study. The final decision about keeping or removing sites was made by closely analyzing the site characteristics using both Google Earth Pro (www.google.com/earth/download/) and the USDA Web Soil Survey (websoilsurvey.nrcs.usda.gov) to discern whether visually or lithologically the site could be classified as a cohesive bluff. Ultimately, 39 of the original 56 sites were eliminated (likely to be dunes) while the remaining 17 sites (likely to be cohesive sediments) were kept for examination. These sites were then plotted on the world imagery base map in ArcGIS 10.6 to show general distribution of the sites along the eastern Lake Michigan coastline (Figure 5). 30 Figure 4: Location of sites from Buckler (1981) plotted on Michigan coastal dune locations. Green points represent sites that were kept for further analysis. Red points are sites that were eliminated from further analysis. The “X” symbols denote sites that had already been field evaluated prior to digital site evaluation. The coastal dunes polygon layer is an open-source data set from the Michigan Environmental Council website. 31 Figure 5: Map showing the location of the seventeen bluff sites that were examined in this study. Map generated in ArcGIS 10.6. RTK-GPS Field Surveys Field surveys to delineate present bluff crest locations were conducted throughout 2019 to 2021. A highly accurate RTK-GPS unit (Trimble R10-2 with GNSS Antenna; geospatial.trimble.com) with centimeter precision was utilized to measure bluff crest position at each site. This unit has 672 channels that can receive data from the U.S. Navstar, the Russian GLONASS, the European Galileo and the Chinese BeiDou Global Navigation Satellite Systems to determine instrument location (Langley 1998). RTK positioning is based on at least two GPS receivers, a base station, and a rover receiver. The local base station makes position estimates from all satellites in view and broadcasts both the station location and the satellite locations to the rover receiver (Langley 1998). The rover receiver simultaneously measures its position 32 relative to the in-view satellites and processes these data in real time with the base station data (Langley 1998). This RTK trilateral positioning technique offers highly accurate position data (8 mm horizontal /15 mm vertical). The rover utilized was the Trimble R10-2 with GNSS antenna and the bluff crest positions were collected and stored on a Trimble TSC7 Controller using the Trimble® Access™ software; both instruments were secured on a 2m rod whose height was accounted for in the data stored on the controller. At each site, bluff and site characteristics were determined and recorded including apparent dominant lithology, apparent bluff height, and presence of armoring/ protective structures. Photos were taken at each site, and the personal experience from landowners was documented. The appropriate northing coordinate corresponding to the section line was located using the Location RTK feature in the Trimble® Access™ application on the TSC7 Controller. This line was then followed to the bluff crest which was determined in the field as the most lakeward edge of the bluff before a drastic change in slope, which then denoted the bluff face. The RTK-GPS instrument was held on the bluff crest location until an accurate position was determined (vertical and horizontal precision under 3.5 cm). At site M43, additional bluff crest positions were taken approximately every 20 m for the length of the property, following the Buckler (1981) method. At the remaining 16 sites, only one data point from the RTK-GPS was recorded. Aerial Image Acquisition Historical aerial imagery was used to fill in time gaps between the GLO, Buckler (1981), and present field survey dates. Aerial images from the 1970s, 1980s, 1990s, 2000s, and 2010s for each site were acquired from open-source data portals, USGS Earth Explorer and the NOAA Data Access Viewer; these data portals provide access to imagery from many different image 33 acquisition programs throughout time. Aerial images from the 1970s were acquired from the Aerial Photography Single Frame Records (APSFR), which is a large collection of imagery collected by Federal organizations from 1937 to present. The APSFR contains over 6.4 million frames of imagery which are available as medium- and high-resolution digital products. Coverage of the APSFR is predominantly over the U.S., but some areas of Central America and Puerto Rico are also available. Individual images are not georeferenced and differ in scale, size, film type, and quality. Aerial images from the 1980s were acquired by the National High-Altitude Photography (NHAP) program, which was conducted by the USGS to obtain cloud-free aerial photographs at an altitude of 40,000 feet above mean ground elevation. Two camera systems were used to capture both black-and-white (BW) and color infrared (CIR) aerial photography over the U.S. The CIR images were taken with an 8.25-inch focal length lens and are at a scale of 1:58,000. The BW images were taken with a 6-inch focal length lens and are at a scale of 1:80,000. The NHAP program operated from 1980 to 1989 and the image archive contains approximately 500,000 images. All photographs were acquired on 9-inch film and centered over USGS 7.5- minute quadrangles. These images are not georeferenced. Aerial images from the 1990s was acquired from Digital Orthophoto Quadrangles (DOQs) generated by the USGS. A DOQ is a computer-generated image of an aerial image in which image displacement from terrain relief and camera tilt has been removed, merging image qualities of the original photograph with georeferenced qualities of a map. Like NHAP imagery, DOQs are available as BW and CIR, but also natural color. DOQs images have a 1-meter ground resolution. The USGS generated two types of DOQs: 3.75-minute (quarter-quad) DOQs and 7.5- 34 minute (full-quad) DOQs. DOQs are in a GeoTIFF format cast onto the UTM projection and referenced to NAD83. The most recent aerial photography from 2005, 2012, and 2018 was collected by the National Agriculture Imagery Program (NAIP) directed by the U.S. Department of Agriculture’s Farm Service Agency (USDA, FSA) through the Aerial Photography Field Office (APFO). The NAIP acquires aerial imagery with a resolution of 1-m ground sample distance (GSD) during agricultural growing seasons in the continental U.S.A.; the goal of this program is to make digital ortho photography available to governmental agencies and the public within a year of collection. The images are orthorectified which merges the image qualities of the aerial image with the georeferenced qualities of a map. Each image tile covers a 3.75-minute (quarter quad) with a 300-meter buffer. The tiles are available as 3-band natural color (red, green, and blue bands) or 4-band color infrared (red, green, blue, and near infrared bands) and could have as much as 10 percent cloud cover per tile. NAIP images are projected to UTM and referenced to NAD83. Aerial Image Georeferencing All aerial imagery that was not available as an orthophoto was georeferenced in ArcGIS 10.6 using the Georeferencing tool. Each image was projected to UTM, referenced to NAD83. For reference images, the corresponding NAIP 2018 imagery was used after being converted to a single raster dataset using the Create Raster Data Set (Data Management) tool in ArcGIS 10.6. For reference, the location of the bluff site in the particular image was added to the map using the ‘Add XY Data’ feature. Using the Georeference tab, control points were created on the un- georeferenced image and then added to the reference image to establish known coordinates for the control points in the un-georeferenced image. At least 7 control points were established, depending on the size of the un-georeferenced image. The control points and their RMSE values 35 were reviewed in the ‘View Link Table’ tab of the Georeferencing toolbar. Once an RMSE of under 0.75 m was achieved, the accuracy was visually cross-referenced using the Swipe tool on the Editor toolbar. In addition, imagery layers were repetitively turned on and off to analyze if the site XY position moved. Once a satisfactory accuracy was achieved, the newly georeferenced image was saved and could then be analyzed. This process was completed for all un- georeferenced aerial imagery. Slight adjustments were also made to the DOQ and 2005 NAIP images whenever georeferencing errors were detected. Digital Shoreline Analysis System The Digital Shoreline Analysis System (DSAS) v 5.0 (www.usgs.gov/centers/whcmsc/science/digital-shoreline-analysis-system-dsas ) was used to calculate rate-of-change statistics over time at each site (Himmelstoss et al. 2018). The DSAS is a software add-in to Esri ArcGIS desktop 10.4 - 10.6 and was developed by the U.S. Geological Survey’s Coastal Change Hazards project (Himmelstoss et al. 2018). The software calculates rate-of-change statistics from multiple historical shoreline positions that are digitized in ArcGIS by the user (Himmelstoss et al. 2018). The DSAS offers a strong selection of regression rates in a consistent and reproducible method that simplifies the shoreline change-calculation process and produces rate-of-change information and statistical data essential to verify the reliability of the computed results (Himmelstoss et al. 2018). The USGS DSAS website states that the “software is also suitable for any generic application that calculates positional change over time, such as assessing change of glacier limits in sequential aerial photos, river edge boundaries, or land- cover changes”, thus it is appropriate for analyzing coastal bluff crest positional change over time. The methodology remains largely constant between applications though there are slight adjustments in user-interpretation and approach which are outlined below. 36 The DSAS was downloaded, installed, and added as a toolbar in Esri ArcGIS 10.6. The appropriate aerial image was then added as a layer to the map in ArcGIS 10.6 (Himmelstoss et al. 2018). In ArcCatalog, a blank polyline shapefile was created for the appropriate image and was projected in UTM, referenced to NAD83, and added to the map. Using the Editor toolbar, the extent of the bluff crest was digitized following two guidelines: (1) the bluff crest is the furthest evident edge before a change in slope denoting the bluff face and (2) if image quality was poor, there was dense vegetation, and/or it was hard to discern where the crest was due to displaced bluff material, the crest position must not exceed the most lakeward extent of the prior timestep that was digitized. The second guideline follows the concept that the bluff crest will never accrete. All bluff crests at the 17 sites were digitized for the years 1975-78, 1986-88, 1999, 2005, 2012, and 2018, depending on the availability of the imagery. After the bluff crest on each aerial image per site was digitized, the shapefile was imported into a personal geodatabase as required by the DSAS (Himmelstoss et al. 2018). From that geodatabase, all digitized crest polylines were brought into a new map in ArcGIS. In the DSAS toolbar, the Attribute Automator was utilized to add Data Field Name (DSAS_date) and Uncertainty Field Name (DSAS_uncy) fields to the polyline attribute tables (Himmelstoss et al. 2018). The collection dates of the aerial images were then added to the attribute table using the editor toolbar and followed the MM/DD/YYYY format (Himmelstoss et al. 2018). Following this, the Merge (Data Management) tool was used to merge all polylines into one layer (Himmelstoss et al. 2018). In ArcCatalog, a new blank line feature class was created with the appropriate coordinate system in the same personal geodatabase and was added to the map (Himmelstoss et al. 2018). Using the Add Data from ArcGIS Online tab, the Michigan PLSS Quarter-Quarter Sections map 37 from the Michigan GIS Open Data Portal (gis-michigan.opendata.arcgis.com) was added. The section corner, GLO, Buckler (1981), and 2020 survey points (if applicable) were added to the map using the Add XY Data feature (Buckler 1981). The PLSS quarter-quarter section map was used as reference to create a DSAS baseline and transects (Himmelstoss et al. 2018). Using the trace function on the Editor toolbar, a baseline was created for the extent of the bluff, following the easting aspect of the PLSS grid (Himmelstoss et al. 2018). Using the Attribute Automator on the DSAS toolbar, an ID Field Name (DSAS_ID), Group Field Name (DSAS_group), and Search Distance Field Name (DSAS_search) fields were added to the baseline attribute table; the DSAS_ID and DSAS_group fields were assigned a value of 1, per the DSAS Manual (Himmelstoss et al. 2018). Following this, the Set Default Parameters tab on the DSAS toolbar was utilized to assign the baseline and shoreline (the merged polyline layer) settings (Himmelstoss et al. 2018). The Cast Transects tab on the DSAS toolbar was then used to establish transects that intersected the shoreline layer in order to calculate rate of change (Himmelstoss et al. 2018). Using the editor toolbar, the transects were arranged so that the central transect line intersected the shoreline layer and the added XY points (GLO, Buckler (1981), and 2020 survey points) (Buckler 1981). Following this adjustment, four transects to the north and south of this survey line were placed approximately 25 m apart, totaling 200 m of bluff crest extent and 9 transects to be analyzed using DSAS (Figure 6) (Himmelstoss et al. 2018). Lastly, the Calculate Rates tab on the DSAS toolbar was utilized to calculate rate-of- change statistics and the table that contained intersection data was converted to an Excel table for further analysis (Himmelstoss et al. 2018). The Excel table was then organized for evaluation and crest position changes were calculated for each of the 9 transects. These values 38 were structured into a summary table and then computed into the average and standard deviation of all the transects for each time step (Table 1). The averages were then divided by the length of time between images in order to normalize the rates. Figure 6: Results of DSAS analysis on site M42 generated in ArcGIS 10.6 showing digitized shorelines (blue), a baseline (black), GLO, Buckler (1981) and 2020 survey points, and section corner (green points, left to right respectively), transects (red), and transect intersection points (red point), superimposed over 2018 NAIP imagery. Time Step Time step Average STDEV +/- Transect 1839-1977 1977-1986 1986-1999 1999-2005 2005-2012 2012-2018 1839-1977 -1.595652174 0 1 -3.054779047 0.957410742 -1.514689028 -0.444495195 -1.681134644 1977-1986 -2.518575889 2.497257828 2 0.318101229 -0.255803827 -0.547850785 -0.51121198 -2.012491775 1986-1999 -1.166620616 0.929228302 3 -0.578880543 -1.200171233 0.309684942 -0.014467094 -0.575774385 1999-2005 -1.043623328 0.556141078 4 -3.007918292 -1.893867 -1.413922042 -0.453753214 -0.595785523 2005-2012 -0.456561194 0.566719094 5 -1.595652174 -8.930192165 -2.380787238 -1.25201636 -0.098473231 -0.594859742 2012-2018 -0.925326617 0.504934858 6 -2.680317966 -1.49563782 -0.939467848 0.587371931 -0.775984059 2018-2020 -1.5785279 0 7 -1.577732072 -1.589273176 -1.283174564 -0.549162551 -0.691618686 8 -1.577732074 -1.106845446 -1.463101755 -1.299638629 -0.807687939 9 -1.577732074 -1.534610543 -1.28807251 -1.325220781 -0.592602796 Table 1: Summary table of rate-of-change at site M42 in Manistee, Michigan. Transect 5 includes the GLO and Buckler (1981) points. To the right is the table denoting the average of all transects per timestep and the associated standard deviation. For the 1839-1977 timestep, the total change calculated by Buckler is normalized by decade to follow the near-decadal pattern of the more recent time steps. For the 1839 - 1977 and 2018 - 2020 time steps, the standard deviation is zero since there is only one data point for each time step. 39 This process was repeated for all 17 bluff sites. Post completion, all data were organized into an Excel spreadsheet showing the magnitude of change and magnitude of change normalized by time. These data were then plotted in Grapher(www.goldensoftware.com). Four plots per site were created: crest position since the GLO survey to present, crest position since the Buckler (1981) survey to present, total magnitude of change, and rate of change normalized by time. Missing data were filled in, as needed, which is discussed below. Further, normalized rates from the 1970s-present were compared to normalized rates from the mid-1800s to 1970s to analyze if rates are increasing, decreasing, or remaining the same over time, relative to those time steps. Then, normalized rates were subdivided into smaller time steps to analyze rate of change at finer time scales. The data were also analyzed and examined for spatial and temporal patterns of bluff recession and were used to explore possible linkages between erosional driving forces and rate of recession. Bluff Crest Extraction from Topobathy LiDAR Not all landowners could be contacted or did not grant permission to conduct the RTK- GPS field surveys. At these three sites (M4, M5, M6), 2020 topobathy LiDAR from the Joint Airborne Lidar Bathymetry Technical Center of Expertise (JALBTCX) of the USACE was utilized to digitize the bluff crest. For the three sites (M4, M5, M6) where RTK-GPS field surveys were not feasible, the appropriate topobathy LiDAR was brought into ArcGIS 10.6 and the 2020 bluff crest was digitized using the same method described in the previous section. These digitized crests were then added to the DSAS analysis for that site and rate of change was computed and added to the graphs. 40 Lithology The fundamental lithology of the bluffs was assessed analyzing Dr. Buckler’s stratigraphic descriptions collected during field surveys in the 1970s (Buckler 1981 Table B2). The according bluff stratigraphy from Buckler 1981 was input into Excel, and layer extents converted to meters; these values were then to describe site lithology and general stratigraphy at the bluff toe, along the face, and crest, depending on the extent of each unit and the total height of the bluff. Lithology was further examined by plotting the bluff site locations on the digital version of the Quaternary Geology of Southern Michigan (Farrand and Bell 1982). Using the identify tool, the polygon containing each bluff site was determined and the Quaternary material was noted from the attribute table. The polygons recorded were the one that overlapped with the bluff site, or the most lakeward, unit that would be representative of the bluff. However, it must be noted that Quaternary Geology of Southern Michigan map was compiled at a scale of 1:500,000, which may influence the accuracy of the results of plotting the bluff sites on the layer, resulting in misclassified lithologies if the bluff site is not plotted in the correct location; small- scale maps preserve less actual detail of the features. The printed Quaternary Geology of Southern Michigan (Farrand and Bell 1982) was utilized for more descriptive unit explanations. Results from Dr. Bucklers stratigraphic analysis and from Farrand and Bell (1982) were then compared to denote extent of similarity, however the analysis from Dr. Buckler was utilized as the main source of data for lithologic analysis, with Farrand and Bell (1982) as a supporting dataset. Hardened Shoreline Analysis To examine the proximity of study bluff sites to shoreline protection structures and other shoreline structures such as marina jetties or breakwaters, a vector data set containing hardened 41 shoreline classifications for the U.S. Great Lakes shoreline was imported into ArcGIS 10.6. This dataset (coast.noaa.gov/digitalcoast/data/hardened-shorelines.html) was created by NOAA in 2019 by digitizing the shoreline using NAIP imagery from 2014 through 2017. These shoreline vectors were compared to large-scale, oblique imagery (Great Lakes Shoreviewer) and shoreline segments are classified as either artificial or natural, and artificial shoreline segments were further classified by structure type and condition. The structure types include revetments, seawalls/bulkheads, groins, jetties, and ad hoc concrete rubble/rip rap. The bluff site locations were displayed on the Great Lakes Hardened Shorelines Classification layer in order to delineate the presence and condition of hardened shoreline segments immediately adjacent to each site. Using the 2018 NAIP imagery and the ArcGIS 10.6 measure tool, the distance of each of the 17 bluff sites to marina structures (jetties, breakwaters, etc.) was measured shore-parallel in order to imitate the dominant longshore current.. Furthermore, the historical aerial imagery described above (NAIP, NHAP, DOQ, APSFR) was also utilized to estimate when the shoreline structures were built. Nearshore Bathymetry and Beach Topography Extraction In order to examine the relationship between nearshore bathymetry and bluff recession, bathymetric data from the Joint Airborne Lidar Bathymetry Technical Center of Expertise of the USACE (JALBTCX) and from the Michigan State University Coastal Lab were utilized. Topobathy LiDAR for 2012 from JALBTCX was downloaded from the open-source NOAA Data Access viewer (coast.noaa.gov/dataviewer/). Preliminary topobathy LiDAR from JALBTCX for 2020 was received through personal communication with the USACE. Nearshore bathymetry for 2012 was examined at all sites, however, due to data extent limitations, the 2020 bathymetry could only be examined at ten of the seventeen sites. For the 42 2012 data set, the appropriate topobathy LiDAR was loaded into the Golden Software’s Surferprogram (www.goldensoftware.com) and nearshore topographic profiles were extracted on the survey line, 100 m north, and 100 m south, totaling 3 nearshore topographic profiles which were spatially in agreement with the DSAS transect methodology (Figure 7). Additional nearshore bathymetry was collected using single-beam sonar at four selected sites (M42, M43, M44, and M45) to supplement the 2020 bathymetry dataset. These data were imported into Surfer where they were cleaned, and profiles were extracted (Figure 8). For site M2, the 2012 topobathy LiDAR data was inadequate, so 2008 data was utilized (also downloaded from the NOAA Data Access viewer). In Grapher, both the 2012 and 2020 profiles were plotted against the mean water level of the Lake Michigan Basin during the time of data collection. The mean water level data from the buoy located off the coast of Ludington, MI (Station 9087023) was obtained from NOAA’s Center for Operational Oceanographic Products and Services (CO- OOPS) tide and water level data website (tidesandcurrents.noaa.gov/products.html). The plots in Golden Software’s Grapher were visually interpreted to assess nearshore morphological change over time and to locate nearshore and beach morphological features such as outer sandbars, inner sand bars, and berms. Though the 2020 bathymetry data were not examined at all sites due to time and data access constraints, bathymetry from single-beam sonar should be collected at all sites in any future studies. 43 Figure 7: Locations of the profiles for site M43 shown on the 2008 topobathy LiDAR (JALBTCX). Figure 8: Nearshore bathymetry map generated from single-beam sonar data collected by the MSU Coastal Lab in October of 2020. The dotted line shows the track of the survey vessel. 44 Beach, Bluff, and Nearshore Morphometric Extraction To examine the role that local morphodynamics plays in bluff recession, all sites were analyzed for various characteristics using the 2012 topobathy LiDAR from JABLTCX. Site parameters examined include nearshore slope, bluff face slope, bluff toe elevation, bluff crest elevation, beach width, beach slope, and maximum perpendicular fetch distance. Nearshore slope was calculated (rise/run) by determining the distance along the profile between the upper elevation limit of 176.1 m and the lower elevation limit taken to be 170 m. Bluff face slope was calculated (rise/run) by comparing the bluff toe position (X,Y,Z) with the bluff crest location (X,Y,Z). Bluff toe and bluff crest elevations were extracted from the profile. Beach width and slope were calculated by measuring the distance between the intersection of the mean water line (176.1 m) with the profile and the bluff toe position (X,Y,Z). These width measurements and slope calculations were repeated for each profile (south, middle, north) at each site and the averages of these values were used to represent the beach width and slope parameters of that site. These 2012 beach morphodynamics were plotted against the 2012- 2020 average annual rates normalized by time (grouping of 2012 - 2018 and 2018 - 2020 values) for that given site to examine the relationship between the 2012 beach morphodynamics and the 2012 - 2020 bluff crest response. The relationship between recession rates and wave energy was analyzed by quantifying the daily mean wave energy for the 2012-2020 period by utilizing nowcast wave data which contains hourly wave climate data at given sites in the Great Lakes region (NOAA GLERL 2021). Equation 1 can be used to calculate wave power where E is mean daily wave energy (m2), N is total number of hours with significant wave height > 0.5 ms-l, H ἰ is significant wave height for hour ἰ, and m is number of days in measurement period (Amin and Davidson-Arnot 1995). 45 Equation 1: Additionally, each site was qualitatively analyzed for presence of nearshore and beach features such as outer sandbars, inner sandbars, and berms by examining the extracted topographic profiles relative to the mean water level at the time of data collection. Again, all sites were analyzed for all features in 2012, select sites were analyzed for nearshore features in 2020, and all sites were examined for the presence of a berm in 2020 using Google Earth Pro and 2018 NAIP imagery. Lastly, shoreline azimuth and associated maximum perpendicular fetch was calculated by using the Linear Directional Mean tool in ArcGIS 10.6 to delineate the approximate azimuth of the shoreline measured clockwise from 270 degrees to avoid magnitude problems associated with using 0 degrees/ 360 degrees as due north. From that azimuth, maximum perpendicular fetch was calculated by measuring the maximum distance before hitting land exactly perpendicular to the azimuth. Linear regression analysis was performed on each morphodynamic characteristic. 46 Results Bluff Recession Rates Rates of bluff recession at the seventeen study sites along the eastern shoreline of Lake Michigan were calculated on near-decadal time scales ranging from the 1970s to present. At each site, bluff recession rates either generally increased, decreased, or remained relatively constant over time. Table 2 summarizes bluff crest position change over time at each of the seventeen sites where the earliest documented bluff crest position from the mid-1800s is represented as the starting point (i.e., zero) and consecutive negative numbers represents the recession, or landward movement of the bluff crest. Table 3: shows the magnitude normalized by time of all sites from GLO survey to Dr. Buckler’s survey (mid-1800s-1977) compared to Dr. Buckler’s survey to present (1977-2020), as well as the difference between the rates and if they are increasing, decreasing, or remaining constant over time based on the threshold of +/- 0.25m of change. From this comparison, 7 sites are increasing over time (M1, M5, M9, M14, M37, M38 and M43), 3 are decreasing over time (M2, M4, and M6), and 7 are remaining constant (M3, M10, M42, M44, M45, M54 and M55); with this, there is large spatial distribution of these groupings along the Lake Michigan shoreline. Figure 9 shows the magnitude normalized by time of all sites from the GLO survey to Dr. Buckler’s survey (mid-1800s-1977) compared to Dr. Buckler’s survey to present (1977-2020) and shows the magnitude of difference between rates. Tables 2 and 3, and figure 10 show results that indicate bluff crests moved landward over time at all seventeen sites, indicating bluff recession as rates never surpassed zero, which would indicate accretion; however, the rates and magnitudes of bluff recession varied spatially and temporally, and these patterns can be analyzed at finer temporal scales and examined for spatial patterns. Figure 10 47 shows the magnitude of change normalized by time at all seventeen bluff sites at near-decadal time steps. M1 M2 M3 M4 M5 M6 Year Crest Position Year Crest Position Year Crest Position Year Crest Position Year Crest Position Year Crest Position 1829 0.000 1829 0.000 1829 0.000 1829 0.000 1830 0.000 1830 0.000 1977 -84.770 1977 -97.390 1977 -55.580 1977 -192.580 1977 -131.030 1977 -105.550 1986 -118.751 1986 -101.988 1986 -57.130 1986 -202.832 1986 -146.315 1986 -109.200 1997 -120.912 1997 -105.193 1997 -63.961 1997 -221.171 1997 -157.783 1997 -114.292 2005 -122.509 2005 -106.344 2005 -65.003 2005 -222.606 2005 -159.604 2005 -116.693 2012 -125.389 2012 -106.986 2012 -66.034 2012 -225.234 2012 -173.216 2012 -120.210 2020 -127.695 2020 -111.370 2020 -67.870 2020 -235.939 2020 -183.557 2020 -122.738 M9 M10 M14 M37 M38 M42 Year Crest Position Year Crest Position Year Crest Position Year Crest Position Year Crest Position Year Crest Position 1831 0.000 1831 0.000 1831 0.000 1831 0.000 1838 0.000 1839 0.000 1977 -59.730 1977 -68.350 1977 -41.180 1977 -33.950 1977 -26.430 1977 -22.020 1986 -67.475 1986 -70.291 1986 -46.966 1986 -41.866 1986 -31.053 1986 -24.539 1997 -81.369 1997 -71.357 1997 -56.628 1997 -62.005 1998 -47.025 1997 -25.705 2005 -85.339 2005 -73.778 2005 -58.302 2005 -65.681 2005 -49.372 2005 -26.749 2012 -87.716 2012 -75.370 2012 -60.625 2012 -69.095 2012 -50.704 2012 -27.205 2020 -89.457 2020 -77.902 2020 -64.375 2020 -73.322 2020 -53.500 2020 -29.709 M43 M44 M45 M54 M55 Year Crest Position Year Crest Position Year Crest Position Year Crest Position Year Crest Position 1847 0.000 1847 0.000 1847 0.000 1850 0.000 1851 0.000 1977 -75.870 1977 -105.030 1977 -39.320 1977 -15.270 1977 -58.540 1986 -84.369 1986 -112.883 1986 -41.441 1987 -17.902 1987 -61.000 1999 -90.775 1999 -118.781 1999 -42.785 1999 -19.699 1999 -62.358 2005 -94.176 2005 -122.767 2005 -43.898 2005 -21.879 2005 -62.876 2012 -95.731 2012 -125.988 2012 -44.624 2012 -24.347 2012 -63.677 2020 -112.720 2020 -127.593 2020 -48.890 2020 -27.218 2020 -65.317 Table 2: Results from DSAS showing bluff crest position change over time in meters. The first documented bluff crest position (mid-1800s) is represented as zero. Negative number indicates landward movement of bluff crest. 48 Magnitude Normalized by Magnitude Normalized by Site no. Time Time Difference Result (mid-1800s-1977) (m/yr) (1977-2020) (m/yr) M1 -0.573 -0.998 0.425 Increasing M2 -0.658 -0.325 -0.333 Decreasing M3 -0.378 -0.286 -0.092 Constant M4 -1.311 -1.008 -0.303 Decreasing M5 -0.890 -1.222 0.332 Increasing M6 -0.728 -0.400 -0.328 Decreasing M9 -0.408 -0.691 0.283 Increasing M10 -0.469 -0.222 -0.247 Constant M14 -0.283 -0.539 0.256 Increasing M37 -0.244 -0.916 0.672 Increasing M38 -0.189 -0.630 0.441 Increasing M42 -0.158 -0.179 0.021 Constant M43 -0.582 -0.857 0.275 Increasing M44 -0.808 -0.686 -0.122 Constant M45 -0.302 -0.223 -0.079 Constant M54 -0.119 -0.292 0.173 Constant M55 -0.009 -0.158 0.149 Constant Table 3: Magnitude normalized by time of all sites from GLO survey to Dr. Buckler’s survey (mid-1800s-1977) compared to Dr. Buckler’s survey to present (1977-2020), as well as the difference between the rates and if they are increasing, decreasing, or remaining constant over time based on the threshold of +/- 0.25m of change. Figure 9: Magnitude normalized by time of all sites from GLO survey to Dr. Buckler’s survey (mid-1800s-1977) compared to Dr. Buckler’s survey to present (1977-2020). 49 Figure 10: Magnitude of change normalized by time in meters per year for all seventeen bluff sites. M44-projected represents projected rate of change for modern time step due to shoreline modification at site. Of the seventeen sites, eight locations had recession rates that generally increased over time (Figure 11). Sites with generally increasing recession rates include M2 (Berrien County), M4 (Berrien County), M5(Berrien County), M10 (Allegan County), M14 (Allegan County), M42 (Manistee County), M43 (Manistee County), and M45 (Manistee County). Five sites (M2, M4, M5, M10, and M14) are located along the southern to lower-middle portion of the eastern Lake Michigan shoreline. The remaining three sites with increasing recession rates (M42, M43, and M45) are located along the upper-middle portion of the eastern Lake Michigan shoreline. Of the eight sites with increasing recession rates, four sites had peaks in recession rates in the 1970s to 1990s (M4, M5, M14, and M43) three of which (M4, M5, and M14) are located along the southern to lower-middle shoreline and one (M43) is located along the upper-middle portion of 50 the eastern Lake Michigan shoreline (Figure 13). The remaining four sites (M2, M10, M42, and M45) with increasing recession rates did not have a peak in recession rates in the 1970s-1990s, but rather began to increase over the past decade (since the late 2000s). Two of these sites (M2 and M10) are located along the lower-middle shoreline, and the other two (M2 and M45) are located along the upper-middle portion of the shoreline (Figure 13). Figure 11: Bluff sites with recession rates in meters per year that are generally increasing over time. Figure 12: Bluff sites with recession rates in meters per year that are generally increasing that had a peak in rate of recession in the 1970s-1990s. 51 Figure 13: Bluff sites with recession rates in meters per year that are generally increasing that did not have a peak in rate of recession in the 1970s-1990s. Of the seventeen sites examined in this study, nine sites had recession rates that are generally decreasing or remaining relatively constant over time (Figure 14). Sites with generally decreasing or relatively constant recession rates include M1 (Berrien County), M3 (Berrien County), M6 (Berrien County), M9 (Allegan County), M37 (Mason County), M38 (Mason County), M44 (Manistee County), M54 (Leelanau County) and M55 (Leelanau County). Four of these sites with decreasing or relatively constant recession rates (M1, M3, M6, and M9) are located along the southern to lower-middle portion of the eastern Lake Michigan shoreline. Three sites with decreasing or relatively constant recession rates (M37, M38, and M44) are located along the upper middle portion of the eastern Lake Michigan shoreline. Two sites with decreasing or relatively constant recession rates (M54 and M55) are located along the northern portion of the eastern Lake Michigan shoreline. Of the sites with decreasing or relatively constant recession rates, six sites had peak recession rates in the 1970s-1990s (M1, M3, M9, M37, M38, and M44) with three located along the southern to lower-middle portion (M1, M3, and M9) and three located in the upper-middle portion (M37, M38, and M44) of the eastern Lake Michigan shoreline (Figure 5). The remaining three sites (M6, M54, and M55) with decreasing or near-constant recession rates did not have a peak in recession rates in the 1970s-1990s but 52 appear to have slightly increasing recession rates during the current period of high lake level. One of these sites (M6) is located along the southern portion and two are located along the northern portion (M54 and M55) of the eastern Lake Michigan shoreline (Figure 5). Figure 14: Bluff sites with recession rates in meters per year that are generally decreasing or near-constant over time. M44-projected represents the projected modern recession rate due to shoreline modification at the site. Figure 15: Bluff sites with recession rates in meters per year that are generally decreasing or remaining constant that had a peak in rate of recession in the 1970s-1990s. 53 Figure 16: Bluff sites with recession rates in meters per year that are relatively constant that did not have a peak in rate of recession in the 1970s-1990s. Lithology To analyze the lithology, Dr. Buckler’s stratigraphic descriptions of the sites were utilized (Buckler 1981 Table B2). Table 4 shows the key and unit descriptions associated with the stratigraphic analysis from Buckler (1981 Table B2) and table 5 shows the redrawn stratigraphic analysis from Buckler (1981 Table B2) with unit extents converted to meters. Table 6 shows the interpreted results of the stratigraphic analysis in terms of bluff toe lithology, bluff face lithology, and bluff crest lithology. Predominantly, the sites are composed of water-laid sand and till (M1, M2, M3, M4, M5, M6, M10, M14, M37, M45, M44, M45, M54, and M55), two sites are composed of water-laid sand and clay (M9 and M38) and one site composed of dune sand (M42). Further, all bluff sites were plotted on the Quaternary Geology of Southern Michigan map (Farrand and Bell 1982) in ArcGIS 10.6 as a supporting dataset to Buckler’s stratigraphic analysis (Figure 17). Table 7 shows the interpreted results of plotting the bluff sites on Quaternary Geology of Southern Michigan map (Farrand and Bell 1982) as the Quaternary material that each site overlapped with. Figures 21 through 37 show oblique aerial imagery of each site to serve as visual aid for lithologic analysis. Overall, the results show comparable lithologies of varying sizes of Quaternary material deposited by glacial and lacustrine activity. 54 However, it should be noted that these stratigraphic and lithologic mappings of the bluff sites and the southern portion of Michigan were completed in the 1970s and 1980s and the stratigraphy may have changed since the time of completion due to bluff erosion. Further, though the lithologies at large are comparable, there are likely small-scale variabilities that are not represented in these descriptions, that likely play a role in dictating how a site evolves. Key Description DS Dune sand; eolian deposits of sand size particles. Water-laid sand; water-deposited sand size particles, with and without pebbles, WS and to include thin interbedded zones with high percentage of clay or silt size particles. CL Clay; water-deposited sediments o f a clay or silty - clay texture. Till; non-stratified, non-sorted glacially deposited sediments which at the study sites T are normally of a clay loam texture and which usually include pebbles and/or cobbles. COV Covered; the bluff stratigraphy is obscured by overburden. Foredune; the sand dune immediately behind the backshore and fronting the primary bluff. FD This feature tends to be ephemeral; during higher water periods it generally undergoes erosion while during lower lake elevations it tends to undergo accretion. Previous foredune; a foredune was present at the beginning of the PREV. FD present high water period in 1968 but had eroded completely by 1976-77. Remnant of a foredune; only the very last portion of a foredune remains and this may be spaced REMM. FD intermittantly along the lakeshore segment between points where bluff toe erosion has begun. Table 4: The key and unit descriptions associated with the stratigraphic analysis of sites as redrawn from Buckler (1981 Table B2). 55 Site no. Bluff Stratigraphy M1 17.6784m WS 2.286m WS/ 3.81m T/ 1.524m WS/ M2 1.2192m T/ 9.4488m WS/ 4.572m COV M3 3.048m WS/ 10.9728m T/ 18.8976m WS M4 3.6576-4.572m T/ 25.908-28.956m WS M5 1.524m WS/ 16.4592m T/ 15.5449 WS 2.1336m WS/ 10.0584m T/ 9.4488m WS/ M6 w/ REMM. FD M9 2.7432m WS/ 3.6576m CL/ 15.5448m WS M10 6.096m WS/ 2.7432m T/ 4.572m WS/ 7.62m T M14 5.7912m WS/ 14.6304m T/ PREV. FD M37 19.812m WS/ 30.48m T M38 34.7472m WS/ 7.0104m CL/ 3.6576m COV M42 8.8392m DS/ w/ FD M43 4.572m T/ 8.2296m COV (prob WS)/ 7.9248m WS M44 4.8768m T/ 14.0208m WS M45 0.6096-3.048m T/ 3.048-12.192m WS/ 3.048-12.192m T/w REMM. FD M54 9.144m T/ w/ REMM. FD & DS veneering slope M55 12.4968m T/ 19.2024m WS Table 5: Bluff stratigraphy of the seventeen sites, redrawn and modified from Buckler (1981 Table B2) 56 Site no. Crest lithology Face lithology Toe lithology M1 Water-laid sand Water-laid sand Water-laid sand M2 Water-laid sand and till Water-laid sand and till Water-laid sand and dislodged material M3 Water-laid sand Till Water-laid sand M4 Till Water-laid sand Water-laid sand M5 Water-laid sand Till Water-laid sand M6 Water-laid sand Till Water-laid sand M9 Water-laid sand Clay Water-laid sand M10 Water-laid sand Water-laid sand and till Till M14 Water-laid sand Till Till M37 Water-laid sand Water-laid sand and till Till M38 Water-laid sand Water-laid sand and clay Clay and dislodged material M42 Dune sand Dune sand Dune sand M43 Till Water-laid sand and dislodged material Water-laid sand M44 Till Water-laid sand Water-laid sand M45 Till Water-laid sand Till M54 Till Till Till M55 Till Water-laid sand and till Water-laid sand Table 6: Lithology of each site at the bluff toe, bluff face, and crest, as interpreted from Buckler (1981 Table B2) 57 Figure 17: Bluff sites plotted on the Quaternary Geology of Southern Michigan map (Farrand and Bell 1982). Inset map site names from south to north: bottom left: sites M1, M2, M3, M4, M5, M6, bottom right: sites M9, M10, M14, top left: sites M37 and M38, top middle: sites M42, M43, M44, and M45, top right: sites M54 and M55. 58 Site no. Farrand and Bell Quaternary Map (1982) M1 Glacial outwash sand and gravel and postglacial alluvium M2 End moraines of coarse-textured till M3 Dune sand M4 End moraines of coarse-textured till M5 End moraines of coarse-textured till M6 Glacial outwash sand and gravel and postglacial alluvium M9 End moraines of fine-textured till M10 Fine-textured glacial till M14 Lacustrine sand and gravel M37 End moraines of fine-textured till M38 End moraines of fine-textured till M42 Lacustrine sand and gravel M43 Lacustrine sand and gravel M44 Lacustrine sand and gravel M45 End moraines of medium-textured till M54 Lacustrine sand and gravel M55 Lacustrine sand and gravel Table 7: The interpreted lithology of the sites from Farrand and Bell (1982). Figure 18: Oblique imagery of site M1. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. 59 Figure 19: Oblique imagery of site M2. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. Figure 20: Oblique imagery of site M3. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. 60 Figure 21: Oblique imagery of site M4. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. Figure 22: Oblique imagery of site M5. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. 61 Figure 23: Oblique imagery of site M6. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. Figure 24: Oblique imagery of site M9. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. 62 Figure 25: Oblique imagery of site M10. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. Figure 26: Oblique imagery of site M14. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. 63 Figure 27: Oblique imagery of site M37. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. Figure 28: Oblique imagery of site M38. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. 64 Figure 29: Oblique imagery of site M42. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. Figure 30: Oblique imagery of site M43. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. 65 Figure 31: Oblique imagery of site M44. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. Figure 32: Oblique imagery of site M45. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. 66 Figure 33: Oblique imagery of site M54. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. Figure 34: Oblique imagery of site M55. Retrieved from: https://toolkit.climate.gov/tool/great- lakes-shoreviewer. Site is represented as red star. 67 Hardened Shoreline Analysis All sites were examined for the extent and type of shoreline protection structures using both the 2019 U.S. Great Lakes Hardened Shoreline Classification dataset generated by NOAA and visual inspection of the historical imagery used to generate bluff recession rates. Utilizing the NOAA Hardened Shoreline Classification dataset yielded a highly accurate assessment of modern along-coast presence or absence of shore protection. Examining the historical imagery for the presence or absence of shore protection provides an estimate of when structures were installed. The majority of the sites (14/17) examined in this study are not armored presently and appear to never have been armored (Figure 36). Table 9 reports whether each site was armored or not and for the three sites that were armored depicts what structures are present, the relative condition of the structure, and approximately when the structure was built. The time of construction is a minimum age estimate because historical aerial imagery was utilized to estimate when the structure first appeared. Since there are gaps in time of collection, ranging from approximately 2 to 12 years, depending on the site and time step, it is possible that the structure could have been built several years prior to the first aerial photograph of it. Three of the sites (M1, M2, and M10) are armored. M1 has sporadically placed concrete rubble and rip rap, as well as slope grading and bluff stabilization measures; the approximate time of construction is 2005 and these structures are labeled by NOAA as poorly constructed and poorly maintained. M2 has a revetment wall that was constructed approximately in 2005, which NOAA classified as moderately engineered, meaning that the structure has an estimated lifespan of 5 - 15 years. M10 has slope grading and bluff stabilization measures, as well as seawalls and bulkheads, that NOAA classified as poorly constructed and poorly maintained; they were constructed 68 approximately in 2015. The remaining 14 sites (M3, M4, M5, M5, M6, M9, M14, M37, M38, M42, M43, M44, M45, M54, and M55) are not considered armored. However, M44 is modified: the bluff site was graded out around 2012 (lowering the slope) and a concrete channel to conduct surface runoff from the street was installed. Slope stabilizing measures were also established (Figure 35). Figure 35: Side by side change of site M44 in Manistee Michigan. The left was taken in 2005, prior to modification, and the right is in 2018 after the bluff was graded out, bluff stabilization measure installed, and concrete-lined channel established. All sites were also analyzed for proximity to shoreline perpendicular structures that could inhibit littoral sediment transport processes. If the site was determined to fit those attributes, the distance to the structure was determined by using the 2018 NAIP image and the measuring tool in ArcGIS 10.6. Table 8 shows the sites that were in proximity to and downdrift of shore- perpendicular structures and the distance between the two. 69 Site no. Distance (Km south of) Structure Distance (Km north of) Structure M1 6.65 Jetty N/A N/A M2 4.8 Jetty N/A N/A M3 N/A N/A 4.6 Jetty M4 N/A N/A 6.3 Jetty M5 N/A N/A 8.3 Jetty M6 N/A N/A 10.3 Jetty M9 N/A N/A 8.9 Jetty M10 N/A N/A 12 Jetty M14 4.5 Jetty N/A N/A M37 1.8 Breakwater 10.7 Jetty M38 4.75 Breakwater/Jetty 1.25 Breakwater M42 2.9 Breakwater/Jetty N/A N/A M43 2.1 Breakwater/Jetty N/A N/A M44 1.3 Breakwater/Jetty N/A N/A M45 9.75 Jetty 3.7 Breakwater/Jetty M54 N/A N/A N/A N/A M55 N/A N/A N/A N/A Table 8: The distance in kilometers (south of or north of) from a shore-perpendicular structure to the local bluff site, and the structure type. 70 Figure 36: Bluff site locations relative to the hardened shoreline classification layer generated by NOAA (coast.noaa.gov/digitalcoast/data/hardened-shorelines.html). 71 Approximate Site no. Armorred time of Type (Primary) Type (Secondary) Condition construction Ad Hoc Concrete Rubble / Rip Poorly constructed and M1 Yes 2005 Slope Grading / Bluff Stabilization Rap poorly maintained M2 Yes 2005 Revetment, moderately engineered N/A Moderately engineered M3 No N/A N/A N/A N/A M4 No N/A N/A N/A N/A M5 No N/A N/A N/A N/A M6 No N/A N/A N/A N/A M9 No N/A N/A N/A N/A Slope Grading / Bluff Poorly constructed and M10 Yes 2015 Stabilization, Seawalls / Bulkheads poorly maintained M14 No N/A N/A N/A N/A M37 No N/A N/A N/A N/A M38 No N/A N/A N/A N/A M42 No N/A N/A N/A N/A M43 No N/A N/A N/A N/A M44 No N/A N/A N/A N/A M45 No N/A N/A N/A N/A M54 No N/A N/A N/A N/A M55 No N/A N/A N/A N/A Table 9: The interpretation of plotting the bluff sites on the hardened shoreline classification layer (coast.noaa.gov/digitalcoast/data/hardened-shorelines.html) showing approximate time of construction, primary and secondary structure type, and relative condition. Nearshore Bathymetry Nearshore bathymetry at each site was analyzed by extracting profiles from topobathymetric LiDAR acquired in 2012 (2008 for M2) and for select sites from single-beam, sonar-derived bathymetry collected in 2020. Google Earth Pro and 2018 NAIP imagery was utilized to assess for presence of a modern berm at sites where 2020 topobathymetric LiDAR data were not available. Table 10 summarizes the morphological characteristics at each site for a given year in terms of presence of an inner bar, outer bar, and berm. Ten sites (M1, M2, M3, M4, M5, M6, M42, M43, M44, and M45) were analyzed using both 2012 and 2020 topobathymetric LiDAR data, except for M2 which utilized 2008 in the place of 2012 LiDAR data. In 2012, only three sites (M1, M2, and M3) showed the presence of an outer sand bar. Only sites M1 and M2 showed the presence of an inner bar in 2012 (2008). At sites M3, M5, M42, and M45 a berm was present in 2012. In 2020, 6 sites (M1, M2, M3, M4, M5, and M6) in the southern portion of the eastern Lake Michigan shoreline exhibited both inner and outer bars. All four sites in the upper 72 middle portion of the eastern Lake Michigan shoreline (M42, M43, M44, and M45) lacked any sand bars in 2012 and 2020. Additionally, in 2012, a berm was present at sites M42 and M45, but were lacking at sites M43 and M44. Notably, berms were absent at all 17 sites in 2020. Seven sites (M9, M10, M14, M37, M38, M54 and M55) were analyzed using just 2012 topobathymetric data because 2020 data were not available yet. Inner sand bars were absent from all of these sites, and only site M14 exhibited an outer sand bar. Berms were present at sites M54 and M55, but absent at sites M9, M10, M14, M37, and M38. Sand bars of either type were absent at the four sites (M9, M10, M37, and M38) in the lower middle portion of the eastern lake Michigan shoreline in 2012; an outer sand bar was present at site M14 in 2012. Berms were present at only two of these sites (M9 and M14) in 2012. Two sites (M54 and M55) in the northern portion of the eastern Lake Michigan shoreline lacked either inner or outer sand bars, however, a berm was present at both sites. For 2020, these sites were analyzed in Google Earth Pro and ArcGIS with 2018 NAIP imagery for presence of a modern berm. Berms were absent at all seventeen sites indicating a wide and stable beach capable of protecting the bluff. 73 Inner Bar Outer Bar Berm Site no. 2012 2020 2012 2020 2012 2020 M1 Yes Yes Yes Yes No No *M2 Yes Yes Yes No No No M3 No Yes Yes Yes Yes No M4 No Yes No Yes No No M5 No Yes No Yes Yes No M6 No Yes No Yes Yes No M9 No N/A No N/A Yes No M10 No N/A No N/A No No M14 No N/A Yes N/A Yes No M37 No N/A No N/A No No M38 No N/A No N/A No No M42 No No No No Yes No M43 No No No No No No M44 No No No No No No M45 No No No No Yes No M54 No N/A No N/A Yes No M55 No N/A No N/A Yes No Table 10: Morphological characteristics of bathymetry at each site in terms of presence of inner bar, outer bar, and berm, relative to the monthly mean water level at the time of data collection. N/A values represent missing data. *M2 utilizes topobathymetric data from 2008 instead of 2012 due to inadequate data. Beach, Bluff, and Nearshore Morphometrics Additional morphology parameters (called morphometrics), including nearshore slope, bluff face slope, bluff crest elevation, bluff toe elevation, beach width, and beach slope, were extracted from the 2012 topographic profiles. Table 11 summarizes the 2012-2020 bluff recession magnitude, normalized by time (meters per year) and the associated nearshore slope, bluff face slope, bluff crest elevation, bluff toe elevation, beach width, and beach slope parameters from 2012 at each site. 74 2012-2020 Magnitude Site no. Normalized 2012 Nearshore Slope 2012 Bluff Face Slope 2012 Bluff Crest Elevation (m) 2012 Bluff Toe Elevation (m) 2012 Beach Width (m) 2012 Beach Slope by Time (m/yr) M1 -0.671 0.008 0.579 194.906 176.983 0.923 0.158 M2 -1.498 0.011 0.608 202.435 177.147 0.919 0.210 M3 -0.492 0.011 0.535 210.957 177.711 1.603 0.054 M4 -3.137 0.011 0.476 213.177 177.628 0.652 0.075 M5 -2.625 0.013 0.571 212.719 178.175 2.806 0.053 M6 -0.852 0.013 0.524 203.294 178.932 2.852 0.077 M9 -0.439 0.012 0.450 201.236 177.376 1.305 0.032 M10 -0.787 0.011 0.681 199.938 176.700 0.620 0.053 M14 -1.401 0.009 0.444 198.141 177.495 1.434 0.050 M37 -1.032 0.014 0.704 226.851 176.663 0.580 0.061 M38 -0.670 0.015 0.615 232.340 176.772 0.702 0.059 M42 -0.893 0.018 0.263 192.088 177.626 1.542 0.111 M43 -4.252 0.013 0.595 200.175 176.611 0.523 0.039 M44 -0.720 0.017 0.325 198.552 176.739 0.653 0.099 M45 -1.612 0.016 0.526 201.564 177.262 1.170 0.008 M54 -0.793 0.015 0.392 189.770 176.777 0.701 0.064 M55 -0.461 0.015 0.609 211.375 177.742 1.664 0.092 Table 11: The 2012-2020 bluff recession magnitude normalized by time in meters per year and the associated nearshore slope, bluff face slope, bluff crest elevation, bluff toe elevation, beach width, and beach slope parameters from 2012 at each site. Relationships Between Beach, Bluff, and Nearshore Morphometrics and Bluff Recession Rates The relationship between beach, bluff, and nearshore morphometrics and bluff recession rates were analyzed using 2012 topobathy data from JALBCTX for all seventeen sites and single-beam sonar data collected in 2020 for ten of the sites. These morphometrics were regressed against bluff recession rates to evaluate their potential impact on the rate of bluff recession. Table 12 summarizes the linear regression analysis results which includes the R- squared value and P-value. Figure 37 shows the nearshore slope of all seventeen sites plotted against the 2012-2020 bluff recession magnitude normalized by time. Figure 38 shows the correlation of 2012-2020 bluff recession magnitude normalized by time and bluff face slope. Figure 39 shows the correlation between the 2012-2020 bluff recession magnitude normalized by time and bluff crest elevation. Figure 40 shows the correlation between the 2012-2020 bluff recession magnitude normalized by time and bluff toe elevation. Figure 41 shows the correlation between 2012-2020 bluff recession magnitude normalized by time and beach width. Figure 42 shows the correlation between 2012-2020 bluff recession magnitude normalized by time and 75 beach slope. In summary, table 12 shows low R-squared values and P-values, showing that none of the parameters tested significantly explain the observed variance of bluff recession rates. Correlation R-Squared P-Value Rates vs Nearshore Slope 0.014 0.652 Rates vs Bluff Face Slope 0.014 0.648 Rates vs Bluff Crest Elevation 0.026 0.921 Rates vs Toe Elevation 0.001 0.907 Rates vs Beach Width 0.008 0.736 Rates vs Beach Slope 0.030 0.505 Table 12: Summarization of linear regression analysis results of relationships between beach, bluff, and nearshore morphometrics and bluff recession rates including the R-squared value and P-value. Figure 37: The 2012-2020 bluff recession magnitude normalized by time of the seventeen sites plotted against the average nearshore slope. 76 Figure 38: The 2012-2020 bluff recession magnitude normalized by time of the seventeen sites plotted against the average bluff face slope. 77 Figure 39: The 2012-2020 bluff recession magnitude normalized by time by time of the seventeen sites plotted against the average bluff crest elevation. 78 Figure 40: The 2012-2020 bluff recession magnitude normalized by time of the seventeen sites plotted against the average bluff toe elevation. Figure 41: The 2012-2020 bluff recession magnitude normalized by time of the seventeen sites plotted against the average beach width. 79 Figure 42: The 2012-2020 bluff recession magnitude normalized by time the seventeen sites plotted against the average beach slope. Fetch, Shoreline Azimuth, and Wave Power Table 13 shows 2012-2020 bluff recession magnitude normalized by time in meters per year, and the azimuth of the shoreline azimuth. The maximum perpendicular fetch distance relative to each site’s shoreline azimuth is shown in Table 14. Table 15 shows the 2012-2020 bluff recession magnitude normalized by time and the daily mean wave energy from 2012-2020 in square meters. Daily mean wave energies vary from approximately 13.78 m2 to 22.52 m2 for the 2012-2020 time step. 80 2012-2020 Bluff Recession Magnitude Site no. Azimuth of Shoreline Normalized by Time (m/yr) M1 -0.671 115.725 M2 -1.498 113.591 M3 -0.492 126.204 M4 -3.137 125.697 M5 -2.625 125.879 M6 -0.852 123.990 M9 -0.439 98.771 M10 -0.787 94.977 M14 -1.401 80.265 M37 -1.032 78.530 M38 -0.670 109.463 M42 -0.893 110.698 M43 -4.252 109.290 M44 -0.720 108.545 M45 -1.612 107.964 M54 -0.793 107.210 M55 -0.461 94.514 Table 13: The 2012-2020 bluff recession magnitude normalized by time in meters per year and the azimuth of the shoreline in degrees measured clockwise from 270. 81 2012-2020 Bluff Recession Magnitude Site no. Normalized by Time (m/yr) Fetch (km) Azimuth of Fetch M1 -0.671 116 25.724909 M2 -1.498 123 23.59132 M3 -0.492 127 36.203698 M4 -3.137 128 35.696769 M5 -2.625 128 35.87947 M6 -0.852 130 33.98952 M9 -0.439 128 8.770763 M10 -0.787 127 4.977165 M14 -1.401 129 350.264626 M37 -1.032 103 348.53047 M38 -0.670 103 19.462566 M42 -0.893 95 20.698257 M43 -4.252 95 19.290046 M44 -0.720 95 18.544711 M45 -1.612 93 17.964358 M54 -0.793 15 17.209558 M55 -0.461 20 4.513988 Table 14: The 2012-2020 bluff recession magnitude normalized by time in meters per year, the maximum perpendicular fetch distance (km) relative to each site’s shoreline azimuth, and the azimuth of the perpendicular fetch orientation. 82 2012-2020 Bluff Recession Magnitude 2 Site no. Mean Daily Wave Energy (m ) Normalized by Time (m/yr) M1 -0.671 18.656 M2 -1.498 17.695 M3 -0.492 19.051 M4 -3.137 18.283 M5 -2.625 18.178 M6 -0.852 18.172 M9 -0.439 21.518 M10 -0.787 19.820 M14 -1.401 22.518 M37 -1.032 15.094 M38 -0.670 13.768 M42 -0.893 17.924 M43 -4.252 19.281 M44 -0.720 17.705 M45 -1.612 19.037 M54 -0.793 17.175 M55 -0.461 18.661 Table 15: The 2012-2020 bluff recession magnitude normalized by time in meters per year and the daily mean wave energy from 2012-2020 in square meters. Relationships Between Fetch, Shoreline Azimuth, Wave Power and Bluff Recession Rates To analyze the relationships between bluff recession rates and fetch, shoreline azimuth, and wave power, linear regression analyses was conducted. Table 16 summarizes the linear regression analysis results which includes the R-squared value and P-value; the results show no significant correlation between the bluff recession rates and the tested variables. Figure 43 shows the correlation between the 2012-2020 bluff recession magnitude normalized by time and maximum perpendicular fetch. Table 14 shows the maximum perpendicular fetch distances relative to each site’s shoreline azimuth used in the linear regression analysis. Figure 44 shows the correlation between the 2012-2020 magnitude normalized by time and bluff azimuth with respect to 270 degrees. Figure 45 shows the correlation between the 2012-2020 magnitude 83 normalized by time and daily mean wave energy. Figure 46 shows the correlation between the 2012-2020 magnitude normalized by time of the seventeen sites plotted against the perpendicular fetch azimuth with respect to 270 degrees. Correlation R-Sqaured P-Value Rates vs Max. Perpendicular Fetch 0.032 0.493 Rates vs Mean Daily Wavey Energy 0.009 0.709 Rates vs Bluff Azimuth 0.065 0.322 Rates vs Perpendicular Fetch Azimuth 0.026 0.552 Table 16: Summary of linear regression analysis results of relationships between maximum perpendicular fetch, mean daily wave energy, bluff azimuth, and maximum perpendicular fetch azimuth and bluff recession rates including the R-squared value and P-value. Figure 43: The 2012-2020 bluff recession magnitude normalized by time for the seventeen sites plotted against the maximum perpendicular fetch distance. 84 Figure 44: The 2012-2020 bluff recession magnitude normalized by time for the seventeen sites plotted against wave energy in meters squared. Figure 45: The 2012-2020 bluff recession magnitude normalized by time for the seventeen sites plotted against bluff azimuth. 85 Figure 46: The 2012-2020 bluff recession magnitude normalized by time for the seventeen sites plotted against the perpendicular fetch azimuth. 86 Discussion The goal of this study was to establish linkages between bluff recession and lake levels to discern if recession rates are increasing, to analyze the response in recession to fluctuating lake levels, and to identify the role of various hydrodynamic and geomorphic variables in driving recession. Furthermore, this study aimed to establish relationships between bluff recession rates and various site characteristics regarding their influence on recession rates. These relationships were analyzed at both long-term and recent temporal scales, as well as spatially, as the sites are distributed along the extent of the eastern Lake Michigan shoreline in Michigan. The overarching question was “How do natural and anthropogenic events that seem to be increasing in magnitude, such as rapid water level fluctuations and urban development, affect coastal bluff erosion rates”. In total, seventeen sites along the eastern shoreline of Lake Michigan were analyzed (Figure 5) and it was hypothesized that as these events increase in intensity and significance, namely fluctuations between high and low lake level periods and strong storm events, progression of bluff erosion will be become more extreme in both magnitude and severity, likely due to the increasing severity of erosional driving processes. Furthermore, as lake levels are projected to remain high or increase, it was thought that bluff recession rates will accelerate in the future due to the lagged response of bluffs to erosional driving forces. This study was an expansion of bluff recession studies done by Dr. Bill Buckler at Michigan State University (Buckler 1973, 1981) in which he documented long-term rates of bluff recession along the Wisconsin and Michigan Lake Michigan coastline by comparing GLO surveys from the mid-1800s to his own field surveys. In his studies, Buckler analyzed sites composed of both eolian and non-eolian material; eolian sites were not included in this current study. Buckler concluded that there was wide variability between recession rates on both sides of 87 Lake Michigan and that the long-term recession rates could not be related in a meaningful way with sediments or sediment arrangement (Buckler 1981). He also concluded that long-term recession rates could not be correlated to groundwater activity, bluff height, or beach width; however, shoreline azimuth and fetch appeared to have an impact on recession rates (Buckler 1981). Buckler also concluded that the progression of shore zone urban development, mainly shoreline protection structures, resulted in the 1970s bluff recession rates being different in magnitude compared to the mid-1800s rates and that these structures may actually have an unintended adverse effect on bluff recession (Buckler 1981). Though eolian sites were excluded from my study, it is important to note that Buckler concluded eolian material sites were receding at significantly lower long-term rates than sites composed of cohesive sediments, due to the accretionary ability of dunes to accumulate and grow over time (Buckler 1981). In contrast, once a cohesive bluff recessions and material is removed, that material will not accrete back and is permanently lost to the nearshore system. Long-term Rates of Bluff Recession Comparison of Historic Recession Rates to Modern Rates Bluff recession rates from 1977 to 2020 generated through DSAS were compared to Dr. Buckler’s rates from the mid-1800s to 1977; these rates are represented as the magnitude normalized by time. The magnitudes normalized by time were compared using a +/-0.25m threshold to delineate if the rate is increasing, decreasing, or remaining constant compared to the past rate. To restate, with the addition of post 1970s data, 7 sites are increasing over time (M1, M5, M9, M14, M37, M38 and M43), 3 are decreasing over time (M2, M4, and M6), and 7 remained constant (M3, M10, M42, M44, M45, M54 and M55). Amongst the sites that are generally increasing, sites M4 and M43 seem to be experiencing modern (2012-2020) recession 88 rates that are particularly higher in magnitude than past rates. Sites M2, M10, M42, and M45 have modern (2012-2020) recession rates that are higher than all of the previous recession rates, however they have not accelerated as rapidly. Despite having period of low recession rates, M5 and M14 are experiencing increasing erosion rates of relatively similar magnitude to past rates. With this, M4 and M43 are experiencing the most severe modern recession rates. All of the sites that have decreasing or constant recession rates in the periods before 2012 have lower or similar recession rates (+/- 0.25m) in the period from 2012 to 2020. As there is large spatial and temporal variability in recession rates between and within sites, it must be the result of some local control or facilitator. If water level were the only facilitator, the large spatiotemporal variability in patterns would not be present as the bluffs would erode similarly as they all experienced the same general fluctuations in lake level. These patterns lead to questions such as what factors or characteristics of a site to lead to modern changes in recession rates? Are there any distinguishable factors or characteristics that lead certain sites to have constant recession rates over time? Also, what causes one site to behave opposite of a neighboring site with respect to patterns of recession? The trend of bluff recession and water level is not linear through time and space, so there are undoubtably other site characteristics or facilitators at work. However, analyzing these patterns at such a large temporal time scale is difficult as the scope is too large to pinpoint certain site characteristics or facilitators of erosion due to the inherent complexity of the system. For example, it is difficult to pinpoint wave climate and nearshore bathymetry to recession rates as those parameters can change in relatively short amounts of time (i.e., due to a storm event). Further, it is difficult to correlate recession rates to larger scale parameters, such as water level, because since the 1970s, there have been periods of both high and low lake levels, and it is difficult to correlate one recession 89 rate to a time period that contained water level variability. (Figure 1). The use of historical aerial imagery has allowed for the large temporal scale to be broken up into near-decadal time steps which allows for better connecting of recession rates to site characteristics and facilitators of erosion. These near-decadal recession rate patterns (Figure 10 to 19) are discussed below and are compared to the various site characteristics and facilitators of erosion. Sites with Generally Increasing Recession Rates Over Time The bluff recession rates determined in this study show substantial spatial and temporal variability. Along the eastern coast of Lake Michigan, bluff recession rates are not increasing linearly over time, in fact, large fluctuations in recession rates at certain sites contrast with recession rates that remain relatively constant or are even decreasing at other sites (Figure 10). Eight of the seventeen study sites (M2, M4, M5, M10, M14, M42, M43, and M45) exhibited accelerated recession rates in the recent time step (2012-2020) relative to past recession rates (1970s and 1980s) (Figure 11). Sites M2, M4, M5, M10, M14, M42, and M43 all experienced increased recession rates during the 2012-2020 timestep relative to the 2005-2012 timestep, which shows the bluff response to the present high lake level. Of the sites with generally increasing recession rates over time, sites M4 and M14 had peaks in recession rates during the 1986–1997 time step, indicating that bluff recession increased after the 1980s high lake levels, which peaked in 1986. The results from these sites are congruent with the idea that there is a delay in bluff response to high lake levels with the highest recession rates experienced after a highstand as the bluff response occurred after peak lake level had occurred (Castedo et al., 2013; Volpano et al., 2020; Roland et al., 2021). Of the sites with generally increasing bluff recession rates, sites M5 and M43 had peaks in their recession rates during the 1978–1986 time step, which was during the rising limb of a 90 high lake level phase. This is not congruent with the notion that there is a delay in bluff response to fluctuating lake levels as the peak in recession rates corresponded with the peak in lake level. Since both sites (M5 and M43), as well as other adjacent sites, were not armored at all during this time, there must be either a lithologic or morphodynamic explanation for the variability in recession. The lithology is only similar at the toe, not the crest it likely does not explain the relationships with bluff recession rates which were measured at the crest (Table 6). a different site characteristic or erosional facilitator at work as the lithology is only similar at the bluff toe. As there were a limited number of high quality historical aerial photographs from the 1970s-1990s to use in this study, it is difficult to make conclusions on the similarities of drivers of long-term recession at sites M5 and M43, however both do, to some extent, follow the patterns of rising and falling lake level. Irrespective of this, recent morphodynamic and geographic conditions can be used to evaluate patterns. In terms of beach, bluff, and nearshore morphometrics, both sites have the same 2012 nearshore slope of 0.013, or 1.3%. Further, both sites have similar steep bluff-face slopes: 0.571 (57.1%) at M5 and 0.595 (59.5%) at M43 (Table 11). M5 has a higher bluff crest elevation than M43 by about 12 meters, and a higher elevation bluff toe by about 1.6 meters, relative to the same lake level at the time (Table 11). Furthermore, both sites have differing beach morphologies where M43 has a significantly narrower beach than M5, though M5 has a steeper-sloped beach than M43 (Table 11). In terms of fetch, shoreline azimuth, and wave power, M43 and M5 differ in their characteristics. Site M5 has a higher measured fetch (128 km) compared to M43 (95 km) (Table 14). Furthermore, M43 is orientated closer to north-south than M5 (table 13). Regarding wave power, M43 experienced a slightly higher daily mean wave energy in the 2012-2020 period compared to M5 [19.3 m2 vs. 18.2 m2] (Table 15). In terms of nearshore bathymetry, M5 91 exhibited a much more dynamic nearshore environment than M43 did since inner and outer sandbars developed between 2012 and 2020 at M5, but not at M43 (Table 10). At both sites, berms were absent in 2020, which shows the gradual erosion of the berm that was present at M5 during 2012 (Table 10). The absence of berms in 2020 may explain the increase in recession rates that were detected at both sites (M5 and M43) after 2012 as there was no barrier to protect the bluff toe from wave action. However, site M5 had a significantly wider (2.5 m vs. 0.5 m) and somewhat steeper (0.053 vs. 0.039) beach than M43 in 2012, thus indicating the presence of some other untested morphodynamic or site characteristic that is impacting recession rates. Additionally, M5 began to experience an increase in recession rates during the 2005-2012 timestep, which continued into the 2012-2020 timestep. Other sites with generally increasing recession rates did not experience an increase in recession rate during the 2005-2012 time step and only showed a significant increase in recession rates during the 2012–2020-time step. This follows a similar peak in rates that M5 showed during the rise in lake level during the 1978-1986 time step. This may be explained by the fact that site M5 did not exhibit an inner or outer sandbar in 2012 but developed these features by 2020 (Table 10). More bathymetric data is needed to further explain this. If the lack of sandbars explained the high recession rates 2012 and beyond, we would expect to see developed sandbars in 2005 when recession rates were low (Table 10). Further, we might expect to see lower or more constant recession rates at M5 into the future as sandbars were present there in 2020. However, with both sites, other drivers and/or mediators not tested may be at work which could explain this pattern. Perhaps groundwater conditions or finer-scale lithological or stratigraphic differences have contributed to these differences. Overall, the only observed similarity between M5 and M43 is the nearshore slope which may have an influence on recession rates, however more data and analysis is needed. 92 Of sites that have generally increasing recession rates over time, sites M2, M10, M42, and M45, did not experience a peak in recession rates between the 1978–1986 and/or the 1986– 1997 time step. The bluff crests at these sites seemingly did not respond to the high lake level period in the late 1980s. However, they are currently experiencing relatively high recession rates during the recent rise in lake level. These sites were experiencing nearly constant recession rates over time until 2012, then the recession rate began to increase in response to rising lake levels after 2014. This pattern raises questions on what caused these sites to not respond to the high lake level during the 1980s and what is making them respond now to high lake levels. All four sites have relatively similar lithologies of coarser-grained, sandy material (Table 6): M2 is composed of water-laid sand and till; M10 is composed of water-laid sand and till; M42 is composed of dune sand, and M45 is composed of water-laid sand and till. This lithologic component may be in part, responsible for the low recession rates observed historically, and a different characteristic or erosional facilitator may have initiated the more recent high recession rates, such as the large fluctuation between low and high lake level periods; however, to conclude this, more detailed stratigraphic analysis is required. Further, surrounding sites with similar lithologies did not experience this pattern, suggesting there is another site characteristic or erosional facilitator controlling recession rates at these sites. In terms of shoreline armoring, both M2 and M10 are armored, whereas M42 and M45 are not. M2 was armored around 2005 with a revetment wall and M10 was armored around 2015 with seawalls and with slope grading/ stabilization methods (Table 9). Furthermore, all four sites have different proximities to shore-perpendicular harbor structures: M2 is south of a harbor jetty, M10 is north of a harbor jetty, M42 is south of harbor jetties and breakwaters, and M45 is in between two harbor jetties and breakwaters to the north and south of the site (Table 8). In 93 accordance with Beletsky and Schwab (2008), the dominant current near these sites is originating from the south, which may trap sediment updrift of any structure. With this, long-term, near- constant recession rates at M42 and M45 (Figure 14) may be explained by possible trapped sediment allowing for beach replenishment. Additionally, the presence of recent, adjacent shoreline armoring may disrupt the natural replenishment of the beach and increase nearshore lakebed downcutting which may be leading to slope instability and subsequent recession at M2 and M10; however, comparatively, M42 and M45 do not have adjacent shoreline armoring present. Overall, the recent installment of shoreline armoring at M2 and M10 may explain the increase in recession rates, however this is not the case at M42 and M45 as they are not armored. The varying lithology also suggests that there are other erosional forces or site characteristics impacting recession rates at these sites. In terms of beach, bluff, and nearshore morphometrics, sites M2, M10, M42, and M45 exhibit different characteristics. Both M2 and M10 have the same 2012 nearshore slope (0.011) and M42 and M45 are slightly steeper, 0.018 and 0.016, respectively (Table 11). Additionally, M2 and M10 have a steep bluff face slope of 0.608 and 0.681, respectively, whereas M45 has a slightly less steep bluff face slope of 0.526 and M42 has a relatively gently bluff face slope of 0.263 (Table 11). The bluff crest elevations at M2, M10, and M45 are similar at approximately at 200 meters, whereas M42 has a somewhat lower bluff crest elevation of approximately 192 meters (Table 11). The bluff toe elevations between these sites are similar, only varying by 1 meter (Table 11). All four sites have different beach widths: M42 had the widest beach 2012 and M10 had the narrowest (Table 11). Furthermore, M2 had the steepest beach slope of these sites, while M10, M42, and M45 exhibited comparatively gentler beach slopes (Table 11). 94 In terms of nearshore bathymetry, the four sites show a variety of presence and evolution of nearshore features (Table 10). In 2012, M2 had both an inner and outer sand bar, but the outer sand bar was no longer present in 2020 (Table 10). M10, M42 and M45 did not have any sandbars in 2012 or 2020 (Table 10). M42 and M45 had berms in 2012, but not in 2020 (Table 10), in fact, none of these sites exhibited a berm in 2020. Note that, due to a lack of 2020 bathymetric data, M10 could not be analyzed for nearshore bathymetry during this time. Despite this, these widely varying bluff, beach, and nearshore morphometrics further suggest that there must be other circumstances at these sites driving this pattern of recession. The fetch, shoreline azimuth, and wave power at sites M2, M10, M42 and M45 are different from one another. Of the four sites, the shoreline at M10 is the closest to a north-south azimuth (5.0ofrom north) while M2, M42, and M45 are oriented more north-northeast (23.6o, 20.7o and 18.0ofrom north, respectively) (Table 13 – note azimuths measured clockwise from 270o in this table). Furthermore, these four sites can be grouped into two classes regarding their magnitudes of maximum possible fetch perpendicular to the shoreline azimuth. M2 and M10 have a slightly great fetch (123 km and 127 km, respectively) compared to M42 and M45 (95 km and 93 km, respectively) (Table 14). Regarding wave power, M10 and M45 experienced somewhat greater mean daily wave energy for the 2012-2020 timestep (19.8 and 19.0, respectively) compared to M2 and M42 which were impacted by comparatively lower wave energies (17.7 and 17.9, respectively) (Table 15). With this, the relatively constant rates of recession followed by a recent increase in recession rates may best be explained by the unintended consequence of recent shoreline armoring (structures may be promoting bluff instability), as well as the lack of a modern berm which could dampen the amount of wave action at the bluff toe. However, there are most likely other site morphodynamics and characteristics 95 that influence the long-term rates of recession at these sites, such as groundwater conditions or finer-scale lithological or stratigraphic differences. In summary, eight of the seventeen study sites (M2, M4, M5, M10, M14, M42, M43, and M45) all exhibited generally increasing recession rates over time. Two sites (M4 and M14) showed a lagged bluff recession response to high lake levels during the 1980s, which is consistent with previous studies (Castedo et al., 2013; Volpano et al., 2020; Roland et al., 2021). Other sites (M5 and M43) experienced an immediate response to the rise in lake level in the 1980s, as well as the present rise in lake level, however it is not obvious based on the metrics studied here why these sites responded differently. The four remaining sites (M2, M10, M42, and M45) experienced nearly constant recession rates over time until the current rise in lake level, which may be explained, in part to the proximity of shoreline armoring, however more research is needed. This recent increase in recession may also be explained by the lack of a berm on the beach during 2020 at these sites. However, there is large variation in bluff, beach, and nearshore morphodynamics, as well as variance in fetch, shoreline azimuth, and wave power, suggesting there are other variables controlling recession rates at these sites. All eight sites showed an increase in recession rates during the modern rise in lake level, which is to be expected, but their response during the 1980s gives an indication of what might happen over the next decade. Sites that peaked after the 1980s high are likely to follow a similar pattern, suggesting that the worst bluff recession is yet to come. Sites that peaked during the rising limb of high lake levels are likely experiencing peak erosion now. It is difficult to predict what might happen to the sites that did not recede until this recent lake level rise, however, due to the inherent erodibility and depositional circumstances of the bluffs, these sites will most likely experience some amount of 96 recession in the future. Based on the data, sites that continue to lack a berm on the beach, may see increased recession rates in the near future. In conclusion, regardless of the erosional drivers, it is evident that not all bluffs experience recession over space and time; furthermore, there is large variability in recession rates at singular sites, even though mediating characteristics (i.e., lithology, azimuth, armoring or lack thereof) remain the same, leading to the conclusion that there must be some other characteristic(s) exerting control at these sites. Sites with Generally Decreasing or Constant Recession Rates Over Time Of the seventeen sites, nine sites (M1, M3, M6, M9, M37, M38, M44, M54, and M55) are experiencing recession rates that are generally decreasing or remaining relatively constant over time (Figure 14). Of these sites, sites M3, M9, M37, and M38 had peaks in recession rates during the 1986–1997 time step, again showing a lagged response to the high lake levels of the 1980s. However, after the high lake level period in the 1980s, recession rates at these sites lowered and then remained constant or did not change significantly enough to be considered increasing. The recession rates appear to be slightly increasing during the recent rise in lake level to record and near-record levels, however, the rates of this increase are much lower than other sites in this study. This leads to questions on what caused these sites to peak in recession rates during the last lake highstand but not during the modern one. The lithologies of these sites are to some extent comparable (Table 6); M1, M3, M6, M37, M44, M45, and M55 are composed of water-laid sand and till, whereas M9 and M38 are composed of water-laid sand and clay. Although these sites differ slightly in clay versus till content, they all contain water-laid sand, which may in part explain the relatively low recession rates over time. However, there may be other site characteristics or erosional facilitators controlling recession at these sites. 97 Regarding shoreline armoring, sites M1, M37, M38, and M44 are near shore-perpendicular harbor structures, whereas M3, M6, M9, M54, and M55 are not (Table 8). Additionally, only M1 had shore-parallel protective structures present, which would be the apparent explanation for its sustained low recession rates (Table 9). These varying lithologies and degree of shoreline armoring do not explain the pattern of recession experienced, suggesting something else is at play influencing the recession rates at these sites. Beach, bluff, and nearshore morphometrics at these sites also vary. M1 had the gentlest 2012 nearshore slope, and as the sites progressed northwards, the nearshore slopes became steeper, until at M54 and M55 the nearshore slope was gentler relative to the adjacent southern sites (Table 11). This gradual steepening of the nearshore system as sites progress north is an interesting pattern; however, it does not seem to have an effect on recession rates. Regarding 2012 bluff face slopes, M37 had the steepest slope of 0.704, whereas the remaining sites had relatively intermediate slopes of between 0.45 and 0.615, and M44 and M54 had the gentlest bluff face slopes of 0.32 and 0.392 respectively (Table 11). All sites had largely varying bluff crest elevations relative to each other, ranging from 232.34 meters at M38, to 189.77 meters at M54 (Table 11). Similarly, the bluff toe elevation of these sites shows large variability, ranging from approximately 176 meters to 178 meters. The beach morphodynamics at all sites showed varying beach widths and slopes (Table 11). Beach widths ranged from 2.852 meters at M6, to 0.58 meters at M37. Beach slopes ranged from 0.158 at M1 to 0.032 at M9,. In terms of nearshore bathymetry, the sites show varying presence and evolution of nearshore features (Table 10). Overtime, M1 retained both an inner and outer sand bar, but did not have a berm in 2012 (Table 10). M3 and M6 lacked either an inner or outer sand bar in 2012 but had both in 2020 (Table 10). Both sites had a berm in 2012 98 (Table 10). Sites M9, M37, and M38 lacked inner or outer sandbars or a berm in 2012; lack of 2020 bathymetric data for these sites prevents analysis for modern bathymetry (Table 10). Over time, M44 lacked an inner or outer sand bar, or a berm (table 10). M54 and M55 lacked either type of sand bar in 2012 but did have berms (Table 10); again, due to lack of 2020 bathymetric data for these sites, analysis for modern bathymetry was not possible at this time. In terms of fetch, shoreline azimuth, and wave power, these sites vary in characteristics. Sites M1, M3, and M6 are oriented to the northeast, sites M37 and 38 are oriented to the northwest, and sites M9, M44, M54, and M55 are oriented more to the north-northeast (Table 13 – note azimuths measured clockwise from 270o in this table). Furthermore, all sites have different magnitudes of maximum possible fetch perpendicular to the shoreline azimuth. M6 has the potential to experience a higher fetch than the remaining sites, with M54 and M55 having particularly low values, as discussed later (Table 14). The wave power at all sites was similar from 2012-2020, except for M37 and M38 which experienced relatively lower mean daily wave energies (Table 15). M1 had a peak in its bluff recession rate from 1979-1986 during the rising limb in lake level, and then reduced to nearly constant rates. This is not consistent with the pattern of having a delayed response to the high lake levels in the late 1980s. The recently low rates of recession may be explained by the presence of shoreline armoring at M1 (Table 9); here, sporadically placed rip rap and slope grading/bluff stabilization measures have been employed, which may be generally lowering recession rates. However, there may be other factors at work such as groundwater conditions, wave climate, or finer-scale lithological or stratigraphic differences than what was observed. 99 M44 has had relatively constant high bluff recession rates over time and recently has ceased receding due to the modification of the site. Before the 2005–2012 time step, M44 was experiencing significant bluff recession in the form of rotational failures. However, in an effort to remedy this, the bluff was graded to stabilize it. Slope stabilizing plantings and concrete-lined channels to control surface runoff were added sometime in the 2005-2012 timestep. These measures may continue to thwart future recession at this site. However, as it is well documented in the literature, wave action will continually act upon whatever material is at the shoreline (Carter and Guy 1988; Vallejo and Degroot 1988; Brown et al. 2005; Earlie et al. 2018; among others). Many modified shorelines tend to ultimately promote erosion instead of reducing it. The armoring at this site may fail in the future, leading to a return of bluff recession, potentially at an even higher magnitude. The sites that have generally decreasing or constant recession rates over time that did not have a peak in recession rates between the 1978-1986 and 1986-1996 timesteps include M6, M54, and M55. In comparison to other sites, the recession rates at these sites have stayed nearly constant over time with only slight fluctuations within the 0 to 1 meter per year range. These sites seemingly did not have a substantial erosional response to the high lake level period in the 1980s and are not receding in response to the current high lake levels. All three sites are not armored, and their lithology consists of fine to medium sand and gravel (Table 6). M6 has comparatively large maximum perpendicular fetch of 130 kilometers relative to M54 and M55, which have fetches of 15 km and 20 km, respectively. The only similarities between these sites are that they are not armored and have a coarser lithology. However, there are most likely other factors at work such as groundwater conditions, wave climate, or finer-scale lithological or stratigraphic differences. It is noteworthy that at M54 and M55 there may be local geographic 100 controls, such as shoreline azimuth, causing consistently low recession rates over time. These sites are in a region where shoreline azimuth varies significantly, and both are moderately sheltered in a large-scale embayment; additionally, to the northeast of M54 is South Manitou Island and to the northeast of M55 is North Manitou Island (Figure 47). As a result, both sites have a low maximum perpendicular fetch distance relative to other sites and have less variability of incoming wave directions that may impact the bluff toe. In other words, only a very specific wave direction can generate large waves, unlike the other sites that are much more open to a range of wave directions (i.e., M1, M2, M3, M4, M5, M6, etc.) (Table 14). These sites may be more protected against large wave action in comparison to other sites due to the regional geography of the shoreline. M6 does not exhibit this same large-scale, sheltered, geographic control, however it does have a wide beach with intermediate steepness relative to other sites, which may hinder wave action from reaching the bluff toe (Table 11). Figure 47: Satellite image of M54, M55, and surrounding geography. Image taken from Google Earth Pro. 101 In summary, nine sites (M1, M3, M6, M9, M37, M38, M44, M54, and M55) are all experiencing generally decreasing or constant recession rates over time and show widely varying bluff, beach, and nearshore morphodynamics, as well as varying shoreline azimuth, fetch, and wave power characteristics. Some sites (M9, M37, and M38) have shown a lagged response to the high lake level during the 1980s, which is consistent with other sites in this study (M4, M5, M14, and M43), as well as with previous studies (Castedo et al., 2013; Volpano et al., 2020; Roland et al., 2021). However, in contrast to the other sites in the study that had a lagged response, recession rates are only slightly increasing during the current high lake level phase, significantly below the increased rates of the other sites, which suggests either that something has changed at these sites and the recession rates will remain low, or that these sites may experience a substantial increase in bluff recession over the next decade after lake levels peak. Future work to evaluate any changes in the nearshore bathymetry over time could help to explain these spatiotemporal patterns. In contrast, M1 experienced an immediate response to the rise in lake level in the 1980s, which may be due to the extent of shoreline armoring at the site. The five remaining sites (M3, M6, M44, M55, and M54) experienced generally decreasing or constant recession rates over time indicating little or no response to fluctuating lake levels. However, no apparent causes were identified for this pattern from the variables tested. Based on the data, the decreasing to constant rates of recession at the southern sites (M1, M3, and M6) may best be explained by the modern presence and development of sandbars in the nearshore zone. The presence of a berm in 2012 may have also reduced recession at some sites (M3, M6, M9). Further, low to constant rates of recession at the most northern sites (M54 and M55) may best be explained by the relative sheltered geography of the region against strong wave action. 102 Sites that exhibit a delayed response to rising lake level in the 1980s (M9, M37, and M38) may have not experienced peak recession during the present high lake level, and may increase in recession rate into the future, as is estimated with M5 and M14. M1, which showed a contemporaneous response to 1980s high lake level, may be experiencing peak erosion rates currently, as is estimated with M5 and M43. Sites that have not showed a response to any fluctuating lake level (M3, M6, M44, M54, and M55) may experience constant or decreasing recession rates into the future. With this, the sites that exhibit decreasing or relatively constant recession rates over time have widely varying morphodynamic and hydrodynamic characteristics that do not immediately suggest strong control of recession rates. The most striking relationship between all of these sites was that none of them had a berm present in 2020. As a result, at certain sites, recession rates are increasing since there is no barrier to protect the bluff toe. However, at sites with constant or decreasing recession rates, it is likely that some other morphodynamic or hydrodynamic controls are interacting to keep recession rates low even with no berm present. Correlation to Site Morphometrics and Hydrodynamics 2012 site morphodynamics and hydrodynamics were compared to 2020 rates of bluff crest recession at all seventeen sites to discern the correlation. The linear regression analyses on site morphodynamic correlation to bluff recession rates suggest that none of the variables tested (nearshore slope, bluff face slope, crest elevation, toe elevation, beach width, and beach slope) from the 2012 site characteristics have a discernable correlation to the 2012-2020 rates of recession at the seventeen sites (Table 12). The linear regression analysis on site hydrodynamics (bluff azimuth, fetch, and wave power) also do not show a strong correlation with recession rates (Table 16). All tested relationships have a significantly low R-squared value and P-value, 103 indicating that there is no correlation between variables (Tables 12 and 16). This is not to say that these morphologic variables are not related to bluff recession, but rather that each do not singularly explain the spatial variability in recession. This may be due to error in the DSAS- generated rates of recession from user interpretation error or georeferencing error, although care was exerted to minimize these. Given the complete lack of relationships between the morphometrics and recession rates these results may suggest that the aforementioned parameters do not have a strong control over bluff recession and that rates of recession may be more closely related to variables not tested, such as specific wave climate metrics and groundwater conditions. Though the correlation is not strong amongst other tested variables, the site-to-site bathymetric characteristics and degree of shoreline armoring seemed to have the strongest control on recession rates, from a qualitative perspective. The presence of inner and outer nearshore bars, amount of natural geographic protection (i.e., sites being located in bays), and degree of shore- parallel protection structures seems to exert the most control over magnitude of recession, given the results of this data set. More data collected at finer temporal and spatial scales is likely necessary to better constrain the role of the tested variables to bluff recession. In other words, the variables tested were not detailed enough to correlate to recession rates since the time scales at which these variables were averaged and may be masking patterns and relationships. The same can be said for the averaging of the recession rates among the nine transects utilized in DSAS as averaging may suppress the spatial variability in recession patterns within a site. Commentary on Methodology The methodology of this study is a mixture of Buckler 1981 coupled with modern, higher accuracy approaches. The two primary methods were: (1) utilizing digitized bluff crests and the DSAS and (2) completing field surveys of bluff crest locations using RTK-GPS. Each of these 104 approaches has pros and cons related to ease of access/convenience and the amount of error introduced. While digitizing bluff crest locations from georeferenced historical aerial imagery is convenient as it is open source and can be done in-office, interpretation errors can be introduced. Georeferencing historical aerial imageries is difficult on two accounts: first, the image quality results in difficulty finding distinguishable features to use as control points and second, urbanization and land use change leads to large differences in the modern, reference images (Moore 2000). In different applications, the amount of error created by georeferencing may be acceptable (i.e., Thieler and Danforth 1994), but is not acceptable when the amount of bluff crest change between time periods is small in magnitude. Additionally, the method of testing for accuracy in georeferencing is not exact as the root mean squared error (RMSE) value of control points is not a good indicator of the accuracy of the whole image since the RMSE just denotes the accuracy of those individual points and does not account for warping due to weighting (Hughes et al. 2006). Also, although the swipe tool in GIS is useful for visually assessing accuracy, coarse resolution makes it difficult to pick out features to use as a reference control points. Additionally, there are inherent georeferencing errors that were noticed in the recent aerial imagery (e.g., 2005 NAIP imagery) that was purportedly already georeferenced when acquired for the study. With the resolution that is needed to analyze fine-scale changes in bluff crest position, any introduced error can skew the results. As these errors compound, it is important to take them into consideration. Further, error in interpretation is introduced during digitizing and is most related to image quality, vegetation cover, and bluff crest interpretation where the position is ambiguous. Granular quality and coarse resolution of some historical aerial imagery can make detecting the 105 bluff crest location extremely difficult. Heavily vegetated sites also create difficulty in delineating the crest location. Often, vegetation is distributed along both the bluff face, crest, and adjacent upland, making digitizing extremely difficult as you must rely on past years’ images and sparse pockets of visible crest on the image you are digitizing. Additionally, most aerial photographs are collected during the summer, leaf-on period, which hides the true crest location. Lastly, as bluffs recede in different ways, the bluff crest location can be somewhat arbitrary. For example, in aerial images for sites that were experiencing slumping, the bluff crest location is difficult to locate as it is unclear where the true crest location is versus what is dislodged material resting on the bluff face. Digitizing this dislodged material would result in erroneous crest recession rates. Overall, digitizing bluff crest locations on aerial photography is most accurate when there is little to no vegetation cover and the bluff crest is distinctive, such as at M43. The field surveys using the RTK-GPS introduced the least error of any of the methods it has centimeter accuracy and determining crest location is easier to do in the field than on an aerial photograph. The predominant issues that arose when using this method was due to vegetation cover and GPS signal problems. Overhead branches will disrupt the GPS signal, even in the absence of leaves, however leaves make it worse; this reduces the accuracy of the positioning system and increases the amount of time needed to stand on the bluff crest waiting for good horizontal and vertical accuracy. As outlined above, both methods offer pros and cons, however the cons can impact the results, which is undesirable for analyzing spatially and temporally fine-scale bluff crest recession. Since the RTK-GPS produces positional data of extremely high accuracy and DSAS bluff crest digitizing introduces many user interpretation errors, the overlap of accuracy (or 106 inaccuracy) can influence the results, leading to inaccurate rates of change over time. Overall, DSAS is beneficial for analyzing change over time for timesteps of 5 years or greater and RTK- GPS is best for shorter timespans since image quality creates difficulty in digitizing subtle change differences. Moving forward, when measuring bluff crest locations and assessing short- term recession rates, aerial photographs should be avoided in order to diminish errors associated with georeferencing and digitizing. Aerial photographs are best used in this application when analyzing longer-term rates of bluff recession. Additionally, more consideration should be put towards implementing LiDAR and UAS methods to enhance bluff recession studies by focusing on three-dimensional geomorphic analysis, as opposed to the strictly two-dimensional analysis in the DSAS, in order to better understand the geomorphic relationship between bluff geometry (i.e., bluff face slope, bluff crest height) and recession rates. Future Work As bluffs in the Great Lakes region are often populated by homeowners, there will be a continual demand to understand how and why bluffs recede, in order to avoid losing more property to the lake. As these cohesive bluffs can only move landward over time, there is a continual urgency to understand recession as these landforms cannot be naturally reconstructed. Further, lake level in Lake Michigan is currently high and is projected to stay relatively high (USACE Detroit District Website). With this, studies on linking bluff recession to erosional facilitators and mediating site characteristics at varying spatial and temporal scales must continue in order to work toward creating a probabilistic model to estimate impending recession rates. Future work should focus on choosing a consistent methodology that will reduce both technological error and user interpretation error in order to measure bluff crest position most accurately. In this regard, RTK-GPS is the most accurate, and user-friendly method of obtaining 107 crest locations. However, this has consequences as it involves more time spent in the field. Nonetheless, physically seeing a bluff site firsthand is a crucial step in understanding the drivers, mediators, and patterns of bluff recession. With this, careful consideration should be taken when choosing bluff sites. In particular, tree cover should be minimized in order to reduce multipath errors which degrade the accuracy of RTK-GPS positions. The use of UAVs should be incorporated more often since bluff crest position can be more easily located by extracting and analyzing topographic profiles from the DEMs produced by structure from motion techniques. In future work, more attention should be paid to processes driving change over time at the bluff toe. The bluff toe is where the majority of bluff recession is initiated by instability due to undercutting. Additionally, as the literature and these results show, there are often no strong relationships between recession rates and bluff site characteristics such as bluff height, bluff face slope, beach width, etc. Again, this is not to say that these parameters do not have an influence, however the results from this study suggest that there are likely some different mediators or characteristics at work. More attention should be paid to the role that specific wave climate characteristics (i.e., wave period, wave direction, deep water wavelength) have in driving bluff recession. These wave characteristics should be analyzed at finer time-scales since analyzing wave influence at large timescales involves averaging of values which may mask the relationship with bluff toe erosion. Further, more understanding is needed on how vegetation mediates bluff recession; for example, a greater understanding of which vegetation types anchors the material best, or how different densities of vegetation impact recession rates, could greatly improve shoreline management and green infrastructure approaches to preventing recession. Without question, the role of shoreline armoring needs to be further assessed. The literature supports that shoreline armoring may negatively impact shorelines by actually promoting erosion, however 108 more understanding is needed on how these structures specifically impact recession rates. The potential negative impacts of shoreline armoring should be a main focus since shoreline armoring is a common shore protection strategy. A recommended future study would be to use UAVs and RTK-GPS to complete routine field surveys following the occurrence of storm events.; additionally, local buoy and weather station data should be utilized so that rates of change might then be correlated to wave and atmospheric parameters. Further, sites should be chosen such that one is downdrift of a shore- perpendicular structure, one has adjacent shore-parallel armoring, and one has no armoring. Bathymetric surveys and RTK-GPS wading surveys should also be conducted to monitor the evolution of the nearshore bathymetry and sediment movement that may play important roles in protecting the bluff toe. Wave climate and extent of shoreline armoring seem to be large determinants of recession rates and such a study would allow for quantifiable results and better correlations to be determined between variables. 109 Conclusion The intent of the study was to establish linkages between bluff recession, lake levels, erosional processes (i.e., wave climate), and site characteristics (i.e., bluff height, lithology, nearshore bathymetry) based on hypotheses that specific geomorphic characteristics and erosional drivers dictate recession rates. However, all variables tested were not meaningfully correlated to recession rates. Additionally, this study aimed to identify if bluff recession rates were increasing over time. Originally, it was thought that it would be evident if bluff sites were experiencing an increase, decrease, or consistency in recession rates over time. However, my thesis findings suggest that it is not so simple to say bluff recession rates are accelerating, decelerating, or staying the same; that is an oversimplification of these dynamic coastal environments. Rather, bluff recession rates vary over time depending on the magnitude of the erosional drivers including wave climate, and a host of mediators like degree of shoreline armoring, and various geologic and geomorphic characteristics. It is crucial to understand how these erosional drivers influence bluffs that have subtle differences in characteristics (i.e., lithology, crest and toe height, beach width) in order to create predictive models of future recession rates. However, at the scale of this study, these fine-scale subtleties were often masked by averaging the data. Put simply, there are many dynamic variables at play that need to be quantified at both large and small spatiotemporal scales to better understand process and bluff recession relationships. The conclusions of this study can be summarized as: 1. Based on the variables tested, there are no immediate similarities between sites to statistically explain the large spatial and temporal variability in recession rates. 110 2. More focus should be paid to quantifying and understand the coupled changes at the bluff toe and crest, to better constrain the three-dimensional geomorphic recession patterns of Great Lakes bluffs. 3. It is difficult to quantify and represent fine-scale bluff recession patterns in aerial photographs. With the record of images used, small changes are masked by the timeframe in consideration. Also, analyzing bluff crest recession from historical images and the DSAS introduces the larger possibility of error than RTK-GPS, LIDAR, and UAVs because of user interpretation and georeferencing errors. More attention should be given toward accurately and thoroughly representing small changes in bluff recession at individual sites, while still attempting to understand bluff recession at large spatiotemporal scales, by using a larger and/or longer record of aerial photographs, coupled with field surveys. 4. At the seventeen sites studied, recession rates do not have a strong correlation with the following variables: nearshore slope, bluff face slope, bluff crest elevation, bluff toe elevation, beach width, beach slope, maximum perpendicular fetch distance, perpendicular fetch azimuth, shoreline azimuth, and wave power. This is not to say that these variables do not have an influence on recession rates. Possibly due to noise introduced by user interpretation error, the data do not show any significant statistical relationships with the tested variables. 5. Lithology does not seem to have a strong impact on the fluctuation of recession rates since the rates fluctuated over time while lithology theoretically remained the same; however, more frequent, and detailed stratigraphic analysis is needed to assess the degree of change in lithology experienced by a bluff after material is eroded away. 111 6. Future studies should focus on quantifying the relationship between wave climate and characteristics and degree of shoreline armoring by completing frequent field surveys surrounding the occurrence of storm events. This should be done at carefully selected sites with minimum vegetation cover. With these conclusions in mind, this study ultimately yields insight into improved and more efficient methods of discerning bluff crest location and quantifying recession rates over varying spatiotemporal scales. Further, this study gathered insight into the spatial and temporal variability of bluff recession along the eastern shoreline of Lake Michigan. This study highlights the necessity for frequent bluff crest location surveys in order to better correlate recession rates with complex erosive forces, namely wave climate parameters. These conclusions may be applied to other cliffed coastlines outside of the Great Lakes region. As these landforms continue to be developed and modified into the future, it is becoming increasingly important to fully understand how they evolve, so as to better manage and protect them. 112 BIBLIOGRAPHY 113 BIBLIOGRAPHY 1. R. P. Allan, Climate change: Human influence on rainfall. Nature. 470 (2011), pp. 344–345. 2. M. Amen, R. Harris, R. C. Baird, Coastal urbanization: the challenge of management lag. Management of Environmental Quality: An International Journal. 20, 371–382 (2009). 3. S. M. N. Amin, R. G. D. Davidson-Arnott, Toe erosion of glacial till bluffs: Lake Erie south shore. Canadian Journal of Earth Sciences. 32, 829–837 (1995). 4. A. J. Bechle et al., Meteotsunamis in the Laurentian great lakes. Scientific Reports. 6 (2016), doi:10.1038/srep378 5. D. Beletsky, D. Schwab, Climatological circulation in Lake Michigan. Geophysical Research Letters. 35 (2008), doi:10.1029/2008GL035773. 6. W. A. Birkemeier, The effect of structures and lake level on bluff and shore erosion in Berrien County, Michigan, 1970-74. US Army Coastal Engineering Research Center, Miscellaneous Report. 80 (1980). 7. P. D. Blanken, C. Spence, N. Hedstrom, J. D. Lenters, Evaporation from Lake Superior: 1. Physical controls and processes. Journal of Great Lakes Research. 37, 707–716 (2011). 8. J. Bramante, Why is wave climate important?. Woods Hole Oceanographic Institution. (September 4th, 2017). Retrieved from: https://web.whoi.edu/big/why-is-wave-climate- important/#:~:text=Wave%20climate%20is%20the%20distribution,a%20given%20location %20and%20time. 9. C. B. Breed and G. L. Hosmer, The Principles and Practice of Surveying, 11th ed. New York: John Wiley & Sons. pp. 100-108. (1077) 10. S. F. Brennan, P. E. Calkin, Analysis of bluff erosion along the southern coastline of Lake Ontario, New York. New York Sea Grant Institute. (1984). 11. E. A. Brown, C. H. Wu, D. M. Mickelson, T. B. Edil, Factors controlling rates of bluff recession at two sites on Lake Michigan. Journal of Great Lakes Research. 31, 306–321 (2005). Buckler, W.R. 1973. Bluff erosion at selected sites along the southeastern shore of Lake Michigan. Unpublished M.A. research paper, Michigan State University, East Lansing, Michigan, 68 pp. 114 12. W. R. Buckler, Rates, and implications of bluff recession along the Lake Michigan shorezone of Michigan and Wisconsin. Ph.D. dissertation, Department of Geography, Michigan State University. (1981). 13. W. R. Buckler, H. A. Winters, Lake Michigan Bluff Recession. Annals of the Association of American Geographers. 73, 89–110 (1983). 14. C. H. Carter, Erosion rates of rock, till, and glaciolacustrine clay along the U.S. shore of Lake Erie. Geological Society of America. Abstract Programs 8:470. 15. C. H. Carter, D. E. Guy Jr., Coastal erosion: Processes, timing and magnitudes at the bluff toe. Marine Geology. 84, 1–17 (1988). 16. R. W. G. Carter, S. C. Jennings, J. D. Orford, Headland erosion by waves. Journal of Coastal Research. 6, 517–529 (1990). 17. R. Castedo, M. Fernández, A. S. Trenhaile, C. Paredes, Modeling cyclic recession of cohesive clay coasts: Effects of wave erosion and bluff stability. Marine Geology. 335, 162– 176 (2013). 18. J. L. Catto et al., The Future of Midlatitude Cyclones. Current Climate Change Reports. 5 (2019), pp. 407–420. 19. R. H. Clemens, The role of vegetation in shoreline management: a guide for Great Lakes shoreline property owners. National Oceanic and Atmospheric Administration. (1970). 20. S. M. Colman et al., Lake-Level History of Lake Michigan for the Past 12,000 Years: The Record From Deep Lacustrine Sediments. Journal of Great Lakes Research. 20, 73–92 (1994). 21. B. D. Collins, N. Sitar, Processes of coastal bluff erosion in weakly lithified sands, Pacifica, California, USA. Geomorphology. 97, 483–501 (2008). 22. R. A. Davis Jr., E. Seibel, W. T. Fox, Coastal erosion in eastern Lake Michigan – causes and effects. (1973). 23. S. A. Dawson, J. E. Evans, Geological causes of local variation in coastal bluff recession rates, northeast Ohio Shoreline of Lake Erie. Environmental Geosciences. 8, 1–10 (2001). 24. M. N. Dethier, J. D. Toft, H. Shipman, Shoreline Armoring in an Inland Sea: Science-Based Recommendations for Policy Implementation. Conservation Letters. 10, 626–633 (2017). 25. A. Dufresne, T. R. Davies, Longitudinal ridges in mass movement deposits. Geomorphology. 105, 171–181 (2009). 115 26. C. Earlie, G. Masselink, P. Russell, The role of beach morphology on coastal cliff erosion under extreme waves. Earth Surface Processes and Landforms. 43, 1213–1228 (2018). 27. T. B. Edil, L. E. Vallejo, Mechanics of coastal landslides and the influence of slope parameters. Engineering Geology. 16, 83–96 (1980). 28. Fox 2 Detroit, Cottage collapses into Lake Michigan after years of erosion. (January 3rd, 2020). Retrieved from: https://www.fox2detroit.com/news/cottage-collapses-into-lake- michigan-after-years-of-erosion 29. G. Esposito, R. Salvini, F. Matano, M. Sacchi, C. Troise, Evaluation of geomorphic changes and recession rates of a coastal pyroclastic cliff in the Campi Flegrei volcanic district, southern Italy. Journal of Coastal Conservation. 22, 957–972 (2018). 30. W.R. Farrand, D. L. Bell, Quaternary Geology of Southern Michigan, Michigan Department of Natural Resources, Michigan Geological Survey, Geological Publication QG-01, scale 1:500,000. (1982). 31. R. J. Geier, P. E. Calkin, Great Lakes coastal geology: stratigraphy and bluff recession along the Lake Erie coast, New York. (1983) 32. Great Lakes Shoreviewer. U.S. Climate Resilience Toolkit. Retrieved from: http://www.greatlakesshoreviewer.org/#/great-lakes. (2019). 33. G. Griggs, THE IMPACTS OF COASTAL ARMORING. Shore and beach, 13–22 (2005). 34. A. D. Gronewold et al., Coasts, water levels, and climate change: A Great Lakes perspective. Climatic Change. 120, 697–711 (2013). 35. M. A. Hampton, G. B. Griggs, Formation, Evolution, and Stability of Coastal Cliffs – Status and Trends. USGS Professional Paper 1693, 129 (2004). 36. L. M. Highland, P. Bobrowsky, The landslide Handbook - A guide to understanding landslides. US Geological Survey Circular, 1–147 (2008). 37. E. A. Himmelstoss, R.E. Henderson, M.G. Kratzmann, A. S. Farris, Digital Shoreline Analysis System (DSAS) version 5.0 user guide: U.S. Geological Survey Open-File Report 2018–1179, 110 p., (2018). https://doi.org/10.3133/ ofr20181179. 38. M. L. Hughes, P. F. McDowell, W. A. Marcus, Accuracy assessment of georectified aerial photographs: Implications for measuring lateral channel movement in a GIS. Geomorphology. 74, 1–16 (2006). 39. M. D. Hursta, D. H. Rood, M. A. Ellis, R. S. Anderson, U. Dornbusch, Recent acceleration in coastal cliff recession rates on the south coast of Great Britain. Proceedings of the National Academy of Sciences of the United States of America. 113, 13336–13341 (2016). 116 40. R. W. Jibson, J. K. Odum, J. M. Staude, Rates and Processes of Bluff Recession Along the Lake Michigan Shoreline in Illinois. Journal of Great Lakes Research. 20, 135–152 (1994). 41. J. Johannessen, Assessing Littoral Sediment Supply (Feeder Bluffs) and Beach Condition in King and Southern Snohomish Counties, Puget Sound, Washington. Coastal Geologic Services Inc. 135-152. (2010). 42. Johnson, B. L. & Johnston, C. A. Relationship of Lithology and Geomorphology to Erosion of the Western Lake Superior Coast. Journal of Great Lakes Research 21, 3–16 (1995). 43. J. W. Kamphuis, Recession Rate of Glacial Till Bluffs. Journal of Waterway, Port, Coastal, and Ocean Engineering. 113, 60–73 (1987). 44. J. P. Keillor, and R. DeGroot, Recent recession of Lake Michigan shorelines in Racine County, Wisconsin. University of Wisconsin Sea Grant College Program Special Report No. 507. University of Madison – Wisconsin. (1978). 45. K. C. Kornelsen, P. Coulibaly, Synthesis review on groundwater discharge to surface water in the Great Lakes Basin. Journal of Great Lakes Research. 40 (2014), pp. 247–256. 46. R. Krueger, L. K. Zoet, J. E. Rawling, Coastal Bluff Evolution in Response to a Rapid Rise in Surface Water Level. Journal of Geophysical Research: Earth Surface. 125 (2020), doi:10.1029/2019JF005428. 47. R. B. Langley, Rtk Gps. GPS World, 70–75 (1998). 48. G. Larson, R. Schaetzl, Origin and evolution of the Great Lakes. Journal of Great Lakes Research. 27 (2001), pp. 518–546. 49. P. L. Lawrence, Natural hazards of shoreline bluff erosion: a case study of Horizon View, Lake Huron. Geomorphology. 10, 65–81 (1994). 50. J. Lewis, A new framework for monitoring the Great Lakes water balance error. U.S. Army Engineer Research and Development Center, Coastal and Hydraulics Laboratory. (2005). 51. D. Lusch, Remounted Survey Section Corner Data. Personal Communication. (2019). 52. S. D. Mackey, Great Lakes Nearshore and Coastal Systems. U.S. National Climate Assessment Midwest Technical Input Report, 14 (2012). 53. Maine Geological Survey, Types of Bluffs Along Maine’s Coast. Maine Department of Agriculture, Conservation & Forestry. (2011) 54. R. M. Maxwell, N. L. Miller, On the development of a coupled land surface and groundwater model. Developments in Water Science. 55, 1503–1510 (2004). 117 55. Michigan Legislature Website, State Survey and Remonumentation Act. Retrieved from: http://www.legislature.mi.gov/(S(sgh1r3aiplfqk1rhtasx0kxt))/mileg.aspx?page=getobject&ob jectname=mcl-act-345-of-1990&queryid=1810308&highlight=. 56. D. M. Mickelson, T. B. Edil, D. E. Guy, Erosion of coastal bluffs in the Great Lakes. US Geological Survey Professional Paper, 107–123 (2004). 57. W. W. Montgomery, Groundwater Hydraulics and Slope Stability Analysis: Elements for Prediction of Shoreline Recession. Ph.D. Dissertation, Western Michigan University. (1998). 58. L. J. Moore, Shoreline mapping techniques. Journal of Coastal Research. 16, 111–124 (2000). 59. A. Morang, M. C. Mohr, C. M. Forgette, Longshore sediment movement and supply along the U.S. shoreline of Lake Erie. Journal of Coastal Research. 27, 619–635 (2011). 60. L. Mortsch et al., Climate change impacts on the hydrology of the great lakes-st. Lawrence system. Canadian Water Resources Journal. 25, 153–179 (2000). 61. B. Neff, J. Nicholas, Uncertainty in the Great Lakes Water Balance Scientific Investigations Report 2004-5100. Water, 42 (2005). 62. National Oceanic and Atmospheric Administration Great Lakes Environmental Research Laboratory. Great Lakes Coastal Forecasting System, Lake Michigan [NOWCAST and FORECAST]. Dataset. https://www.glerl.noaa.gov/res/glcfs/ (2021). 63. Osher Map Libraries, Nineteenth-Century Surveying. Smith Center for Catagraphic Education. Retrieved from: https://oshermaps.org/exhibitions/american-treasures/section-4. 64. G. Pierre, Processes and rate of recession of the clay and sandstone sea cliffs of the northern Boulonnais (France). Geomorphology. 73, 64–77 (2006). 65. R. M. Quigley, P. J. Gelinas, W. T. Bou, R. W. Packer, (1976), pp. 10–27. 66. C. J. Roland, L. K. Zoet, J. E. Rawling, M. Cardiff, Seasonality in cold coast bluff erosion processes. Geomorphology. 374 (2021), doi:10.1016/j.geomorph.2020.107520. 67. P. Ruggiero, Impacts of shoreline armoring on sediment dynamics. Puget Sound Shorelines and the Impacts of Armoring—Proceedings of a State of the Science Workshop. U.S. Geological Survey Scientific Investigations Report., 179 – 186 (2009). 68. E. Seibel, E. 1972. Shore erosion at selected sites on Lake Michigan and Lake Huron. Ph.D. dissertation, Department of Oceanography, University of Michigan. (1972). 118 69. H. Shipman, A Geomorphic Classification of Puget Sound Nearshore Landforms. Washington Department of Ecology. (2008). 70. R. M. Sorensen, Basic coastal engineering. (1978), doi:10.1016/0378-3839(79)90016-4. 71. T. Sunamura, Processes of sea cliff and platform erosion. Japanese Geomorphological Union. (1983) 72. H. Stanchev, R. Young, M. Stancheva, Integrating GIS and high resolution orthophoto images for the development of a geomorphic shoreline classification and risk assessment-a case study of cliff/bluff erosion along the bulgarian coast. Journal of Coastal Conservation. 17, 719–728 (2013). 73. J. A. Steers, J. L. Davies, Geographical Variation in Coastal Development. The Geographical Journal. 139, 350 (1973). 74. M. J. Swenson, C. H. Wu, T. B. Edil, D. M. Mickelson, Bluff recession rates and wave impact along the Wisconsin coast of Lake Superior. Journal of Great Lakes Research. 32, 512–530 (2006). 75. E. R. Thieler, W. W. Danforth, Historical shoreline mapping (II): application of the digital shoreline mapping and analysis systems (DSMS/DSAS) to shoreline change mapping in Puerto Rico. Journal of Coastal Research. 10, 600–620 (1994). 76. T. A. Thompson, S. J. Baedke, Beach-ridge development in Lake Michigan: shoreline behavior in response to quasi-periodic lake-level events. Marine Geology. 129, 163–174 (1995). 77. T. A. Thompson, S. J. Baedke, Strand-plain evidence for late Holocene lake-level variations in Lake Michigan. Bulletin of the Geological Society of America. 109, 666–682 (1997). 78. G. W. Thorson, Soil bluffs + rain = slide hazards. Washington Geologic Newsletter. 3-11 (1987). 79. Trimble Website, What is RTK?. Retrieved from: https://www.trimble.com/OEM_ReceiverHelp/V4.44/en/PositionModes_WhatisRTK.html 80. University of Oregon Libraries Website, About GLO Maps. Retrieved from: https://library.uoregon.edu/map/map_resources/about_glo 81. US Army Corps of Engineers, Great Lakes Water Level Data. USACE Detroit District Website. Retrieved from: https://www.lre.usace.army.mil/Missions/Great-Lakes- Information/Great-Lakes-Information-2/Water-Level-Data/ 82. US Army Corps of Engineers, Experimental 5-year Outlook of Great Lakes Water Levels. USACE Detroit District Website. Retrieved from: 119 https://www.lre.usace.army.mil/Missions/Great-Lakes-Information/Great-Lakes-Water- Levels/Water-Level-Forecast/Long-Term-Forecast/ 83. L. E. Vallejo, R. Degroot, Bluff response to wave action. Engineering Geology. 26, 1–16 (1988). 84. S. Vavrus, J. Van Dorn, Projected future temperature and precipitation extremes in Chicago. Journal of Great Lakes Research. 36, 22–32 (2010). 85. Virtual Museum of Surveying, Changing Chains. Surveyors Historical Society. Retrieved from: http://www.surveyhistory.org/changing_chains.htm. 86. C. A. Volpano, L. K. Zoet, J. E. Rawling, E. J. Theuerkauf, R. Krueger, Three-dimensional bluff evolution in response to seasonal fluctuations in Great Lakes water levels. Journal of Great Lakes Research. 46, 1533–1543 (2020). 87. Z. Wang et al., Coupled model of surface runoff and surface-subsurface water movement. Advances in Water Resources. 137 (2020), doi:10.1016/j.advwatres.2019.103499. 88. 1. J. A. Winkler, in Geography and Technology (Springer Netherlands, 2004), pp. 461–490. 89. J. A. Winkler, R. W. Arritt, S. C. Pryor, Climate Projections for the Midwest: Availability, Interpretation and Synthesis. US National Climate Assessment Midwest Technical Input Report, 24 (2012). 90. W. L. Wood, Effects of Seawalls on Profile Adjustment Along Great Lakes Coastlines. Journal of Coastal Research, 135–146 (1988). 91. D. C. Wyllie, C. W. Mah, Rock slope engineering: Civil and mining, 4th edition (CRC Press, 2017). 92. Y. T. Lin, C. H. Wu, A field study of nearshore environmental changes in response to newly- built coastal structures in Lake Michigan. Journal of Great Lakes Research. 40, 102–114 (2014). 93. A. P. Young, S. A. Ashford, Application of airborne LIDAR for seacliff volumetric change and beach-sediment budget contributions. Journal of Coastal Research. 22, 307–318 (2006). 94. A. P. Young, J. E. Carilli, Global distribution of coastal cliffs. Earth Surface Processes and Landforms. 44, 1309–1316 (2019). 95. Y. Zhong, M. Notaro, S. J. Vavrus, M. J. Foster, Recent accelerated warming of the Laurentian Great Lakes: Physical drivers. Limnology and Oceanography. 61, 1762–1786 (2016). 120 96. L. K. Zoet, J. E. Rawling, Analysis of a sudden bluff failure along the southwest Lake Michigan shoreline. Journal of Great Lakes Research. 43, 999–1004 (2017). 121