LAKE TROUT HABITAT SELECTION AT DRUMMOND ISLAND SPAWNING REEFS : PARADIGM OR PARADOX ? By Steven A. Farha A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Fisheries and Wildlife Master of Science 2018 ABSTRACT LAKE TROUT HABITAT SELECTION AT DRUMMOND ISLAND SPAWNING REEFS : PARADIGM OR PARADOX? By Steven A. Farha Progress towards r estoration of self - sustaining populations of lake trout in the Laurentian Great Lakes has bee n slow. Adequate survival of stocked yearlings to maturity suggests potential recruitment bottlenecks exist between spawning and the yearling life stage , possibly related to an inability to select spawning habitats suitable for egg incubation. However, little is known about the fundamental features of lake trout spawning habitats or the criteria by which it is selected. T his information gap was addressed using a nov el acoustic telemetry - based approach whereby sampling effort was apportioned among habitats known to have been encountered during the spawning period based on behavioral da ta from tagged adult lake trout . Among sites e ncoun t er ed by lake trout on three spaw ning reefs in northern Lake Huron , p hysical characteristics and egg survival w ere compared between habitats positive and negative for egg deposition . Widespread egg deposition was confirmed on all three reefs, but variables measured (s ubstrate diameter, ho mogeneity, interstitial depth , and slope) were unable to fully explain the habitat selection observed, suggesting other variables may have been important. Based on an in situ habitat bioassay, egg survival did not differ between habitats that received egg deposition and those that did not , suggesting the supply of spawning habitat may exceed demand in this area. Further, lake trout spawned in habitats inconsistent with the commonly - accepted lake trout spawning habitat paradigm , including previously und escribed boulder - associated habitats . T hese unconventional habitat s not only received egg deposition , but also produced viable juveniles , forcing us to rethink, adapt, and expand our conceptual understanding of suitable trout spawning habitat. iii AC KNOWLEDGEMENTS Funding and administrative support for this research was provided by the Great Lakes Fishery Commission (Grant ID #2013_BIN_44024) via the Great Lakes Restoration Initiative (Grant ID #GL - 00E23010). A hearty thank you to the numerous indivi duals that provided field and lab support throughout the project; this thesis is testament to your dedication and hard work . The journey behind this thesis was largely made possible due to the unwavering support of my colleagues, friends and family . I am exceedingly grateful for the wonderful opportunity provided by my Co - Dr. Chuck Krueger and Dr. Tom Binder to challenge myself to grow and learn. Th ank you for teaching me to never stop look ing at my work with a critical eye, no matter how difficu lt. Your determined leadership and repeated pushing was the steady current that got us here . Thank you for sharing your time and energy , and of course, the innumerable edits along the way . Above all, thank you for your belief in me even when my own faltere d . I also thank Dr. Dan Hayes, for his seemingly limitless insight and uncanny ability to simplify the most complicated of analyses. Thank you for assuring me that no matter how dim it may seem, there is, in fact, a light at the end of the tunnel. To the professors of the D epartment of F isheries and W ildlife at MSU , thank you for challenging me to view issues from multiple perspectives. The dedication with which each of you have shared your expertise was inspiring, and I am richer for it; thank you. I forever grateful to m y colleagues at the USGS - Great Lakes Science Center for the opportunity to chase a dream . T h ank you for granting me the freedom to broaden my perspective and the means to apply it. To the deepwater team, thank you for helping me throug h the iv countless roadblocks; all the check - ins, well - wishes and reviews were more helpful than you know. I am also grateful for the warm hospitality of staff of the USGS - Hammond Bay Biological station, who welcomed me as one of their family. Thank you for your support, and the great lunch time pinochle memories . To my family, thank you for instilling in me a n insatiable desire to question everything and the passion to chase the answers . Dearest Ale x, my love, thank you for your seemingly infinite patience and unwavering vision . Your steadfast belief in me and willingness to chase headlong into the unknown has redefined the meaning of love , thank you. O be foolish not to mention how exceptionally grateful I am for the countless packed lunches and late - night coffee breaks (snacks are important!) ; you really are the best. Ludo, thank you for joyfully sticking beside me through all the travel these last few same without you, pup. Lastly, this project would not have been possible without the warm hospitality and kindness of the residents of Drummond Island. A special thanks to the Drummond Island Volunteer Fire and Rescue Team, Galen Terry of Vacation Real Estate, Thor and Wendy at the Country Store, Esther and her wonderful Authentic Mexican Cuisine, Julie Covert of the Drummond Island Digest, and the staff at the Drummond Island Resort, Sunes Home Center and T he Northwood for their support and for share laughs, and fill a belly or two. Cheers ! v TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ...................... vii LIST OF FIGURES ................................ ................................ ................................ ....................... x KEY TO ABBREVIATIONS ................................ ................................ ................................ ... xiii Cha pter 1 : Introduction ................................ ................................ ................................ ............... 1 REFERENCES ................................ ................................ ................................ ....................... 10 Chapter 2 : Evaluation of lake trout habitat selection at Drummond Island spawning reefs ................................ ................................ ................................ ................................ ........................ 16 Ab stract ................................ ................................ ................................ ................................ ... 16 Introduction ................................ ................................ ................................ ............................. 17 Methods ................................ ................................ ................................ ................................ ... 20 Study Area ................................ ................................ ................................ ........................ 20 Habitat Use Index ................................ ................................ ................................ ............. 23 Surveys fo r Egg Deposition ................................ ................................ .............................. 24 Physical Characterization (Objective 1) ................................ ................................ ........... 25 Relative Importance of Physical Characteristics (Objective 1) ................................ ........ 28 Incubation Success (Objective 2) ................................ ................................ ...................... 30 Results ................................ ................................ ................................ ................................ ..... 32 Egg Deposition ................................ ................................ ................................ .................. 32 General Spawning Reef Characteristics ................................ ................................ ............ 33 Relative Importance of Physical Characteristics (Objective 1) ................................ ........ 36 Incubation Success (Objective 2) ................................ ................................ ...................... 39 Discussion ................................ ................................ ................................ ............................... 42 Acknowledgements ................................ ................................ ................................ ................. 50 REFERENCES ................................ ................................ ................................ ....................... 51 Chapter 3 : Atypical spawning habitat selection by lake trout in the Drummond Island Refuge, Lake Huron: spawning in substrates at the base of giant boulders .......................... 61 Abstract ................................ ................................ ................................ ................................ ... 61 Introduction ................................ ................................ ................................ ............................. 62 Methods ................................ ................................ ................................ ................................ ... 67 Site Selection ................................ ................................ ................................ .................... 67 Surveys for Eg g Deposition (Objective 1) ................................ ................................ ........ 70 Physical Characterization (Objective 2) ................................ ................................ ........... 70 Abundance of Free Embryos and Alevins (Objective 3) ................................ .................. 72 Statistical Comparisons ................................ ................................ ................................ ..... 76 Results ................................ ................................ ................................ ................................ ..... 77 Surveys for Egg Deposition (Objective 1) ................................ ................................ ........ 77 Physical Characterization (Objective 2) ................................ ................................ ........... 78 Abundance of Free Embryos and Alevins (Objective 3) ................................ .................. 83 Discussion ................................ ................................ ................................ ............................... 88 vi Acknowledgments ................................ ................................ ................................ ................... 94 REFERENCES ................................ ................................ ................................ ....................... 95 Chapter 4 : Conclusion ................................ ................................ ................................ .............. 102 REFERENCES ................................ ................................ ................................ ..................... 112 APPENDICES ................................ ................................ ................................ ............................ 116 APPENDIX A ................................ ................................ ................................ ........................ 117 APPENDIX B ................................ ................................ ................................ ........................ 121 vii L IST OF TABLES Table 2.1: Models and parameter de finitions used to predict egg deposition on Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ) ranked by AIC corrected for small sample sizes ( AICc ). All models include a spatial autocorrelation term (distance) that was omitted from this table to aid in clarity. Dashed line represents the information theoretic cutoff for models with moderate support ( < 5.0). Categorical substrate classifications w ere based on a Modified Wentworth scale ( Marsden et al. , 1995 ) sites with >33.3% of measured substrates in more than one category. ................................ . 37 Table 2.2: Unstandardized and standardized (denoted by an *) model - averaged parameter estimates and relative importance of physical characteristics for the selection of habitats for egg deposition by lake trout on Horseshoe Reef ( HSR ). An unconditional variance estimator was used to calculate 95% confidence intervals on model - averaged parameter estimates. Furthermore, the relative importance of each physical characteristic was expressed i n two ways; first, as a ratio of the ith parameter Akaike weight to the maximum Akaike weight , and second, as an evidence ratio of the i th standardized model - averaged parameter estimate to the maximum standardized parameter estimate Based on these criteria, slope magnitude was the only measured characteristic found to be important for the selection of spawning habitats on HSR . ........... 38 Table 3.1: Summary of boulder, overhang, and substrate characteristics. Substrates were characterized using a categorical classification based on the percentage of measured substrate within size bins using a modified Wentwort h scale (Marsden et al., 1995a, Farha, Ch. 1, this volume) . The size bin with the largest proportion of measured substrate was usually considered the predominant substrate size, but sites with multiple categories having greater than 33.3% of measured subst rates were given a mixed classification. Substrates at Boulder Alley ( BLD ) located around the base and away from boulders were smaller (gravel - rubble mix; 0.2 - 25.6 cm) than substrates on Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ) (rubble; 6.5 25.6 cm ; Chi - square test of independence; 2000 simulations, 2 = 22.8, df = 6, p < 0.01, expected frequencies in parentheses). Boulder - associated substrates on BLD also had less interstitial depth than substrates at HSR , but greater than those at SCS (One - way ANO VA , f (2, 113) = 41.8 , p < 0.01; Tukey HSD , p <0.01 for all comparisons). Variability was expressed as the coefficient of variation (CV) for comparison among parameters with differing units. ................................ ................................ ............................. 81 Table 3.2: Weekly relative abundance (CPUE: fish · trap - 1 · d - 1 ) of lake trout free embryos and alevins caught in three habitat classes using custom rectangular traps at Boulder Alley ( BLD ) and traditional round and custom rectangular traps on typical habitats on Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ) from May 28 June 24, 2015. Habitat classes for viii BLD included: habitats beneath overhanging edges of boulder s ( beneath boulder), habitats adjacent to a boulder ( beside boulder), and nearby habitats at least 2x the diameter of a boulder away ( away from boulder). N indicates the weekly number of traps of each type deployed at each site. CV represents the coefficie nt of variation (%) for comparing weekly variability among reefs. ................................ ................................ ................................ ......... 86 Table A.1: Unstandardized and standardized (denoted by an *) model - averaged parameter estimates and relative importance of physical characteristics for the selection of habitats for egg deposition by lake trout on Horseshoe Reef (HSR). An unconditiona l variance estimator was used to calculate 95% confidence intervals on model - averaged parameter estimates. Furthermore, the relative importance of each physical characteristic was expressed in two ways; first, as a ratio of the ith parameter Akaike weight to the maximum Akaike weight , and second, as an evidence ratio of the ith standardized model - averaged parameter estimate to the maximum standardized parameter estimate . Based on these criteria, none of the measure d characteristics were found to be important for the selection of spawning habitats on SCS . ................................ .............. 118 Table A.2: Complete dataset used for analyses on Horseshoe Reef ( HSR) . Logistic and incubators columns are binary dummy variables indicating whether site data was used (1) or not (0) in the respective analyses presented in Chapter 1. On HSR , site 37 was ex cluded due to technical issues that prevented complete analysis and all sites not - encountered by lake ................................ ... 119 Table A.3: Complete dataset used for analyses on Scammon Shoal ( SCS) . Logistic and incubators columns are binary dummy variables indicating whether site data was used (1) or not (0) in the respective analyses p resented in Chapter 1. On SCS , all sites not - encountered ...................... 120 Table B.4: Complete boulder characterization data for all boulders characterized on Boulder Alley ( BLD ). In total, 40 boulders were surveyed consisting of 86 discrete overhangs. .... 122 Table B.5: Complete substrate characterization data for boulder - associated substrates adjacent to boulders at Boulder Alley ( BLD ) and beneath overhangs vs. the most popular sites on Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ). Substrate was classified based on the proportion of measured substrate in size bins using a modified Wentworth scale (Marsden et al., 1995a, Farha et al. Ch. 1) . The size bin with the largest proportion of measured substrate was usually considered the predominant substrate size, but sites with multiple categories having greater than 33.3% of measured substrates were given a mixed classification. ................................ ................................ ................................ ....................... 123 ix Table B.6: Relative survival to hatch for incubators deployed at Boulder Alley ( BLD ) and controls ( CON ) using methodology described in Chapter 1 (See Chapter 1 Methods In cubation Success (Objective 2) for complete details). This analysis was removed from Chapter 3 for the sake of clarity. Habitat classes were (1) areas of clean gravel (< 6 cm diameter) associated with an overhang , (2) areas of gravel (< 6 cm diam.) not ass ociated areas of clean rubble (6 - 25 cm diam.) associated with an overhang , and (4) areas of rubble (6 - 25 cm diam.) not associated with boulder s . buried in the substrate ( Orientation = 1 moved ( Orientation = 2 ; either tipped over or completely dislodged from substrate). .................. 124 x LIST OF FIGURES Figure 2.1: Lake Huron study site (grey box in map inset) adjacent to Drummond Island with primary spawning reefs, Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ). Habitat Use Index ( HUI ), an index of relative amount of time tagged lake trout spent in various habitats over the 2012 autumn spawning period, was calculated across 20 m x 20 m cells within reef polygons and was used to select sampling sites. Egg deposition was observed at 17 sites on HSR and eight sites on SCS . However, eight cells on HSR changed status between 2013 and 2014. ................................ ................................ ................................ ................................ ...... 22 Figure 2.2: Physical characteristics of sites that were surveyed to be positive for egg deposition vs. those that were negative at Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ), Lake Huron. Dark shading represents habitats in which lake trout spawned, light shading represents habitats in which no eggs were found. Numbers beside bars in top panel indicate number of sites characterized in a given predominant substrate size class. O n HSR , slope magnitude was the only measured parameter that differed between habitats positive for egg deposition and those that were negative (Welch two - sample t - test; df = 26.1, t = 4.13, p < 0.01). On SCS , no difference between physical characteristics of habitats selected for egg deposition and those that were not was detected. ................................ ................................ .. 35 Figure 2.3: Relative in situ incubation success (proportion survival to hatch) of sites positive (+) and negative ( - ) for egg deposition on Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ) vs. control incubators ( CON ) in 2013 (light bars) and 2014 (dark bars). No evidence was observed for a difference in me an incubation success of habitats positive for egg deposition vs. those that were negative (+ = 73.8%, - = 69.8%; two - way repeated measures ANOVA, f = 0.06, p = 0.81), but mean incubation success differed between reefs ( HSR : 78.2%, SCS : 64.4%; two - way repe ated measures ANOVA, f = 5.65, p = 0.03). Furthermore, mean in situ incubation success of positive and negative sites at SCS were both significantly lower than control incubators (letters, one - way repeated measures ANOVA, f = 4.71, p < 0.01; Tukey HSD, SCS - vs. CON , t - ratio = 3.80, p < 0.01, SCS+ vs. CON , t - ratio = 3.36, p = 0.02). .... 41 Figure 3.1 : Study site located within Drummond Island Refuge ( DIR ; grey polygon; inset) with primary spawning reefs: Horsesehoe Reef ( HSR ), Scammon Shoal ( SCS ), and Boulder Alley ( BLD ). Grey shading represents the relative activity by lake trout as estimated by acoustic telemetry standardized to the reef - specific maximum. White squares indicate sites with the highest activity on HSR and SCS , as estimated by aco ustic telemetry detections. ................ 66 Figure 3.2: Boulder sites selected for sampling at Boulder Alley ( BLD ) on the south shore of Drummond Island, Lake Huron. Potential boulders were identified within the 0.63 km 2 study area ( black polygon ) based on comparison of the depth of 1m cells to the 75 th percentile depth of the surrounding 100 m 2 . Cells in which the mean depth was least 0.5 m greater (depths were negative) than the 75 th percentile depth were identified as potential xi boulders ( black pixels ). Boulder sites ( blue shaded circles ; enlarged for clarity) were selected for sampling to include sites containing varying number of boulder s and differing levels of activity by lake trout. Red shading represents the relative activity by lake trout as estimated by acoustic telemetry such that the darkest areas are those with the highest activity. Blue lines with black digits represent 5m depth contours (10 m and 15 m pictured). ................................ ................................ ................................ ................................ ............... 69 Figure 3.3: Schematic drawing and pictures of custom rectangular traps for the collection of early life stages of lake trout from beneath large boulders. Rectangular traps were deployed to estimate the abundance of early life stages of lake trout beneath boulders at Boulder Alley ( bottom - left ) and adjacent to round traps at Horseshoe Reef and S cammon Shoal for trap comparisons ( bottom - right ). Rectangular traps were constructed from the same 3 mm galvanized mesh and had the same coverage area (~ 0.30 m 2 ) as round traps but have a long shallow slope for accessing boulder - associated habitats ( botto m - left ). Collection cups were fastened to trap flange using elastic cords. ................................ ................................ ............ 75 Figure 3.4: Representation of three types of overhanging edges observed in boulder - associated spawning habitats: a) boulders, b) adjacent boulders, and c) boulders. Sample photographs of each type of overhang are ordered by row, increasing in zoom from left to right (i.e., pictu re farthest to the right depicts habitat located beneath an overhanging edge of the boulder pictured in a given row). Parallel reference lasers ( green dots ) were used to indicate 10 cm in the image regardless of focal length. All habitats pictured were pos itive for egg deposition by lake trout. ................................ ....................... 82 Figure 3.5: Relative abundance (CPUE: # of fish · run - 1 ) of lake trout free embryos and alevins caught via SeaRay ROV modified for electroshocker sampling in boulder - associated habitats at Boulder Alley ( BLD ) and typical habitats with the most activity at Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ). CPUE was calculated as the number of fish was triggered). Boulder sites are represented by single letters ( A - M ) and include fish collected around the base ( dark shading ) and top ( medium shading ) of boulders , while single sampling events were conducted in 20 m x 20 m cells on each HSR and SCS ( light shading ). ................................ ................................ ................................ ................................ 87 Figure 4.1: C onceptual diagram of primary physical characteristics, slope magnitude (%), interstitial depth (cm), and substrate diameter (cm), of habitats selected for egg deposition on Horseshoe Reef ( red dots ), Scammon Shoal ( blue dots ) and Boulder Alley ( green dots ) vs. the - yellow box ). This model states habitats suitable for spawning have substrates between 8 - 100 cm in diameter, with interstitial depth greater than 30cm, and slopes between 8 - 100% (5 - 45°; yell ow box ). Nearly all habitats selected for egg deposition at Drummond Island were outside of the intersection of characteristics described in the conceptual model. In general, Drummond Island habitats had smaller diameter substrates with less interstitial d epth than that described in the model. ................................ ................................ ................................ ........................ 111 xii Figure B.1: Weekly relative abundance (CPUE: fish · trap - 1 · d - 1 ) of lake trout free embryos and alevins caught in three habitat classes using custom rectangular traps at Boulder Alley ( BLD; Left ) vs. traditional round and custom rectangular traps on typical habitats on Horseshoe Reef ( HSR; Right ) and Scammon Shoal ( SCS; Right ) from May 28 June 24, 2015. Habitat classes for BLD included: habitats beneath overhanging edges of boulders ( beneath boulder), habitats adjacent to a boulder ( beside boulder), and nearby habitats at least 2x the diameter of a boulder away ( away from boulder). ................................ ........... 125 xiii KEY TO ABBREVIATIONS The following table describes abbreviations commonly referenced throughout this thesis and provides t he page on which each abbreviation is first defined. Abbreviation Meaning Page AIC c , corrected for small sample size 29 ANOVA Analysis of variance 32 BLD Boulder A lley 8 Model - aver aged parameter estimate 29 Standardized model - averaged parameter estimate 29 Standardized evidence ratio 29 CI C onfidence interval 29 CON Control 32 CPUE Catch per unit effort 73 CV Coefficient of variation 84 DI Detection interval 23 DIR Drummond Island R efuge 6 F Unique fish 23 HSD honest significant difference 32 HSR Horseshoe R eef 7 HUI Habitat use index 23 IS Interstitial depth 25 Mean difference 46 ROV Remotely operated vehicle 72 SCS Scammon S hoal 7 SCUBA Self - contained underwater breathing apparatus 7 w Akaike weight 29 1 Chapter 1 : I ntroduction Until the mid - 20 th century, lake trout, Salvelinus namaycush, was the dominant native keystone predator within the Laurentian Great Lakes, occurring in great abundance in Lake Superior, Lake Huron, Lake Michigan, Lake Ontario, and the eastern basin of Lake Erie. Early explorers of these vast inland seas, including J acques Marquette, Louis Jolliet, and Louis Agassiz recounted the immense size and productivity of lake trout and commented on the dependence of aboriginal peoples on them for sustenance and trade (Agassiz and Cabot, 1850; Cleland, 1982). While lake trout w ere relatively inaccessible for much of the year due to their preference for great depths, they were available in almost limitless quantities throughout the spawning season ranging from August to December, supporting a crucial late season subsistence fishe ry (Zedeño eloped a productive commercial fishery for lake trout with sustained combined annual harvests well over 6 million kg in lakes Michigan, Huron and Superior from 1900 to 1940 (Baldwin and Saalfeld, 1962; Berst and Spangler, 1973). Coastal communities and their economies flourished, driven largely by the seemingly limitless lake trout abundance and successive generations of immigrant - origin commercial fisherman (Bogue, 2001). With European colonization came drastic environmental changes at scales never before experienced in the Great Lakes basin. Timber harvest and agriculture that supported rapid population growth drastically altered Great Lakes watersheds (Berst and Spangler, 1973; Bogue, to allow shipping from the rich Vermillion, Gogebic, and Menominee ranges in Minnesota, Wisconsin, and Michigan to eastern steel mills in Ohio, Pennsylvania, an d New York. This increased navigational connectivity opened a migratory corridor around Niagara Falls to the 2 upper Great Lakes for invasion by non - native species, such as sea lamprey, Petromyzon marinus, and alewife, Alosa psuedoharengus, causing sweeping ecological changes to the trophic structure of the Great Lakes food web. In particular, the predatory and parasitic activity of sea lamprey , following their invasion in the 1920s, drastically altered growth and condition of lake trout , decimating native st ocks (e.g., Smith and Tibbles, 1980 ; Cobl e et al. , 1990; Eshenroder, 1992 ). At the same time, increased manufacturing capacity resulting from rapid industrialization altered water quality and caused widespread near - shore habitat degradation that potentiall y hindered lake trout reproductive capacity (e.g., Zint et al., 1995; Tillitt et al., 2008; Marsden et al., 2016). In combination with expanding commercial fisheries, these anthropogenic stressors led to a catastrophic crash of lake trout populations in th e Great Lakes throughout the 1940 s. By 1950, total lake trout yield had declined from 2.7 million to 0.18 million kg in Lake Huron, from 3.1 million to 0.16 million kg in Lake Michigan, and from 2.1 million to 0.23 million kg in Lake Superior (Hansen, 1999 ). Today, rehabilitation of lake trout in the Laurentian Great Lakes represents one of North - scaled native species recovery endeavors. The task of reestablishing naturally reproducing populations of the keystone piscivore from near extin largest freshwater ecosystems is unparalleled in ecological and socio - poli tical complexity (Krueger and Ebener , 2004; Muir et al., 2012b). As a comparison of spatial scale among ecological restoration sites in the U.S., Lake Hur on (59,600 km 2 ) alone is over six times the area of Yellowstone National Park (8,982 km 2 ), and nearly 30 times the area of the Florida Everglades (1,901 km 2 ), where the recovery of native predators, the grey wolf, Canis lupis , and Florida panther, Puma con color coryi , are national priorities (e.g., Johnson et al., 2010; Ripple and Beschta, 2012). Moreover, the watershed of the Laurentian Great Lakes includes the borders 3 of two Canadian provinces (Ontario and Quebec), eight US states (New York, Pennsylvania, Ohio, Michigan, Indiana, Illinois, Wisconsin, and Minnesota), and three tribal authorities (Chippewa - Ottawa Resource Authority, 1854 Treaty Authority, Great Lakes Indian Fish and Wildlife Authority) . When c ombined with federal authorities of Canada and U. S. , lake trout management is greatly complicated due to multiple jurisdictions and the possibility of conflicting interests. Remarkably, i nter - jurisdictional management ha s been effectively conducted voluntarily by subnational governments, as opposed to using legislated federal jurisdictional authority (e.g., Gaden et al., 2008, 2009 ). To f acilitate coordination among various entities, the binational Great Lakes Fishery Commissio n (GLFC) was formed by conventi on in 1955, with emphasis on guidance of research to support fishery rehabilitation and coordination of sea lamprey control efforts (Gaden et al., 2009). Initial management efforts aimed at restoring self - sustaining populatio ns of lake trout in the Great Lakes were focused on reducing total mortality through fishery regulation, control of the invasive sea lamprey, and the re - introduction of lake trout through supplemental stocking (Hansen, 1999; Muir et al., 2012b). Despite a verage annual stockings of over 4 million lake trout in each of the upper lakes from 1970 to 2000, progress towards the establishment of self - sustaining, naturally - reproducing populations of lake trout in the Great Lakes has been s low (e.g., Krueger and Eb ener, 2004 ; Muir et al., 2012b ). Numerous impediments to successful rehabilitation have been hypothesized, including continued excessive predation from sea lamprey, low genetic diversity, habitat degradation, low survival of early life stages, and poor spa wning habitat selection caused by maladaptive hatchery propagation and stocking prac tices (e.g., Bronte et al., 2003, 2008 ). Furthermore, continued in vasions of non - native species (e.g., Dreissenid mussels, Dreissenia spp.; round goby, Neogobius melanostom as ; Spiny water flea, Bythotrephes spp.) have resulted 4 in profound ecological changes that greatly complicated rehabilitation efforts from the 1960s to present (Muir et al., 2012b). Today, after nearly 60 years of stocking and sea lamprey control, evidence of recovery (i.e., recruitment of wild origin adults) has been limited to Lake Superior, portions of Lake Huron and, most recently, a few isolated regions in Lake Michigan (Hansen, 1999; Riley et al., 2007; Hanson et al., 2013 ). Only Lake Superior has bee n declared restored. While these successes have been encouraging, basin - wide restoration has remained elusive as many areas continue to rely heavily on stocking of hatchery - reared lake trout, typically yearling age fish. Because survival of stocked yearli ngs appears to be sufficie nt to establish spawning stocks, mortality affecting natural recruitment most likely occur s between spawning (deposition of eggs) and the yea rling life stage ( Selgeby et al., 1995 ; Krueger and Ebener, 2004; Muir et al., 2012b ). As a result, the reproductive ecology of lake trout has long been identified as a critical research priority for their recovery in the Great Lakes (Eshenroder et al., 1984, 1999; Bronte et al., 2003). Hypothesized recruitment bottlenecks include low egg viab ility due to contaminant loading ( e.g., Zint et al., 1995) and thiamine deficiency ( e.g., Fitzsimons et al., 2010; Riley et al., 2011), excessive early mortality due to early life - stage predation, particularly by non - native species ( e.g., Krueger et al., 1 995; Claramunt et al., 2005; Jonas et al ., 2005), asynchrony between free embryo emergence and feeding and food availability ( e.g., Casselman, 1995; Edsall and Cleland, 2000; Houde, 2008), and egg and free embryo mortality due to low quality and quantity o f suitable incubation habitats ( e.g., Marsden et al., 1995; Muir et al., 2012a). In the absence of appropriate cues (e.g., geographic imprinting, olfactory homing, or pheromones), hatchery origin lake trout may be unable to successfully locate spawning ree fs, find spawning aggregations, or select appropriate locations for spawning where eggs can successfully incubate 5 and hatch. Unfortunately, understanding of the fundamental characteristics of spawning habitat and the criteria by which lake trout evaluate s ites for spawning is limited. Previous research aimed at understanding critical features of lake trout spawning habitat has generated an evolving conceptual framework describing spawning habitat suitability (e.g., Marsden et al., 1995, 2016; Muir et al., 2012a ) , recently - ( Simard 2017 ). In this model, lake trout spawn on multi - layered, rounded to sub - angular rubble - cobble (8 cm 3 m diameter) substrates with clean interstitial spaces (typically > 30 cm) that are associated with sharp bathymetric features, such as a steeply sloped lake bottom (reported slope = 8 - 100%; e.g., Marsden and Krueger, 1991; Casselman, 1995; Fitzsimons, 1996). Interactions among rubble - cobble substrates with deep interstitial spaces with lake curr ents and bath ymetric relief (i.e., slope) have been thought to provide ideal incubation conditions that entrain eggs, provide refuge from predation, and buffer physical forces associated with wave action and ice scour, while allowing adequate flow to preve nt siltation and provide oxygen for incubation (e.g., Marsden and Chotkowski, 2001; Callaghan et al., 2016; Riley et al., 2017). This model has received support from studies conducted in several North American systems (e.g., Fitzsimons, 1995; Marsden et al ., 1995; Muir et al., 2012a), but not all reported lake trout spawning sites conform to this model (e.g., Beauchamp et al., 1992; Callaghan et al., 2016). - that examin e habitat selection at fine spatial scale s occurred through direct observation of spawnin g behaviors or egg deposition. Even when direct observation has occurred, most sampling for eggs, free embryos, and alevins has been targeted at spawning habitats that match those predicted to be used based on the existing model rather than at sites elsewhere (e.g., Gunn, 1995; Marsden et al., 1995). As a result, the model fails to explain 6 why some habitats are selected for egg deposition while other adjacent habitats with seemingly suitable characteristics are not. Consequently, little is known about how habitats selected for egg deposi tion differ from those that were e ncountered but not selected for egg deposition, nor is it known to what degree suitable habitats may not be used simply because they are not encountered by lake trout (Binder et al., 2018). Further, egg deposition by lake trout in spawning sites that do not fit model predictions is unexplained. Decades of stocking of hatchery reared lake trout and sea lamprey control in Lake Huron have established spawning stocks of sufficient size to sustain natural reproduction ( Selgeby et al., 1995 ; Ebener, 1998; He et al., 2012 ) and recent declines in alewife throughout Lake Huron have lessened predatory pressure and reduced the incidence of thiamine deficiency (Krueger et al., 1995; Madenjian et al., 2008b; Fitzsimons et al., 2010; Ril ey et al., 2011). As a result, survival and recruitment of early life stages of wild - spawned lake trout has greatly improved in Lake Huron, such that by 2013, greater than 50% of main basin lake trout age - 7 or younger were naturally produced fish ( Riley et al., 2007 ; He et al., 2012; Johnson et al., 2015 ). Successful natural reproduction by lake trout has recently been reported in Lake Huron (He et al., 2012; Johnson et al., 2015), particularly near the Drummond Island Refuge (DIR; Ebener 1998; Riley et al. , 2007 ; Madenjian et al. , 2008a ). Created in 1985, the DIR covers 55,000 ha following the southern shore of Drummond Island eastward to the US - Canada border at False Detour C hannel of lake trout has been restricted within the DIR in support of efforts to reestablish self - sustaining populations in the Great Lakes. As such, this area provides a unique opportunity to better understand the spawning behavior and habitat use of a recoveri ng population of successfully reproducing lake trout. 7 Recent acoustic - telemetry based research conducted within a ~25 km 2 portion of the DIR monitored the spawning behavior of ~390 tagged wild (likely primarily F1 generation) and hatchery - reared lake trou t over five consecutive spawn ing periods (Binder et al., 2016a, 2016b, 2018 ; Grimm et al., 2016). From 2010 - 2014, at least five discrete spawning sites in which lake trout consistently congregated were identified, including several that did not conform to the commonly - accepted conceptual model ( e.g., Binder et al., 2018). Subsequent SCUBA - based surveys revealed egg deposition on each of the previously identified reefs in all years of the study, but no detailed analysis was done on the physical characteristi cs of habitats selected for spawning nor how they differ from habitats at sites that were encountered by lake trout, but not selected for egg deposition. This thesis builds upon the acoustic telemetry - based work conducted at the DIR from 2010 - 2014 to explo re fine - scale spawning habitat selection on three of the six previously identified spawning sites. In chapter one, fine - scale habitat selection was investigated on the two most popular spawning reefs within in the telemetry receiver array, Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ), to determine whether habitats selected for egg deposition differed in either physical characteristics or incubation success (egg survival) from habitats on the same reefs that were known to be encountered by spawning lake tro ut, but not selected for egg deposition. Specifically, the objectives of chapter one were to compare : 1) the physical characteristics of sites selected for egg deposition vs. those encountered but not selected, and (2) the relative survival of fertilized e ggs incubated over winter at these same sites. I hypothesized that: 1) lake trout select spawning habitats based on specific measurable physical substrate characteristics (e.g., substrate diameter, interstitial depth, substrate homogeneity, and slope), and 2) lake trout select spawning habitats that maximize egg incubation success relative to other nearby available 8 habitats. Improving understanding of the relative quality of spawning habitats being used by lake trout and the criteria by which they are selec ted is a fundamental component of identifying and addressing possible limitations to recruitment. In chapter two, spawning habitat selection by lake trout and physical description of sites chosen was studied within - spawning habitats associated with giant (2 - 7 m diameter) boulders found at a third site, Boulder Alley ( BLD ). Congregations of lake trout were consistently observed between 2010 and 2014 within this 0.63 km 2 area located to the west of HSR , but the a rea appeared to lack any classic reef structure in terms of depth contours or cobble substrate in comparison to other spawning sites (Binder et al., 2018). Preliminary observations revealed the area was comprised of numerous broadly - spaced boulders greater than 1 m in diameter, with some as great as 3 - 4 m in diameter. Further surveys prior to my research confirmed that egg deposition occurred in clean gravel - rubble substrate under a handful of these boulders, and a single larva was caught adjacent to one of the boulders. However, t he fu ll extent to which boulder - associated spawning occurs within the array was unknown. Multi - beam bathymetry of BLD indicated that likely hundreds of these large boulders occurred in this area, so the combined contribution of this habitat type to lake trout r ecruitment in northern Lake Huron could be substantial. To better understand the potential for boulder - associated spawning to contribute to lake trout recruitment in northern Lake Huron, the specific objectives of chapter two were to: 1) describe the spati al extent of lake trout spawning in association with boulders at Drummond Island, 2) determine whether use of boulder - associated substrates for egg deposition by lake trout could be predicted by measurable physical characteristics of boulders or their adja cent substrates, and 3) compare the abundance of free embryos and alevins located in boulder - associated spawning habitats to 9 those in other highly - used nearby spawning habitats within the acoustic telemetry array. An accurate accounting of the total amount of suitable spawning habitat available to recovering populations improves understanding of the variability in recruitment and thus, limitations imposed by th e carrying capacity of a system . 10 REFERENCES 11 REFERENCES Agassiz, L., and Cabot, J.E., 1850. Lake Superior: its physical character, vegetation, and animals, compared with those of other and similar regions. Gould, Kendall and Lincoln. Boston, MA. Baldwin, N.S., and Saalfeld, R.W., 1962. Commercial fish production in the Great Lak es, 1867 - 19 60. Great Lakes Fishery Commission. Ann Arbor, MI. Beauchamp, D.A., Allen, B.C., Richards, R.C., Wurtsbaugh, W.A., and Goldman, C.R., 1992. Lake trout spawning in Lake Tahoe: egg incubation in dee pwater macrophyte beds. N. Am. J. Fish. Manag. 12, 442 449. Berst, A., and effects on it. Technical report, Great Lakes Fishery Commission. Ann Arbor, MI. 1973 - 21. Binder, T.R., Holbrook, C.M., Hayden, T.A., and Krueger, C.C., 2016a. Spatial and temporal variation in positioning probability of acoustic telemetry arrays: fine - scale variability and complex interactions. Anim. Biotelemetry 4 (1). doi: 10.1186/s40317 - 016 - 0097 - 4 Binder, T.R., Riley, S.C., Holbrook, C.M. , Hansen, M.J., Bergstedt, R.A., Bronte, C.R., He, J., and Krueger, C.C., 2016b. Spawning site fidelity of wild and hatchery lake trout ( Salvelinus namaycush ) in northern Lake Huron. Can. J. Fish. Aquat. Sci. 73, 18 34. doi: 10.1139/cjfas - 2015 - 0175 Binder, T.R., Farha, S.A., Thompson, H.T., Holbrook, C.M., Bergstedt, R.A., Riley, S.C., Bronte, C.R., He, J., and Krueger, C.C., 2018. Fine - scale acoustic telemetry reveals unexpected lake trout, Salvelinus namaycush , spawning habitats in northern Lake Huron, Nor th America. Ecol. Freshw. Fish 27, 594 605. doi: 10.1111/eff.12373 Bogue, M.B., 2001. Fishing the Great Lakes: an environmental history, 1783 1933. Univ ersity of Wisconsin Press. Madison, WI. Bronte, C.R., Krueger, C.C., Holey, M., Toneys, M., Eshenroder, R ., and Jonas, J., 2008. A guide for the rehabilitation of lake trout in Lake Michigan. Miscellaneous p ublication , Great Lakes Fishery Commission. Ann Arbor , MI. 2008 - 01. Bronte, C.R., Jonas, J., Holey, M.E., Eshenroder, R.L., Toneys, M.L., McKee, P., Brei dert, B., Claramunt, R.M., Ebener, M. P., Krueger, C.C., and Wright, G . , 2003. Possible impediments to lake trout restoration in Lake Mich igan. Lake trout task group report to the Lake Michigan Committee, G reat Lakes Fish ery Commission. Ann Arbor, MI. Avail able from http://www.sealampreycontrol.net/lakecom/lmc/ltrestore.pdf [accessed 30 October 2014]. 12 Callaghan, D.T., Blanchfield, P.J., and Cott, P.A., 2016. Lake trout ( Salvelinus namaycush ) spawni ng habitat in a northern lake: t he role of wind and physical characteristics on habitat quality. J. Gt. Lakes Res. 42, 299 307. doi: 10.1016/j.jglr.2015.07.001 Casse lman, J.M., 1995. Survival and development of lake trout eggs and f ry in Eastern Lake Ontario in situ i ncubation, Yorkshir e Bar, 1989 1993. J. Gt. Lakes Res. 21, Supple . 1, 384 399. doi: 10.1016/S0380 - 1330(95)71112 - 1 Claramunt, R.M., Jonas, J.L., Fitzsimons, J.D., and Marsden, J.E., 2005. Influences of spawning habitat characteristics and interstitial predators on lake trout e gg deposition and mortality. Trans. Am. Fish. Soc. 134 (4) , 1048 1057. doi: 10.1577/T04 - 075.1 Cleland, C.E., 1982. The inland shore fishery of the northern Great Lakes: i ts development and importance in prehistory. Am. Antiq. 47, 761 784. doi: 10.2307/280281 Coble, D.W., Bruesewitz, R.E., Fratt, T.W., and Scheirer, J.W., 1990. Lake trout, sea lampreys, and overfishing in the upper Great Lakes: a review and reanalysis. Trans. Am. Fish. Soc. 119, 985 995. Available from http://www.tandfonline.com/doi/abs/10.1577 /1548 - 8659(1990)119%3C0985%3ALTSLAO%3E2.3.CO%3B2 [accessed 31 October 2014]. Ebener, M.P. [ED.] , 1998. A lake trout rehabilitation guide for Lake Huron. Miscellaneous publications, Great Lakes Fishery Comm ission. Ann Arbor, MI. 1998 : 1 44. Available from h ttp://www.glfc.org/pubs/misc/LkTroutRepHuron.pdf [accessed 13 October 2013]. Edsall, T.A., and Cleland, J. , 2000. Optimum temperature for growth and preferred temperatures of age - 0 lake trout. N. Am. J. Fish. Manag. 20(3): 804 809. Available from http://ww w.tandfonline.com/doi/abs/10.1577/1548 - 8675(2000)020%3C0804%3AOTFGAP%3E2.3.CO%3B2 [accessed 30 October 2014]. Eshenroder, R.L., Poe, T.P., and Olver, C.H. , 1984. Strategies for rehabilitation of lake trout in the Great La kes: p roceedings of conference on lake trout research, August 1983. Great Lakes Fishery Commission. Ann Arbor, MI. 40: 0 - 63 . Eshenroder, R.L., 1992. Decline of lake trout in Lake Huron. Trans. Am. Fish. Soc. 121, 548 554. Available from http://www.tandfonline.com/doi/pdf/10.1577/1548 - 8659 - 121.4.548 [accessed 31 October 2014]. Eshenroder, R.L., Peck, J.W., and Olver, C.H. , 1999. Research priorities for lake trout rehabilitation in the Great Lakes: a 15 - year retrospective. Technical report, Great Lakes Fishery Commission. Ann Arbor, MI. 64: 1 40. Fitzsim ons, J.D., 1995. Assessment of lake trout spawning habitat and egg deposition and s urvival in Lake Ontario. J. Gt. Lakes Res. 21, Supple . 1, 337 347. doi: 10.1016/S0380 - 1330(95)71108 - X Fitzsimons, J.D., 1996. The significance of man - made structu res for lake trout spawning in the Great Lakes: are they a viable alternative to natural reefs? Can. J. Fish. Aquat. Sci. 53, 142 151. 13 Fitzsimons, J.D., Brown, S., Brown, L., Honeyfield, D., He, J., and Johnson, J.E., 2010. Increase in lake trout reproduct ion in Lake Huron following the collapse of alewife: relief from thiamine deficiency or larval predation? Aquat. Ecosyst. Health Manag. 13, 73 84. doi: 10.1080/14634980903581252 Gaden, M., Krueger, C.C., Goddard, C., and Barnhart, G., 2008. A joint strategi c plan for management of Great Lakes fisheries: a cooperative regime in a multi - jurisdictional setting. Aquat. Ecosyst. Health Manag. 11, 50 60. doi: 10.1080/14634980701877043 Gaden, M., Krueger, C.C., and Goddard, C., 2009. Managing across jurisdictional b oundaries: fishery governance in the Great Lakes and Arctic - Yukon - Kuskokwim regions. Am. Fish. Soc. Symp. 70, 941 960. Grimm, A.G., Brooks, C.N., Binder, T.R., Riley, S.C., Farha, S.A., Shuchman, R.A., and Krueger, C.C., 2016. Identification of lake trout Salvelinus namaycush spawning habitat in northern Lake Huron using high - resolution satellite imagery. J. Gt. Lakes Res. 42, 127 135. doi: 10.1016/j.jglr.2015.11.011 Gunn, J.M., 1995. Spawning behavior of lake trout: effects on colonization ability. J. Gt. L akes Res. 21, Supple . 1, 323 329. doi: 10.1016/S0380 - 1330(95)71106 - 6 Hansen, M.J., 1999. Lake trout in the Great Lakes: basin - wide stock coll apse and binational restoration. Pages 417 454 in Great Lakes Fisheries Policy and Management. Edited by W . W . Taylor and C . P . Ferreri. Michigan State University Press, East Lansing, MI. Hanson, S.D., Holey, M.E., Treska, T.J., Bronte, C.R., and Eggebraaten, T.H., 2013. Evidence of wild juvenile lake trout recruitment in western Lake Michigan. N. Am. J. Fish. Manag. 33, 186 191. doi: 10.1080/02755947.2012.754804 He, J.X., Ebener, M.P., Riley, S.C., Cottrill, A., Kowalski, A., Koproski, S., Mohr, L., and Johnson, J.E., 2012. Lake trout status in the main basin of Lake Huron, 1973 2010. N. Am. J. Fis h. Manag. 32, 402 412. doi: 10.1080/02755947.2012.675947 Houde, E.D hadow. J. Northwest Atl. Fish. Sci. 41, 53 70. doi: 10.2960/J.v41.m634 Johnson, J.E., He, J.X., and Fielder, D.G., 2015. Rehabilitation stocking of walleyes an d lake trout: r estoration of reproducing stocks in Mich igan waters of Lake Huron. N. Am. J. Aquac. 77, 396 408. doi: 10.1080/15222055.2014.993488 Johnson, W.E., Onorato, D.P., Roelke, M.E., Land, E.D., Cunningham, M., Belden, R.C., McBride, R., Jansen, D., Lotz, M., Shindle, D., Howard, J., Wildt, D.E., Penfold, L.M., Hostetler, J.A., Oli, M.K., and p anther. Science 329, 1641 1645. doi: 10.1126/science.1192891 Jonas, J.L., Claramunt, R.M., Fitzsimons, J. D., Marsden, J.E., and Ellrott, B.J., 2005. Estimates of egg deposition and effects of lake trout ( Salvelinus namaycush ) egg predators in three regions of the Great Lakes. Can. J. Fish. Aquat. Sci. 62, 2254 2264. doi: 10.1139/f05 - 141 14 Krueger, C.C., Perkins, D.L., Mills, E.L., and Marsden, J. E. , 1995. Predation by alewives on lake trout fry in Lake Ontario: role of an exotic species in preventing restoration of a native species. J. Gt. Lakes Res. 21, Supple . 1, 458 469. doi: 10.1016/S0380 - 1330(95)71117 - 0 Krueger, C.C., and Ebener, M., 2004. Rehabilitation of lake trout in the Great Lakes: p ast lessons and future challenges. In Boreal Shield Watersheds: Lake Trout Ecosystems in a Changing Environment. Edited by J.M. Gunn, R.J. Steedman, and R.A. Ryder. CRC Press, Boca Raton, Florida. pp. 37 56. Madenjian, C.P., Ebener, M.P., and Desorcie, T.J., 2008a. Lake trout population dynamics at Drummond Island Refuge in Lake Huron: implications for future rehabilitation. N. Am. J. Fish. Manag. 28(1), 979 992. doi: 10.1577/M07 - 083.1 Stockwell , J.D., and Stapanian, M.A. , 2008b. Adverse effects of a lewives on Laure ntian Great Lakes fish c ommunities. N. Am. J. Fish. Manag. 28(1) , 263 282. doi: 10.1577/M07 - 012.1 . Marsden, J.E., and Krueger, C.C., 19 91. Spawning by hatchery - origin lake trout ( Salvelinus namaycush ) in Lake Ontario: data from egg collections, substrate analysis, and diver observations. Can. J. Fish. Aquat. Sci. 48, 2377 2384. Marsden, J.E., Casselman, J.M., Edsall, T.A., Elliott, R.F., Fitzsimons, J.D., Horns, W.H., Manny, B.A., McAughey, S.C., Sly, P.G., and Swanson, B.L., 1995. Lake trout spawning habitat in the Great Lakes a review of current knowledge. J. Gt. Lakes Res. 21, Supple . 1, 487 497. doi: 10.1016/S0380 - 1330(95)71120 - 0 Mars den, J.E., and Chotkowski, M.A., 2001. Lake trout spawning on artificial reefs and the effect of zebra mussels: fatal attraction? J. Gt. Lakes Res. 27, 33 43. Marsden, J. E., Binder, T.R., Johnson, J., He, J., Dingledine, N., Adams, J., Johnson, N.S., Buch inger, T.J., and Krueger, C.C., 2016. Five - year evaluation of habitat remediation in Thunder Bay, Lake Huron: c omparison of constructed reef characteristics that attract spawning lake trout. Fish. Res. 183, 275 286. Muir, A. M., Blackie, C.T., Marsden, J.E ., and Krueger, C.C., 2012a. Lake charr , Salvelinus namaycush , spawning behaviour: new field observations and a review of current knowledge. Rev. Fish Biol. Fish. 22, 575 593. doi: 10.1007/s11160 - 012 - 9258 - 6 Muir, A.M., Krueger, C.C., and Hansen, M.J. , 2012b. Re - establishing lake trout in the Laurentian Great Lakes: past, present, and future. In Great Lakes fishery policy and management: A binational perspective, 2 nd edition. Edited by W.W. Taylor. Michigan State University Press, East Lansing. pp. 533 588. Available from https://www.researchgate.net/profile/Michael_Hansen18/publication/268207669_Re - establishing_lake_trout_in_the_Laurentian_Great_Lakes_past_present_and_future/links/ 5463aa9c0cf2cb7e9da997fb.pdf [accessed 9 March 2017]. 15 Riley, S.C., He, J. and Schaeffer, J.S., 2007. Evidence of widespread natural reproduction by lake trout, Salvelinus namaycush , in the Michigan waters of Lake Huron. J. Gt. Lakes Res. 33, 917 921. doi: 10.3394/0380 - 1330(2007)33[917:EOWNRB]2.0. CO;2 Riley, S.C., Rinchard, J., Honeyfield, D.C., Evans, A.N., and Begnoche, L., 2011. Increasing thiamine concentrations in lake trout eggs from Lakes Huron and Michigan coincide with l ow alewife abundance. N. Am. J. Fish. Manag. 31, 1052 1064. doi: 10.108 0/02755947.2011.641066 Riley, S.C., Binder, T.R., Tucker, T.R., Menzies, J., Eyles, N., Janssen, J., Muir, A.M., Esselman, P.C., Wattrus, N.J., and Krueger, C.C., 2017. Islands in the ice stream: were spawning habitats for native salmonids in the Great Lakes created by paleo - ice streams? Fish and Fish . 18, 347 359. doi: 10.1111/faf.12173 Ripple, W.J., and Beschta, R.L., 2012. Trophic cascades in Yello wstone: th e first 15 years after wolf reintroduction. Biol. Conserv. 145, 205 213. doi: 10.1016/j.biocon.2011.11.005 Selgeby, J.H., Bronte, C.R., Brown Jr., E.H., Hansen, M.J., Holey, M.E., VanAmberg, J.P., Muth, K.M., Makauskas, D.B., McKee, P., Anderson, D.M., Paola Ferreri, C., and Schram, S.T., 1995. Lake trout restoration in the Great Lakes: stock - size criteria for natural reproduction. J. Gt. Lakes Res. 21, Supple . 1, 498 504. doi: 10.1016/S0380 - 1330(95)71121 - 2 Simard, L.G., 2017. Spawning site selection a nd fry development of invasive lake t rout in Yellowstone Lake, Yellowstone National Park, Wyoming (PhD Thesis). The University of Vermont and State Agricultural College. Smith, B.R., and Tibbles, J.J., 1980. Sea l amprey ( Petromyzon marinus ) in L akes Huron, Mich igan, and Superior: h istory of invasion and c ontrol, 1936 78. Can. J. Fish. Aquat. Sci. 37, 1780 1801. doi: 10.1139/f80 - 222 Tillitt, D.E., Cook, P.M., Giesy, J.P., Heideman, W., and Peterson, R.E. , 2008. Reproductive impairment of Great Lake s lake trout by dioxin - like chemicals. Pages 819 - 875 in The toxicology of fishes. Edited by R.T. Di Giulio and D.E. Hinton. CRC Press, Taylor and Francis Group, Boca Raton, FL. Zedeño, M. N., Stoffle, R. W., Pittaluga, F., Dewey - Hefley, B. , Basaldú, R. C., and Porter, M. , 2001 . Traditional Ojibway Resources in the Western Great Lakes: An ethnographic inventory in the s tates of Michigan, Minnesota, and Wisconsin. Final r eport prepared for Michael J. Evans, National Park Service, Midwest Region. Bureau of Applied Research in Anthropology, The University of Arizona in Tucson. Zint, M.T., Taylor, W.W., Carl, L., Edsall, C.C., Heinrich, J., Sippel, A., Lavis, D., and Schaner, T., 1995. Do toxic substances pose a threat to rehabilitation of lake trout in the Gr eat Lakes? A review of the literature. J. Gt. Lakes Res. 21, Supple . 1, 530 546. doi: 10.1016/S0380 - 1330(95)71124 - 8 16 Chapter 2 : Evaluation of lake trout habitat selection at Drummond Island spawning reefs Abstract In the Laurentian Great Lakes, slow recovery of lake trout, Salvelinus namaycush , suggests potential recruitment bottlenecks exist between spawning and the yearling life stage. Determining the critical features and quality of sites selected for embryo incubation may help understand limitatio ns to lar val recruitment. I used a novel acoustic - telemetry based sampling approach to compare physical characteristics and egg survival between sites receiving egg deposition and those that did not on two reefs in northern Lake Huron. Egg deposition was c onfirmed at 25 of 58 sites (20m x 20m) , but the physical characteristics of sites receiving egg deposition varied between reefs. Slope magnitude was associated with habitat selection on one reef but not the other. No evidence existed to support a relations hip between other substrate characteristics (i.e., substrate size, interstitial depth, and substrate homogeneity) and egg deposition on either reef. Based on an in situ habitat bioassay , n o difference in survival to hatch was observed between sites selecte d for egg depo sition and nearby sites that were not. However, s o me sites with adequate survival to hatch were not used , suggesting supply of suitable spawning habitat may exceed demand in this area. 17 Introduction Lake trout, Salvelinus namaycush , were once the dominant native predator in the Laurentian Great Lakes, supporting a large and valuable fishery. These fish were diverse in diet and habitat use, and existed in several morphotypes, representing alternative life history strategies (e.g., Hansen et al. , 2016) . Populat ions collapsed in the late 1940 s due to a combination of commercial exploitation and predation from invasive sea lamprey, Petromyzon marinus (e.g., Hile , 1949; Coble et al. , 1990; Eshe nroder , 1992) . Since their decline, restoration of self - sustaining populations of lake trout in the Great Lakes has been the focus of one of the largest native species recovery endeavors in North America (Hansen , 1999; Krueger and Ebener, 2004; Muir et al. , 2012b) . However, after nearly 60 years of stocking and sea lamprey control, evidence of recovery (i.e., recruitment of wild origin adults) has been limited to Lake Superior, portions of Lake Huron and, most recently, a few isolated regions in Lake Michig an (Hansen , 1999; Krueger and Ebener, 2004; Muir et al. , 2012b; Hanson et al. , 2013) . While these successes are encouraging, basin - wide restoration has remained elusive as many areas continue to rely heavily on stocking of hatchery - reared lake trout, typic ally yearling age fish. Numerous possible impediments to lake trout recovery have been identified since rehabilitation efforts began, though none solely explain the slow recovery experienced over the last 60 years (Krueger and Ebener , 2004) . Lack of succes s has occurred even though survival to maturity by hatchery - reared lake trout has yielded sufficient spawning stock abundances to sustain natural reproduction and year - class recruitment in some areas (Selgeby et al. , 1995; Bronte et al. , 2007; He et al. , 2012) . Because substantial survival occurs from when yearlings are stocked to adult life stages, limitations to natural recruitment most likely occur between 18 spawning and the yearling life stage (Bronte et al. , 2003; Krue ger and Ebener, 2004; Muir et al. , 2012b) . As a result, the reproductive ecology of lake trout has long been identified as a critical research priority (Eshenroder et al. , 1984, 1999; Bronte et al. , 2003) . Hypothesized recruitment bottlenecks include low egg viability due to contaminant lo ading ( e.g., Zint et al. , 1995) or thiamine deficiency ( e.g., Madenjian et al. , 2008b; Fitzsimons et al. , 2010; Riley et al. , 2011) , excess ive mortality due to early life stage predation, particularly by non - native species ( e.g., Krueger et al. , 1995; Clar amunt et al. , 2005; Jonas et al. , 2005) , asynchrony between commencement of exogenous feeding and food availability ( e.g., Casselman , 1995; Edsall and Cleland , 2000; Houde , 2008) , and excessive mortality of eggs, free em b r yos and alevins due to low quality or quantity of suitable incubation habitats ( e.g., Marsden et al. , 1995; Muir et al. , 2012a) . The ability of hatchery origin adults to select spawning habitats where early life stages can develop successfully could also affect egg survival, including pred ation vulnerability and food asynchrony. Optimum lake trout spawning habitat is thought to have two critical components; suitable habitat must 1) facilitate aggregation and mate selection by spawning fish, and 2) provide adequate protection and interstiti al water quality for successful incubation of eggs from the autumn spawn ing period through hatch and alevin dispersal the following spring (e.g., Marsden et al. , 1995) . The interaction of broad - scale physical characteristics such as water depth, distance to shore, lake bathymetry, and reef orientation with fine - scale habitat characteristics such as substrate size, shape, and abundance must provide adequate flow conditio ns to oxygenate eggs, protect them from physical forces such as waves and ice, and reduce their vulnerability to predation (Dorr III et al. , 1981; Sly , 1988; Fitzsimons , 1994) . In the absence of appropriate cues (e.g., geographic imprinting, olfactory homi ng, or pheromones), hatchery origin lake trout may 19 be unable to successfully locate spawning reefs, find aggregations of adults, or select appropriate locations for spawning where eggs can successfully incubate and hatch. Decades of research focused on lak e trout spawning habitat has yielded a conceptual framework for spawning habitat suitability (e.g., Marsden et al. , 1995 , 2016 ; Muir e t al., 2012a) , - ( Simard , 2017 ) . In this model, lake trout spawn on multi - layered , rounded to sub - angular rubble - cobble (8 cm 3 m diameter) substrates with clean interstitial spaces (typically > 30 cm in depth ) that are associated with sharply changing bathymetric features often linked with a steeply sloped lake bottom ( reported slop e = 8 - 100%; e.g., Marsden and Krueger , 1991; Casselman , 1995; Fitzsimons , 1996) . The model has received support from studies conducted in several North American systems (e.g., Fitzsimons , 1995; Marsden et al. , 1995; Muir et al. , 2012a) , but not all repor ted lake trout spawning habitat conforms to this model , including spawning on aquatic vegetation and in rivers (e.g., Beauchamp et al. , 1992; Callaghan et al. , 2016 , Jones et al. , in press ) . Furthermore, the model fails to explain why some habitats are sel ected by lake trout for egg deposition while other seemingly similar habitats are not. Verification of spawning through direct observation of spawning behaviors or egg deposition ha s been rare and most sampling for early life stages has been limited to spa wning habitats that fit this model (e.g., Gunn , 1995; Marsden et al. , 1995) , creating a potential bias in support of the model . As a result, little is known about how the characteristics of habitats selected for egg deposition differ from those that are not selected for egg deposition, nor is it known to what degree suitable habitats may not be used because they are not encountered by lake trout. The goal of this study was to determine if habitats selected for egg deposition differed from those that were en countered by lake trout but not selected for egg deposition. The specific 20 objectives were to compare : 1) the physical characteristics of sites selected for egg deposition vs. those known to be encountered by lake trout but were not selected, and (2) the relative survival of fertilized eggs incubated over winter deployed at these same sites. I hypothesized that: 1) lake trout select spawning habitats based on specific measurable physical habitat characteristics (substrate diameter, interstitial depth, subs trate homogeneity, and slope), and 2) lake trout select spawning habitats that maximize egg incubation success relative to other available habitats. For hypothesis one, physical habitat measurements were compared between sites that received egg deposition and those that were encountered by lake trout but did not. Logistic regression, m odel selection and multi model inference were then used to estimate the relative importance of physical characteristics to predicting egg deposition. For hypothesis two, a habi tat bioassay that included seeding incubators with hatchery fertilized eggs to compare the relative in situ survival of fertilized eggs overwinter to hatching between habitats that received egg deposition and those that were encountered by lake trout but d id not. Methods Study Area The Drummond Island Refuge (DIR) is located in northern Lake Huron, covering 55,000 ha along the south shore of Drummond Island eastward to the USA - Canada border at False itude ( Figure 2 . 1 ) . Created in 1985, the DIR restricts commercial and sport fishing of lake trout in support of efforts to reestablish and rehabilitate self - sustaining populations in northern Lake Huron . Successful natural reproduction by lake trout in Lake Huron has recently been reported (Johnson and VanAmberg , 1995; He et 21 al. , 2012; Johnson et al. , 2015) , particularly near the DIR (Ebener , 1998; Riley et al. , 2007; Madenjian et al. , 2008a) . A large - scale study of lake trout movements associated with spawning was c onducted within the DIR from 2010 to 2014, using fine - scale, 2D positional acoustic telemetry within a ~25km 2 region along the southern shore of Drummond Island (Riley et al. 2014; Binder et al. 2016a, 2016b, 2018) . This multi - agency collaborative effort u sed a VEMCO (Halifax, NS, Canada) positioning system (VPS) array consisting of between 108 and 152 autonomous acoustic telemetry receivers (VR2W receivers) to monitor the fine - scale movements (i.e., 5 - 10 m accuracy) of 390 acoustically - tagged (V16 transm itters) adult lake trout during late summer and autumn (see Binder et al. 2016a, 2016b, 2018 for complete tagging and acoustic array details). Throughout the course of the Drummond Island telemetry study, lake trout were concentrated consistently in six ar eas (Binder et al. 2018). Spawning at these areas was confirmed by diver - based egg surveys; however, spawning activity was greatest at two locations, Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ), located adjacent to one another approximately 1.0 to 1.85 km away from the southern shore of Drummond Island in the northwest corner of the array ( Figure 2 . 1 ). 22 Figure 2 . 1 : Lake Huron study site ( grey box in map inset) adjacent to Drummond Island with primary spawning reefs, Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ). Habitat Use Index ( HUI ), an index of relative amount of time tagged lake trout spent in various habitats over th e 2012 autumn spawning period, was calculated across 20 m x 20 m cells within reef polygons and was used to select sampling sites. Egg deposition was observed at 17 sites on HSR and eight sites on SCS . However, eight cells on HSR changed status between 201 3 and 2014. 23 Habitat Use Index The Drummond Island acoustic telemetry study provided a unique opportunity to advance understanding of lake trout spawning habitat characteristics by evaluating habitat selection using behavior of the fish. Fine - scale acoustic telemetry position data were used to concentrate sampling effort at HSR and SCS on habitats known to have been encountered by lake trout and, ultimately, to compare the physical characteristics ( Objective 1 ) and relative survival of fertiliz ed eggs to hatch ( Objective 2 ) between specific locations that received egg deposition and those that did not. To accomplish this comparison, an index of the amount of time lake trout spent in various habitats on each reef ( Habitat Use Index: HUI ) was deve loped based on telemetry data collected in 2012 and was subsequently used to select sampling sites . Polygons surrounding the boundaries of each reef were demarcated along contour lines based on high - resolution bathymetric survey data (1 m 2 resolution) coll ected via multi - beam sonar in 2010 and 2011 ( See Riley et al. , 2014 for complete bathymetric survey details ) . Reef polygons then were divided into a grid of 20 m x 20 m cells and a HUI value was calculated for each cell ( HSR = 1449 cells; SCS = 2038 cells). HUI was calculated as the sum of the number of 1 - h intervals in which a fish was detected at least once within cell q on reef r, termed detection intervals ( DI q,r ), divided by the total number of unique fish ( F ) that were detected on reef r over the spawning period (Eq. 1 ). In 2012, 144 unique fish were detected within the telemetry array and the spawning period ranged from October 6 to November 18 (Binder et al. , 2016b) . Of these, 104 fish were detected on HSR and 93 were detected on SCS . Detection intervals were used in place of raw number of positions as an estimate of habitat use to limit potential bias associated with positioning pro babilities that 24 varied both spatially (e.g., across reefs) and temporally ( e.g., between years; See Binder et al. , 2016a) . Equation 1 ) HUI q = The fundamental sampling units in this study were the 20 m x 20 m cells selected for for sampling , cells on each reef were ranked in descending order by their HUI and the 40 highest - ranking non - adjacent cells were identified. From these , the top 20 ranked non - adjacent cells and 10 random non - adjacent cells from rank 21 - 40 were selected (totaling 30 cells) as sampling sites for each reef. No two adjacent cells were selected to ensure that sampling sites were at least 40m apart and thus, r educe the potential for spatial autocorrelation. Surveys for Egg Deposition SCUBA diver - based egg surveys were conducted within a portion of all 30 sites on each reef during 2013 and 2014 to assign sites as having eggs present ( positive = 1 ) or absent ( negative = 0 ). Egg surveys were conducted within a 10 m diameter circle at the center of each site so that they corresponded with substrate photographs (see Physical Characterization below). Two divers searched independently for a minimum of 10 minutes or until both divers independently confirmed egg deposition by flipping rocks and flushing water into the substrate by hand to suspend eggs. Surveys commenced after peak spawning was estimated to have occurred. All sites identified as negative during the firs t survey were surveyed a second time later in the season to confirm egg absence. In both years, egg surveys began by Nov. 1, and all secondary surveys were complete by Nov. 18. 25 For statistical analyses, egg survey results from 2013 and 2014 were pooled suc h that sites in which eggs were observed in either year were considered positive for egg deposition (1), while sites in which eggs were not observed in both years were considered negative for egg deposition (0). Additionally, two sites on HSR were excluded from the analyses due to logistical issues in the field that prevented complete sampling, resulting in a final count of 28 sites on HSR and 30 sites on SCS . Physical Characterization (Objective 1) To determine the relative importance of physical charac teristics on selection of spawning habitats (Objective 1), bathymetric features (e.g., water depth and slope) and measurements of the bottom substrate (e.g., diameter, homogeneity, and interstitial depth) were characterized at each site. Estimates of slope and water depth at each site were calculated using the multi - beam bathymetry by dividing each site into a grid of 400 - 1 m 2 sub - cells and then calculating the mean of 20 deepest sub - cells and 20 shallowest sub - cells as an estimate of the maximum and mini mum water depth within each site. The predominant slope (%) for each site was calculated as the difference between the maximum and minimum mean water depth divided by the distance between the single shallowest and deepest sub - cell within the site, multiplied by 100. Substrat e at each site was characterized using a combination of underwater photographs and direct measurement by divers in 2013. A colored brick was placed at the center of each site and coordinates were confirmed using a taut surface line and Trimble ® GeoXH diffe rential GPS unit with sub - meter accuracy. Substrate was photographed by divers using an 8.1 megapixel digital camera (Panasonic Lumix DMC - FX55) mounted on an aluminum frame set 1.5 m above the lake bottom. Five ~1 m 2 images were collected within each site, one centered at the central 26 identifier brick and four others centered ~5 m from the center in each cardinal direction. Divers made three independent measurements as a surrogate of interstitial depth ( IS ) within each of the five imaged areas by inserting a 25.4 mm wide x 3 mm thick aluminum ruler into the spaces between substrates to its maximum depth (i.e., until bedrock was reached, or substrate did not permit further penetration). Interstitial depth for a site was estimated as the mean of fifteen interst itial depth measurements. To estimate substrate size, shape, and homogeneity at each site, substrate photos were processed using a semi - automated segmentation macro developed using the FIJI package of ImageJ (Rasband , 1997; Schindelin et al. , 2012) . Images were calibrated using a 173 mm measure on a portion of the camera frame that was visible in each photo such that the total imaged area was 0.816 m 2 per photo for a total of 4.08 m 2 per site. To standardize for light differences between photos due to varyi ng water depths and times of collection , images were altered using preprocessing filters and an iterative auto - thresholding technique was applied prior to segmentation. Segmentation results were reviewed individually by a single reader. Over - segmented regi ons were deleted, and poorly segmented or misaligned particles were manually outlined using a free - hand edge tool. All substrates near the border of the image in which at least 50% of its surface area was estimated to be included in the image were segmente d and measured. Estimates of the maximum Feret diameter (i.e., the diameter along the particles longest axis; Merkus , 2009) and surface area were calculated for all visible substrates in each photo. The predominant surficial substrate for each site was cl assified using a categorical classification scheme based on the proportion of the total measured surface area within various substrate size categories using a modified Wentworth scale (e.g., Marsden et al. , 1995; Table 2 . 1 , bottom ). Measured substrate at each site was first assigned to a size category (i.e., gravel, rubble, 27 cobble) based on its maximum Feret diameter. The surface area of all substrate within each size category was then summed and expressed as a percentage of total measured area for each site. Usually, the substrate category with the highest percentage of total measured area was selected as the predominant substrate category for a given site. However, an y site with greater than 33.3% combined substrate categories (e.g., gravel - rubble mix, rubble - cobble mix, or gravel - cobble mix). Sites with > 50% of the imaged area not Collectively, this category represented spaces between individual substrates which include d fines such as sand or silt and bedrock or were indistinguishable substrates that could not be reliably segmented from the remainder of the image due to image quality. Using this classification scheme, all sites were assigned to one of seven possible predominant substrate categories ( Table 2 . 1 , b otto m ). Relative substrate homogeneity was expressed as a continuous variable representing the maximum proportion of the total imaged area that occurred in a running 15 cm substrate diameter size bin. First, measured substrates in each image were assigned to 5 cm bins from 0 cm to 100 cm based on maximum Feret diameter. Next, the surface areas of all substrates in each 5 cm bin were summed and expressed as the percent age of the total imaged area (4.08 m 2 ) within a site. Finally, the 5 cm diameter bin comprising the greatest proportion of the total imaged area was identified and summed with the proportions in the two adjacent 5 cm size bins. In this manner, the higher t he proportion of imaged area in this 15 cm range of substrate diameters, the more homogeneous the substrate. 28 Relative Importance of Physical Characteristics (Objective 1) While HSR and SCS were only located ~1 km apart from one other, the range and vari ance of measured variables differed substantially between reefs ( Figur e 2 . 2 ; Binder et al. , 2018 ). As such, model selection procedures were run on each reef independen tly and multimodel inference was used to compare the relative importance of measured physical habitat characteristics within and between reefs. Initial data exploration revealed minor negative correlation between substrate homogeneity and interstitial dept h on SCS (Pearson product - moment correlation: Homogeneity - IS; r = - 0.36, p = 0.01). Furthermore, slope and interstitial depth estimates were spatially autocorrelated on HSR I = - 0.09, p < 0.01, IS; I = - 0.14, p < 0.01 ) and egg deposition and slope estimates were spatially autocorrelated on SCS I = - 0.12, p < 0.01, slope; I = - 0.15, p < 0.01). As such, an autocorrelation distance f adjacency in habitat selection (e.g., Dormann et al. , 2007) distance geospatial distance of each site from the highest ranked cell by HUI on that reef using the distMeeus function in the geosphere package in the R p rogramming environment (Hijmans et al. , 2014; R Core Team, 2015) . This method assumes spatial stationarity, i.e., the effect of proximity is constant across sites within a reef. Therefore, due to the unique shape of HSR and the presence of two distinct are as of condensed activity, the distance parameter for sites on HSR was calculated as the geospatial distance of each site from the nearest of the top - ranked cells on each arm (East and West) of the reef ( Figure 2 . 1 ). A set of 16 competing additive fixed effect logistic regression models was used to relate egg deposition in 2013 and 2014 to measured habitat characteristics on each reef ( Objective 1; Table 2 . 1 ). For each reef, models containing all possible combinations of measured physical 29 characteristics were evaluated, including the global model (all variables) and null model distance Table 2 . 1 ). To simplify interpretation of the model outputs, an intercept term was not estimated, resulting in separate coefficients for each level of surficial substrate classification. Logistic regression models were fit using the glm function ( family the base stats package in R (R Core Team 2015) small sample sizes ( AICc ) was use d to rank competing models on each reef by the maximum log likelihood, including a penalty for model complexity (e.g., Burnham and Anderson , 2002; Anderson , 2007) . Multimodel inference and model averaging were then used to estimate the relative importance of measured physical habitat characteristics to predicting egg deposition among sites encountered by tagged lake trout on each reef (e.g., Burnham and Anderson , 2002 ; Grueber et al. , 2011 ) . Model - averaged parameter estimates ( ) were calculated for each r eef using a weighted average across all models, substituting zero for models in which the parameter did not appear, and standardized ( ) relative to their partial standard deviation (adjusted for multiple correlations of input variables; See Grueber et al., 2011) to aide interpretation between parameters with differing scales and compensate for potentia l correlations among predictors. Because cumulative Akaike weights ( w i ) were diffi cult to interpret among parameters with differing units, the relative importance of each physical characteristic was also expressed as an evidence ratio of the j th standardized model - averaged parameter estimate to the maximum standardized parameter estimat e for comparison between reefs . An unconditional variance estimator, including both within and between model uncertainty estimates (e.g., Burnham and Anderson , 2002) , was used to calculate 95% confidence intervals (CI) on the model - averaged parameter coefficients and parameters in which this 95% CI included zero were considered uninformative to habitat selection. All AICc - based analyses were conducted 30 following procedures described by Burnham and Anderson (2002) and Grueber et al. (2011) usin g the MuMIN package in R , 2013; R Core Team , 2015) . Incubation Success (Objective 2) Relative in situ incubation success was compared between sites positive for egg deposition and those that were not using plexiglass incubators (e.g., Manny et al. , 1989; Eshenroder et al. , 1995) seeded with Seneca Lake wild (SLW) strain embryos stripp ed and fertilized at the Sullivan Creek National Fish Hatchery (U.S. Fish and Wildlife Service; Brimley, MI). Incubation success was assessed at 10 sites on each reef that were previously surveyed for egg deposition; five sites positive for egg deposition, including the highest ranked site by HUI , and five sites negative for egg deposition. Three incubators were attached along a 7.6m chain bisecting the central identifi er and buried in substrate near the end of the 2013 and 2014 spawning seasons at each of the 10 sites (a total of 30 incubators per reef). Incubators were buried vertically such that the uppermost long edge of the incubator was approximately flush with the substrate surface. In this configuration, incubators consisted of five layered rows of cells ranging in depth from 1 - 2 cm to 10 - 12 cm beneath the surface. E ach incubator was inspected immediately prior to deployment and eggs that had died since loading were recorded. Incubators were recovered the following spring and the proportion of live eggs at the time of deployment that were found to have developed to free embryos at the time of retrieval was used as an estimate of relative in situ incubation success. To establish a baseline for comparisons of egg incubation success between reefs, a minimum of 18 incubators were retained as laboratory controls each year. Control incubators underwent identical handling procedures a s the treatment groups but were subsequently 31 transferred to the USGS - Hammond Bay Biological Station (HBBS; Millersburg, MI) where incubators were placed in Heath incubation trays and supplied with ambient water from Lake Huron at 9 - 12 L/min throughout th e incubation period. The temperature profile experienced by all control and in situ incubators was monitored over the incubation period using tethered Onset HOBO Pendant (UA - 002) temperature loggers. Upon recovery in the spring, each string of incubators w as photographed to record its orientation prior to transport to the surface. The orientation of each incubator was assigned one buried in the substrate (maintained vertical orientation), moved ither tipped over or completely dislodged from substrate). Recovered incubators were transported to the lab in coolers filled with ambient lake water and were examined within 24 - 48 hours of retrieval. The contents of each incubation chamber were classifi ee embryo or fr ee embryos were present and their condition at the time of inspection. Incubation success for a given site was calculated as the mean proportion of eggs survi ving to hatch (i.e., sum of live eggs, live fr ee embryos , and dead fr ee embryos ) among stayed buried in substrate at time of recovery at that site (i.e., each site was considered a replicate). On HSR , 26 of 30 (86.7%) incubators deployed in 2013 and 30 of 30 (100%) of incubators deployed in 2014 remained buried in substrate until recovery. Of the four incubators on HSR that moved , three were from sites positive for egg deposition, including one incubator from the highest ranked site by HUI . On SCS , 18 of 30 (60%) incubators deployed in 2013 and 24 of 30 (80%) incubators deployed in 2014 remained buried in the substrate. Of the 18 incubators on SCS that moved, 10 were from sites positive for egg deposition, including two incubators from the highest ranked site by HUI. In situ had significantly 32 stayed excluded from subsequent analyses ( stayed = 68.9%, moved = 39.0%; Welch two - sample t - test, t = 5.49, p < 0.01). Comparisons of mean in situ incubation success between reefs and egg present/absent sites were made using two - way repeated measures Analysis of Variance (ANOVA). In this model, egg deposition ( present/absent ) and reef ( HSR, SCS ) were fixed effects, and site ID was a random effect to account for repeated measures between years. Comparison of mean incubation t - test in R (R Core Team 2015) . Comparisons of mean incubation success between in situ and control ( CON ) incubators were made using a one - significant difference) test. In this model, reef ( HSR, SCS ) and eggs ( present = +, absent = - ) were combined as a categorical grouping factor ( HSR+, HSR - , SCS+, SCS - , and CON ) with fixed effect, and Site ID was a random effect to handle repeated measures between years. In all cases, the response variable, mean proportion hatch, was arcsine (square root) transformed to improve normality and control heteroscedasticity of the model residuals (e.g., Sokal and Rohlf , 1995) . All repeated measures models were evaluated using the lme4 package in R (Bates et al. , 2015; R Core Team , 2015) . Results Egg Deposi tion Over 2013 and 2014, egg deposition was observed at 17 of 28 (60.7%) surveyed sites on HSR and 8 of 30 (26.6%) surveyed sites on SCS ( Figure 2 . 1 ). However, egg de position at some sites on HSR was not consistent between years; four sites near the shallowest ridge of the reef 33 switched from positive in 2013 to negative in 2014, while four sites on the outside boundary of the reef switched from negative in 2013 to posi tive in 2014 ( Figure 2 . 1 ). Most sites selected for egg deposition by lake trout occurred in three distinct clusters; one on each arm of HSR and a third within a singl e area in the northwest corner of SCS ( Figure 2 . 1 ). General Spawning Reef Characteristics Overall physical characteristics differed substantially between HSR and SCS ( Figur e 2 . 2 ). In general, HSR was a 0.535 km 2 sediments ranging in size from silt to boulders. The reef was distinct from the surrounding bat hymetry, rising abruptly from fine sediments at its base at 16 m depth to a small 0.038 km 2 portion of the eastern arm that rose above the water surface. HSR had steep slopes around its outer edges, particularly on the northern tip of the eastern arm. By c omparison, SCS was a broad, relatively flat 0.758 km 2 shoal comprised of layers of sorted sediments ranging in size from silt to boulders lying atop bedrock that was exposed in some places. SCS was less distinct from its surrounding bathymetry than HSR , ri sing gradually from its base at 11 m depth to a small 0.003 km 2 portion lying just beneath the water surface (~0.5 m depth). While this lack of steep slopes at SCS was the most striking difference between the two reefs in this study, predominant substrates at sampled sites on SCS were also classified in smaller size classes (especially gravel - rubble class) than those on HSR ( Figur e 2 . 2 ; Fishers exact test; 2000 replica tes, p < 0.01). The three distinct clusters of egg deposition noted above aligned closely with the areas of highest activity as estimated by HUI using acoustic telemetry data ( Figure 2 . 1 ; Binder et al. , 2018) . However, the physical characteristics of these three areas differed from each other ( Figur e 2 . 2 ). The predominant spawning area located at the tip of the east arm of HSR had features that 34 most closely aligned with the commonly - accepted conceptual model of lake trout spawning habitat. This area had steep slopes around its northern edg e and was composed of rubble - cobble substrates with deep (> 15 cm) interstitial spaces. By contrast, the predominant spawning area on SCS had little to no slope and was composed of predominantly gravel - rubble substrates with shallow (< 10 cm) interstitial space. The predominant spawning area on the west arm of HSR , was between these two extremes, with more gradual slope than the east arm of HSR , and rubble substrates that were larger than those at SCS , but smaller than those on the east arm of HSR . 35 Figur e 2 . 2 : Physical characteristics of sites that were surveyed to be positive for egg deposition vs. those that were negative at Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ), Lake Huron. Dark shading represents habitats in which lake trout spawned, light shading represents habitats in which no eggs were found. Numbers beside bars in top panel indicate number of sites characterized in a given predominant substrate size class. On HSR , slope magnitude was the only measured parameter that differed between habitats positive for egg deposition and those that were negative (Welch two - sample t - test; df = 26.1, t = 4.13, p < 0.01). On SCS , no difference between physical characteristic s of habitats selected for egg deposition and those that were not was detected. 36 Relative Importance of Physical Characteristics (Objective 1) (i.e., w i > 0.90 ) w as observed on either reef ( Table 2 . 1 ). On HSR , 10 of 15 habitat models evaluated ranked higher than the null model, indicating that physical parameters other than mer e proximity to the highest ranked site by HUI were important for predicting egg deposition ( Table 2 . 1 ). However, on SCS , the null model ranked highest ( Table 2 . 1 ), indicating that the addition of measured habitat characteristics could not predict egg deposition any better than adjacency alone. As a result, measured ph ysical characteristics were not deemed to be important to predicting egg deposition on SCS and no additional model inference was conducted on this reef. On HSR , the highest - ranked habitat model only included slope , and carried 54.1% of the weight of evidence as expressed by Akaike weights ( w i ; Table 2 . 1 ) . Two other habitat models, both including the slope parameter, were ranked above the information theoretic cut off ( AICc < 5) and thus were moderately supported by the data. Collectively, these top three models accounted for 85.4% of the weight of evidence. However, the 2 nd and 3 rd ranked models were hierarchically more - complex versions of the top - ranked model, an d thus inclusion of additional parameters did not improve model fit enough to overcome the parametrization penalty inherent to AICc . As such, these additional parameters were also considered uninformative to habitat selection (e.g., Arnold , 2010) . Model - av eraged parameter estimates further supported slope being the only important measured characteristic for predicting egg deposition on HSR ( Table 2 . 2 ). The standardized, model - slope was 1.90, nearly six times distance ( * = - 0.34), and was the only parameter estimate, whose 95% confidence interval did not include zero ( Table 2 . 2 ). 37 Table 2 . 1 : Models and parameter definitions used to predict egg deposition on Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ) ranked by AIC corrected for small sample sizes ( AICc ). All models include a spatial autocorrelation term (distance) that was omitted from this table to aid in clarity. Dashed line represents the information theoretic cutoff for models with moderate support ( < 5.0). Categorical substrate classifications were based on a Modified Wentworth scale (Marsden et al. , 1995) substrate categories for sites with >33.3% of measured substrates in more than one category. 38 Table 2 . 2 : Unstandardized and standardized (denoted by an *) m odel - averaged parameter estimates and relative importance of physical characteristics for the selection of habitats for egg deposition by lake trout on Horseshoe Reef ( HSR ). An unconditional variance estimator was used to calculate 95% confidence intervals on model - aver aged parameter estimates. Furthermore, the relative importance of each physical characteristic was expressed in two ways; f irst, as a ratio of the ith parameter Akaike weight to the maximum Akaike weight , and second, as an evidence ratio of th e ith standardized model - averaged parameter estimate to the maximum standardized parameter estimate Based on these criteria, sl ope magnitude was the only measured characteristic found to be important for the selection of spawning habitats on HSR . Unstandardized Estimates Standardized Estimates Relative Importance Parameter 95% Confidence Interval * * 95% Confidence Interval Distance (m) - 0.006 0.005 - 0.015 0.003 - 0.339 0.254 - 0.836 0.158 1.00 0.178 Physical Habitat Characteristics Slope Magnitude (%) 0.218 0.106 0.011 0.426 1.904 0.911 0.118 3.690 0.937 1.00 Substrate Homogeneity (%) 0.001 0.016 - 0.307 0.033 0.008 0.071 - 0.131 0.148 0.238 0.003 Interstitial Depth (cm) 0.008 0.059 - 0.108 0.124 0.025 0.166 - 0.301 0.351 0.233 0.010 Predominant Substrate Category Rubble 0.045 0.527 - 0.988 1.077 0.006 0.071 - 0.132 0.144 0.053 0.002 Rubble - Cobble 0.031 0.724 - 1.388 1.450 0.005 0.127 - 0.244 0.253 0.053 0.002 Uncharacterized - 0.897 671.41 - 947.14 945.34 - 0.262 195.95 - 384.56 384.04 0.053 0.103 39 Incubation Success (Objective 2) No evidence of a significant difference in mean proportion of eggs surviving to hatch (in situ incubation success) was observed between sites selected for egg deposition and those that were not ( POS = 73.8%, NEG = 69.8%; two - way repeated measures ANOVA, f = 0.06, p = 0.81; Figure 2 . 3 ). However, mean incubation success differed significantly between reefs ( HSR: 78.2%, SCS: 64.4%; two - way repeated measures ANOVA, f = 5.65, p = 0.0 3) . No evidence occurred for an interaction between egg deposition and reef (two - way repeated measures ANOVA, f = 0.32, p = 0.58). Site ID was a significant random effect in this model (i.e., variance dif ferent from 0), suggesting intra - annual variability be tween sites was high, but overall in situ estimates of incubation success appeared to be consistent between years at the reef scale. Incubation success on HSR ranged from 46.0 92.7% ( = 76.7%, s = 18.2%) in 2013 and from 65.3 93.3% ( = 79.7%, s = 9 .0%) in 2014. Similarly, incubation success on SCS ranged from 36.7 84.0% ( = 65.3%, s = 14.6%) in 2013 and 36.0 84.7% ( = 63.7%, s = 16.3%) in 2014. Control incubators held at Hammond Bay Biological Station (HBBS) showed lower survival of h atchery - reared eggs in 2014 than in 2013 ( CON ; 2013 = 92.6%, 2014 = 86.1%; - test, t = 4.24, p < 0.01 ). After accounting for this variability, mean in situ incubation success differed from that of control incubators (one - way repeated measures ANOVA, f = 4.71, p < 0.01). Mean incubation success of sites negative for egg deposition on SCS ( SCS - ) and sites positive for egg deposition on SCS ( SCS +) were both sig nificantly lower than that of control incubators (Tukey HSD, SCS - vs. CON , t - ratio = 3.80, p < 0.01, SCS + vs. CON , t - ratio = 3.36, p = 0.02). In contrast, no difference in mean incubation success between HSR + or HSR - sites and controls ( HSR - vs. CON , t - rat io = 1.56, p = 0.54, HSR + vs. CON , t - ratio 40 = 1.93, p = 0.33) was observed, nor were any differences observed among in situ incubators ( Figure 2 . 3 ). 41 Figure 2 . 3 : Relative in situ incubation success (proportion survival to hatch) of sites positive (+) and negative ( - ) for egg deposition on Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ) vs. control incubators ( CON ) in 2013 (light bars) and 2014 (dark bars). No evidence was observed for a difference in mean incubation success of habitats positive for egg deposition vs. those that were negative (+ = 73.8%, - = 69.8%; two - way repeated measures ANOVA, f = 0.06, p = 0.81), b ut mean incubation success differed between reefs ( HSR : 78.2%, SCS : 64.4%; two - way repeated measures ANOVA, f = 5.65, p = 0.03). Furthermore, mean in situ incubation success of positive and negative sites at SCS were both significantly lower than control i ncubators (letters, one - way repeated measures ANOVA, f = 4.71, p < 0.01; Tukey HSD, SCS - vs. CON , t - ratio = 3.80, p < 0.01, SCS+ vs. CON , t - ratio = 3.36, p = 0.02). 42 Discussion Physical characteristics of spawning substrates on two adjacent spawnin g reefs in the Drummond Island Refuge varied considerably from one another, yet both were used for spawning by lake trout. On HSR , bathymetric slope was greater in habitats in which lake trout deposited eggs than on those that they did not. However, on SCS , habitats in which eggs were observed were indistinguishable from those where no eggs were found based on the habitat characteristics measured. While slope appears to have been an important predictor of spawning habitats on HSR , it was not a feature of every site chosen for spawning and some sites not receiving egg deposition also occurred on slopes greater than 8%. Moreover, inter - annual variability in sites selected for egg deposition on HSR suggest ed that some variables used by lake trout to select spawning habitats may vary over time. Due to environmental stochasticity, optimal conditions for egg incubation on a given spawning reef likely vary in both space and time (e.g., Fitzsimons and Marsden , 2014) . Indeed, inter - annual v ariability in in situ incubation success appeared greater at the site scale than at the reef scale, so it is possible lake trout may respond to these changes by shifting their spawning locations within reefs between years. Alternatively, in environments wi th an abundance of suitable habitat, lake trout may spawn over a variety of physical conditions as a bet - hedging strategy to overcome this variability in incubation suitability ( e.g., Fitzsimons and Marsden , 2014; Callaghan et al. , 2016) . Nonetheless, the physical characteristics measured in this study were unable to completely explain the spawning habitat selection observed , so other unmeasured variables may have been important. No evidence was found to support a relationship between egg deposition and pr edominant substrate diameter . In both years of the study, egg deposition was observed on habitats in each of the predominant substrate categories measured, including small gravel 43 substrates (0.2 - 6.4 cm diameter) not typically included in the commonly - acc epted conceptual model of lake trout spawning habitat suitability (e.g., Marsden et al. , 1995) . Admittedly, I found accurate characterization of substrates to be challenging. While surveys for egg deposition occurred in the same areas as were physically ch aracterized, mean and median diameter estimates were likely biased due to spatial limitations inherent in using an image - based characterization method. Therefore, the use of categorical classifications for predominant substrate diameter required estimation of additional parameters ( K +3), while parameters on a bounded continuous scale, such as slope, interstitial depth, and homogeneity only required a single parameter estimate. As such, the parameterization penalty inherent to AIC may have been exaggerated f or models including substrate diameter. Furthermore, analyses were limited by small sample sizes and an unbalanced design between habitats positive and negative for egg deposition. Sites were intentionally selected to avoid use of adjacent cells within ree fs, but portions of the reefs were broadly similar enough that not all combinations of habitat parameters were available at all levels of egg deposition. For example, because most sites were classified as predominantly rubble substrate ( HSR : 21/28 sites, S CS : 17/30 sites), the three remaining sites classified as predominantly rubble - cobble substrate on SCS did not exist in both high and low levels of homogeneity or interstitial depth to determine if these parameters influenced the odds of egg deposition whi le accounting for other characteristics. Nonetheless, this study suggests that the specific size and shape of substrates are likely less critical for lake trout spawning habitat selection than previously thought (e.g., Sly , 1988) . Rather, when placed on th e same scale with other parameters, a single unit (%) change in slope increased the odds of selection for egg deposition on HSR by 12.8 (CI = [1.9, 63.4]) times. In contrast, slope did not appear to be important for predicting egg deposition on SCS . Howeve r, nearly all estimates of slope were 44 below that thought to be required (8%) for spawning described in the commonly - accepted conceptual model (e.g., Marsden et al., 1995). Therefore, slopes greater than 8% may not be required for spawning in all locations. Alternatively, slope may have an indirect relationship with other unmeasured correlated variables that influence habitat selection, such as current. Slope or bathymetric contour has consistently been associated with lake trout spawning habitat (Marsden a nd Krueger , 1991; Bronte et al. , 1995; Casselman , 1995; Marsden et al. , 1995) . While bathymetric contour may play a physical role in guiding and congregating spawning fish, the association of spawning habitats with areas of pronounced slope also may be related to the currents experienced above and through these habitats rather than the slopes themselves. Currents flowing interstitially benefit egg incubation by preventing siltation of eggs while ensuring eggs remain in oxygenated water (Gunn and Keller , 1984; Sly , 1988) . Consistent water currents are critical to maintaining suitable interstitial water quality for successful development of incubating eggs and early life stages . Currents may also transport emergent lake trout from spawning to nursery habitats (Bronte et al. , 1995; Casselman, 1995) . However, i ncubating eggs are suscepti ble to mechanical stresses and therefore experience decreased survival when exposed to strong wave action (Eshenroder et al. , 1995) . Based on the Bernoulli principle and its corollary the Venturi effect, as velocity increases in an area of constriction, th e static pressure exerted must decrease to maintain equilibrium (Vogel , 1994) . Therefore, a s strong, potentially damaging waves and currents course around reefs and through the interstices of the substrate that forms them, low pressure eddies of swirling w ater are formed, thus reducing the mechanical energy of this flow to a level that supports successful incubation (e.g., Thibodeaux and Boyle , 1987; Tonina and Buffington , 2007) . In this study, many sites receiving egg deposition, including the most highly - used sites on HSR and SCS appeared to be on the 45 leeward (NW) side of the reefs, on slopes facing opposite the predominant current direction (SE to NW). In this manner, I hypothesize that water currents, bathymetric contour, shoreline shape, and substrate c haracteristics interact to buffer the physical forces of strong currents, providing reliable, consistent interstitial flow to oxygenate incubating eggs without damaging them. Consistent interstitial flow during egg incubation is a common requirement for a ll salmonids (Chapman , 1988) . The hyporheic exchange of groundwater and surface waters has been found to be a primary source of interstitial flow for the selection of spawning habitats in several closely related salmonid species (e.g., Power et al. , 1999) . For example, in brook trout, Salvelinus fontinalis, spawning occurs almost exclusively in close association with groundwater upwellings both in stream (e.g., Curry and Noakes , 1995) and lacustrine habitats (e.g., Ridgway and Blanchfield , 1998) . In bull tr out, Salvelinus confluentus , spawning habitats are strongly associated both with areas of upwelling and areas of downwelling within the same stream (e.g., Baxter and Hauer , 2000) . Arctic char, Salvelinus alpinus, and chum salmon, Oncorhynchus keta, appear to favor groundwater upwellings for redd construction (e.g., Cunjak et al. , 1986; Burril et al. , 2010) , while brown trout, Salmo trutta, and Atlantic salmon, Salmo salar, have been shown to preferentially select areas of downwelling for spawning (e.g., Zim mer and Power , 2006; Louhi et al. , 2008) . Nonetheless, lacustrine groundwater exchange has yet to have been implicated as a characteristic of lake trout spawning habitats. Given the potential overlap of areas of groundwater discharge and nearshore spawning habitats, the chemistry and predictable flow regimes associated with groundwater upwellings may be attractive to spawning lake trout (e.g., Gunn and Keller , 1984) . Alterna tively, hyporheic exchange of water as lake currents are forced through substrates due to sharp changes in bathymetric contour may resemble the flow regimes caused by groundwater upwellings (e.g., Gunn , 1995) . While the degree to which lacustrine 46 groundwat er discharge may influence lake trout spawning habitat selecti on is currently speculative, observations in this study suggest that factors other than those measured may be important. Therefore, I hypothesize that localized flow characteristics in close pro ximity to the lake bottom are likely to be an important component influencing the selection of habitats for egg deposition. Consequently , f uture studies seeking to identify the critical physical habitat requirements for successful lake trout spawning shoul d focus on fine - scale hydrology and characterization of interstitial flow dynamics rather than characteristics of substrate alone. I found no evidence to support a difference between the mean in situ incubation success of habitats on which lake trout depo sited eggs and those that did not receive egg deposition. However, small sample sizes and high variability within reefs meant that power for detecting a significant difference was low. For example, assuming an = 0.05 and a mean difference of 10% ( diff 0.10 ) in incubation success between habitats selected for egg deposition and those that were not, a simulation - based power analysis using the powerSim function of the simr package in R (1000 simulations; R Core Team , 2015; Green and MacLeod , 2016) estimat ed power (1 - at 42.5% (95% CI = [39.41, 45.63]). To achieve reasonable power (i.e., 1 - analysis, the mean difference in incubation success between habitats selected for egg deposition and those that were not would have had to exceed 1 8% ( diff ). While no mean difference in incubation success was observed between habitats selected for egg deposition and those that were not, the fact that 12 of 18 (66%) sites selected for egg deposition had estimates of mean incubation success > 75% suggests lake trout are capable of finding habitats with suitable chemical and physical properties to support incubation on both reefs. Conversely, some sites that were not selected for spawning had survival estimates greater than those that did receiv e egg deposition and therefore may be equally suitable from this 47 perspective. For example, some habitats where eggs were not observed had survival rates as high as 90%, while survival estimates as low as 40% were observed in habitats that did receive egg d eposition. These survival estimates do not include all potential sources of mortality such as predation (e.g., Claramunt et al. , 2005) , dislodgement (e.g., Marsden and Krueger , 1991) , or physical shock (e.g., Fitzsimons , 1994) . Nonetheless, based on this information and the variable egg deposition observed on HSR , I conclude that the supply of suitable spawning habitat may exceed demand at Drummond Island. However, the precise explanation as to why some habitats were chosen for egg deposition while other seemingly suitable habitats were not remains unclear. Despite similar deployment and handling methodologies, the mean in situ incubation success reported in this study for habitats in which lake trout deposited eggs (73.8% survival to hatch; 95% CI = [65.9%, 81.7%]) was greater than the 10 to 55% reported in five previous studies using the same habitat bio - assay throughout the Great Lakes (Casselman , 1995; Edsall et al. , 1995; Eshenroder et al. 1995; Manny et al. , 1995; Perkins and Krueger , 19 95) . Differences in survival to hatch between this study and previous studies may partially be explained by intrinsic characteristics such as the genetic makeup of the eggs used in this study, or extrinsic factors such as improved water quality could have also contributed. Non - native Dreissenid mussels ( Dreissena bugensis and D. polymorpha ) have been hypothesized to threaten egg incubation by altering interstitial flow regimes and smothering incubating eggs with feces and psuedofeces (Ackerman , 1999; Marsde n and Chotkowski , 2001; Barbiero et al. , 2011) . However, lakewide oligotrophication associated with their rapid spread has greatly increased water clarity, particularly in nearshore benthic environments (Noonburg et al. , 2003; Hecky et al. , 2004; Evans et al. , 2011) . As a result, the overall incubation success of reef habitats in Lake Huron may have 48 improved over the last decade, lending further support to the idea that lake trout recruitment at Drummond Island does not appear to be limited by availability of suitable habitat. The underlying architecture and features that distinguishes this study from previous studies of lake trout spawning habitat selection is the use of acoustic telemetry to guide sampling of habitats known to be encountered by lake trout . Selecting sampling sites based on behavior of the fish freed this study from biases associated with the apriori broad characterization of spawning shoals based on the capture of ripe fish, or from the sampling of habitats presumed to have adequate suitab ility. As a result, this study approached habitat selection at a scale not previously possible, allowing me to distinguish habitat use among closely - located sites based on whether habitats were deemed less preferable by fish known to have encountered them. Moreover, this methodology identified novel spawning habitats not typically included in the current conceptual model and therefore, accommodated behavioral plasticity as a possible explanation for spawning habitat selection. While the use of acoustic tele metry has considerable start - up costs associated with labor and equipment, behavior - based methods for habitat assessment probably hold the gr eatest promise for advancing conceptual understanding of the mechanisms underlying spawning behavior and habitat se lection. In this study, I observed widespread egg deposition in habitats with a wide range of physical characteristics and found no difference in mean relative incubation success between habitats in which lake trout deposited eggs and those they did not . Based on the inability of the measured substrate characteristics to ful ly explain habitat selection, I conclude d substrate size was not a critical attribute and other, unmeasured, possibly correlated c haracteristics may influence s election of lake trout spawning habitats. Further, l ocalized flow characteristics in close proximity to the lake bottom may be a critical variable influencing habitat selection in lake 49 trout . Based on in situ est imates of incubation success, I concluded that lake trout spawning at Drummond Island were capable of locating habitats that support successful incubation, and that habitat suitability does not appear to be limiting. Therefore, the greatest limitation to natural recruitment of lake trout in Lake Huron was probably the com bination of low abundance of adult lake trout and poor survival of early life stages due to predation (e.g., Krueger et al. , 1995; Chotkowski and Marsden , 1999; Jonas et al. , 2005) . The implications of these findings to management of lake trout in the Great Lakes are at least three - fold. First, lake trout are highly plastic in terms of spawning habitat requirements and in variables used to select habitats for egg deposition. Thus, l ake trout are adaptive to a variety of spawning habitats a nd these habitats are not likely to restrain successful restoration efforts in the Great Lakes but could impede control efforts where lake trout have become invasive (Crossman , 1995; Hansen et al. , 2008; Martinez et al. , 2009) . Second, interstitial flow dynamics represent a substantial knowledge gap in understanding of the critical characteristics of lake trout spawning habitat. With a better understanding of interstitial flow, whether due to lake cu rrents, upwellings, or both, and its influence on habitat selection, managers can better understand recruitment dynamics and increase the efficiency of recovery or control efforts by matching stocking or fishing efforts to specific habitat requirements. Th ird, lake trout recruitment at Drummond Island does not appear to be limited by habitat availability or restricted by selection of poor incubation habitats. Therefore, I predict that recruitment from this location may continue to increase as the abundance of spawning adults increases. 50 Acknowledgements This work was largely possible thanks to the technical expertise and dedication of H. Thompson and C. Wright who provided crucial dive support in less than ideal conditions throughout the study. I hear tily thank Z. Wickert, E. Larson, D. Operhall, B. Lamoreux, H. Mcmath, P. Wigren, K. Smith, R. Darnton, J. Hinderer, and M. Faust for their essential field and vessel support. I also appreciate C. LeGault - Anderson and J. Anderson for support from the USFWS Sullivan Creek National Fish Hatchery. This work was funded by the Great Lakes Fishery Commission (Grant ID #2013_BIN_44024) via the Great Lakes Restoration Initiative (Grant ID #GL - 00E23010). Th is paper is c ontribution X of the Great Lakes Acoustic Telem etry Observation System (GLATOS). All sampling and ha ndling of fish during research we re carried out in accordance with guidelines for the care and use of fishes by the American Fisheries Society ( http://fisheries.org/docs/wp/Guidelines - for - Use - of - Fishes.p df ) , the U.S. Geological Survey ( https://pubs.er.usgs.gov/publication/2001108 ) and the MSU Institutional Animal Care and Use Committee (MSU IACUC; See AUF exemption) . The findings and conclusions in this article are those of the authors and do not necessar ily represent those of the U.S. Geological Survey or the U.S. Fish and Wildlife Service. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. government. 51 R EFERENCES 52 REFERENCES Ackerman, J.D. , 1999. Effect of velocity on the filter feeding of dreissenid mussels ( Dreissena polymorpha and Dreissena bugensis ): implications for trophic dynamics. Can. J. Fish. Aquat. Sci. 56(9): 1551 1561. Anderson, D.R. , 2007. Model based inference i n the life sciences: a primer on e vidence. Springer Science & Business Media, New York, NY. Arnold, T.W. , criterion. J. Wildl. Manag. 74(6): 1175 1178. doi:10.2193/2009 - 367. Barb iero, R.P., Schmude, K., Lesht, B.M., Riseng, C.M., Warren, G.J., and Tuchman, M.L. , 2011. Trends in Diporeia populations across the Laurentian Great Lakes, 1997 2009. J. Gt. Lakes Res. 37(1): 9 17. doi:10.1016/j.jglr.2010.11.009. , 2013. MuMIn: multi - model inference. R Package Version 1.42.1. https://CRAN.R - project.org/package=MuMIn. Bates, D., Mächler, M., Bolker, B. , and Walker, S. , 2015. Fitting linear mixed - effects models u sing lme4. J. Stat. Softw. 67(1): 1 48. doi:10.18637/jss.v067.i01. Bax ter, C.V., and Hauer, F.R. , 2000. Geomorphology, hyporheic exchange, and selection of spawning habitat by bull trout ( Salvelinus confluentus ). Can. J. Fish. Aquat. Sci. 57(7): 1470 1481. Beauchamp, D.A., Allen, B.C., Richards, R.C., Wurtsbaugh, W.A., and G oldman, C.R. , 1992. Lake trout spawning in Lake Tahoe: e gg incubation in deepwater macrophyte beds. N . Am. J. Fish. Manag. 12(3): 442 449. doi:10.1577/1548 - 8675(1992)012<0442:LTSILT>2.3.CO;2. Binder, T.R., Holbrook, C.M., Hayden, T.A., and Krueger, C.C. , 2 016a. Spatial and temporal variation in positioning probability of acoustic telemetry arrays: fine - scale variability and complex interactions. Anim. Biotelemetry 4(1). doi:10.1186/s40317 - 016 - 0097 - 4. Binder, T.R., Riley, S.C., Holbrook, C.M., Hansen, M.J., Bergstedt, R.A., Bronte, C.R., He, J., Krueger, C.C. , 2016b. Spawning site fidelity of wild and hatchery lake trout ( Salvelinus namaycush ) in northern Lake Huron. Can. J. Fish. Aquat. Sci. 73(1): 18 34. doi:10.1139/cjfas - 2015 - 0175. Binder, T.R., Farha, S.A ., Thompson, H.T., Holbrook, C.M., Bergstedt, R.A., Riley, S.C., Bronte, C.R., He, J., and Krueger, C.C. , 2018. Fine - scale acoustic telemetry reveals unexpected lake trout, Salvelinus namaycush , spawning habitats in northern Lake Huron, North America. Ecol . Freshw. Fish 27(2): 594 605. doi:10.1111/eff.12373. 53 Bronte, C.R., Selgeby, J.H., Saylor, J.H., Miller, G.S., and Foster, N.R. , 1995. Hatching, dispersal, and bathymetric distribution of age - 0 wild lake trout at the Gull Island Shoal complex, Lake Superio r. J. Gt. Lakes Res. 21 , Supple. 1 : 233 245. Available from http://www.sciencedirect.com/science/article/pii/S0380133095710966 [accessed 31 October 2014]. Bronte, C.R., Jonas, J., Holey, M.E., Eshenroder, R.L., Toneys, M.L., McKee, P., Breidert, B., Claram unt, R.M., Ebener, M. P., Krueger, C.C., and Wright, G . , 2003. Possible impediments to lake trout restoration in Lake Mich igan. Lake trout task group report to the Lake Michigan Committee, G reat Lakes Fish ery Commission. Ann Arbor, MI. Available from http://www.sealampreycontrol.net/lakecom/lmc/ltrestore.pdf [accessed 30 October 2014]. Bronte, C.R., Holey, M.E., Madenjian, C.P., Jonas, J.L., Claramunt, R.M., McKee, P.C., Toneys, M.L., Ebener, M.P., Breidert, B., Fleischer, G.W., Hess, R., Martell, A.W. , and Olsen, E.J. , 2007. Relative abundance, site fidelity, and survival of adult lake trout in Lake Michiga n from 1999 to 2001: i mplications for future restoration strategies. N. Am. J. Fish. Manag. 27(1): 137 155. doi:10.1577/M05 - 214.2. Burnham, K.P., an d Anderson, D.R. , 2002. Model selection and multimodel inference: a practical information - theoretic approach. In 2nd ed. Springer Science & Business Media, New York, NY. Burril, S.E., Zimmerman, C.E., and Finn, J.E. , 2010. Characteristics of fall chum salm on spawning habitat on a mainstem river in interior Alaska. United States Geological Survey Report No. 2010 - 1164. Available from https://pubs.er.usgs.gov/publication/ofr20101164 Callaghan, D.T., Blanchfield, P.J., and Cott, P.A. , 2016. Lake trout ( Salvelin us namaycush ) spawni ng habitat in a northern lake: t he role of wind and physical characteristics on habitat quality. J. Gt. Lakes Res. 42(2): 299 307. doi:10.1016/j.jglr.2015.07.001. Casselman, J.M. , 1995. Survival and development of lake trout eggs and fr y in eastern Lake Ontario in situ incubation, Yorkshire Bar, 1989 1993. J. Gt. Lakes Res. 21, Supple. 1: 384 399. doi:10.1016/S0380 - 1330(95)71112 - 1. Chapman, D. , 1988. Critical review of variables used to define effects of fines in redds of large salmoni ds. Trans. Am. Fish. Soc. 117(1): 1 21. Chotkowski, M.A., and Marsden, J. E., 1999. Round goby and mottled sculpin predation on lake trout eggs and fry: field predictions from laboratory ex periments. J. Gt. Lakes Res. 25, Supple. 1 : 26 35. Available from ht tp://www.sciencedirect.com/science/article/pii/S0380133099707148 [accessed 28 October 2014]. Claramunt, R.M., Jonas, J.L., Fitzsimons, J.D., and Marsden, J.E. , 2005. Influences of spawning habitat characteristics and interstitial predators on lake trout eg g deposition and mortality. Trans. Am. Fish. Soc. 134(4): 1048 1057. doi:10.1577/T04 - 075.1. 54 Coble, D.W., Bruesewitz, R.E., Fratt, T.W., and Scheirer, J.W. , 1990. Lake trout, sea lampreys, and overfishing in the upper Great Lakes: a review and reanalysis. T rans. Am. Fish. Soc. 119(6): 985 995. Available from http://www.tandfonline.com/doi/abs/10.1577/1548 - 8659(1990)119%3C0985%3ALTSLAO%3E2.3.CO%3B2 [accessed 31 October 2014]. Crossman, E.J. , 1995. Introduction of the lake trout ( Salvelinus namaycush ) in areas outside its native distribution: a review. J. Gt. Lakes Res. 21, Supple. 1: 17 29. doi:10.1016/S0380 - 1330(95)71081 - 4. Cunjak, R., Power, G., and Barton, D. , 1986. Reproductive habitat and behaviour of anadromous Arctic char ( Salvelinus alpinus ) in the Kor oc River, Quebec. Nat. Can. 113(4): 383 387. Curry, R.A., and Noakes, D.L. , 1995. Groundwater and the selection of spawning sites by brook trout ( Salvelinus fontinalis ). Can. J. Fish. Aquat. Sci. 52(8): 1733 1740. Available from http://www.nrcresearchpress .com/doi/abs/10.1139/f95 - 765 [accessed 30 October 2014]. Dormann, C.F., McPherson, J. M., Araújo, M. B. , Bi vand, R., Bolliger, J., Carl , Davies, R. G. , Hirzel, A., Jetz, W., Kissling, W. D. , Kühn, I., Ohlemüller, R., R. Peres - Neto, P., Reineking, B., Schröder, B., Schurr, F. M. , and Wilson, R. , 2007. Methods to account for spatial autocorrelation in the analysis of species distributional data: a review. Ecography 30(5): 609 628. doi:10.1111/j.2007.0906 - 7590.05171.x. and Jude, D.J. , 1981. Substrate conditions and abundance of lake trout eggs in a traditional spawning area in southeastern Lake Michiga n. N. Am. J. Fish. Manag. 1(2): 165 172. Available from http://www.tandfonline.com/doi/abs/10.1577/1548 - 8659(1981)1%3C165 %3ASCAAOL%3E2.0.CO%3B2 [accessed 31 October 2014]. Ebener, M.P. , 1998. A lake trout rehabilita tion guide for Lake Huron. Miscellaneous p ublications, G reat Lakes Fishery Commission , Ann Arbor, MI. (1998): 1 44. Available from http://www.glfc.org/pubs/misc/L kTroutRepHuron.pdf [accessed 13 October 2013]. Edsall, T.A., Holey, M.E., Manny, B.A., and Kennedy, G.W. , 1995. An evaluation of lake trout reproductive habitat on Clay Banks Reef, northwestern Lake Michigan. J. Gt. Lakes Res. 21, Su pple. 1: 418 432. doi:1 0.1016/S0380 - 1330(95)71114 - 5. Edsall, T.A., and Cleland, J. , 2000. Optimum temperature for growth and preferred tempera tures of age - 0 lake trout. N. Am. J. Fish. Manag. 20(3): 804 809. Available from http://www.tandfonline.com/doi/abs/10.1577/1548 - 8675(200 0)020%3C0804%3AOTFGAP%3E2.3.CO%3B2 [accessed 30 October 2014]. Eshenroder, R.L., Poe, T.P., and Olver, C.H. , 1984. Strategies for rehabilitation of lake trout in the Great Lakes: p roceedings of conference on lake trout research, August 1983. Great Lakes Fi shery Commission. Ann Arbor, MI. Eshenroder, R.L. , 1992. Decline of lake trout in Lake Huron. Trans. Am. Fish. Soc. 121: 548 554. Available from http://www.tandfonline.com/doi/pdf/10.1577/1548 - 8659 - 121.4.548 [accessed 31 October 2014]. 55 Eshenroder, R.L., Br onte, C.R., and Peck, J.W. , 1995. Comparison of lake trout - egg survival at inshore and offshore and shallow - water and deepwater sites in Lake Superior. J. Gt. Lakes Res. 21, Supple. 1: 313 322. doi:10.1016/S0380 - 1330(95)71105 - 4. Eshenroder, R.L., Peck, J.W ., and Olver, C.H. , 1999. Research priorities for lake trout rehabilitation in the Great Lakes: a 15 - year retrospective. Tech nical Report, G reat Lakes Fishery Commission. Ann Arbor, MI. 64: 1 40. Evans, M.A., Fahnenstiel, G., and Scavia, D. , 2011. Incidental oligotrophication of North American Great Lakes. Environ. Sci. Technol. 45(8): 3297 3303. doi:10.1021/es103892w. Fitzsimons, J.D. , 1994. Survival of lake trout embryos after receiving physical shock. Progress. Fish - Cult. 56(2): 149 151. d oi:10.1577/1548 - 8640(1994)056<0149:SOLTEA>2.3.CO;2. Fitzsimons, J.D. , 1995. Assessment of lake trout spawning habitat and egg deposition and survival in Lake Ontario. J. Gt. Lakes Res. 21, Supple. 1: 337 347. doi:10.1016/S0380 - 1330(95)71108 - X. Fitzsimons, J.D. , 1996. The significance of man - made structures for lake trout spawning in the Great Lakes: are they a viable alternative to natural reefs? Can. J. Fish. Aquat. Sci. 53 , Supple. 1 : 142 151. Available from http://www.nrcresearchpress.com/doi/pdf/10.1139 /f96 - 008 [accessed 28 October 2014]. Fitzsimons, J.D., Brown, S., Brown, L., Honeyfield, D., He, J., and Johnson, J.E. , 2010. Increase in lake trout reproduction in Lake Huron following the collapse of alewife: relief from thiamine deficiency or larval pre dation? Aquat. Ecosyst. Health Manag. 13(1): 73 84. doi:10.1080/14634980903581252. Fitzsimons, J.D., and Marsden, J.E. , 2014. Relationship between lake trout spawning, emb ryonic survival, and currents: a case of bet hedging in the face of environmental sto chasticity? J. Gt. Lakes Res. 40(1): 92 101. doi:10.1016/j.jglr.2013.12.014. Green, P., and MacLeod, C.J. , 2016. simr: an R package for power analysis of generalised linear mixed models by simulation. Methods Ecol. Evol. 7(4): 493 498. doi:10.1111/2041 - 210 X.12504. Grueber, C.E., Nakagawa, S., Laws, R.J., and Jamieson, I.G. , 2011. Multimodel inference in ecology and evolu tion: challenges and solutions. J. Evol. Biol. 24(4): 699 711. doi:10.1111/j.1420 - 9101.2010.02210.x. Gunn, J.M. , 1995. Sp awning behavior of lake trout: e ffects on colonization ability. J. Gt. Lakes Res. 21, Supple. 1: 323 329. doi:10.1016/S0380 - 1330(95)71106 - 6. Gunn, J.M., and Keller, W. , 1984. Spawning site water chemistry and lake trout ( Salvelinus namaycush ) sac fry survival during spring snowmelt. Can. J. Fish. Aquat. Sci. 41(2): 319 329. Available from http://www.nrcresearchpress.com/doi/abs/10.1139/f84 - 036 [accessed 29 October 2014]. 56 Hansen, M.J., 1999. Lake trout in the Great Lakes: basin - wide stock coll apse and binational restoration. Pages 417 454 in Great Lakes Fisheries Policy and Management. Edited by W . W . Taylor and C . P . Ferreri. Michigan State University Press, East Lansing, MI. Hansen, M.J., Horner, N.J., Liter, M., Peterson, M.P., and Maiolie, M.A. , 2008. Dynamics of an increasi ng lake trout population in Lake Pend Oreille, Idaho. N. Am. J. Fish. Manag. 28(4): 1160 1171. doi:10.1577/M07 - 149.1. Hansen, M.J., Nate, N.A., Muir, A.M., Bronte, C.R., Zimmerman, M.S., and Krueger, C.C. , 2016. Life history variation among four lake trout morphs at Isle Royale, Lake Superior. J. Gt. Lakes Res. 42(2): 421 432. doi:10.1016/j.jglr.2015.12.011. Hanson, S.D., Holey, M.E., Treska, T.J., Bronte, C.R., and Eggebraaten, T.H. , 2013. Evidence of wild juvenile lake trout recruitment in western Lake Mi chigan. N. Am. J. Fish. Manag. 33(1): 186 191. doi:10.1080/02755947.2012.754804. He, J.X., Ebener, M.P., Riley, S.C., Cottrill, A., Kowalski, A., Koproski, S., Mohr, L., and Johnson, J.E. , 2012. Lake trout status in the main basin of Lake Huron, 1973 2010. N. Am. J. Fish. Manag. 32(2): 402 412. doi:10.1080/02755947.2012.675947. Hecky, R.E., Smith, R.E., Barton, D.R., Guildford, S.J., Taylor, W.D., Charlton, M.N., and Howell, T. , 2004. The nearshore phosphorus shunt: a consequence of ecosystem engineering by dreissenids in the Laurentian Great Lakes. Can. J. Fish. Aquat. Sci. 61(7): 1285 1293. doi:10.1139/f04 - 065. Hijmans, R.J., Williams, E., a nd Vennes, C. , 2014. geosphere: s pherical Trigonometry. R package version 1.3 - 11. Hile, R. , 1949. Trends in the lake trout fishery of Lake Huron through 1946. Trans. Am. Fish. Soc. 76(1): 121 147. Available from http://www.tandfonline.com/doi/abs/10.1577/1548 - 8659(1946)76%5B121%3ATITLTF%5D2.0.CO%3B2 [accessed 29 October 2014]. Houde, E.D. , dow. J. Northwest Atl. Fish. Sci. 41: 53 70. doi:10.2960/J.v41.m634. Johnson, J.E., and VanAmberg, J.P. , 1995. Evidence of natural reproduction of lake trout in western Lake Huron. J. Gt. Lakes Res. 21, Supple. 1: 253 259. Available from http://www.science direct.com/science/article/pii/S038013309571098X [accessed 28 October 2014]. Johnson, J.E., He, J.X., and Fielder, D.G. , 2015. Rehabilitation stocki ng of walleyes and lake trout: r estoration of reproducing stocks in Michigan waters of Lake Huron. N. Am. J. Aquac. 77(3): 396 408. doi:10.1080/15222055.2014.993488. Jonas, J.L., Claramunt, R.M., Fitzsimons, J.D., Marsden, J.E., and Ellrott, B.J. , 2005. Estimates of egg deposition and effects of lake trout ( Salvelinus namaycush ) egg predators in three reg ions of the Great Lakes. Can. J. Fish. Aquat. Sci. 62(10): 2254 2264. doi:10.1139/f05 - 141. 57 Jones, N.E., Parna, M., Parna, S., and Chong, S., in press. Evidence of lake trout ( Salvelinus namaycush ) spawning and spawning habitat use in the Dog River, Lake Su perior. J. Gt. Lakes Res. 2018. doi: 10.1016/j.jglr.2018.07.017 Krueger, C.C., Perkin s, D.L., Mills, E.L., and Marsden, J. E. , 1995. Predation by alewives on l ake trout fry in Lake Ontario: r ole of an exotic species in preventing restoration of a native species. J. Gt. Lakes Res. 21, Supple. 1: 458 469. doi:10.1016/S0380 - 1330(95)71117 - 0. Krueger, C.C., and Ebener, M. , 2004. Rehabilitation of lake trout in the Great Lakes: past lessons and futu re challenges. Pages 37 - 56 i n Bor eal Shield Watersheds: lake trout ecosystems in a changing e nvironment. Edited by J.M. Gunn, R.J. Steedman, and R.A. Ryder. CRC Press, Boca Raton, FL. Louhi, P., Mäki - Petäys, A., and Erkinaro, J. , 2008. Spawning habitat of Atlantic salmon and brown trout: general criteria and intragravel factors. River Res. Appl. 24(3): 330 339. doi:10.1002/rra.1072. Madenjian, C.P., Ebener, M.P., and Desorcie, T.J. , 2008a. Lake trout population dynamics at Drummon d Island Refuge in Lake Huron: implications for future rehabilitation. N. Amer. J. Fish. Manage. 28(4): 979 992. doi: 10.1577/M07 - 083.1 . .F., Warner, D.M., Stockwell, J.D., and Sta panian, M.A. , 2008b. Adverse effects of a lew ives on Laurentian Great Lakes fish c ommunities. N. Am. J. Fish. Manag. 28(1): 263 282. doi: 10.1577/M07 - 012.1 . Manny, B .A., J ude, D.J., and Eshenroder, R.L., 1989. Field test of a bioassay procedure for assessing habitat quality on fish spawning grounds. Trans. Am. Fish. Soc. 118(2): 175 182. doi:10.1577/1548 - 8659(1989)118<0175:FTOABP>2.3.CO;2. Manny, B.A., Edsall, T.A., P eck, J.W., Kennedy, G.W., and Frank, A.M. , 1995. Survival of lake trout eggs on reputed spawning grounds in Lakes Huron and Superior: in situ incubation, 1987 1988. J. Gt. Lakes Res. 21, Supple. 1: 302 312. doi:10.1016/S0380 - 1330(95)71104 - 2. Marsden, J.E., and Krueger, C.C. , 1991. Spawning by hatchery - origin lake trout ( Salvelinus namaycush ) in Lake Ontario: d ata from egg collections, substrate analysis, and diver observations. Can. J. Fish. Aquat. Sci. 48(12): 2377 2384. doi:10.1139/f91 - 278. Marsden, J.E., Casselman, J.M., Edsall, T.A., Elliott, R.F., Fitzsimons, J.D., Horns, W.H., Manny, B.A., McAughey, S.C., Sly, P.G., and Swanson, B.L. , 1995. Lake trout spawning habitat in the Great Lakes a review of current knowledge. J. Gt. Lakes Res. 21, Supple. 1: 487 497. doi:10.1016/S0380 - 1330(95)71120 - 0. Marsden, J.E., and Chotkowski, M.A. , 2001. Lake trout spawning on artificial reefs and the effect of zebra mussels: fatal attraction? J. Gt. Lakes Res. 27(1): 33 43. Available from http://www.sciencedirect.com/sc ience/article/pii/S0380133001706211 [accessed 30 October 2014]. 58 Marsden, J.E., Binder, T.R., Johnson, J., He, J., Dingledine, N., Adams, J., Johnson, N.S., Buchinger, T.J., and Krueger, C.C. , 2016. Five - year evaluation of habitat remediati on in Thunder Bay , Lake Huron: c omparison of constructed reef characteristics that attract spawning lake trout. Fish. Res. 183: 275 286. Martinez, P.J., Bigelow, P.E., Deleray, M.A., Fredenberg, W.A., Hansen, B.S., Horner, N.J., Lehr, S.K., Schneidervin, R.W., Tolentino, S .A., and Viola, A.E. , 2009. Western lake trout woes. Fisheries 34(9): 424 442. Available from http://www.tandfonline.com/doi/abs/10.1577/1548 - 8446 - 34.9.424 [accessed 30 October 2014]. Merkus, H.G. , 2009. Particle size measurements: fundamentals, practice, quality. Vol. 17. Springer Science & Business Media, New York, NY. Muir, A.M., Blackie, C.T., Marsden, J.E., and Krueger, C.C. , 2012a. Lake charr , Salvelinus namaycush , spawning behaviour: new field observations and a review of current knowledge. Rev. Fish Biol. Fish. 22(3): 575 593. doi:10.1007/s11160 - 012 - 9258 - 6. Muir, A.M., Krueger, C.C., and Hansen, M.J. , 2012b. Re - establishing lake trout in the Laurentian Great Lakes: past, present, and future. Pages 533 - 588 i n Great Lakes fishery policy and management: a binational perspective, 2 nd edition. Edited by W.W. Taylor. Michigan State University P ress, East Lansing. Available from https://www.researchgate.net/profile/Michael_Hansen18/publication/268207669_Re - establishing_lake_trout_in_the_Laurentian_Great_Lake s_past_present_and_future/links/ 5463aa9c0cf2cb7e9da997fb.pdf [accessed 9 March 2017]. Noonburg, E.G., Shuter, B.J., and Abrams, P.A. , 2003. Indirect effects of zebra mussels ( Dreissena polymorpha ) on the planktonic food web. Can. J. Fish. Aquat. Sci. 60(11 ): 1353 1368. doi:10.1139/f03 - 116. Perkins, D.L., and Krueger, C.C. , 1995. Dynamics of reproduction by hatchery - origin lake trout ( Salvelinus namaycush ) at Stony Island Reef, Lake Ontario. J. Gt. Lakes Res. 21, Supple. 1: 400 417. doi:10.1016/S0380 - 1330(95 )71113 - 3. Power, G., Brown, R., and Imhof, J. , 1999. Groundwater and fish insights from northern North America. Hydrol. Process. 13(3): 401 422. R Core Team, 2015. R: A language and environment for statistical computing. R Foundation for Statistical Comput ing, Vienna, Austria. ISBN 3 - 900051 - 07 - 0. Rasband, W. , 1997. ImageJ software. US National Institute of Health , Bethesda, MD, USA . Ridgway, M., and Blanchfield, P. , 1998. Brook trout spawning areas in lakes. Ecol. Freshw. Fish 7(3): 140 145. Riley, S.C., He , 2007. Evidence of widespread natural reproduction by lake trout , Salvelinus namaycush , in the Michigan 59 waters of Lake Huron. J. Gt. Lakes Res. 33(4): 917 921. doi:10.3394/0380 - 1330(2007)33[917:EOWN RB]2.0.CO;2. Riley, S.C., Rinchard, J., Honeyfield, D.C., Evans, A.N., and Begnoche, L. , 2011. Increasing thiamine concentrations in lake trout eggs from Lakes Huron and Michigan coincide with low alewife abundance. N. Am. J. Fish. Manag. 31(6): 1052 1064. doi:10.1080/02755947.2011.641066. Riley, S.C., Binder, T.R., Wattrus, N.J., Faust, M.D., Janssen, J., Menzies, J., Marsden, J.E., Ebener, M.P., Bronte, C.R., He, J.X., Tucker, T.R., Hansen, M.J., Thompson, H.T., Muir, A.M., and Krueger, C.C. , 2014. Lake trout in northern Lake Huron spawn on submerged drumlins. J. Gt. Lakes Res. 40(2): 415 420. doi:10.1016/j.jglr.2014.03.011. Schindelin, J., Arganda - Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch, T., Preibis ch, S., Rueden, C., Saalfeld, S., Schmid, B., and Tinevez, J.Y . , 2012. Fiji: an open - source platform for biological - image analysis. Nat. Methods 9(7): 676 682. Selgeby, J.H., Bronte, C.R., Brown Jr., E.H., Hansen, M.J., Holey, M.E., VanAmberg, J.P., Muth, K.M., Makauskas, D.B., McKee, P., Anderson, D.M., Paola Ferreri, C., and Schram, S.T. , 1995. Lake trout restoration in the Great Lakes: stock - size criteria for natural reproduction. J. Gt. Lakes Res. 21, Supple. 1: 498 504. doi:10.1016/S0380 - 1330(95)71121 - 2. Simard, L.G., 2017. Spawning site selection a nd fry development of invasive lake t rout in Yellowstone Lake, Yellowstone National Park, Wyoming (PhD Thesis). The University of Vermont and State Agricultural College. Sly, P.G. , 1988. Interstitial water qu ality of lake trout spawning habitat. J. Gt. Lakes Res. 14(3): 301 315. doi:10.1016/S0380 - 1330(88)71560 - 9. Sokal, R.R., and Rohlf, F.J. , 1995. Biometry: the principles of statistics in biological research. WH Freeman and Co. New York, NY. Thibodeaux, L.J., and Boyle, J.D. , 1987. Bedform - generated convective transport in bottom sediment. Nature 325(6102): 341. Tonina, D., and Buffington, J.M. , 2007. Hyporheic exchange in gravel bed river s with pool - riffle morphology: l aboratory experiments and three - dimensio nal modeling. Water Resour. Res. 43(1). doi:10.1029/2005WR004328. Vogel, S. , 1994. Life in moving fluids: the physical biology of flow. Princeton University Press. Boston, MA. Zimmer, M.P., and Power, M. , 2006. Brown trout spawning habitat selection prefe rences and redd characteristics in the Credit River, Ontario. J. Fish Biol. 68(5): 1333 1346. Zint, M.T., Taylor, W.W., Carl, L., Edsall, C.C., Heinrich, J., Sippel, A., Lavis, D., and Schaner, T. , 1995. Do toxic substances pose a threat to rehabilitation of lake trout in the Great 60 Lakes? A review of the literature. J. Gt. Lakes Res. 21, Supple. 1: 530 546. doi:10.1016/S0380 - 1330(95)71124 - 8. 61 Chapter 3 : Atypical spawning habitat selection by lake trout in the Drummond Island Refuge, Lake Huron: spawning in substrates at the base of giant boulders Abstract Evidence of widespread lake trout spawning was observed in close association with large (>1 m diameter) boulders deposited by glaciers. In 2013, telemetry - based behavioral data and high - resolution multi - beam bathymetric survey data identified congregations of tagged lake trout in a 0.63 km 2 area consisting of numerous (>100) large boulders (2 - 7 m diameter) laying on top of glacial till within the Drummond Island Refuge , Lake Huron . Key features of habitat associated with these boulders were areas of clean gravel, most often located in undercut areas beneath overhanging edges of the boulders and in narrow spaces between adjacent (< 1 m separation ) boulders, which were likely zones of scour by wind - driven currents. Diver - based surveys for egg deposition adjacent to these boulders revealed widespread egg deposition (40 out of 40 boulders surveyed) in gravel - rubble substrates ranging in diameter from 0 .8 to 21.9 cm with interstitial depth from 2.8 to 21.1 cm. Presence of f ree embryos and alevins were confirmed in these habitats using specialized traps and surveys with a modified ROV - based electroshocker, though they appeared to occur at lower densities than in more typical nearby shoal habitats . F ree embryos and alevins were also caught in locations where eggs were not observed previously on top of boulders and along the lake bottom away from boulders . Possibly, these early life stages may have the capab ility to move more than previously thought. The s e observation s of extensive use of boulder - associated habitats for spawning , egg incubation , and early life history growth suggests that th is previously undescribed habitat may provide an important contribution to total available lake trout spawning habitat and recruitment in northern Lake Huron, and possibly elsewhere in the Great Lakes . 62 Introduction Lake trout, Salvelinus namaycush, of the Laur entian Great Lakes once supported a large and valuable freshwater fishery (Smith, 1968; Cleland, 1982; Eshenroder et al., 1995). By t he 1950 s, however, mortality resulting from extensive commercial exploitation and predation from non - native sea lamprey, Pe tromyzon marinus, outpaced the reproductive capacity of lake trout and caused widespread population collapse throughout the Great Lakes (Coble et al., 1990; Eshenroder, 1992; Muir et al., 2012b). Rehabilitation efforts have focused on reestablishing self - s ustaining populations of lake trout in each of the Great Lakes (Hansen, 1999; Krueger and Ebener, 2004 ; Muir et al., 2012b); however, after nearly 60 years of supplemental stocking, sea lamprey control, and fishery regulation, full recovery has only been a chieved in Lake Superior (Hansen, 1999; Krueger and Ebener, 2004 ; Muir et al., 2012b; Hansen and Bronte, in press ). Recently, widespread reproduction has been observed in Lake Huron and portions of Lake Michigan (Riley et al., 2007; He et al., 2012; Hanson et al., 2013), but these systems are not yet fully restored. Decades of stocking of hatchery - reared lake trout and sea lamprey control in Lake Huron have established spawning - stocks of sufficient size to sustain natural reproduction (Madenjian et al., 20 08a; He et al., 2012), and recent declines in alewife, Alosa psuedoharengus, populations throughout Lake Huron have lessened predatory pressure on juvenile lake trout and reduced the incidence of thiamine deficiency (Krueger et al., 1995; Madenjian et al., 2008b; Fitzsimons et al., 2010; Riley et al., 2011). As a result, survival and recruitment of early life stages of wild - spawned lake trout has greatly improved in Lake Huron, such that by 2013, greater than 50% of main basin lake trout age - 7 or younger we re naturally produced fish (He et al., 2012; Johnson et al., 2015). Further, increased capture rates of wild juvenile lake trout in northern Lake Huron 63 adjacent to the Drummond Island Refuge suggests substantial natural reproduction is occurring nearby (DI R; Ebener 1998; Riley et al. , 2007 ; Madenjian et al., 2008a ). Consequently, this area provides an opportunity to identify spawning habitats being successfully used by lake trout. The reproductive ecology of lake trout is a critical research priority (e.g ., Eshenroder et al., 1984 , 1999 ; Muir et al., 2012b). Previous research aimed at identifying the fundamental physical characteristics of lake trout spawning habitat has yielded a conceptual model f or spawning habitat suitability (e.g., Marsden et al., 1995a, 2016; Muir et al., 2012a ) , recently - ( Simard , 2017 ). This model states that lake trout spawn non - randomly on habitats consisting of multi - layered, rounded to sub - angular rubble - cobble (8 99 cm diameter) substrates with clean interstitial spaces (usually >30 cm) typically associated with a strong bathymetric feature, such as a rapid change in depth contour or pronounced ridge on the lake bottom (typically >8% slope; e.g., Marsden and Krueger, 1991; Marsden et al., 19 95a). Rubble - cobble substrates are thought to increase egg survival by providing deep interstitial spaces that entrain developing eggs, buffer physical forces associated with wave action and ice scour and provide refuge from predation. Furthermore, current s directed by steeply sloped lake beds may interact with this three - dimensional structure to provide flow to prevent siltation and provide oxygen for incubation (Marsden and Chotkowski, 2001; Callaghan e t al., 2016; Riley et al., 2017; Farha, Ch. 1, this v olume ). - ng spawning habitat suitability but fails to explain why some habitats are chosen for egg deposition while other adjacent habitats with seemingly suitable characteristics are not (Joh nson et al., 2017; Binder et al., 2018; Farha, Ch. 1, this volume ). Further refinement of this conceptual model has largely been limited by the scale at which habitat selection has been evaluated. While numerous putative 64 historical lake trout spawning site s have been reported based on capture of sexually mature adults in the area, few studies have identified specific spawning habitats by the presence of reproductive end - products such as eggs, free embryos, or alevins (Thibodeau and Kelso, 1990; Edsall and K ennedy, 1995). In addition, most sampling for eggs, free embryos, and alevins has - model (Marsden et al., 1995a; Ellrott & Marsden, 2004; Claramunt et al., 2005 ). Furthermore, because observations of spawning in habitats that do not match the model are assumed to be rare occurrences, the current model does not accommodate the potential for behavioral plasticity as an explanation for the selection of alternative s pawning habitats (Beauchamp et al., 1992; Simard, 2017 ; Jones et al. in press ). A recent positional acoustic telemetry study of lake trout spawning behavior and habitat use in the Drummond Island Refuge, a region of Lake Huron that now has extensive wild lake trout recruitment (Ebener, 1998; Riley et al., 2007), identified several consistently - used spawning locations that do not fit the commonly - accepted lake trout spawning habitat model (Binder et al., 2018; Farha, Ch. 1, this volume ). Within the ~25 km 2 portion of the refuge encompassed by the positional acoustic telemetry array, activity during the spawning season was most concentrated at two sites, Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ), located adjacent to one another, approximately 1.5 km south of the Drummond Island shoreline ( Figure 3 . 1 ). However, congregations of lake trout were also consistently observed in other areas of the array, including a 0.63 km 2 area to the west of HSR that appeared to lack any typical reef charact eristics in terms of substrate or depth contours in comparison to other spawning sites (Binder et al., 2018). Brief visual scouting via a drop camera found that the area, subsequently named Boulder Alley ( BLD ; Figure 3 . 1 ), was comprised of numerous broadly - spaced boulders 65 greater than 1 m in diameter, with some as large as 3 - 4 m in diameter. Beyond their size, key features of these boulder sites were areas of clean g ravel - rubble substrate located at their bases. Egg deposition was confirmed in autumn 2012 in the clean substrate under a handful of these boulders at the north end of the boulder field and a single alevin was caught adjacent to the boulders in spring 2013 . The full extent to which boulder - associated spawning occurs within the array was unknown, but multi - beam bathymetry of BLD indicated that likely hundreds of these large boulders may exist in this area, so the contribution of this habitat to lake trout re cruitment in northern Lake Huron could be substantial. To better understand the potential for boulder - associated spawning to contribute to lake trout recruitment in northern Lake Huron, I sought to: 1) describe the spatial extent of lake trout spawning in association with boulders at Drummond Island, 2) determine whether egg deposition in boulder - associated habitats by lake trout could be predicted by measurable physical characteristics of the boulders or their adjacent substrates, and 3) compare the abunda nce of free embryos and alevins located in boulder - associated spawning habitats to those in other highly - used nearby spawning habitats within the acoustic telemetry array. To accomplish these objectives, 14 sampling locations on BLD were physically charact erized, surveyed for egg deposition, assessed for free embryo and alevin abundance, and compared to the most highly used sites, based on telemetry detections, on nearby HSR and SCS . 66 Figure 3 . 1 : Study site located within Drummond Island Refuge ( DIR ; grey polygon; inset) with primary spawning reefs: Horsesehoe Reef ( HSR ), Scammon Shoal ( SCS ), and Boulder Alley ( BLD ). Grey shading represents the relative activity by lake trout as estimated by acoustic telemetry standardized to the reef - specific maximum. White squares indicate sites with the highest activity on HSR and SCS , as estimated by acoustic telemetry detection s. 67 Methods Site S election S election of sampling locations at BLD was accomplished in two steps; 1) identification of potential boulders based on high - resolution bathymetry, and 2) selection of a subset of boulders to sample from among all potential boulders. Potential boulders located within the previously defined 0.63 km 2 study area were identified using high resolution (1 m 2 ) bathymetric survey data collected via multi - beam sonar in 2010 and 2011 (see Riley et al., 2014 for complete bathymetric survey details). The mean depth of e ach 1 m 2 cell in the study area was co mpared against the 75 th percentile depth of the surrounding 100 m 2 ; all cells in which the mean depth was at least 0.5 m greater (depths were negative) than the surrounding 75 th percentile depth were considered potential boulders. Potential boulders locate d in adjacent cells were merged, and remaining pixels were quantified using binary image analysis (Rasband, 1997; Schindelin et al., 2012) . Using this methodology, I estimate d that a t least 269 potential boulders we re present within the 0.63 km 2 study area ( Figure 3 . 2 ). Diver surveys were used to ground truth the bathymetric analysis and revealed that many identified using the previously described protocol were actually collections of several boulders in close proximity to one another, often with overhanging edges around their oulders at each site, ranging from sites with a single boulder to complexes consisting of five or more boulders. Boulder sites on BLD were also selected to include areas with relative lake trout activity categorized as high, medium, and low activity based on acoustic telemetry detection data collected in 2013 (see Binder et al., 2018, 2016 and Farha, Ch. 1, this volume for complete tagging and acoustic array 68 details). Additionally, some boulder sites were selected that visually appeared unsuitable for egg d eposition based on boulder embeddedness or the size (e.g., fine gravel) or interstitial depth (e.g., minimal interstitial depth or occluded with fine materials) of boulder - associated substrates. In total, 14 discrete boulder sites, encompassing 40 total bo ulders, were selected for physical characterization and surveyed for egg deposition on BLD for comparison to the most highly used 20 m x 20 m site on each HSR and SCS (one site on each HSR and SCS , 16 total sites among BLD , HSR , and SCS ; Figure 3 . 2 ). Lastly, to determine whether it was the substrate s beneath boulders , or the boulders themselves that attracted lake trout, at least one 3 - 5 m diameter area with similar substrate characteristics (e.g., predominant substrate diameter, interstitial depth) was selected for comparison purposes near each boulder site (i.e., located at least 2x the diameter of the boulder away; min. = 4 m, max. = 12 m). 69 Figure 3 . 2 : Boulder sites selected for sampling at Boulder Alley ( BLD ) on the south shore of Drummond Island, Lake Huron. Potential boulders were identified within the 0.63 km 2 study area ( black polygon ) based on comparison o f the depth of 1m cells to the 75 th percentile depth of the surrounding 100 m 2 . Cells in which the mean depth was least 0.5 m greater (depths were negative) than the 75 th percentile depth were identified as potential boulders ( black pixels ). Boulder sites ( blue shaded circles ; enlarged for clarity) were selected for sampling to include sites containing varying number of boulders and differing levels of activity by lake trout. Red shading represents the relative activity by lake trout as estimated by acousti c telemetry such that the darkest areas are those with the highest activity. Blue lines with black digits represent 5m depth contours (10 m and 15 m pictured). 70 Surveys for Egg D eposition (Objective 1) To describe the extent of boulder - associated sp awning at Drummond Island, egg deposition ( presence or absence ) was assessed at each of 14 boulder sites at BLD as well as the most highly used sites on two nearby lake trout spawning reefs, HSR and SCS , during the weeks of October 27 and November 9, 2014. Egg surveys were conducted via independent visual inspection by paired divers after peak spawning occurred. Within each boulder site, egg surveys were conducted around the perimeter of each boulder, in cluding substrates beneath overhanging edges of boulders and in adjacent selected habitats paired with boulder sites. For the most highly used sites on HSR and SCS, egg surveys were conducted within a 10 m diameter circle at the center of each 20 m x 20 m site. In both cases, divers searched independently, turning rocks over and moving water into the substrate by hand to suspend eggs. Egg surveys determined presence/absence only, but specific locations of eggs within each boulder site were recorded on site maps during each dive. Physical C haracterization (Objective 2) To determine whether egg deposition by lake trout could be predicted based on measurable physical characteristics of a boulder or its associated substrates, the 40 boulder s at BLD were physic ally characterized using a combination of direct measurement by divers and video based image processing. Overall diameter, height, relative compass orientation of the primary axis of each boulder, and number, location, and size of overhanging edges were re corded by divers and used to generate site maps. Overhanging edges were defined as areas extending beneath the boulder judged accessible to adult lake trout (minimum height of 10 cm). The relative area of available substrate present around the perimeter an d beneath overhanging 71 edges of boulders was estimated using a combination of manual measurements and digital measurements using still images taken from video collected with a GoPro Hero III camera mounted on an extension pole with parallel reference lasers separated by a distance of 10 cm ( Figure 3 . 4 ). In instances where the substrate spilled out beyond the overhanging edge of a boulder, the area was measured following the boundary of clean substrate that was usually visible where an overhanging edge stopped. However, the depth of areas beneath overhanging edges was difficult to measure effectively due to low lighting and the shallow angle of the camera required to acces s beneath the boulders. As a result, t otal area estimates of available substrate beneath overhanging edges were likely biased low. Characterization of substrates beneath and adjacent to boulder s on BLD also used a combination of direct measurement by dive rs and still images taken from the GoPro footage . For sites on HSR and SCS, substrate was characterized using five ~1 m 2 images /site taken by divers using an 8.1 - megapixel digital camera (Panasonic Lumix DMC - FX55) mounted on an aluminum frame set 1.5 m above the lake bottom (See Farha, Ch. 1, this volume for complete details) . Substrates were classified based on the modified Wentworth scale described by Marsden et al. ( 1995a) using scaled video imagery and image analysis tools in the FIJI package of Imag eJ (Rasband, 1997; Schindelin et al., 2012) . A categorical classification representing the predominant surficial substrate was assigned for each site based on the maximum proportion of the total surface area comprised of measured substrates categorized as gravel (0.2 - 6.4 cm) , rubble (6.5 - 25.6 cm) and cobble (25.7 - 99.9 cm) . Furthermore, any site with greater than combined substrate categories (e.g., gravel - rub ble mix, rubble - cobble mix, or gravel - cobble mix ; See Farha, Ch. 1, this volume ). As estimates of interstitial depth, t hree independent 72 measurements were taken in overhanging areas and in area s of substrate adjacent to boulders on BLD , as well as in each area photographed on HSR and SCS using a thin, flexible, aluminum meter rule to penetrate the substrate until bedrock was reached or substrate was too tight to permit passage. In cases where the height of an overhanging edge prevented access for direct measurement, interstitial depth was visually estimated by excavating loose substrates until maximum depth was reached. Abundance of Free Embryos and A levins (Objective 3) To determine whether the relative abundance of free embryos and a levins at BLD differed from spawning habitats at HSR and SCS , a subset of sites on BLD and the sites with the highest amount of activity on HSR and SCS were sampled through a combination of targeted trapping using specialized custom - built traps and electro shocker surveys conducted via a remotely - operated vehicle (ROV; Janssen et al., 2006; Olson and Janssen, 2017) during spring 2015. Not all sites on BLD were sampled with both traps and the ROV. Instead, 10 sites were randomly selected for each sampling met hod, such that six of the 10 boulder sites sampled with traps were also sampled via ROV, ensuring each site was sampled using at least one method. For this study, juvenile lake trout caught via traps or ROV were classified as either free embryos, fish havi ng an externally visible yolk - sac, or alevins, fish in which the y olk - sac had been fully adsorbed (Marsden et al., in review ) . Custom 100.5 cm by 30 cm rectangular traps ( Figure 3 . 3 ) constructed specifically to fit under boulders were deployed beneath overhanging edges of boulders and their surrounding substrates at 10 randomly - selected boulder sites. Four traps were deployed at each site, on three classes of habitat; 1 ) habitats associated with overhanging edges of boulders (i.e., beneath 73 boulder; 2 traps), 2) habitats immediately adjacent to a boulder but not associated with an overhanging edge (i.e., beside boulder; 1 trap), and 3) habitats with similar substrate not associated with boulders (i.e., away from boulder; 1 trap). For comparison to HSR and SCS , a combination of 10 rectangular traps and 10 more t ypical 62 cm diameter by 20 cm tall conical traps ( e. g., Stauffer, 1981; Marsden et al., 1988) were deployed on each of these reefs. Both trap types covered an area of ~ 0. 3 015 m 2 and had collection cups mounted as described by Marsden et al. (1988). Traps on all three reefs were deployed on May 20, 2015 an d checked by divers weekly until they were recovered on June 24, 2015. Divers changed collection cups for each trap on the lake bottom without dislodging traps. Therefore, within reefs, individual traps remained in their original location, and catches were tracked using site specific location IDs throughout the entirety of sampling to account for spatial variability in comparisons between trap types (See Statistical comparisons below). The relative abundance of free embryos and alevins (CPUE) was calculated as the total number of fish caught per trap per day ( CPUE: fish·trap - 1 ·d - 1 ) and summarized by habitat class and reef for comparisons. As a complement to trapping efforts, 10 randomly - selected boulder sites and the sites on HSR and SCS were surveyed with a modified VideoRay Pro 3 GTO ROV outfitted with an ETS ABP - 3 electrofishing backpack unit (pulsed DC; 1 - 1000 Hz and up to 250v) and suction sampler consisting of a MiniRover thruster attached to a 50mm diameter clear acrylic collection tube that protruded ~10 cm from the bottom of the ROV ( e.g., Janssen et al., 2006; Olson & Janssen, 2017). To access undercut spaces beneath boulders, twin cathodes were mounted on semi - rigid poles on each side of the ROV (0.3 m wide) and extended ~1 m into the viewing space of the ROV from a single anode mounted on the collection tube, near the main body of the ROV. In 74 this configuration, the general fishing area between cathodes and anode consisted of a 0.15 m 2 triangle (1 m high x 0 .3 m base). All ROV sampling was recorde d digitally for visual quantification of stunned fish that were not collected. Relative abundance (CPUE) of free embryos and alevins via ROV was estimated as the mean number of fish observed from recorded video footage per electrofishing CPUE = fish ·run - 1 ) and was summarized by site ( HSR , SCS , 10 boulder sites within BLD ) and sampling location ( base or top of boulder ). A small led visible in the frame of the video indicated when the electrofishing unit was triggered. Each run was limited to 1 - 2 seconds, during which the ROV would hold as still a position as possible, usually resting on the bottom briefly. T he ROV would t hen quickly maneuver to attempt to collect stunned fish , before conducting another run in a new location . ROV - based sampling occurred from June 1 - 4, 2015. At sites on BLD , the ROV moved around the perimeter of all boulders and across habitats located aw ay from the boulder, electrofishing the substrate at random intervals and attempting to collect any stunned fish. A t the beginning of each dive , the ROV would stop briefly at the top of each boulder to conduct pre - survey systems checks . Free embryos and al evins were observed responding to these checks on top of the boulders , and therefore t he top s of all boulders at sites selected for ROV sampling were also electrofished . At HSR and SCS , the ROV sampled each 20 m x 20 m (400 m 2 ) area, intermittently elect ro fishing the substrate at random intervals and collecting stunned free embryos and alevins. ROV configuration and electroshocker settings used on HSR and SCS were identical to those used at BLD but sampling was limited to 20 min for each site . In comparison , mean sampling time at BLD was 35 min due to habitat complexity. 75 Figure 3 . 3 : Schematic drawing and pictures of custom rectangular traps for the collection of early life stages of lake trout from beneath large boulders. Rectangular traps were deployed to estimate the abundance of early life stages of lake trout beneath boulders at Boulder Alley ( bottom - left ) and adjacent to round traps at Horseshoe Reef and Scammon Shoal for trap comparisons ( bottom - right ). Rectangular traps were constructed from the same 3 mm galvanized mesh and had the same coverage area ( ~ 0.30 m 2 ) as round traps but had a long shallow slope for accessing boulder - associated habitats ( bottom - left ). Collection cups were fastened to trap flange using elastic cords. 76 Statistical C omparisons Substrate size classification among boulder sites for comparison purposes was challenging. The angle divers used to photograph substrate and the number of photographs at each site were not consistent due to the confined spaces associated with overhanging edges. Thus, the total area photographed differed based on the specific conditions at each site. Further , more small substrates fit in a given image than large substrates, causing low mean and median diameter estimates, a consistent low bias. Therefore, substrates adjacen t to boulders and beneath overhanging edges were characterized using a categorical classification scheme based on the predominant substrate diameter. Statistical comparisons of substrate size among reefs and between substrates found at the base and away fr om boulders were conducted using a two - way contingency table and a chi - square test of independence with a simulated p - value (2000 simulations) in the R programming environment . Statistical comparisons of mean interstitial depth among boulder sites at BLD and the two spawning reefs ( HSR and SCS ) were done using a one - - hoc test in R (R Core Team, 2015; Wickham, 2015). For trap - based CPUE estimates, t hree distinct statistical comparisons using lin ear mixed models fit with restricted maximum likelihood were performed using the lme4 and lmertest packages in R (Bates et al., 2015 ; Kuznetsova et al., 2015; R Core Team, 2015). First, to determine whether CPUE differed by trap shape ( round vs. square ) fo r traps deployed on HSR and SCS , trap shape was treated as a fixed effect, reef and week were treated as simple, scalar random effects to account for repeat measures and location ID:reef as a multilevel random effect with location ID nested within reef to account for spatial variability. Second, to determine whether CPUE differed by habitat class ( under boulder, beside boulder, vs. away from boulder) 77 for traps deployed on BLD , class was treated as a fixed effect, while boulder site and week were treated as simple, scalar random effects to account for repeat measures. Lastly, to determine whether CPUE differed by reef ( BLD vs. HSR vs. SCS ), reef was treated as a fixed effect while week was treated as a simple, scalar random effect to account for repeat measur es. For all cases, random effects were assessed as random intercepts only and the response variable, CPUE, was log transformed to improve normality and control heteroscedasticity of the model residuals. To handle this transformation, which cannot accommoda te zeros, a small value (1.0e - 15) was added to all trap - based CPUE estimates. Statistical comparisons of ROV sampling among BLD , HSR and SCS were prevented by single sampling events on HSR and SCS . To determine whether ROV CPUE differed between the base and top of boulders within BLD , a linear mixed model predicting CPUE with Location (i.e., top or base ) as a fixed effect and SiteID as a simple, scalar random effect was fit with restricted maximum likelihood and evaluated using the lme4 and lmertest packages in the R programming environment (Bates et al., 2015; Kuznetsova et al., 2015; R Core Team, 2015). Results Surveys for Egg Deposition (Objective 1) Across 14 sites at BLD , egg deposition was c onfirmed beneath or adjacent to all 40 boulders surveyed, and always occurred in association with overhanging edges of boulders. In total, eggs were found in substrates beneath all 86 discrete overhanging edges among the 40 boulders surveyed. While eggs we re not observed in any of the areas not associated with overhanging edges, they were often observed in continuous patches that extended up to 1m away 78 observed on top of one boulder ~1.6 m above the lake bottom during the week of November 9, 2014 . Surveys on HSR and SCS confirmed widespread egg deposition at each of the most highly used sites, extending well beyond the arbitrary 20 m x 20 m site boundaries. While s urveys for egg deposition were not formally quantitative, egg densities in boulder - associated substrates visually appeared to rival or exceed those at sites on HSR and SCS . However, because boulder - associated habitats on BLD were discontinuous, egg deposit ion appeared to occur in small patches under the boulders vs. the spatially broad egg deposition observed at HSR and SCS . Physical Characterization (Objective 2) Boulders at the 14 sites on BLD varied from large pieces of angular bedrock with sharp edges to round boulders with smooth edges and ranged from 0.9 - 7.3 m in diameter and 0.5 2.2 m in height ( Table 3 . 1 ). Mean water depth of the 14 boulder sites ranged fro m 8.5 - 10.9 m, with a mean of 9.6 m. Among 40 boulders characterized, 86 discrete overhanging edges were observed and classified as one of three distinct types ( Figure 3 . 4 ). At single boulders, the interface between the substrate and the edges of the boulder appeared to have been eroded away undercut perimeter ( Figure 3 . 4 a). In some instances, a boulder edge appeared embedded in the substrate on the lake bottom creating a tunnel, while the opposite edge jutted out of the sub strate, one another appeared to have the space between them eroded, creating overhanging edges on one or both boulders ( Figure 3 . 4 b). Instances in which the overhanging edges of two adjacent boulders faced one another were generally considered a single overhang. Finally, boulders were also observed atop one another, creating space between the bottom of the upper boulder and the 79 lake bottom ( Figure 3 . 4 define, as por tions of an undercut area were sometimes visible from multiple edges of the boulder. To determine whether a single continuous overhanging edge or separate spaces occurred , a diver would shine a flashlight in one side, while the other diver filmed with an e xtension pole to determine if light shined through from the opposite side . In each type of overhanging edge, the boulder itself provided cover for the substrate beneath it, and the substrate within was visibly cleaner (i.e., little to no algae growth, mini mal Dreissena spp. colonization, light siltation) than the surrounding substrate. The space beneath overhanging edges ranged from 59 300 cm in length (distance along perimeter of boulder) , 21 191 cm in width (depth extending beneath boulder) , and 9.4 - 97.4 cm in height ( Table 3 . 1 ). The area of available substrate beneath overhanging edges ranged from 0. 2 5 2.38 m 2 and summed to a total estimate of 73.38 m 2 over the 40 boulders surveyed. Overhanging edges of boulders were not consistent in their compass orientation between boulders, and thus did not appear to be oriented facing the predominant current. However, algae growth appeared to be minimal on northern faces of boulders, likely due to reduced sunlight due to shad ing by the boulder ( at latitude 46 o N). As a result, clean substrate generally extended farther out from the boulder when overhanging edges faced north vs. other directions. In general, boulder - associated substrates within and around overhanging edges vi sually appeared smaller in diameter and had less interstitial depth than those found on the most highly used sites on HSR and SCS ( Table 3 . 1 ). Substrate located benea th overhanging edges and around the perimeter of boulders was predominantly gravel - rubble mix and ranged in diameter from 0.8 - 21.9 cm. Substrates located in boulder - associated habitats at BLD were significantly smaller than those on HSR and SCS , but no c lear differentiation in physical characteristics was observed 80 between substrates found around the edges of boulders vs. those beneath overhanging edges ( Table 3 . 1 ; Chi - square test of independence ; 2000 simulations, 2 = 22.8, df = 6, p < 0.01) . By design, substrates selected for characterization at BLD that were not associated with a boulder were similar in size and composition (gravel - rubble mix, 1.3 - 24.1 cm diame ter) to boulder - associated substrates. In contrast, substrates at sites on HSR and SCS were larger than those at BLD, being predominantly rubble, and ranged in diameter from 1.3 - 39.5 cm on HSR , and 1.2 - 36.8 cm on SCS . Mean interstitial depth of substra tes was greatest at HSR (24.6 cm), followed by boulder - associated substrates at BLD (12.8 cm), and finally SCS (5.9 cm; One - way ANOVA , f (2, 113) = 41.8 , p < 0.01; Tukey HSD , p <0.01 for all comparisons). 81 Table 3 . 1 : Summary of boulder, overhang , and substrate characteristics. Substrates were characterized using a categorical classification based on the percentage of measured substrate within size bins using a modified Wentworth scale (Marsden et al., 1995a, Farha, Ch. 1, this volume ) . The size bin with the largest proportion of measured substrate was usually considered the predominant substrate size, but sites with multiple categories having greater than 33.3% of measured substrates we re given a mixed classification. S ubstrates at Boulder Alley ( BLD ) located around the base and a way from boulders were smaller ( gravel - rubble mix; 0.2 - 25.6 cm) than substrates on Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ) (rubble; 6.5 25.6 cm ; Chi - sq uare test of independence; 2000 simulations, 2 = 22.8, df = 6, p < 0.01, expected frequencies in parentheses ). Boulder - associated substrates on BLD also had less interstitial depth than substrates at HSR , but greater than those at SCS ( One - way ANOVA , f (2 , 113) = 41.8 , p < 0.01; Tukey HSD , p <0.01 for all comparisons ) . Variability was expressed as the coefficient of variation (CV) for comparison among parameters with differing units. 82 Figure 3 . 4 : Representation of three types of overhanging edges observed in boulder - associated spawning habitats: a) boulders, b) adjacent boulders, and c) boulders. Sample photographs of each type of overhang are ordered by row, increasing in zo om from left to right (i.e., picture farthest to the right depicts habitat located beneath an overhanging edge of the boulder pictured in a given row). Parallel reference lasers ( green dots ) were used to indicate 10 cm in the image regardless of focal leng th. All habitats pictured were positive for egg deposition by lake trout . 83 Abundan ce of Free Embryos and A levins (Objective 3) Most free embryos and alevins caught at BLD were collected from traps deployed in undercut habitats; however, free embryos and alevins were caught from each of the three habitat classes, including traps placed away from the boulder where egg deposition was never observed ( Table 3 . 2 ). Boulder site and week were both significant random effects (i.e., variance different from 0) in the model ( site : R 2 = 0.069, week : R 2 = 0.073), but no significant difference in mean CPUE ( fish·trap - 1 ·d - 1 ) was found among habitat classes ( Table 3 . 2 ; Linear mixed model; f (2, 134) = 2.04, p = 0.135). Trap catches on BLD were low and highly variable, which limited statistical power to detect a significant difference amo ng habitat classes. At HSR and SCS , trap - ba s ed CPUE estimates varied by week and reef ( week : R 2 = 0.063, reef : R 2 = 0.057). More free embryos and alevins were caught in total in rectangular traps (1930 fish caught) than round traps (1224 fish caught; Table 3 . 2 ), but catches were highly variable based on trap location ( locationID:reef : R 2 = 0.467). As a result, mean CPUE did not differ significantly between rectang ular and traditional round traps deployed on HSR and SCS ( Table 3 . 2 ; Linear mixed model; f (1, 39.6) = 0.18, p = 0.672), and no evidence occurred for an interaction bet ween reef and trap shape (Linear mixed model; f (1, 38.6) = 0.21, p = 0.652). Therefore, data from the two trap types were pooled for comparison against CPUE at BLD . Trap - based CPUE estimates of lake trout free embryos and alevins varied from 0 - 17.47 fish·trap - 1 ·d - 1 and peaked near the week of June 9 on all reefs. Prior to the week of June 9, fish captured in traps were almost exclusively free embryos as evidenced by visible yolk sacs. In contrast, all fish captured in the final week of sampling (June 24) lacked a visible yolk sac and were therefore considered alevins. Mean CPUE differed significantly among the three reefs ( BLD , HSR , and SCS ; Table 3 . 2 ; Linear mix ed model; f (2, 380.4) = 197.28, p < 0.001). Mean 84 CPUE was highest on SCS (2.74 fish·trap - 1 ·d - 1 ), followed by HSR (2.59 fish·trap - 1 ·d - 1 ), and lastly BLD (0.02 fish·trap - 1 ·d - 1 ). Comparatively few free embryos and alevins were caught on BLD (27 caught) relati ve to habitats at HSR (1570 caught) and SCS (1584 caught). However, catch was highly variable on all reefs. Catches were most variable among traps deployed on BLD ( Coefficient of Variation, CV = 304.1), followed by SCS ( CV = 152.4), and HSR ( CV = 115.8). Results from ROV - based electrofishing for free embryos and alevins ( Figure 3 . 5 ) were consistent with the results from trapping. While no formal statistical comp arison s were possible due to single sampling events at HSR and SCS , the relative abundance of free embryos and alevins at these sites was double that of the single highest boulder site and over seven times greater than the mean CPUE at BLD ( Figure 3 . 5 ). In total , 55 fish were observed over 717 runs at BLD (CPUE = 0.077 fish · run - 1 ), while 31 fish were observed over 53 runs at HSR (CPUE = 0.585 fish · run - 1 ) and 23 fish were observed over 42 runs at SCS (CPUE = 0.548 fish · run - 1 ). Nonetheless, ROV - based sampling provided secondary confirmation of natural production of wild - spawned free embryos and alevins at all boulders sampled. Notably, mean CPUE by electro fishing on the tops of boulders was significantly greater than that at the base of boulders where eggs were observed ( Top = 0.134 fish · run - 1 , Base = 0.024 fish · run - 1 , Figure 3 . 5 ; Linear mixed model; f (1,10) = 10.01, p = 0.01). Moreover, juvenile lake trout were observed around the base of boulders at only six of 10 boulder sites while at least one juvenile lake trout was observed on the top of every boulder sampled ( Figure 3 . 5 ). Over the 10 sites sampled, 10 juvenile lake trout were observed (six caught) over 396 discrete electrofishing runs around the base of boulders and 45 juv enile lake trout were observed (19 caught) over 321 runs on the tops of boulders. Four of the six juvenile lake trout caught around the base of boulders and 11 of the 19 caught on the top of boulders were free embryos and the re maining were alevins. Of the 10 85 juvenile lake trout observed at the base of boulders, six were from undercut habitats, three were in habitats adjacent to boulders, and one was observed responding to the electrical field ~2 - 3 m away from a boulder. Juvenile lake trout were primarily o bserved in gravel (0.2 - 6.4 cm) and rubble (6.5 - 25.6 cm) substrates, including areas where interstitial depth was minimal, but no discernable preference for substrate size was identified. 86 Table 3 . 2 : Weekly relative abundance (CPUE: fish · trap - 1 · d - 1 ) of lake trout free embryos and alevins caught in three habitat classes using custom rectangular traps at Boulder Alley ( BLD ) and traditional round and custom rectangular traps on more typical habitats on Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ) from May 28 June 24, 2015. Habitat classes for BLD included: habitats beneath overhanging edges of boulders ( beneath boulder), habitats adjacent to a boulder ( beside boulder), and nearby h abitats at least 2x the diameter of a boulder away ( away from boulder). N indicates the weekly number of traps of each type deployed at each site. CV represents the coefficient of variation (%) for comparing weekly variability among reefs. 87 Figure 3 . 5 : Relative abundance (CPUE: # of fish · run - 1 ) of lake trout free embryos and alevins caught via SeaRay ROV modified for electroshocker sampling in boulder - associated habitats at Boulder Alley ( BLD ) and typical ha bitats with the most activity at Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ). CPUE was calculated as the number of fish observed on recorded video represent ed by single letters ( A - M ) and include fish collected around the base ( dark shading ) and top ( medium shading ) of boulders, while single sampling events were conducted in 20 m x 20 m cells on each HSR and SCS ( light shading ). 88 Discussion In this study, widespread egg deposition by lake trout was observed in boulder - associated habitats, in consistent with the commonly - accepted conceptual model of habitat suitability. Notably, eggs were observed in substrates at the bases of all 40 boulders s urveyed in this study. Egg deposition occurred exclusively in association with overhanging edges of the boulders and occurred at all 86 discrete overhangs identified. To my knowledge, this observation is the first report of widespread spawning by lake trou t in substrates associated with overhanging edges of boulders >1 m in diameter. Further, this observation is also the first report of free embryos and alevins captured from the tops of large boulders. Possibly, boulder - associated spawning is common in lake trout, but had been previously overlooked due to non - conformity of this habitat type to the commonly - - al., 2016, 1995a; Muir et al., 2012a). Alternatively, boulder - associated spawning may be a local adaptation shown only by lake trout at the Drummond Island reefs and shoals in northe rn Lake Huron, similar to reports of atypical spawning habitat use in other populations (Peck, 1986; Beauchamp et al., 1992; Marsden et al., 1995b). Regard less, over the 40 boulders surveyed, the mean area of available substrate associated with overhanging edges was 1.8 m 2 per boulder (95% CI = [1.5, 2.3]). Extrapolated to the 269 potential boulders identified by bathymetric analysis, I estimated the existen ce of about 488 m 2 (95% CI = [403.5, 618.7]) of potential boulder - associated spawning habitat in the 0.63 km 2 study area at BLD . However, because many of the potential boulders identified from bathymetric analysis were in fact collections of multiple bould ers in close proximity, this estimate was biased low. Nonetheless, the above observations suggest a potentially important areal contribution to habitat being successfully used fo r spawning 89 by lake trout not currently accounted for by the commonly - accepted conceptual model of habitat suitability. Although widespread egg deposition was observed in boulder - associated habitats, no consistent pattern in the diameter or interstitial depth of the substrates in which eggs were deposited was discernable. Habitats s elected for egg deposition varied in substrate size and interstitial depth, including gravel substrates with minimal interstitial depth that appeared to be unsuitable for successful egg incubation. Furthermore, none of the surveyed areas that were away fro m a boulder received egg deposition, despite having similar substrate composition to habitats beneath overhanging edges that were selected for egg deposition. Large boulders may serve as physical features for aggregating lake trout in spawning condition si milar to the role proposed for steep contours in bathymetry (Marsden & Krueger, 1991; Marsden et al., 1995a). However, the observation that egg deposition at BLD only occurred in close proximity to distinct overhanging edges suggests that these undercut fe atures were critical for the selection of boulder - associated habitats for spawning. While the characteristics that makes boulders attractive to spawning lake trout remain uncertain, I hypothesize that the selection of boulder - associated habitats for spawn ing was most likely driven by flow dynamics present in these habitats rather than the specific physical characteristics of the boulder or its associ ated substrates. Further, I posit that the undercut spaces beneath boulders were most likely formed through mechanical erosion resulting from the interaction of lake currents and the boulders, possibly in association with storm events and surface waves. Though the flow characteristics of boulder - associated habitats were not directly measured in this study, diver s reported feeling stronger currents while sampling at BLD than on adjacent reefs and marked c oncrete blocks (20 cm x 20 c m x 41 c m; ~ 14 kg) that were placed on 90 top of boulders for site identification were often swept off the tops of boulders after storm events. Based on the Bernoulli principle and its corollary the Venturi effect, lake currents would be accelerated as they are forced under and around boulders, potentially eroding under the boulder and cleaning fine sediments from the associated substrates over time (Thibodeaux & Boyle, 1987; Vogel, 1994). Local circulations around obstructions, such as differing bed topography in stream channels, can induce hyporheic exchange (Buffington & Tonina, 2009) which also may be attractive to spawning lake trout ( Farha, Ch. 1, this volume ). Furthermore, these same flow dynamics probably prevent siltation and help oxygenate developing embryos during incubation, and therefore satisfy two common requirements for all salmonid spawning habitats (Gunn & Keller, 1984; Cha pman, 1988; Sly, 1988). Successful incubation of wild - spawned embryos to the free embryo and alevin stages was confirmed in boulder - associated habitats using two complimentary methods. However, despite the appearance of similar egg densities in the substrate, the abundance of ear ly life stages in boulder - associated habitats was low compared to those in habitats at HSR and SCS . This result could be explained at least partially by the inability of the sampling techniques I used to access complex boulder - associated habitats as effici ently as more typical spawning habitats. However, given the combination of current acceleration and small gravel - rubble substrates with minimal interstitial space, eggs deposited in boulder - associated habitats may also experience higher rates of dislodgeme nt and predation ( e.g., Claramunt et al., 2005; Marsden & Krueger, 1991). Small - bodied interstitial predators such as crayfish ( Cambrus spp. and Orconectes spp.) , round goby ( Neogobius melanostomus ), and sculpin ( Cottus cognattus and Cottus bairdi ) appeare d to be abundant in boulder - associated habitats and are known to prey on early life stages of lake trout (Jonas et al., 2005; F itzsimons et al., 2007). In early November , stationary underwater cameras 91 used to monitor lake trout spawning behaviors around bo ulders record ed bufflehead ducks ( Bucephala albeola ) diving and feeding on unprotected eggs lying atop gravel substrates with minimal interstitial depth beneath overhanging edges of boulders in 10 m of water depth. Alternatively, the productivity of boulde r - associated habitats may be higher than I estimated due to high rates of emigration of exposed free embryos and alevins that quickly move from these habitats to find refuge from predation. These observations of both free embryos and alevins caught in habi tats where eggs had not been observed previously, such as on top of boulders and in habitats away from boulders, suggests that these early life stages may have the capability to move more than previously thought. Spawning in association with boulders was consistent with recent linkages between lake trout spawning habitats and glacially - derived bedforms, such as drumlins, eskers, and moraines (Riley et al., 2014, 2017). The boulders described herein appear to have been most dens ely located in a 300m wide by 500m long band oriented in a northeast - southwest direction, consistent with the adjacent submerged drumlin field described by Riley et al. (2014). These boulders were likely to have been deposited along with fine glacial till following a junction line bet ween lobes of the Laurentide Ice Sheet during the Wisconsinan glaciation (Karrow, 1987; Eyles & Doughty, 2016). Likely, additional boulders were deposited throughout the drumlin fields at Drummond Island and west following along the mainland Upper Peninsul a of Michigan, including the Les Cheneaux Islands (Eschman & Mickelson, 1986; Karrow, 1987). Indeed, lake trout eggs were observed in gravel substrate (0.5 - 3 cm diameter) at the base of a single boulder (1.5 m diameter) located near the center of SCS dur ing surveys related to a separate study (See Farha, Ch. 1, this volume ). If boulder fields exist throughout the Great Lakes and spawning in 92 boulder - associated habitats is widespread, these habitats may make an important contribution to total lake trout rec ruitment, even if net production per boulder of juvenile lake trout is low. In this study, the use of acoustic telemetry to inform sampling led to the identification of spawning habitats not previously included in the commonly - accepted conceptual model of spawning habitat suitability. Furthermore, the discovery of widespread boulder - associated spawning suggested that flow dynamics was likely to be an important variable governing spawning habitat selection by lake trout. The implications of these findings to management of lake trout are at least three - fold. First, an accurate accounting of the total amount of spawning habitat available to recovering populations allows for better understanding of variability in recruitment, and thus, the speed of a given po pulations recovery. Combined with population modelling, this information could be used to adjust stocking numbers to increase the likelihood of successful rehabilitation efforts and minimize wasted effort. Second, identification of the critical features of lake trout spawning habitat could be used to identify prime habitats elsewhere for protection or restoration, as well as help specify construction features to incorporate into new artificial habitats (e.g., Marsden et al., 2016) . Finally, where lake trout are an invasive species and have caused negative interactions with native salmonids, this information could be helpful in designi ng control efforts in areas that include boulder - associated habitats (e.g., Crossman, 1995; Hansen et al., 2008). Based on the potential contribution of boulder - associated habitats to total recruitment in the Great Lakes, I advocate for high resolution bathymetric mapping to determine locations of boulder fields and for additional comparisons of juvenile lake trout production and relative incubation success of boulder - associated and traditional habitats. Moreover, to further improve conceptual understanding of lake trout spawning habitat selection, future studies seeking to identify critical habitat requirements should focus on ch aracterizing fine - scale flow 93 regimes within and immediately adjacent to substrates in both traditional and boulder - associated habitats. 94 Acknowledgments This work was largely possible due to the technical expertise and dedication of H. Thompson and C. Wright, who provided crucial dive support in less than ideal conditions throughout the study. Collections via ROV were possible thanks to the piloting skills of Dylan Olson. Essential field and vessel support was provided by Z. Wickert, E. Larson, B. La moreux, and H. Mcmath. This work was funded by the Great Lakes Fishery Commission (Grant ID #2013_BIN_44024) by way of Great Lakes Restoration Initiative appropriations (Grant ID #GL - 00E23010). This paper is c ontribution X of the Great Lakes Acoustic Telemetry Observation System (GLATOS). All sampling and ha ndling of fish during research we re carried out in accordance with guidelines for the care and use of fishes by the American Fisheries Society ( http://fisheries.org/docs /wp/Guidelines - for - Use - of - Fishes.pdf ) , the U.S. Geological Survey ( https://pubs.er.usgs.gov/publication/2001108 ) and the MSU Institutional Animal Care and Use Committee (MSU IACUC; See AUF exemption). The findings and conclusions in this article are those of the authors and do not necessarily represent those of the U.S. Geological Survey or the U . S . Fish and Wildlife Service. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. government. 95 REFERENCES 96 REFERENCES Bates, D., Mächler, M., Bolker, B., and Walker, S., 2015. Fitting linear mixed - effects models using lme4. J. Stat. S oftw. 67, 1 48. doi: 10.18637/jss.v067.i01 Beauchamp, D.A., Allen, B.C., Richards, R.C., Wurtsbaugh, W.A., and Goldman, C.R., 1992. Lake trout spawning in Lake Tahoe: egg incubation in deepwater macrophyte beds. N. Am. J. Fish. Manag. 12, 442 449. doi:10.1577/1548 - 8675(1992)012<0442:LTSILT>2.3.CO;2. Binder, T.R., Riley, S.C., Holbrook, C.M., Hansen, M.J., Bergstedt, R.A., Bronte, C.R., He, J., and Krueger, C.C., 2016. Spawning site fidelity of wild and hatchery lake trout ( Salvelinus namaycush ) in northern Lake Huron. Can. J. Fish. Aquat. Sci. 73, 18 34. doi: 10.1139/cjfas - 2015 - 0175 Binder, T.R., Farha, S.A., Thompson, H.T., Holbrook, C.M., Bergstedt, R.A., Riley, S.C., Bronte, C.R., He, J., and Krueger, C.C., 2018. Fine - scale acoustic telemetry reveals unexpected lake trout, Salvelinus namaycush , spawn ing habitats in northern Lake Huron, North America. Ecol. Freshw. F ish 27, 594 605. doi: 10.1111/eff.12373 Buffington, J.M., and Tonina, D., 2009. Hyporheic e xchange in mountain rivers II: e ffects of channel morphology on mechanics, scales, and rates of exc hange. Geogr. Compass 3, 1038 1062. doi: 10.1111/j.1749 - 8198.2009.00225.x Callaghan, D.T., Blanchfield, P.J., and Cott, P.A., 2016. Lake trout ( Salvelinus namaycush ) spawni ng habitat in a northern lake: t he role of wind and physical characteristics on habit at quality. J. Gt. Lakes Res. 42, 299 307. doi: 10.1016/j.jglr.2015.07.001 Chapman, D., 1988. Critical review of variables used to define effects of fines in redds of large salmonids. Trans. Am. Fish. Soc. 117, 1 21. Claramunt, R.M., Jonas, J.L., Fitzsimons , J.D., and Marsden, J.E., 2005. Influences of spawning habitat characteristics and interstitial predators on lake trout egg deposition and mortality. Trans. Am. Fish. Soc. 134, 1048 1057. doi: 10.1577/T04 - 075.1 Cleland, C.E., 1982. The inland shore fishery of the nor thern Great Lakes: it s development and importance in prehistory. Am. Antiq. 47, 761 784. doi: 10.2307/280281 Coble, D.W., Bruesewitz, R.E., Fratt, T.W., and Scheirer, J.W., 1990. Lake trout, sea lampreys, and overfishing in the upper Great Lakes: a review and reanalysis. Trans. Am. Fish. Soc. 119, 985 995. Available from http://www.tandfonline.com/doi/abs/10.1577/1548 - 8659(1990)119%3C0985%3ALTSLAO%3E2.3.CO%3B2 [accessed 31 October 2014]. Crossman, E.J., 1995. Introduction of the lake trout ( Salvel inus namaycush ) in areas outside its native distribution: a review. J. Gt. Lakes Res. 21, Supple. 1, 17 29. doi: 10.1016/S0380 - 1330(95)71081 - 4 97 Ebener, M.P., 1998. A lake trout rehabilita tion guide for Lake Huron. Miscellaneous p ublications, G reat Lakes Fish ery Commission . Ann Arbor, MI. (1998): 1 44. Available from http://www.glfc.org/pubs/misc/LkTroutRepHuron.pdf [accessed 13 October 2013]. Edsall, T.A., and Kennedy, G.W., 1995. Availability of lake trout reproductive habitat in the Great Lakes. J. Gt. Lakes Res. 21, Supple. 1, 290 301. doi: 10.1016/S0380 - 1330(95)71103 - 0 Ellrott, B.J., and Marsden, J.E., 2004. Lake trout reproduction in Lake Champlain. Trans. A m. Fish. Soc. 133, 252 264. doi: 10.1577/02 - 165 Eschman, D.F., and Mickelson, D.M., 1986. Correlation of glacial deposits of the Huron, Lake Michigan and Green Bay lobes in Michigan and Wisconsin. Quat. Sci. Rev. 5, 53 57. Eshenroder, R.L., 1992. Decline of lake trout in Lake Huron. Trans. Am. Fish. Soc. 121, 548 554. Available from http://www.tandfonline.com/doi/pdf/10.1577/1548 - 8659 - 121.4.548 [accessed 31 October 2014]. Eshenroder, R.L., Poe, T.P., and Olver, C.H., 1984. Strategies for rehabilitation of la ke trout in the Great Lakes: p roceedings of conference on lake trout research, August 1983. Great Lakes Fishery Commission . Ann Arbor, MI . Eshenroder, R.L., Robert Payne, N., Johnson, J.E., Bowen II, C., and Ebener, M.P., 1995. Lake trout rehabilitation in Lake Huron. J. Gt. Lakes Res. 21, Supple. 1, 108 127. doi: 10.1016/S0380 - 1330(95)71086 - 3 Eshenroder, R.L., Peck, J.W., and Olver, C.H., 1999. Research priorities for lake trout rehabilitation in the Great Lakes: a 15 - year retrospective. Tech nical Report, G reat Lakes Fishery Commission. Ann Arbor, MI. 64: 1 40. Eyles, N., and Doughty, M., 2016. Glacially - streamlined hard and soft beds of t he paleo - Ontario ice stream in s outhern Ontario and New York state. Sediment. Geol. 338, 51 71. doi: 10.1016/j.sedgeo.2016.01.019 Farha, S.A., This volume. Chapter 1: Evaluation of lake trout habitat selection at Drummond Island spawning reefs. thesis. Michigan State University. Fitzsimons, J.D., Jonas, J.L ., Claramunt, R.M., Williston, B., Williston, G., Marsden, J.E., Ellrott, B.J., and Honeyfield, D.C., 2007. Influence of egg predation and physical disturbance on lake trout, Salvelinus namaycush , egg mortality and implications for life - history theory. J. Fish Biol. 71, 1 16. doi: 10.1111/j.1095 - 8649.2007.01437.x Fitzsimons, J.D., Brown, S., Brown, L., Honeyfield, D., He, J., and Johnson, J.E., 2010. Increase in lake trout reproduction in Lake Huron following the collapse of alewife: relief from thiamine def iciency or larval predation? Aquat. Ecosyst. Health Manag. 13, 73 84. doi: 10.1080/14634980903581252 98 Gunn, J.M., and Keller, W., 1984. Spawning site water chemistry and lake trout ( Salvelinus namaycush ) sac fry survival during spring snowmelt. Can. J. Fish. Aquat. Sci. 41, 319 329. Available from http://www.nrcresearchpress.com/doi/abs/10.1139/f84 - 036 [accessed 29 October 2014]. Hansen, M.J., 1999. Lake trout in the Great Lakes: basin - wide stock coll apse and binational restoration. Pages 417 454 i n Great Lak es Fisheries Policy and Management. Edited by W . W . Taylor and C . P . Ferreri. Michigan State University Press, East Lansing, MI. Hansen, M.J., Horner, N.J., Liter, M., Peterson, M.P., and Maiolie, M.A., 2008. Dynamics of an increasing lake trout population i n Lake Pend Oreille, Idaho. N. Am. J. Fish. Manag. 28, 1160 1171. doi: 10.1577/M07 - 149.1 Hansen, M.J., and C.R. Bronte. In press. Restoration of Lake Trout in Lake Superior through interagency cooperation. P ages XXX - XXX in From catastrophe to recovery: stories of fishery management success. Edited by C.C. Krueger, W .W. Taylor, and S. Youn . American Fisheries Society, Bethesda, MD . Hanson, S.D., Holey, M.E., Treska, T.J., Bronte, C.R., and Eggebraaten, T.H., 2013. Evidence of wild juvenile lake trout recruitment in western Lake Michigan. N. Am. J. Fish. Manag. 33, 186 191. doi: 10.1080/02755947.2012.754804 He, J.X., Ebener, M.P., Riley, S.C., Cottrill, A., Kowalski, A., Koproski, S., Mohr, L., and Johnson, J.E. , 2012. Lake trout status in the main basin of Lake Huron, 1973 2010. N. Am. J. Fish. Manag. 32, 402 412. doi: 10.1080/02755947.2012.675947 Janssen, J., Jude, D.J., Edsall, T.A., Paddock, R.W., Wattrus, N., Toneys, M., and McKee, P., 2006. Evidence of lake - lake reef complex. J. Gt. Lakes Res. 32, 749 763. doi: 10.3394/0380 - 1330(2006)32[749:EOLTRA]2.0.CO;2 Johnson, J.E., He, J.X., and Fielder, D.G., 2015. Rehabilitation stocki ng of walleyes and lake trout: r estoration of reproducing stocks in Mich igan waters of Lake Huron. N. Am. J. Aquac. 77, 396 408. doi: 10.1080/15222055.2014.993488 Johnson, N.S., Higgs, D.M., Binder, T.R., Marsden, J.E., Buchinger, T.J., Brege, L., Bruning, T., Farha, S.A., and Krueger, C.C., 2018 . E vidence of sound production by spawning lake trout ( Salvelinus namaycush ) in lakes Huron and Champlain. Can. J. Fish. Aquat. Sci. 75(3), 429 - 438, doi: 10.1139/cjfas - 2016 - 0511 Jonas, J.L., Claramunt, R.M., Fitzsimons, J.D., Marsden, J.E., and Ellrott, B.J., 2005. Estimates of egg deposition and effects of lake trout ( Salvelinus namaycush ) egg predators in three regions of the Great Lakes. Can. J. Fish. Aquat. Sci. 62, 2254 2264. doi: 10.1139/f05 - 141 Jones, N.E., Parna, M., Parna, S., and Chong, S., in press. E vidence of lake trout ( Salvelinus namaycush ) spawning and spawning habitat use in the Dog River, Lake Superior. J. Gt. Lakes Res. 2018. doi: 10.1016/j.jglr.2018.07.017 99 Karrow, P., 1987. Glacial and glaciolacustrine events in northwestern Lake Huron, Michiga n and Ontario. Geol. Soc. Am. Bull. 98, 113 120. Krueger, C.C., Perkins, D.L., Mills, E.L., and Marsden, J. E. , 1995. Predation by alewives on lake trout fry in Lake Ontario: role of an exotic species in preventing restoration of a native species. J. Gt. La kes Res. 21, Supple. 1, 458 469. doi: 10.1016/S0380 - 1330(95)71117 - 0 Krueger, C.C., and Ebener, M. , 2004. Rehabilitation of lake trout in the Great Lakes: past lessons and future challenges. Pages 37 - 56 i n Bor eal Shield Watersheds: lake trout ecosystems in a changing e nvironment. Edited by J.M. Gunn, R.J. Steedman, and R.A. Ryder . CRC Press, Boca Raton, FL . pp. 37 56. Version 2. Madenjian, C.P., Ebener, M.P., and Desorci e, T.J., 2008a. Lake trout population dynamics at Drummond Island Refuge in Lake Huron: implications for future rehabilitation. N . Am. J. Fish. Manag. 28(1), 979 992. doi: 10.1577/M07 - 083.1 Madenjian, C.P., Stockwell, J.D., and Stapanian, M.A. , 2008b. Adverse effects of a lew ives on Laurentian Great Lakes fish c ommunities. N. Am. J. Fish. Manag. 28(1): 263 282. doi: 10.1577/M07 - 012.1 . Marsden, J.E., Krueger, C.C., and Schneider, C.P., 1988. Evidence of natural reproduction by stocked lake trout in Lake Ontario. J. Gt. Lakes Res. 14, 3 8. doi: 10.1016/S0380 - 1330(88)71526 - 9 Marsden, J.E., and Kr ueger, C.C., 1991. Spawning by hatchery - origin lake trout ( Salvelinus namaycush ) in Lake Ontario: data from egg collections, substrate analysis, and diver observations. Can. J. Fish. Aquat. Sci. 48, 2377 2384. doi:10.1139/f91 - 278. Marsden, J.E., Casselman, J.M., Edsall, T.A., Elliott, R.F., Fitzsimons, J.D., Horns, W.H., Manny, B.A., McAughey, S.C., Sly, P.G., and Swanson, B.L., 1995a. Lake trout spawning habitat in the Great Lakes a review of current knowledge. J. Gt. Lakes Res. 21, Supple. 1, 487 497. d oi: 10.1016/S0380 - 1330(95)71120 - 0 Marsden, J.E., Perkins, D.L., and Krueger, C.C., 1995b. Recognition of spawning areas by lake trout: deposition and survival of eggs on small, man - made rock piles. J. Gt. Lakes Res. 21, Supple. 1, 330 336. doi: 10.1016/S0380 - 1330(95)71107 - 8 Marsden, J.E., and Chotkowski, M.A., 2001. Lake trout spawning on artificial reefs and the effect of zebra mussels: fatal attraction? J. Gt. Lakes Res. 27, 33 43. Marsden, J.E., Binder, T.R., Johnson, J., He, J., Dingledine, N ., Adams, J., Johnson, N.S., Buchinger, T.J., and Krueger, C.C., 2016. Five - year evaluation of habitat remediati on in Thunder Bay, Lake Huron: c omparison of constructed reef characteristics that attract spawning lake trout. Fish. Res. 183, 275 286. 100 Marsden , J.E., Muir, A.M., and Krueger, C.C., In review . Terminology issues in lake charr early development. Pages XXX - XXX in Lake c harr Salvelinus namaycush : Biology, Ecology, Distribution, and Management. Edited by A.M. Muir, C.C. Krueger, M.J. Hansen, Mitchell editors. Springer Fish & Fisheries Series Series Ed.: D. Noakes . Muir, A.M., Blackie, C.T., Marsden, J.E., and Krueger, C.C., 2012a. Lake charr , Salvelinus namaycush , spawning behaviour: new field observations and a review of current knowledge. Rev. Fish Biol. Fish. 22, 575 593. doi: 10.1007/s11160 - 012 - 9258 - 6 Muir, A.M ., Krueger, C.C., and Hansen, M.J. , 2012b. Re - establishing lake trout in the Laurentian Great La kes: past, present, and future. Pages 533 588 i n Great Lakes fishery policy and management: a binational perspective, 2 nd edition. Edited by W.W. Taylor. Michigan State University Press, East Lansing. Available from https://www.researchgate.net/profile/Michael_Hansen18/publication/268207669_Re - establishing_lake_trout_in_the_Laurentian _Great_Lakes_past_present_and_future/links/ 5463aa9c0cf2cb7e9da997fb.pdf [accessed 9 March 2017]. Olson, D.S., and Janssen, J., 2017. Early feeding of round goby ( Neogobius melanostomus ) fry. J. Gt. Lakes Res. 43(4), pp.728 - 736. doi: 10.1016/j.jglr.2017.04.0 06 Peck, J.W., 1986. Dynamics of reproduction by hatchery lake trout on a man - made spawning reef. J. Gt. Lakes Res. 12, 293 303. R Core Team, 2015. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3 - 900051 - 07 - 0. Rasband, W. , 1997. ImageJ software. US National Institute of Health , Bethesda, MD, USA . and Schaeffer, J.S., 2007. Evidence of widespread natural reproduction by lake trout, Salvelinus namaycush , in the Michigan waters of Lake Huron. J. Gt. Lakes Res. 33, 917 921. doi: 10.3394/0380 - 1330(2007)33[917:EOWNRB]2.0.CO;2 Riley, S.C., Rinchard, J., Honeyfield, D.C., Evans, A.N., and Begnoche, L., 2011. Increasing thiamine concentration s in lake trout eggs from Lakes Huron and Michigan coincide with low alewife abundan ce. N. Am. J. Fish. Manag. 31, 1052 1064. doi: 10.1080/02755947.2011.641066 Riley, S.C., Binder, T.R., Wattrus, N.J., Faust, M.D., Janssen, J., Menzies, J., Marsden, J.E., E bener, M.P., Bronte, C.R., He, J.X., Tucker, T.R., Hansen, M.J., Thompson, H.T., Muir, A.M., and Krueger, C.C., 2014. Lake trout in northern Lake Huron spawn on submerged drumlins. J. Gt. Lakes Res. 40, 415 420. doi: 10.1016/j.jglr.2014.03.011 Riley, S.C., Binder, T.R., Tucker, T.R., Menzies, J., Eyles, N., Janssen, J., Muir, A.M., Esselman, P.C., Wattrus, N.J., and Krueger, C.C., 2017. Islands in the ice stream: were spawning habitats for native salmonids in the Great Lakes created by paleo - ice streams? Fis h. 18, 347 359. doi: 10.1111/faf.12173 101 Schindelin, J., Arganda - Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch, T., Preibisch, S., Rueden, C., Saalfeld, S., Schmid, B., and Tinevez, J.Y. , 2012. Fiji: an open - source platform for biological - image a nalysis. Nat. Methods 9(7): 676 682. Selgeby, J.H., 1995. Introduction to the proceedings of the 1994 international conference on restoration of lake trout in the Laurentian Great Lakes. J. Gt. Lakes Res. 21, Supple. 1, 1 2. doi: 10.1016/S0380 - 1330(95)71078 - 4 Simard, L.G., 2017. Spawning site selection and fry development of invasive Lake Trout in Yellowstone Lake, Yellowsto ne National Park, Wyoming. PhD Thesis . The University of Vermont and State Agricultural College. Sly, P.G., 1 988. Interstitial water quality of lake trout spawning habitat. J. Gt. Lakes Res. 14, 301 315. doi: 10.1016/S0380 - 1330(88)71560 - 9 Smith, S.H., 1968. Species succession and fishery exploitation in the Great Lakes. J. Fish. Board Can. 25, 667 693. Stauffer, T .M., 1981. Collecting gear for lake trout eggs and fry. Progress. Fish - Cult. 43, 186 193. doi: 10.1577/1548 - 8659(1981)43[186:CGFLTE]2.0.CO;2 Thibodeau, M., and Kelso, J., 1990. An evaluation of putative lake trout ( Salvelinus namaycush ) spawning sites in th e Great Lakes. Can. Tech. Rep. Fish. Aquat. Sci. 1739. Thibodeaux, L.J., and Boyle, J.D., 1987. Bedform - generated convective transport in bottom sediment. Nature 325 (6102): 341. Vogel, S., 1994. Life in moving fluids: the physical biology of flow. Princeto n University Press. Boston, MA. Wickham, H., 2015. plyr: Tools for splitting, applying and combini ng data. R package version 1.8. 1. R Foundation for Statistical Computing, Vienna, Austria 102 Chapter 4 : C onclusion From 2013 201 4 , lake trout spawning in the Drummond Island refuge deposited eggs in a wide variety of habitats, including several habitats with physical characteristics that did not match those described in the commonly - accepted - conceptual model of spawning habit at suitability. In chapter one , habitats in which lake trout deposited eggs on two adjacent reefs were described. Slope magnitude was found to be an important predictor of egg deposition at one reef, but not the other. Differences in substrate composition did not differentiate betwe en habitats positive for egg deposition and those negative on either reef. Further, a few of the habitats selected for egg deposition varied between years on one reef but remained consistent between years on the other. In chapter two , widespread egg deposi tion was detected in association with boulder - associated habitats at a third location within the Drummond Island refuge and this site differed substantially in physical characteristics from the most highly used habitats at the other two reef s examined in t his thesis . Collectively, these observations of variable spawning habitat use at several locations in close proximity (< 3 km apart) suggest ed that a wide variety of habitats fit the requirements for successful spawn ing, and therefore, lake trout we re not highly specific in terms of substrate requirements for spawning habitat . Additionally , the variables used to select spawning habitats can vary both temporally and spatially. Common characteristics among all sites was the use of clean stony substrates and t he likelihood that water current was an important variable. These observations are consistent with previous studies that suggest lake trout are highly adaptive to shifting conditions in both their native range and areas where they have been introduced (Eva ns and Olver , 1995; McAughey and Gunn , 1995; Simard , 2017) . 103 Based on three of the most - commonly referenced physical characteristics in the cobble - contour model of lake trout spawning habitat suitability, nearly all habitats characterized in this study were outside of the intersection o f ranges thought to be suitable ( Figure 4 . 1 ) . In general, habitats surveyed at Drummond Island were associated with substrates that were smaller in diameter and had less interstitial depth than those generally considered appropriate for lake trout spawning, particularly in the Great Lakes ( Figure 4 . 1 ) . - model of habitat suitability, as currently described, did not sufficiently encompass the range of habitats bei ng successfully used by lake trout spawning in the DIR. Further, the physical habitat characteristics measured in this study did little to advance and modify the conceptual model of spawning habitat suitability, and therefore, other unmeasured, possibly co rrelated variables, are likely involved in habitat selection. Therefore, I hypothesize that water flow is likely to be a primary variable governing spawning habitat selection. Moreover, because optimum conditions for egg incubation are variable spatially b oth within and between reefs and temporally season to season (e.g., Fitzsimons and Marsden , 2014; Callaghan et al. , 2016) , physical habitat characteristics for spawning likely also vary in both space and time, such that nearly all combinations of stony sub strates, slope, and interstitial depth can be suitable for egg deposition, depending on the site - specific flow characteristics. For example, i n conditions with a predictable strong current that may damage incubating eggs, a sharp break in bathymetry with l arge substrate and deep interstitial spaces may be required to buffer the physical forces involved to a level suitable for incubation. In contrast, small substrates with minimal interstitial depth may also be suitable for incubation in areas where currents are less likely to damage eggs, such as around the base of boulders in water deeper than waves and ice penetrate. 104 Interstitial flow is a common requirement of all salmonid spawning habitats (Chapman , 1988) . Currents flowing interstitially benefit egg inc ubation by preventing suffocation of eggs through siltation while ensuring eggs remain well - oxygenated (Gunn and Keller , 1984; Sly , 1988) . Consistent water currents are critical in maintaining suitable interstitial water quality for successful development of early life stages and currents may also transport emergent lake trout from spawning to nursery habitats (Bronte et al. , 1995; Casselma n , 1995) . In chapter one , four sites near the shallowest ridge of the reef switched from positive for egg deposition in 2013 to negative in 2014, while four sites on the outside boundary of the reef switched from negative in 2013 to positive in 2014 ( Figure 2 . 1 ). This variability in habitats selected for egg deposition on HSR between years may be explained by an abrupt change in water levels. The water level in northern Lake Huron fluctuated from 0.43 m below the long - term average for October (1918 2012; 176.43 m above sea level) in 2013 to 0.10 m above the long - term average for October in 2014. Possibly, this change in water levels altered the interaction of lake curr ents and the slope present on HSR , shifting the flow dynamics such that some sites on the shallowest central portion of the reef that were previously attractive to lake trout for spawning became less attractive, and others around the outer edges of the ree f, where prominent slopes existed, became more attractive. In contrast, egg deposition on SCS was consistent between years and did not appear to be influenced by possible changes in flow conditions due to differing water levels, likely due to the limited s lope present on that reef. Collectively, these observations h ighlight the importance of flow, its possible interaction with bathymetric features, and their collective influence on spawning habitat selection. While the current - of lake trout spawning habitat suitability has been instrumental in identifying habitats important to lake trout rehabilitation in the Great Lakes, 105 an accurate accounting of the quality and quantity of habitat available is critical to understanding the limita tions affecting the speed of population restoration (Marsden et al. , 1995; Bronte et al. , 2007; Muir et al. , 2012) . Therefore, further improvement of the model is a key component of ensuring the success of rehabilitation efforts and could also be useful in developing methods to control populations in systems where lake trout are invasive . As currently defined, at least two limitations to advancing the conceptual model of lake trout spawning habitat suitability exist and should be addressed in future studies of lake trout spawning habitat selection. The next two paragraphs will explore these limitations in turn. First, I propose that an overemphasis on substrate size has occurred in the past. Historically, much of the work characterizing lake trout spawning habitats has focused on substrate size and shape, but the size of substrates in which lake trout spawned was often variable, leading to uncertainty regarding its influence on habitat selection (e.g., Dorr III et al. , 1981; Wagner , 1982; Fitzsimons , 1995; S chreiner et al. , 1995) . Substrates, particularly those in the rubble size class (6.5 25.6 cm), were thought to be important for habitat selection based on their ability to buffer potentially damaging lake currents, while excluding egg predators and preve nting dislodgement of early life stages (Gunn , 1995; Marsden et al. , 1995; Sly and Evans , 1996) . However, the characterization of substrate is challenging and often not consistent between studies (Edsall et al. , 1992; Marsden et al. , 1995) . Further, surpri singly little information about structure at the lake bottom exists for the Great Lakes, and that which does exist is generally limited to high traffic near - shore areas and often only describes hardness (i.e., rock vs. sand or silt). High - resolution charac terization is labor - intensive and typically req uires specialized equipment and , therefore, can be cost - prohibitive. As such, researchers often only broadly characterize surficial substrates, and the actual terminology used to describe them frequently 106 diffe rs between studies. Broad classification of substrates typically assumes some degree of uniformity, but egg deposition by lake trout, including that described in this thesis, is generally patchily distributed and the total area used for spawning is commonl y < 10% of the total available habitat on a given reef (e.g., Marsden and Krueger , 1991; Gunn , 1995 ; Kelso et al. , 1995; Binder et al. , 2018) . In reality, the lake bottom is composed of variable patches of an intricate mix o f multiple substrate sizes in 3D space that interact with water currents to provide suitable incubation conditions. Broad characterization focused on the diameter of surficial substrate alone greatly oversimplifies this complex relationship. Moreover, this issue is further complicated be cause the size ranges of substrate classes are not equal (e.g., gravel: 2 - 6.4 cm vs. cobble: 25.7 - 99.9 cm), and therefore habitats are more likely to be classified in the larger size categories, potentially over - inflating the relative - importance of the se sizes. Improvements in the resolution of remote characterization technologies, such as multibeam sonar and 3 - d stereoscopic imaging, along with enhanced processing capabilities, such as machine learning, hold great promise for improving understanding of the 3 - d structure of the lake bottom, and lowering the costs of doing so. Nonetheless, this study suggests that substrate size alone was not informative to predicting egg deposition. Therefore , future efforts to improve understanding of t he role of substrate size on the selection of spawning habitats should focus on the interaction of various size substrates and their influence on flow conditions at fine spatial scales, using these enhanced methodologies as they become more accessible, rat her than the broad characterization of surficial substrate alone. Second, several parameters in the 'cobble - contour' model tend to be correlated and are therefore difficult to assess independent from one another. Substrate size and shape, interstitial dep th, and substrate homogeneity are all attempts to quantify the interstitial volume between 107 substrates and are therefore, correlated. In this study, I observed positive correlation between interstitial depth and mean diameter (Pearson product - moment correla tion; r = 0.78, p < 0.01) and negative correlation between substrate homogeneity and interstitial depth (Pearson product - moment correlation; r = - 0.43, p = 0.01). Moreover , this issue is often further compounded by spatial autocorrelation. In this study, mo del residuals did not appear to be spatially autocorrelated, but the degree of spatial autocorrelation among predictors varied between reefs such that interstitial depth was autocorrelated on one reef but not the other. With little information on the area evaluated by spawning lake trout or the degree to which lake trout spawn in multiple locations in a single season and statistical methods to account for spatial autocorrelation are limited (e.g., Dormann et al. , 2007). Lastly , the degree to which these parameters influence the selection of spawning habitats is likely dependent on site - specific flow conditions and therefore difficult to generalize. As such, future studies should focus on characteriz ation of flow dynamics and their interaction with the complete interstitial volume, rather than on these parameters independently. Successful spawning at Drummond Island does not appear to be limited by availability of suitable incubation habitats or by l ake trout selecting unsuitable habitats for egg deposition. In chapter one , mean in situ incubation success of habitats in which lake trout deposited eggs was estimated to be 73.8% (95% CI = [65.9%, 81.7%]), which was higher than the 10 to 55% reported in five previous studies using the same habitat bio - assay throughout the Great Lakes (Casselman , 1995; Edsall et al. , 1995; Eshenroder et al. , 1995; Manny et al. , 1995; Perkins and Krueger , 1995) . While these estimates did not include all possible sources of mortality, such as predation or dislodgment, habitats at Drummond Island appear to support successful egg incubation, and lake trout appeared to be capable of finding suitable spawning habitat s. 108 Moreover, overlap in incubation success between habitats positive for egg deposition and those that were negative indicated that some habitats not being used may be equal ly suitable for incubation based on interstitial water q u a lity and the physicochemi cal conditions experienced . T herefore, from this perspective suitable habitats did not appe ar to be saturated. In chapter two , natural production of wild - spawned juvenile lake trout was confirmed within boulder - associated habitats using two methods , albeit at lower abundance than the most highly used sites on HSR and SCS . Because boulder - associated substrates were generally smaller in diameter and had minimal interstitial depth, eggs deposited beneath boulders were hypothesized to have experienced greater r ates of predation and dislodgement than eggs deposited on nearby HSR and SCS . Lake trout may diversify egg deposition in differing habitats as a bet - hedging strategy to optimize the probability of success in the face of environmental uncertainty (Fitzsimon s and Marsden , 2014) . Unfortunately, little is known about the extent to which a single fish may evaluate or spawn on multiple reefs in the same season (Pinheiro et al. , 2017) . As such, the precise explanation as to why some habitats were chosen for egg de position while other seemingly suitable habitats were not remains unclear. Nonetheless, recruitment at Drummond Island does not appear to be limited by habitat availability or the selection of inappropriate habitats, and therefore, spawning activity in thi s area and year - class recruitment contribution is likely to increase as the abundance of spawning adults increases. The underlying architecture and feature that distinguishes this study from others was the use of acoustic telemetry to evaluate habitat sel ection from the perspective of the fish. Selecting sampling sites based on the behavior of the fish freed this study from biases associated with the apriori broad characterization of spawning shoals based on the capture of ripe fish. As a result, this stud y evaluate d habitat selection at a scale not previously possible and identified use of 109 habitats not suspected to be important that may have been overlooked otherwise. Perhaps most interesting, not only was egg deposition observed in these habitats, they produce d viable naturally s pawned juvenile lake trout each year, forcing us to rethink, adapt, and expand our conceptual understanding of lake trout spawning habitat suitability. Whether or not spawning obser ved within the DIR represents the new paradigm for recovering populations in Lake Huron or if behaviors observed here were single, potentially inconsistent observations (paradox) is uncertain ; more research is needed . Further investigations should be focused on areas being successfully us ed by spawning lake trout, such as the DIR, using technologies such as high - resolution bathymetric mapping and acoustic telemetry as these methods likely hold the gr eatest promise for advancing conceptual understanding of the mechanisms underlying habitat selection by lake trout. Improvement in understanding the fundamental features of spawning habitats and the criteria by which habitats are selected may benefit the rehabilitation of lake trout in the Great Lakes in at least three ways. First, with b etter knowledge of the features of habitats being successfully used by recovering populations, limitations imposed by the carrying capacity of a system can be better identified, and the ability of recovering fish to select suitable habitats can be better e valuated. Second, improved understanding of the habitat requirements of spawning lake trout may enhance the effectiveness of stocking programs through targeted stocking of early life stages on high - quality substrates for imprinting. Third, a better underst anding of the habitat requirements of lake trout will help to identify prime habitats on which to focus conservation and rehabilitation efforts. Further, remediation of historic spawning reefs or the construction of artificial reefs in areas where habitat is limited can be enhanced to match habitat characteristics that increase the probability of use and therefore, extend population recovery to areas that may 110 not have experienced it otherwis e. Finally, while broadening conceptual understanding of lake trout spawning behavior will support rehabilitation efforts in the species native range, this information also will benefit control efforts where lake trout have become invasive (e.g., Crossman , 1995; Hansen et al. , 2008; Martinez et al. , 2009) , by identifying potential habitats that previously may have been deemed unsuitable . 111 Figure 4 . 1 : C onceptual diagram of primary physical characteristics, slope magnitude (%), interstitial depth (cm), and substrate diameter (cm), of habitats selected f or egg deposition on Horseshoe R eef ( red dots ), Scammon S hoal ( blue dots ) and Boulder A lley ( green dots ) vs. the - yellow box ). This model states habitats suitable for spawning have substrates between 8 - 100 cm in diameter, with interstitial depth greater than 30cm, and slopes between 8 - 100% (5 - 45°; yellow box ). Nearly all habitats selected for egg deposition at Drummond Island were outside of the intersection of cha racteristics described in the conceptual model. In general, Drummond Island habitats had smaller diameter substrates with less interstitial depth than that described in the model. 112 REFERENCES 113 REFERENCES Binder, T.R., Farha, S.A., Thompson, H.T., Holbrook, C.M., Bergstedt, R.A., Riley, S.C., Bronte, C.R., He, J., and Krueger, C.C. , 2018. Fine - scale acoustic telemetry reveals unexpected lake trout, Salvelinus namaycush , spawning habitats in northern Lake Huron, North America. Ecol. Freshw. Fish 27, 594 605. doi: 10.1111/eff.12373 Bronte, C.R., Holey, M.E., Madenjian, C.P., Jonas, J.L., Claramunt, R.M., McKee, P.C., Toneys, M.L., Ebener, M.P., Breidert, B., Fleischer, G.W ., Hess, R., Martell, A.W., and Olsen, E.J. , 2007. Relative abundance, site fidelity, and survival of adult lake trout in Lake Michigan from 1999 to 2001: implications for futu re restoration strategies. N. Am. J. Fish. Manag. 27, 137. doi:10.1577/M05 - 214.1 . Bronte, C.R., Selgeby, J.H., Saylor, J.H., Miller, G.S., and Foster, N.R. , 1995. Hatching, dispersal, and bathymetric distribution of age - 0 wild lake trout at the Gull Island Shoal complex, Lake Superior. J. Gt. Lakes Res. 21, Supple. 1: 233 245. Callagh an, D.T., Blanchfield, P.J., and Cott, P.A. , 2016. Lake trout ( Salvelinus namaycush ) spawni ng habitat in a northern lake: t he role of wind and physical characteristics on habitat quality. J. Gt. Lakes Res. 42, 299 307. doi:10.1016/j.jglr.2015.07.001. Casse lman, J.M. , 1995. Survival and development of lake trout eggs and fry in eastern Lake Ontario in situ incubation, Yorkshire Bar, 1989 1993. J. Gt. Lakes Res. 21, Supple. 1: 384 399. doi:10.1016/S0380 - 1330(95)71112 - 1. Chapman, D. , 1988. Critical review of variables used to define effects of fines in redds of large salmonids. Trans. Am. Fish. S oc. 117, 1 21. Crossman, E.J. , 1995. Introduction of the lake trout ( Salvelinus namaycush ) in areas outside its native distribution: a review. J. Gt. Lakes Res. 21, S upple. 1: 17 29. doi:10.1016/S0380 - 1330(95)71081 - 4. Dormann, C.F., McPherson, J. M., Araújo, M. B. , Bi vand, R., Bolliger, J., Carl , Davies, R.G. , Hirzel, A., Jetz, W., Kissling, W. D. , Kühn, I., Ohlemüller, R., R. Peres - Neto, P., Reineking, B., Schröder, B., Schurr, F. M. , and Wilson, R. , 2007. Methods to account for spatial autocorrelation in the analysis of species distributional data: a review. Ecography 30, 609 628. doi:10 .1111/j.2007.0906 - 7590.05171.x. Jude, D.J. , 1981. Substrate conditions and abundance of lake trout eggs in a traditional spawning area in s outheastern Lake Michigan. N. Am. J. Fish. Manag. 1, 165 172. Edsall, T.A., Brown, C.L., Kennedy, G.W., and French III, J.R.P. , 1992. Surficial subst rates and bathymetry of five historical lake trout spawning reefs in near - shore waters of the Great Lakes. Tech nical r ep ort, Great Lakes Fishery Commission. Ann Arbor, MI. 58, 1 - 56. 114 Edsall, T.A., Holey, M.E., Manny, B.A., and Kennedy, G.W. , 1995. An evalua tion of lake trout reproductive habitat on Clay Banks Reef, northwestern Lake Michigan. J. Gt. Lakes Res. 21, Supple. 1: 418 432. doi:10.1016/S0380 - 1330(95)71114 - 5. Eshenroder, R.L ., Bronte, C.R., and Peck, J.W., 1995. Comparison of lake trout - egg survival at inshore and offshore and shallow - water and deepwater sites in Lake Superior. J. Gt. Lakes Res. 21, Supple. 1: 313 322. Evans, D.O., and Olver, C.H. , 1995. Introduction of lake trout ( Salvelinus namaycush ) to inland lakes of Ontario, Canada: factors cont ributing to successful colonization. J. Gt. Lakes Res. 21, Supple. 1: 30 53. Fitzsimons, J.D. , 1995. Assessment of lake trout spawning habitat and egg deposition and survival in Lake Ontario. J. Gt. Lakes Res. 21, Supple. 1: 337 347. doi:10.1016/S0380 - 1330 (95)71108 - X. Fitzsimons, J.D., and Marsden, J.E. , 2014. Relationship between lake trout spawning, emb ryonic survival, and currents: a case of bet hedging in the face of environmental stochas ticity? J. Gt. Lakes Res. 40, 92 101. doi:10.1016/j.jglr.2013.12.0 14. Gunn, J.M. , 1995. Spawning behavior of lake trout: effects on colonization ability. J. Gt. Lakes Res. 21, Supple. 1: 323 329. doi:10.1016/S0380 - 1330(95)71106 - 6. Gunn, J.M., and Keller, W. , 1984. Spawning site water chemistry and lake trout ( Salvelinus namaycush ) sac fry survival during spring snowmelt. Can. J. Fish. Aquat. Sci. 41, 319 329. Hansen, M.J., Horner, N.J., Liter, M., Peterson, M.P., and Maiolie, M.A. , 2008. Dynamics of an increasing lake trout population in Lake Pend Oreille, Idaho. N. Am. J. Fish. Manag. 28, 1160 1171. doi:10.1577/M07 - 149.1. Kelso, J.R.M., MacCallum, W.R., and Thibodeau, M.L. , 1995. Lake trout spawning at five sites in Ontario waters of Lake Superior. J. Gt. Lakes Res. 21, Supple. 1: 202 211. doi:10.1016/S0380 - 1330(95) 71092 - 9. Manny, B.A., Edsall, T.A., Peck, J.W., Kennedy, G.W., and Frank, A.M. , 1995. Survival of lake trout eggs on reputed spawning grounds in Lakes Huron and Superior: in situ incubation, 1987 1988. J. Gt. Lakes Res. 21, Supple. 1: 302 312. doi:10.1016/ S0380 - 1330(95)71104 - 2. Marsden, J.E., Casselman, J.M., Edsall, T.A., Elliott, R.F., Fitzsimons, J.D., Horns, W.H., Manny, B.A., McAughey, S.C., Sly, P.G., and Swanson, B.L. , 1995. Lake trout spawning habitat in the Great Lakes a review of current knowled ge. J. Gt. Lakes Res. 21, Supple. 1: 487 497. doi:10.1016/S0380 - 1330(95)71120 - 0. Marsden, J.E., and Krueger, C.C. , 1991. Spawning by hatchery - origin lake trout ( Salvelinus namaycush ) in Lake Ontario: data from egg collections, substrate analysis, and diver observations. C an. J. Fish. Aquat. Sci. 48, 2377 2384. 115 Martinez, P.J., Bigelow, P.E., Deleray, M.A., Fredenberg, W.A., Hansen, B.S., Horner, N.J., Lehr, S.K., Schneidervin, R.W., Tolentino, S.A., and Viola, A.E. , 2009. Western l ake trout woes. Fisheries 3 4, 424 442. doi:10.1577/1548 - 8446 - 34.9.424. McAughey, S.C., and Gunn, J.M. , 1995. The behavioral response of lake trout to a loss of traditional spawning sites. J. Gt. Lakes Res. 21, Supple. 1: 375 383. doi:10.1016/S0380 - 1330(95)71111 - X. Muir, A.M., Blacki e, C.T., Marsden, J.E., and Krueger, C.C. , 2012. Lake charr Salvelinus namaycush spawning behaviour: new field observations and a review of current knowled ge. Rev. Fish Biol. Fish. 22, 575 593. doi:10.1007/s11160 - 012 - 9258 - 6. Perkins, D.L., and K rueger, C.C . , 1995. Dynamics of reproduction by hatchery - origin lake t rout ( Salvelinus namaycush ) at Stony Island Reef, Lake Ontario. J. Gt. Lakes Res. 21, Supple. 1: 400 417. doi:10.1016/S0380 - 1330(95)71113 - 3. Pinheiro, V.M., Stockwell, J.D., and Marsden, J.E. , 2017 . Lake trout ( Salvelinus namaycush ) spawning site use in Lake Champlain. J. Gt . Lakes Res. 43, 345 351. doi:10.1016/j.jglr.2016.12.005. Schreiner, D.R., Bronte, C.R., Robert Payne, N., Fitzsimons, J.D., and Casselman, J.M. , 1995. Use of egg traps to invest igate lake trout spawning in the Great Lakes. J. Gt. Lakes Res. 21, Supple. 1: 433 444. doi:10.1016/S0380 - 1330(95)71115 - 7. Simard, L.G. , 2017. Spawning site selection and fry development of invasive lake trout in Yellowstone Lake, Yellowstone National Park , Wyoming. PhD Thesis, The University of Vermont and State Agricultural College. Sly, P., and Evans, D. , 1996. Suitability of habitat for spawning lake trout. J. Aquat. Ecosyst. Health 5, 153 175. Sly, P.G. , 1988. Interstitial water quality of lake trout s pawning h abitat. J. Gt. Lakes Res. 14, 301 315. doi:10.1016/S0380 - 1330(88)71560 - 9. Wagner, W.C. , 1982. Lake trout spawning habitat in the Great Lakes. Fisheries Research Report, Michigan Dept. of Natural Resources, Fisheries Division, Lansing, MI. Rep. 190 4. 116 APPENDICES 117 APPENDIX A Chapter 2: Supplementary M aterial The following section contains supplementary materials related to analyses presented in Chapter 2 but are not immediately essential to understanding of the main text. As presented, Chapter 2 is intended for submission to the Canadian Journal of Fisheries and Aquatic Sciences, and these materials will not be included in the submission. They are included here as an archival record for the sake of posterity. 118 Table A . 1 : Unstandardized and standardized (denoted by an *) m odel - averaged parameter estimates and relative importance of physical characteristics for the selection of habitats for egg deposition by lake trout on Scammon Shoal ( SCS ). An unconditional variance estimator was used to calculate 95% confidence intervals on model - averaged parameter estimates. Furthermore, the relative importance of each physical characteristic was expressed in two ways; f irst, as a ratio of the i th parameter Akaike weight to the maximum Akaike weight , and second, as an evidence ratio of the i th standardized model - averaged parameter estimate to the maximum standardized parameter estimate Based on these criteria, none of the measured characteristic s were found to be important for the selection of spawning habitats on SCS . 119 Table A . 2 : Complete dataset used for analyses on Horseshoe Reef ( HSR) . Logistic and incubators column s are binary dummy variables indicating whether site data was used (1) or not (0) in the respective analyses presented in Chapter 1. On HSR , site 37 was excluded due to technical issues that prevented complete analysis and all sites not - encountered by lake trout (i.e., 120 Table A . 3 : Complete dataset used for analyses on Scammon Shoal ( SCS ) . Logistic and incubators columns are binary dummy variables indicating whether site data was used (1) or not (0) in the respective analyses presented in Chapter 1. On SCS , all sites not - encountered by lake 121 A PPENDIX B Chapter 3: Supplementary M aterial The following section contains supplementary materials related to analyses presented in Chapter 3 but are not immediately essential to understanding of the main text. As presented, Chapter 3 is intended for submission to the Journal of Great Lakes Research, and these materials will not be included in the submission. They are included here as an archival record for the sake of posterity. 122 Table B . 4 : Complete boulder characterization data for all boulders characterized on Boulder Alley ( BLD ). In total, 40 boulders were surveyed consisting of 86 discrete overhangs. 123 Table B . 5 : Complete substrate characterization data for boulder - associated substrates adjacent to boulders at Boulder Alley ( BLD ) and beneath overhangs vs. the most popular sites on Horseshoe Reef ( HSR ) and Scammon Shoal ( SCS ). Substrate was classified based on the proportion of measured substrate in size bins using a modified Wentworth scale (Marsden et al., 1995a, Farha et al. Ch. 1) . The size bin with the largest proportion of measured substrate was usually considered the predominant substrate size, but sites wit h multiple categories having greater than 33.3% of measured substrates were given a mixed classification . 124 Table B . 6 : Relative survival to hatch for incubators deployed at Boulder Alley ( BLD ) and controls ( CON ) using method ology described in Chapter 1 ( See Chapter 1 Methods Incubation Success (Objective 2 ) for complete details ). This analysis was removed from Chapter 3 for the sake of clarity. Habitat classes were (1) areas of clean gravel (< 6 cm diameter) associated with an overhang , (2) areas of gravel (< 6 cm diam.) 2x boulder diameter away), (3) areas of clean rubble (6 - 25 cm diam.) associated with an overhang , and (4) areas of rubble (6 - 25 cm diam.) not a ssociated with boulder s . At recovery, incubators were classified either as buried in the substrate ( Orientation = 1 ; moved Orientation = 2 ; either tipped over or completely dislodged f rom substrate). 125 Figure B . 1 : Weekly relative abundance (CPUE: fish · trap - 1 · d - 1 ) of lake trout free embryos and alevins caught in three habitat classes using custom rectangular traps at Boulder Alley ( BLD ; Left ) vs. traditional round and custom rectangular traps on typical habitats on Horseshoe Reef ( HSR ; Right ) and Scammon Shoal ( SCS ; Right ) from May 28 June 24, 2015. Habitat classes for BLD included: habitats beneath overhanging edges of boulders ( beneath boulder ), habitats adjacent to a boulder ( beside boulder), and nearby habitats at least 2x the diameter of a boulder away ( away from boulder).