THE IMPACT OF CLIMATE CHANGE ON BROOK TROUT (SALVELINUS FONTINALIS) THERMAL HABITAT IN THEIR NATIVE RANGE IN THE UNITED STATES By Kelsey Maggan Schlee A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Fisheries and Wildlife—Master of Science 2014 ABSTRACT THE IMPACT OF CLIMATE CHANGE ON BROOK TROUT (SALVELINUS FONTINALIS) THERMAL HABITAT IN THEIR NATIVE RANGE IN THE UNITED STATES By Kelsey Maggan Schlee Within their native range in the United States (U.S.), brook trout (Salvelinus fontinalis) are valued for the recreational opportunities they provide to anglers and for their utility as indicators of the environmental health of the habitats in which they are found. For brook trout, stream temperatures within their viable thermal range (0°-25°C) are vital to ensuring growth, reproduction, and survival. Changes in air temperatures related to climate change will influence stream temperatures, likely resulting in alterations in the distribution and quantity of thermal habitat available for brook trout, leading to changes in this fish’s range and productivity. I examined the effects of changing air temperature on brook trout thermal habitat availability for 51 streams across 30 subbasins spanning the latitudinal and longitudinal gradient of the brook trout’s native range in the U.S. To determine the impact of air temperature changes driven by climate change for these streams, I converted air temperatures projections for the subbasin level from three coupled climate models into stream temperatures using two linear regression models and then rated the quality of stream habitat for brook trout growth and survival based on the results. According to both linear regression models, all 30 subbasins were predicted to increase in temperature by between 0.94°C and 4.16°C from 2006 to 2056, resulting in reduced thermal habitat quality for brook trout in 20 subbasins according to the first model and 15 according to the second model. To mitigate the effects of climate change related increases in water temperatures due to increased air temperature over the course of the next half-century, fisheries managers must focus their efforts on keeping streams as cool as possible by protecting and planting riparian zone shading and by guarding against additional sources of stream warming such as influxes of runoff during storm events and by maintaining other natural stream cooling mechanisms, like groundwater inputs. Copyright by KELSEY MAGGAN SCHLEE 2014 To my family For all your support, love, and good advice And To Lauri For making my dreams come true. iv ACKNOWLEDGEMENTS I would like to extend my thanks to my committee members Dr. Bill Taylor, Dr. Dana Infante, and Dr. Doug Bead for their guidance and encouragement throughout my time at Michigan State University. I would like to express my sincere gratitude to my major professor and mentor, Dr. Taylor, who encouraged me to pursue my passion and gave me the opportunity to explore the world. I thank Dr. Infante for her enthusiastic lessons about the aquatic world, her excellent problem solving capabilities, and her listening ear. I also thank Dr. Beard for his belief in my project, encouraging me to expand my understanding of science, and for taking the time to answer all of my questions in depth. I earnestly appreciate Dr. Abigail Lynch for her mentorship and advice during both my thesis project and my educational experience as a whole. I thank all the members of Dr. Dana Infante’s lab for over two years of much appreciated technical support. Special thanks in particular to Kyle Herreman for his brilliance with GIS and for his patience. Thank you also to Dr. Yin-Phan Tsang for helping me track down all my data. Thank you very much to Dr. Ernie Hain at North Carolina State University for his help with data collection and GIS. I would also like to extend my gratitude to the Fisheries Division of the Michigan Department of Natural Resources for their support of my project and my professional development. I would like to thank fisheries chief Jim Dexter in particular for teaching me about fisheries policy and management and for facilitating my interactions with the Natural Resources Commission. Thank you to the Michigan Natural Resources Commission for inviting me to present my research to them and to the commissioners for their excellent questions and feedback. I would also like to acknowledge Patricia Stewart for teaching me about the art and science of communication in the natural resources world. I also appreciate Professor Ian Cowx for his friendship, scientific v guidance, and his excellent sense of humor. I would like to extend a sincere thank you to the past and present members of the Berkley Conservation Leaders Advisory Team: Jim Martin, Ken Haddad, Ian Cowx, John Doerr, Becky Humphries, and Noreen Clough for their constant support, for broadening my perspective of Fisheries and natural resources, and for providing me with some of the best memories of my graduate school experience. Thank you also for one lifechanging trip to Portugal! I would also like to extend my gratitude to Dr. Shawn Riley for his excellent guidance and for supporting me in my next steps. I would like to acknowledge the Connecticut Department of Energy and Environmental Protection, the Maine Department of Inland Fisheries and Wildlife, the Michigan Department of Natural Resources, the Minnesota Department of Natural Resources, the North Carolina Wildlife Resources Commission, the New York Department of Environmental Conservation, the Pennsylvania Fish and Boat Commission, the Wisconsin Department of Natural Resources, and the West Virginia Department of Natural Resources for providing me with information about high quality brook trout streams and for granting me permission access to stream temperature information. A special thanks to Fred Henson with the New York Department of Environmental Conservation and Diana Day with the Pennsylvania Fish and Boat Commission for going above and beyond the call of duty. I would to thank my funding sources for their contributions to my project: The United States Geological Survey; the William W. and Evelyn M. Taylor Endowed Fellowship for International Engagement in Human and Natural Systems; the Schrems West Michigan Trout Unlimited Graduate Fellowship; and the Red Cedar Fly Fishers Graduate Fellowship in Fellowship in Fisheries Management. vi TABLE OF CONTENTS LIST OF TABLES ..................................................................................................................... viii LIST OF FIGURES ...................................................................................................................... x THE IMPACT OF CLIMATE CHANGE ON BROOK TROUT (SALVELINUS FONTINALIS) THERMAL HABITAT IN THEIR NATIVE RANGE IN THE UNITED STATES ……………………………...…………………………………………………………. 1 Introductory Summary............................................................................................................... 1 Introduction................................................................................................................................ 2 Brook Trout Life History......................................................................................................... 2 The Importance of Stream Temperature ................................................................................. 3 Effect of Air Temperature on Stream Temperature ................................................................ 6 Climate Change ...................................................................................................................... 6 Climate Change and its Impact on Stream Temperature........................................................ 8 Goals and Objectives ............................................................................................................ 11 Methods .................................................................................................................................... 11 Site Selection ......................................................................................................................... 11 Air Temperature Projections ................................................................................................ 23 Air Temperature to Stream Temperature Conversion .......................................................... 24 Stream Temperature Projections .......................................................................................... 27 Brook Trout Thermal Habitat Suitability.............................................................................. 27 Current Stream Temperatures .............................................................................................. 28 Baseflow & Stream Order ..................................................................................................... 29 Individual Models ................................................................................................................. 29 Results....................................................................................................................................... 30 Changes in Stream Temperature and Thermal Habitat Suitability for Brook Trout ............ 30 Stream Temperature Model Validation ................................................................................ 39 Individualized Air-Stream Temperature Linear Regression Models .................................... 40 Discussion................................................................................................................................. 59 Changing Stream Temperatures and Brook Trout Thermal Habitat .................................... 59 Stream Temperature Model Validation ................................................................................ 62 Factors influencing stream temperature ............................................................................... 66 Baseflow and Stream Order .................................................................................................. 67 Considerations for the future ................................................................................................ 69 Management Implications ..................................................................................................... 73 LITERATURE CITED .............................................................................................................. 77 vii LIST OF TABLES TABLE 1. List of high quality brook trout streams and number corresponding to each subbasin (HUC8) included in the national study site map (figure 1). ..……...……………………………………..... 10 TABLE 2. List of high quality brook trout streams and number corresponding to each subwatershed (HUC12) included in the state study site maps (figures 2- 9). …………………. 16 TABLE 3. The names of state agencies that designated each study stream as high quality for each state: Michigan Department of Natural Resources, Wisconsin Department of Natural Resources, Minnesota Department of Natural Resources, Maine Department of Inland Fisheries and Wildlife, Pennsylvania Fish and Boat Commission, North Carolina Wildlife Resources Commission. *Indicates states with publically-available lists………………………………….. 18 TABLE 4. Descriptive data for each high quality stream site including: stream name (HQ Stream Name)1, state, subbasin (HUC8), subbasin name (HUC8 Name), subwatershed (HUC12), NorEast gage identification number for the gage representing each stream (NorEast ID), area weighted average of percent groundwater contribution to streams within each HUC12 (GW%), Strahler stream order for each gaged stream (Stream Order) ..……..………………………...... 21 TABLE 5. Thermal habitat status designations (thermal habitat status), their corresponding temperature range (temperature range), and the brook trout growth rate they represent (growth status) (Raleigh, 1982, Power, 1980, Fry et al., 1946). …………………………...…………… 28 TABLE 6. Average stream temperature (Average °C) for the warmest month (Month) projected by the three coupled climate models (CCMs) and the resulting thermal habitat status designation (Status) during the years 2006, 2012, 2026, and 2056 for each high quality HUC8. The air-water temperature conversion was conducted using Stefan and Preud’homme’s (1993) equation (SP)………………………………………………………………………………………………35 TABLE 7. Average stream temperature (Average °C) for the warmest month (Month) projected by the three coupled climate models (CCMs) and the resulting thermal habitat status designation (Status) during the years 2006, 2012, 2026, and 2056 for each high quality HUC8. The air-water temperature conversion was conducted using Krider et al.’s (2013) equation (KEA) …….….. 38 TABLE 8. Average projected stream temperatures for the years 2006 and 2012 using both the air-stream temperature conversion equations, Stefan and Preud’homme’s (1993) equation (Pro. SP) and et al. (2013) equation (Pro. KEA), compared with field measured stream temperatures for both years (Actual). The month (Month) projected and actual gage temperatures are provided for is also included. “x” represents no data…………………………………………………....... 49 viii TABLE 9. Average projected stream temperatures for the years 2006 and 2012 using both the air-stream temperature conversion equations, Stefan and Preud’homme’s (1993) equation (Pro. SP) and et al. (2013) equation (Pro. KEA), compared with gage measured stream temperatures for both years (Actual). The month (Month) projected and actual gage temperatures are provided for is also included. “x” represents no data....……………………………………….....……….. 50 TABLE 10. Individualized linear regression models for eight streams created using August monthly air temperature averages collected from NOAA’s National Climatic Data Center (http://www.ncdc.gov) and August monthly stream temperature averages from the USGS’s National Water Information System (http://waterdata.usgs.gov/nwis). All available air and stream temperature data from the year 2000 to 2014 was used. R2 values for each individualized linear regression model are also included in the table.………………………………...……..… 52 TABLE 11. Comparison between stream temperatures and thermal habitat ratings projected by the individualized models (I-models), the Stephan and Preud’homme (1993) model (SP model), and field measured (Actual) temperature for 2006 and 2012 .…….…………………………… 60 TABLE 12. Comparison between stream temperautres and thernmal habitat ratings projected by the individualized models (I-models) and the Stephan and Preud’homme (1993) model (SP model) for the years 2006 and 2056……...…………………………………………...………... 60 ix LIST OF FIGURES FIGURE 1. The brook trout’s native U.S. range divided into its Midwestern and Eastern portions. The subbasins (HUC8s) containing each study stream are numbered and their corresponding identification numbers and names are listed in Table 1. ………………..……….. 9 FIGURE 2. Location of each subwatershed (HUC12) containing a high quality study stream in the state of Minnesota………………………………………...……………………………….... 12 FIGURE 3. Location of each subwatershed (HUC12) containing a high quality study stream in the state of Wisconsin……………………………………………………………………....…... 12 FIGURE 4. Location of each subwatershed (HUC12) containing a high quality study stream in the state of Michigan……………………………………………………………………….....… 13 FIGURE 5. Location of each subwatershed (HUC12) containing a high quality study stream in the state of Maine. …………………………………………………………………………….... 13 FIGURE 6. Location of each subwatershed (HUC12) containing a high quality study stream in the state of New York.………………………………………………………………………...... 14 FIGURE 7. Location of each subwatershed (HUC12) containing a high quality study stream in the state of Connecticut.............................................................................................................… 14 FIGURE 8. Location of each subwatershed (HUC12) containing a high quality study stream in the state of West Virginia…………………………………………………………………….… 15 FIGURE 9. Location of each subwatershed 12 (HUC12) containing a high quality study stream in the state of North Carolina (NC)……………………………………………………………... 15 FIGURE 10. Stream temperatures projected for the years 2006 and 2056 for the 51 study streams using both the Stefan and Preud’homme (1993) (SP model) and Krider et al. (2013) (KEA model) air-stream temperature linear regression equations. ……………………......…..…….... 31 FIGURE 11. Change in thermal habitat status for brook trout between the years 2006 and 2056 predicted by the SP model. Thermal habitat status designations are described in Table 5.….… 33 FIGURE 12. Change in thermal habitat status for brook trout between the years 2006 and 2056 predicted by the KEA model. Thermal habitat status designations are described in Table 5…... 37 FIGURE 13. Comparison between field measured (actual) stream temperature (°C) data and stream temperature projections made using the Stefan and Preud’homme (1993) (SP) model for the year 2006. Data is sorted by percent groundwater contribution to baseflow for each stream from least to greatest…………………………………………...………………………...……... 41 x FIGURE 14. Comparison between field measured (actual) stream temperature (°C) data and stream temperature projections made using the Stefan and Preud’homme (1993) (SP) model for the year 2006. Data is sorted by the Strahler stream order for each gaged stream, from highest to lowest order.………………………………………………………….…………………...…….. 42 FIGURE 15. Comparison between field measured (actual) stream temperature (°C) data and stream temperature projections made using the Stefan and Preud’homme (1993) (SP) model for the year 2012. Data is sorted by percent groundwater contribution to baseflow for each stream from least to greatest.…………………………………………………………………..……...... 43 FIGURE 16. Comparison between field measured (actual) stream temperature (°C) data and stream temperature projections made using the Stefan and Preud’homme (1993) (SP) model for the year 2012. Data is sorted by the Strahler stream order for each gaged stream, from highest to lowest order.………………………………………………………………………….…….....… 44 FIGURE 17. Comparison between field measured (actual) stream temperature (°C) data and stream temperature projections made using the Krider et al. (2013) (KEA) model for the year 2006. Data is sorted by percent groundwater contribution to baseflow for each stream from least to greatest.……………………………………………………………………………………..... 45 FIGURE 18. Comparison between field measured (actual) stream temperature (°C) data and stream temperature projections made using the Krider et al. (2013) (KEA) model for the year 2006. Data is sorted by the Strahler stream order for each gaged stream, from highest to lowest order.………………………………………………………………………….……………….... 46 FIGURE 19. Comparison between field measured (actual) stream temperature (°C) data and stream temperature projections made using the Krider et al. (2013) (KEA) model for the year 2012. Data is sorted by percent groundwater contribution to baseflow for each stream from least to greatest. ………………………………………………………...………………………….… 47 FIGURE 20. Comparison between field measured (actual) stream temperature (°C) data and stream temperature projections made using the Krider et al. (2013) (KEA) model for the year 2012. Data is sorted by the Strahler stream order for each gaged stream, from highest to lowest order. …………………………………………………………………...…………………....… 48 FIGURE 21. Individualized air-stream temperature (°C) linear regression model for Broad Swamp Creek, Connecticut (CT). ………………………………………………………...……. 53 FIGURE 22. Individualized air-stream temperature (°C) linear regression model for Meduxnekeag River, Maine (ME).………………………..………………………….……....… 53 FIGURE 23. Individualized air-stream temperature (°C) linear regression model for Silver River, Michigan (MI).………………………………………………………………………………..… 54 xi FIGURE 24. Individualized air-stream temperature (°C) linear regression model for Escanaba River, Michigan (MI).………………………………………..……………………..………...… 54 FIGURE 25. Individualized air-stream temperature (°C) linear regression model for Au Sable River, Michigan (MI). ………...……………………………………………..…...………..…… 55 FIGURE 26. Individualized air-stream temperature (°C) linear regression model for Deep Creek, North Carolina (NC).…………………………………………………………………………… 55 FIGURE 27. Individualized air-stream temperature (°C) linear regression model for Embarrass River, Wisconsin (WI). ……………………………………………...………...………..……… 56 FIGURE 28. Individualized air-stream temperature (°C) linear regression model for Red River, Wisconsin (WI).……………..………………………………...…………………………..……. 56 xii THE IMPACT OF CLIMATE CHANGE ON BROOK TROUT (SALVELINUS FONTINALIS) THERMAL HABITAT IN THEIR NATIVE RANGE IN THE UNITED STATES Introductory Summary In the United States (U.S.), the brook trout, (Salvelinus fontinalis), is a highly valued sportfish, contributing to the recreational fishing industry and to the well being of participating anglers. In addition to the excitement they provide at the end of a fishing line, this species is also valued for its utility as an indicator of environmental health, due to their sensitivity to changes in water quality. Therefore, maintaining brook trout in their native waters is for the economic, social, and environmental benefit of the nation. Maintaining this species in its native range will prove challenging for fisheries managers, however, in part because this species is reliant on a narrow range of cold water temperatures for growth, reproduction, and survival, making stream warming a threat to their persistence in U.S. streams. Air temperature, one of the drivers of stream temperature, is projected to change as greenhouse gas emissions continue to contribute to global climate change in the coming years (Melillo et al., 2014). As such, air temperature changes and concurrent changes in stream temperature, will affect the availability of thermal habitat for brook trout, making future strategies for the management of this species based on an understanding of the potential risk and magnitude of stream warming a necessity. The goal of this thesis study was to improve this understanding by assessing the potential impacts of changing air temperatures, resulting from climate change, on stream temperature and therefore, on brook trout thermal habitat across their native U.S. range. Specifically, I evaluated changes in summer stream temperature between the years 2006, 2012, 2026, and 2056, based on projected changes in air temperatures for selected streams currently 1 supporting productive brook trout populations and fisheries. Then, I used these findings to project changes in the future thermal habitat suitability of these streams for this species. Lastly, this study provides a methodology for managers to predict current and future stream temperatures and resultant thermal habitat suitability from air temperatures to aid in developing future management plans for brook trout. Introduction Brook Trout Life History The brook charr, Salvelinus fontinalis, or as it is colloquially known in the United States, the brook trout, is native to a large portion of North America. Its range extends from the Canadian province of Manitoba in the west, north to Newfoundland and Labrador, through the Great Lakes Basin, and south along the Appalachian Mountain Chain (MacCrimmon and Campbell, 1969, Waco, 2009). Within this range, brook trout exhibit three distinct life-history forms. Two of these forms exist entirely in freshwater and one is anadromous, indicating time spent in both fresh and salt water. The anadromous form, commonly called “the salter,” can be found along the east cost of North America and migrates to the ocean most likely for feeding opportunities (Karas, 1997). The salter becomes silver in color while living in marine waters, but returns to typical freshwater brook trout coloration upon its return to rivers and streams (Karas, 1997). The two entirely freshwater forms of brook trout are the stream variety and the larger potamodromous “coaster.” The stream brook trout is typically found in coldwater streams and is small and short lived, with a lifespan of approximately 3-5 years and a typical length of between 20 and 25cm (Waco, 2009). Relative to the stream variety, the coaster form is larger, later to mature, and longer lived. The coaster, which inhabits large lakes and rivers, as well as streams, 2 may live to between eight and ten years old and attain a weight of between four and six kg (Power, 1980, Waco, 2009, Raleigh, 1982). While brook trout do exhibit distinct differences in external appearance depending on location (MacCrimmon and Campbell, 1969), in freshwater brook trout typically have an olive green to black body with red spots, some of which are surrounded by blue halos on their sides (Page and Burr, 1991, MI DNR, 2012). The belly is typically white and wavy light green or cream-colored vermiculations run along the back and dorsal fin (MI DNR, 2012). The caudal fin is only very slightly forked to nearly straight-edged and the lower fins are edged in white (MI DNR, 2012, Page and Burr, 1991). A brook trout’s mouth is large and extends past the eye (MI DNR, 2012). Breeding males have black bellies and bright red-orange sides (MI DNR, 2012, Page and Burr, 1991). The Importance of Stream Temperature The brook trout is a stenothermic cold-water obligate fish, therefore maintaining stream temperatures within its optimal thermal range (Table 5) is vital to ensuring the efficient functioning of life processes critical to survival and productivity. This is because the internal temperature of poikilothermic organisms, like brook trout, is directly controlled by the surrounding water temperature, which ultimately influences their metabolic processes, reproduction, behavior (Helfman et al., 2009), and distribution both in a broad geographic sense and within a single river system (Hynes, 1970). Fish metabolism increases with increasing temperature, resulting in a rising demand for oxygen intake and consumed energy to fuel cellular processes (Helfman et al., 2009). As a component of anabolic metabolism, growth is also closely related to stream temperature (Hynes, 1970, Allan, 1995). Feeding activity and digestion rate increase as temperatures rise, allowing 3 fish to grow faster provided that the demand for food and oxygen to fuel this process can still be met (Allan, 1995). Dissolved oxygen concentration in water decreases as temperature increases, however, and at some point fish can no longer meet the oxygen demands of a heightened metabolism at high temperatures. Even when dissolved oxygen concentration is not the limiting factor for supporting a heightened metabolism, limited food availability could be. Trout, for example, are not typically found in warm water streams in part because at approximately 20°C they reach a thermal threshold above which they are unable to meet their energetic demands (Allan, 1995). Additionally, at temperatures near a fish’s upper thermal limit, more energy must be utilized for metabolic processes more crucial to survival than growth, which explains why growth is typically reduced as fish approach their upper thermal limits (Allan, 1995). Gonadal growth and development is also influenced by water temperature in the same way that somatic growth is, therefore, water temperatures play a role in fish reproduction (Helfman et al., 2009, Strussmann et al., 1998). Additionally, fecundity is also influenced by the effect of water temperature on growth because body size is positively correlated with the number of eggs produced (Beldade et al., 2012, Hislop, 1988). Water temperatures also affect fish reproduction by acting as a behavioral cue for spawning events (Hynes, 1970). For example, peak spawning activity was coincident with drops in lake temperature to below 11°C for lakedwelling brook trout in Scott Lake Ontario, Canada over a period of two years (Blanchfield and Ridgway, 1997). As such, in order to maximize growth, survival, and reproduction, fish tend to select environmental temperatures where they function most efficiently (Coutant, 1987, Helfman et al., 2009). Therefore, reduction in the availability of optimal thermal habitat has potentially significant impacts on fish survival and productivity. 4 The brook trout’s upper and lower thermal limits are 25.3°C and 0°C, respectively (Fry et al., 1946). These fish are most productive and exhibit the highest rate of growth between 11°C and 16°C (Raleigh, 1982), where internal physiological processes for this species function most efficiently (Brett, 1956). Outside this thermal range, brook trout experience reduced growth and reproductive potential and are also more vulnerable to external factors, such as predation and disease, which often result in increased mortality (Brett, 1956). For this reason, high temperatures during the summer months in the brook trout’s native U.S. range are limiting to trout biomass and abundance (Bowlby and Roff, 1986). During the height of summer, brook trout can be found in high densities in 15°C-19°C water (Wehrly et al., 2003), suggesting congregation around cool-water refugia like groundwater springs (Helfman et al., 2009). In extreme cases, fish reliance on small pockets of thermally suitable water can result in overcrowding, which may lead to an increase in the prevalence of disease, locally depleted food resources, and overfishing, impacting growth, survival, and reproductive capacity (Helfman et al., 2009). As such, ambient stream temperatures can be a limiting factor to brook trout productivity and distributional range (McCormick et al., 1972); constraining them to cooler locations (Helfman et al., 2009, Meisner, 1990a). Maintaining thermal conditions suitable for brook trout within their native U.S. range is important because this species is valued for the socioeconomic benefits it provides local communities and for its utility as an indicator of environmental health (Eastern Brook Trout Joint Venture, 2008). In 2006, the United States’ trout fishery, which includes the brook trout, provided recreational fishing opportunities for an estimated 6.8 million anglers who spent US$4.8 billion (Harris, 2010). In 2011, participation in trout angling increased to an estimated 7.2 million participants (U.S. Department of Interior et al., 2011). In addition to their value as a 5 recreationally important fish, brook trout are also considered a gauge of environmental condition due to their sensitivity to changes in water quality (Karas, 1997). Therefore, maintaining stream temperature within the brook trout’s thermal range is for the economic, social, and environmental benefit of the nation. Effect of Air Temperature on Stream Temperature When managing thermally sensitive fish species, like brook trout, into the future, it is critical to understand the drivers of stream temperature and the influence that changes in these drivers may have. Stream temperatures are determined by climate, elevation, streamside (riparian) vegetation, runoff, groundwater inputs, and urban and agricultural land use (Allan, 1995, Nelson and Palmer, 2007). For this study, I focused on air temperature, a component of climate, because projected changes in climate are likely to influence this stream temperature driver, resulting in future impacts to brook trout populations. Climate Change Since air temperature, which is ultimately controlled by climate, is a driver of water temperature, which in turn influences brook trout survival and productivity, to manage their populations into the future it is important to understand how changing climate is likely to influence air temperature and therefore, stream temperatures. Climate change is generally understood as long term changes in weather patterns, like precipitation and air temperature, resulting from increasing concentrations of greenhouse gasses (GHGs), most notably carbon dioxide (CO2), water vapor, methane, nitrous oxide, and chlorofluorocarbons, in the atmosphere (Henson, 2011, Houghton, 2009). These gasses essentially blanket the Earth, absorbing the heat emitted by the planet, while only releasing a portion of that heat back into space (Henson, 2011). As the concentration of GHGs in the atmosphere rises, its heat capacity to sequester heat 6 increases, influencing Earth’s climate (Henson, 2011, Houghton, 2009). Much of the increase in atmospheric GHGs since the beginning of the Industrial Revolution around the year 1750 has been attributed to changes in human activities including, but not limited to reliance on fossil fuels for energy, agriculture, and land use change (Henson, 2011). GHG emissions have continued to rise over time with humans adding an estimated 31 gigatons of CO2 into the atmosphere in 2009, as opposed to approximately 26 gigatons in 2002, and 15 gigatons in 1970 (Henson, 2011). To predict future greenhouse gas emission, a driver of climate change, in 2001 The International Panel on Climate Change (IPCC) developed a Special Report on Emissions Scenarios (SRES), which included descriptions of 35 likely futures for GHG emissions (Houghton, 2009). The 35 scenarios were developed based on differing sets of assumptions about the future, including factors like economic growth, human population growth, technological innovations, and the evolution of thought regarding social and environmental sustainability (Houghton, 2009). To illustrate how climate may change, climate-influencing variables, like the projected GHG emissions scenarios described above, are incorporated into global climate models (GCMs). GCMs are highly sophisticated computer models used to simulate the interactions between the atmosphere, ocean, land, ice, and biosphere, which ultimately govern Earth’s climate (Houghton, 2009). Many GCMs, called coupled models, incorporate two or more sub-models describing interactions between these different climate drivers (Henson, 2011). Such models can project future conditions for climatic variables, like air temperature and precipitation patterns. Since GCMs typically provide projections of climate at a very large scale, regional models (RCMs) or statistical downscaling methods can be used to apply climate change 7 projections at more appropriate resolution for scientific endeavors involving the progression of climate change in a fixed area (Henson, 2011, Houghton, 2009). Climate Change and its Impact on Stream Temperature The brook trout’s native range within the U.S. includes the Midwestern states of Minnesota, Iowa, Wisconsin, Michigan, and Ohio and extends along the Appalachian mountain chain from Maine in the north to northeastern Georgia in the south (Figure 1) (MacCrimmon and Campbell, 1969). Within this range, climate change is predicted to result in rising air temperatures and longer, more frequent, and more severe summer heat events (Karl et al., 2009). In the Midwest, regional average air temperatures are expected to rise by between 3.1°C and 4.7°C by the end of the century (2081-2100) ((Melillo et al., 2014). In the Northeast, air temperatures are expected to rise by between 1.7°C to 5.6°C by the 2080s (Kunkel et al. 2012; (Horton and coworkers, (Melillo et al., 2014). In the Southeast, air temperatures are expected to rise by 5.6°C by 2100 (Kunkel et al. 2012). These changes in air temperature are expected to directly influence stream temperatures (Karl et al., 2009) because, as mentioned above, water temperatures typically fluctuate in synch with air temperature (Pilgrim et al., 1998), resulting in changes in brook trout thermal habitat. Rising water temperatures are of concern because many fish species, including brook trout, are often currently found living near their upper thermal limit (Magnuson and Destasio, 1997, Helfman et al., 2009), which means that even minimal changes in stream temperature could result in fish extirpation from the stream. Rising air temperatures are particularly problematic during the summer months, as annual air and water temperatures reach their maximums during mid to late summer, making brook trout vulnerable to heat-related stress during this season (Meisner, 1990b). Longer, hotter summers predicted with climate change (Karl et al., 2009) are likely to result in an increase in the 8 FIGURE 1. The brook trout’s native U.S. range divided into its Midwestern and Eastern portions. The subbasins (HUC8s) containing each study stream are numbered and their corresponding identification numbers and names are listed in Table 1. 9 TABLE 1. List of high quality brook trout streams and number corresponding to each subbasin (HUC8) included in the national study site map (figure 1). National Site Map Key State CT CT CT CT CT CT ME ME ME ME MI MI MI MI MI MN NC NY NY NY NY NY WI WI WI WI WI WV WV WV Watershed Name Housatonic Lower Connecticut Shetucket Thames Quinnipiac Saugatuck Aroostook Meduxnekeag Maine Coastal Lower Kennebec Dead-Kelsey Escanaba Black Thunder Bay Au Sable Buffalo-Whitewater Lower Little Tennessee Great Chazy-Saranac Ausable Chenango Upper Genesee Southern Long Island Wolf Red Cedar Lower Chippewa La Crosse-Pine Lower Wisconsin North Branch Potomac Elk Gauley Map ID 16 18 20 22 24 26 1 2 3 4 5 7 9 12 13 15 30 6 8 17 21 25 10 11 14 19 23 27 28 29 incidence and severity of heat related stress events on brook trout, impacting their distribution (Meisner, 1990a, Meisner, 1990b, Magnuson et al., 1990, Shuter and Post, 1990), and productivity (Hokanson, 1977, Drake and Taylor, 1996) making future strategies for the 10 management of this species based on an understanding of the potential risk and magnitude of stream warming a necessity. Goals and Objectives The goal of this study was to assess the potential impacts of changing air temperatures, resulting from climate change, on brook trout thermal habitat across their native U.S. range. The specific objectives of this study were: 1) evaluate changes in summer stream temperature between the years 2006, 2012, 2026, and 2056, based on projected changes in air temperatures for selected streams currently supporting productive brook trout populations and fisheries; 2) to project changes in future thermal habitat suitability of these streams for brook trout in terms of growth and survival; and 3) to provide a methodology for managers to predict current and future stream temperatures and resultant thermal habitat suitability for brook trout from air temperatures. Methods Site Selection I chose 52 high quality brook trout streams (Figures 1-9; Table 2) distributed across their native U.S. range to evaluate the potential effect of climate change related shifts in air temperature on thermal habitat suitability for this species. I selected streams from states spanning the latitudinal and longitudinal gradient this range, in an effort to capture current and predicted differences in climate along this expanse. States represented in this study include: Minnesota (MN), Wisconsin (WI), and Michigan (MI) in the midwestern region and Maine (ME), New York (NY), Connecticut (CT), and North Carolina (NC) in the Eastern region (Figures 1-9; Table 2). 11 FIGURE 2. Location of each subwatershed (HUC12) containing a high quality study stream st in the state of Minnesota. FIGURE 3. Location of each subwatershed (HUC12) containing a high quality study stream in the state of Wisconsin. 12 FIGURE 4. Location of each subwatershed (HUC12) containing a high quality study stream in the state of Michigan. FIGURE 5. Location of each subwatershed (HUC12) containing a high quality study stream in the state of Maine. 13 FIGURE 6. Location of each subwatershed (HUC12) containing a high quality study stream in the state of New York. FIGURE 7. Location of each subwatershed (HUC12) containing a high quality study stream in the state of Connecticut. 14 FIGURE 8. Location of each subwatershed (HUC12) containing a high quality study stream in the state of West Virginia. FIGURE 9. Location of each subwatershed 12 (HUC12) containing a high quality study stream in the sate of North Carolina (NC). 15 TABLE 2. List of high quality brook trout streams and number corresponding to each subwatershed (HUC12) included in the state study site maps (figures 2- 9). State Site Map Key State CT CT CT CT CT CT CT CT CT CT CT CT CT CT CT CT CT CT CT CT CT CT CT ME ME ME ME ME Stream Name Ames Brook Ballymahack Brook North Branch Hamlin Brook Mallory Brook West Brook Good Hill Brook Humaston Brook Jericho Brook Sutliffe Brook Riggs Street Brook North Branch West Branch Saugatuck River Cemetery Brook West Swamp Brook Broad Swamp Brook Lake Pond Brook Watermans Brook Bunker Hill Brook Gulf Brook Brooksvale Stream Bryant Brook Gilbert Bennett Brook Woods Pond Brook Morehouse Brook Meduxnekeag River Salmon Brook Sandy River Old Stream Little Mopang/Mopang Stream 16 Map ID 1 2 3 4 5 6 7 7 7 8 9 10 11 12 13 14 15 16 17 18 18 18 19 1 2 3 4 5 Table 2 (cont’d). ME MI MI MI MI MI MI MN NC NY NY NY NY NY WI WI WI WI WI WI WV WV WV Pleasant River Escanaba River Au Sable River Comstock Creek Black River Silver River Salmon Trout River Snake Creek Deep Creek East Branch Ausable River True Brook Otselic River Bush Brook Carmans River Black Earth Creek Red River Embarrass River Silver Creek Cady Creek Connors Creek North Fork Cranberry River Laurel Fork Johnnycake Run 17 6 1 2 3 4 5 6 1 1 1 2 3 4 5 1 2 3 4 5 6 1 2 3 I chose to investigate the impacts of changing air temperatures resulting from climate change on high quality brook trout streams specifically, because if the thermal habitat suitability of these streams is impacted, then changing air temperatures will also affect the habitat suitability of more thermally marginal streams within the same regions. Therefore, to select the study streams, I examined streams designated by state fisheries management agencies as high quality recreational brook trout fishing waters with naturally reproducing populations (Table 3). The lists of high quality streams were either obtained from publically available state agency websites, or via direct contact with the agencies. Naturally reproducing populations of brook trout and the presence of a recreational fishery generally indicate high fish productivity, which I assumed reflected the presence of preferred thermal habitat conditions. Therefore, I designated streams supporting a naturally reproducing population of brook trout, a recreational fishery, and therefore, optimal thermal habitat as “high quality” streams for the purposes of this study. TABLE 3. The names of state agencies that designated each study stream as high quality for each state: Michigan Department of Natural Resources, Wisconsin Department of Natural Resources, Minnesota Department of Natural Resources, Maine Department of Inland Fisheries and Wildlife, Pennsylvania Fish and Boat Commission, North Carolina Wildlife Resources Commission. *Indicates states with publically-available lists. Source of Designated High Quality Brook Trout Stream Lists State CT ME MI MN NC NY WI WV Source Connecticut Department of Energy and Environmental Protection (CDEEP) Maine Department of Inland Fisheries and Wildlife (MDIFW)* Michigan Department of Natural Resources (MDNR)* Minnesota Department of Natural Resources (MNDNR)* North Carolina Wildlife Resources Commission (NCWRC)* New York Department of Environmental Conservation (NYDEC) Wisconsin Department of Natural Resources (WDNR)* West Virginia Department of Natural Resources (WVDNR) I used the National Hydrography Dataset Plus Version 1 (NHDPlusV1) (U.S. Geological Survey and U.S. Environmental Protection Agency, 2005) Watershed Boundary Dataset to 18 determine the eight and 12-digit hydrologic unit codes, HUC8 (Table 2; Figure 1) and HUC12 (Figures 2-9) respectively, containing each study stream selected using the method above. These hydrologic unit codes are used to catalog the hydrologic drainage system in the United States, with unit boundaries defined by hydrographic and topographic features (U.S. Department of Agriculture, Natural Resource Conservation Service). The system is nested hierarchically into six levels based on the size of the drainage unit; the levels from largest to smallest are: region, subregion, basin, subbasin, watershed, and subwatershed ( U.S. Department of Agriculture, Natural Resource Conservation Service). The HUC8s and HUC12s used in this study correspond to the subbasin (HUC8) and subwatershed (HUC12) levels. The Subcommittee on Spatial Water Data, including members of several United States federal agencies, is responsible for the development of the Watershed Boundary Dataset, which contains the delineations for HUCs (U.S. Department of Agriculture, Natural Resource Conservation Service). Then, I used NHDPlusV1 ( U.S. Geological Survey and U.S. Environmental Protection Agency, 2005) to locate stream temperature-recording gages within the same subwatersheds (high quality HUC12s) as each of the high quality streams. High quality streams without any gages within their HUC12 were removed from the study. Once I had identified all the stream temperature-recording gages located within each high quality HUC12, I chose one gage to represent each remaining high quality stream. For the purposes of this study, the gages chosen had to be included in the network of gages contributing data to the Northeast Climate Center’s (NECSC) database2 and must have had stream temperature data available to produce monthly averages for at least one summer month, June, July, or August, in either the year 2006, 2012, or 2 NECSC project title: “A Stream Temperature Inventory Network and Decision Support Metadata Mapper Evaluating the Resources to Understanding Climate Change Effects on Streams in New England and the Great Lakes States” 19 both3. Gages that did not meet these criteria were removed from the study. In the event that multiple gages meeting these criteria were available in a high quality HUC12, a gage located directly on the high quality stream was preferentially chosen. If there were multiple gages located on the high quality stream, the most upstream gage was chosen. Similarly, if no gages were available directly on the high quality stream, than the most upstream gage on the most proximal gaged stream within the same subwatershed (HUC12) was used. The rationale behind choosing the most upstream gage was that stream temperatures are typically coolest at their most upstream reaches, warming as the water moves downstream, therefore the most optimal habitat for cold water obligate brook trout would likely be in the most upstream reaches. Gages located directly below dams were disqualified from the study, as the temperature of the water discharged from dams may be artificially altered (Lessard, 2000). At the end of the selection process, I had collected an average monthly water temperature for at least one summer month (June, July, or August), in the year 2006, 2012, or both from 51 stream temperature-recording gages located within the same HUC12 as a state agency designated high quality brook trout stream. Table 4 summaries all metrics collected for all 51 study sites including: high quality stream name, state, subbasin (HUC8), subbasin name, subwatershed (HUC12), NorEast gage identification number for the gage representing each stream (NorEast ID), area weighted average of percent groundwater contribution to streams within each HUC12 (GW%), and Strahler stream order for each gaged stream (Stream Order). 3 Downscaled air temperature projections were available for the years 2006, 2012, 2026, and 2056. 20 TABLE 4. Descriptive data for each high quality stream site including: stream name (HQ Stream Name)4, state, subbasin (HUC8), subbasin name (HUC8 Name), subwatershed (HUC12), NorEast gage identification number for the gage representing each stream (NorEast ID), area weighted average of percent groundwater contribution to streams within each HUC12 (GW%), Strahler stream order for each gaged stream (Stream Order). High Quality Stream Information 4 Stream Name State HUC8 HUC8 Name HUC12 NorEast ID GW % Stream Order Good Hill Brook CT 01100005 Housatonic 011000050903 CT_DEP_922 47.34 4 Humaston Brook CT 01100005 Housatonic 011000051105 CT_DEP_194 47.30 3 Jericho Brook CT 01100005 Housatonic 011000051105 CT_DEP_197 47.30 4 Mallory Brook CT 01100005 Housatonic 011000050703 CT_DEP_1455 48.79 4 Riggs Street Brook CT 01100005 Housatonic 011000051206 CT_DEP_5991 47.93 3 Sutliffe Brook CT 01100005 Housatonic 011000051105 CT_DEP_725 47.30 4 West Brook CT 01100005 Housatonic 011000050802 CT_DEP_333 49.44 4 Broad Swamp Brook CT 01080205 Lower Connecticut 010802050905 USGS_01194750 48.01 6 W. Swamp Brook CT 01080205 Lower Connecticut 010802050603 CT_DEP_453 49.57 2 Brooksvale Stream CT 01100004 Quinnipiac 011000040301 CT_DEP_1807 48.99 2 Bunker Hill Brook CT 01100004 Quinnipiac 011000040202 CT_DEP_1853 47.95 2 Gulf Brook CT 01100004 Quinnipiac 011000040206 CT_DEP_68 48.14 2 N. Branch Hamlin Brook CT 01100004 Quinnipiac 011000040101 CT_DEP_488 50.49 2 Watermans Brook CT 01100004 Quinnipiac 011000040105 CT_DEP_289 49.24 4 Bryant Brook CT 01100006 Saugatuck 011000060202 CT_DEP_245 47.58 4 Cemetery Brook CT 01100006 Saugatuck 011000060102 CT_DEP_319 47.88 3 Gilbert Bennett Brook CT 01100006 Saugatuck 011000060202 CT_DEP_235 47.58 3 Morehouse Brook N. Branch W. Branch Saugatuck River CT 01100006 Saugatuck 011000060302 CT_DEP_1810 47.71 2 CT 01100006 Saugatuck 011000060103 CT_DEP_1288 49.11 2 Woods Pond Brook CT 01100006 Saugatuck 011000060202 CT_DEP_236 47.58 4 Ames Brook CT 01100002 Shetucket 011000020206 CT_DEP_5955 49.25 1 Ballymahack Brook CT 01100002 Shetucket 011000020303 CT_DEP_5968 50.86 1 Stream names indicating directionality have been abbreviated. For example, N. Fork Cranberry River represents North Fork Cranberry River and etc. 21 Table 4 (cont’d). Lake Pond Brook CT 01100003 Thames 011000030304 USGS_011277905 51.18 2 Salmon Brook ME 01010004 Aroostook 010100041001 ME_DMR_14_1SALMON6_40 53.75 2 Meduxnekeag River ME 01010005 Meduxnekeag 010100050407 USGS_01018035 53.35 4 Sandy River ME 01030003 Lower Kennebec 010300030902 ME_DMR_9SANDYR97_66 44.01 3 Mopang Stream ME 01050002 Maine Costal 010500020901 ME_DMR_5MOPANG34_81 52.82 2 Old Stream ME 01050002 Maine Costal 010500020803 ME_DMR_5OLDSTR27_86 56.23 3 Pleasant River Silver River ME 01050002 Maine Costal 010500021202 ME_DMR_4MAINST11_00 53.50 4 MI 04020105 Dead-Kelsey 040201050605 USGS_04043150 58.02 4 Salmon Trout River MI 04020105 Dead-Kelsey 040201050401 USGS_04043238 61.59 2 Escanaba River MI 04030110 Escanaba 040301100308 USGS_04059000 58.77 5 Black River MI 04070005 Black Watershed 040700050201 MI_IFR_38 80.18 1 MI 04070006 Thunder Bay 040700060502 MI_IFR_1189 69.14 3 Comstock Creek Au Sable River MI 04070007 Au Sable 040700070501 USGS_04136000 78.29 5 Snake Creek MN 07040003 070400030602 MN_DNR_03LM002 69.76 2 Deep Creek NC 06010204 Buffalo-Whitewater Lower Little Tennessee 060102040107 USGS_0351706800 63.92 NA E. Branch Ausable River NY 02010004 Ausable 020100040103 NY_DEC_Ausable River, East Br._31 47.95 3 True Brook NY 02010006 Great Chazy-Saranac 020100060501 NY_DEC_True Brook_107 57.97 2 Carmans River NY 02030202 Southern Long Island 020302020302 NY_DEC_Carmans River_12 85.02 1 Otselic River NY 02050102 Chenango 020501020301 NY_DEC_Otselic River_137 47.73 3 Bush Brook NY 04130002 Upper Genesee 041300020705 NY_DEC_Wiscoy Creek_239 36.25 4 Embarrass River WI 04030202 Wolf 040302021005 USGS_0407809265 70.69 3 Red River WI 04030202 Wolf 040302020505 USGS_04077630 66.76 4 Silver Creek WI 07040006 La Crosse-Pine 070400060203 USGS_05382284 59.67 3 Cady Creek WI 07050005 Lower Chippewa 070500051003 WI_WDNR_ZZCady1 60.30 2 Connors Creek WI 07050007 Red Cedar 070500070501 WI_WDNR_Zzsfxhay 60.30 4 Black Earth Creek WI 07070005 Lower Wisconsin 070700050501 USGS_05406500 59.74 4 Johnnycake Run WV 02070002 North Branch Potomac 020700020204 WV_DEP_PNB-00014-10.9 40.93 2 N. Fork Cranberry River WV 05050005 Gauley 050500050201 WV_DEP_KG-00212-0.2 36.38 2 Laurel Fork WV 05050007 Elk 050500070301 WV_DEP_KE-00203-0.5 30.27 2 22 Air Temperature Projections I used three coupled climate models (CCMs) to project future summer air temperatures for the years 2026 and 2056 and to backcast air temperatures for 2006 and 2012: the Third Generation Coupled Global Climate Model (CGCM3, Canadian Centre for Climate Modelling and Analysis) the CM2 Global Coupled Climate Model (CM2, Geophysical Fluid Dynamics Laboratory at the National Oceanic and Atmospheric Administration), and the Hadley Centre Coupled Model version 3 (HadCM3, Met Office, United Kingdom’s National Weather Service), to project future summer air temperatures for the years 2026 and 2056 and to backcast air temperatures for 2006 and 2012. All three models, based on the World Climate Research Programme’s (WRCP) Coupled Model Intercomparison Project phase 3 (CMIP3), were spatially downscaled using the Bias-Correction Spatial Disaggregation (BCSD) approach (Maurer et al., 2007). This is a statistical downscaling method where the spatial resolution of the climate model (approximately 200km x 200km) is adjusted to a finer resolution (12km x 12km) more relevant to the scale of the study with differences between observed and modeled variables (i.e., air temperature and precipitation) used to adjust or correct future simulations for each time step and grid cell (Maurer et al., 2007). I ran all three CCMs with the WCRP’s CMIP3 and chose to assume the Special Report on Emission Scenarios (SRES) A1B climate forcing scenario. SRES A1B is considered the middle emissions path, predicting atmospheric CO2 concentrations of 700ppm by 2100 (Meehl et al., 2007). I chose to use the A1B climate-forcing scenario because I considered it to be the most moderate of the three SRES scenarios, thereby avoiding extreme possible future climate conditions. 23 The United States Forest Service’s (USFS) Eastern Forest Environmental Threat Assessment Center (EFETAC) in North Carolina calculated average monthly air temperatures for the subbasin level based on the predictions provided by the three coupled climate models (CGCM3, CM2, and HadCM3) using area-weighted means for the years 2006, 2012, 2026, and 2056. EFETAC provided us with the monthly summer (June, July, and August) air temperature averages for each HUC8 containing one of the selected high quality brook trout streams. I chose to predict stream temperature for these months because they are typically the warmest months in the studied region and therefore pose the highest risk of heat stress to brook trout. Air Temperature to Stream Temperature Conversion Air temperature is, in many cases, an accurate predictor of current and future stream temperatures, as several studies have cited a nearly 1:1 linear relationship between water an air temperature for weekly and monthly data (Stefan and Preud'homme, 1993, Pilgrim et al., 1998). For example, near 1:1 linear relationships between air and stream temperature have been described for 39 Minnesota streams (Pilgrim et al., 1998) and 11 Mississippi basin streams (Stefan and Preud'homme, 1993).Therefore, air temperatures can be used to approximate stream temperature using linear regression models (Crisp and Howson, 1982, Stefan and Preud'homme, 1993, Pilgrim et al., 1998, Mohseni and Stefan, 1999, Krider et al., 2013). I chose to use two linear equations estimating weekly air-water temperature relationships to convert the HUC8 level air temperature projections I received from EFETAC into stream temperatures for the four study years. The two linear regression equations used were Stefan and Preud’homme’s (1993) weekly stream temperature estimation model and Krider et al.’s (2013) composite model. 24 For Stefan and Preud’homme’s (1993) model, weekly stream temperature is estimated by: Tw (t) = 2.9 + 0.86 Ta. Where Tw is water temperature (°C) and Ta is air temperature (°C). Stefan and Preud’homme (1993) developed both generalized daily and weekly temperature empirical models to convert air temperatures into water temperatures for 11 rivers and streams within the Mississippi River basin to investigate the practice of using air temperatures to predict water temperatures. Stream and air temperatures are more strongly correlated at weekly and monthly timescales than at the daily timescale (Stefan and Preud'homme, 1993, Pilgrim et al., 1998, Erickson and Stefan, 2000). This principle is reflected by the standard deviations produced for Stefan and Preud’homme’s (1993) daily and weekly models. The standard deviation for the generalized daily model is 2.70°C, whereas the weekly model produces more accurate predictions, with an average standard deviation of 2.16°C (Stefan and Preud'homme, 1993). Therefore, I chose to use the weekly temperature model for this study. For Krider et al.’s (2013) model, weekly stream temperature is estimated by: Tw (t) = 6.63 + 0.38 Ta. Where Tw is water temperature (°C) and Ta is air temperature (°C). This model was developed specifically to predict weekly stream temperatures from air temperatures for streams with substantial groundwater inputs (Krider et al., 2013). The 40 streams used to develop this composite linear model were from groundwater fed streams from southeastern Minnesota, a region dominated by karst topography (Krider et al., 2013). The standard deviation for this equation is 4.80°C. 25 I chose to pair these two equations because at summer air temperatures, Stefan and Preud’homme’s (1993) equation produces stream temperatures close to air temperature, indicating an equation that should predict temperature well for surface water dominated streams, with temperatures primarily driven by air temperatures. On the other hand, because Krider et al.’s (2013) equation was developed for groundwater dominated streams that fluctuate in temperature less with changing air temperature, due to continuous inputs of groundwater which remain at a relatively stable temperature throughout the year, the equation produces stream temperatures which are much cooler than air temperature. Essentially, the two equations form a range of projected stream temperatures based on the prominence of either air temperature or groundwater input as drivers of stream temperature. The Stefan and Preud’homme (1993) equation predicts temperatures for surface water dominated streams, those that are likely most reactive to changes in air temperature, while the Krider et al. (2013) equation predicts temperatures for groundwater dominated streams, which are likely less reactive to changes in air temperatures. I hypothesize that future stream temperatures for the high quality brook trout streams will fall within the range of temperatures predicted by the Stefan and Preud’homme (1993) equation as a maximum and the Krider et al. (2013) equation as a minimum, based on the relative importance of groundwater inputs and air temperature as drivers of stream temperature. Since there is nearly a 1:1 relationship between daily and monthly air and water temperatures (Ozaki et al., 2003), I assumed the weekly and monthly relationships to be comparable and substituted the projected monthly air temperature data from EFETAC in place of weekly averages called for by the Stefan and Preud’homme (1993) and Krider et al. (2013) equations when preforming the air-stream temperature conversion. 26 I considered the standard deviations of 2.16°C from Stefan and Preud’hommes’s (1993) equation and 4.80°C weekly air to water conversion model when analyzing the results of the study. Changes in temperature between years were compared to the equation’s standard deviation when analyzing the results produced by both Stefan and Preud’hommes’s (1993) (SD 2.16°C) equation and Krider et al.’s (2013) (SD 4.80°C) equation. Stream Temperature Projections I used an ensemble approach to report the stream temperatures predicted by the three CCMs to best reflect the range of possible changes in air temperature and to account for the uncertainty inherent in all climate models. The projected stream temperatures reported for 2006, 2012, 2026, and 2056 were based on the average of the air temperatures predicted by the three CCMs. For the sake of clarity and to emphasize the maximum stream temperatures brook trout are likely to face in each of the study years, only the stream temperature from the month with the warmest projected water temperature is reported. Since stream temperatures were predicted based on air temperature projections for the subbasin level, the projected stream temperatures reported reflect the temperatures of all streams within the same subbasin. Brook Trout Thermal Habitat Suitability I assigned a rating (status 1-4) of thermal habitat suitability to each subbasin for all four study years based on the reported projected stream temperatures. Habitat suitability ratings were based on the brook trout’s capacity for growth and survival at different temperatures (Table 5). Streams rated as “ status 1” were predicted to be between 11°C and 16.5°C, the optimal thermal range for brook trout growth (Raleigh, 1982). Streams assigned a rating of “ status 2” were predicted to be between 16.5°C and 20°C, where brook trout experience reduced rates of growth compared to the optimal growth range. Streams designated as “3” were predicted to be between 27 20.5°C and 25.3°C, within this temperature range brook trout are capable of survival, but the energetic demands of living above their optimal thermal range inhibit all growth (Baldwin, 1957). Streams designated as “4” were predicted to be 25.3°C or higher, indicating that brook trout were no longer capable of survival within the stream (Fry et al., 1946). TABLE 5. Thermal habitat status designations (thermal habitat status), their corresponding temperature range (temperature range), and the brook trout growth rate they represent (growth status) (Raleigh, 1982, Power, 1980, Fry et al., 1946). Brook Trout Thermal Habitat Status Designation Chart Thermal Habitat Status 1 2 3 4 Temperature Range 11≤°C≤16.5 16.5<°C≤20.5 20.5<°C≤25.3 °C≥25.3 Growth Status Highest rate of growth Reduced rate of growth No growth Extirpation Current Stream Temperatures To determine the accuracy of the model, I compared the reported projected5 stream temperatures for the years 2006 and 2012 to field measured (actual) average monthly stream temperatures for the same month and year. The field measured stream temperature data were provided by the NECSC database. Each gage within this database was identified by a NorEast ID. The NECSC database does not include gage data for the state of North Carolina, however. Therefore, the gage data used for Deep Creek, NC was retrieved from the United States Geological Survey’s (USGS) National Water Information System (http://waterdata.usgs.gov/nwis). 5 When field measured data were not available for the month projected to produce the warmest stream temperatures (reported stream temperature), I used the next warmest summer month for which field measured data was available. 28 Baseflow & Stream Order I hypothesized that since baseflow contribution and stream order can influence stream temperature, the results of the air temperature based SP model were likely to deviate from field measured stream temperatures. The field measured temperatures for streams with high levels of groundwater input would cooler than predicted by the model, and high order streams were likely to be more accurately predicted by the model. Therefore, I collected data on percent groundwater contribution within each high quality subwatershed (HUC12) and the Strahler stream order of each gaged stream. Percent groundwater contribution for each subwatershed was calculated as an area weighted average. Stream segments with known percent groundwater contribution to baseflow were multiplied by the local catchment area and the sum of these values was divided by the total area of the subwatershed. Strahler stream order for each gaged stream was determined using the ArcMap tool in ArcMap 10.2.1 software. Individual Models To further determine the accuracy of the models, individualized linear regression models were created using field measured stream temperature and air temperature data for a small subset of the high quality stream sites. I created linear regression models only for those streams represented by USGS operated stream temperature gages and retrieved available average monthly stream temperatures (°C) for the month of August between the years 2000 and 2014 using the USGS’s National Water Information System. I selected August because the three CCMs predict August to be the warmest month in most of the studied subbasins by 2056 and the outputs of the individualized linear models should reflect the most strenuous future thermal conditions for brook trout. 29 Average August air temperatures from 2000 to 2014 were retrieved from The National Oceanic and Atmospheric Administration’s (NOAA) National Climatic Data Center (http://www.ncdc.gov). This website allows access to weather gage stations across the United States and allows the user a variety of different methods in searching for the desired information. I chose to search for weather stations by subbasin. Since each subbasin included several weather gages, I chose one gage from each with collected monthly average air temperature data from between 2000 and 2014. If more than one gage fit these criteria, I chose the first gage listed. With field-measured air and stream temperatures, I created linear regression models to predict stream temperatures from air temperatures for each stream. Those with fewer than four data points were removed from this portion of the study. I input air temperatures projected by the three CCMs into each individualized linear regression model to predict stream temperatures for 2006 and 2056. Results Changes in Stream Temperature and Thermal Habitat Suitability for Brook Trout Stream temperatures predicted by the Stefan and Preud’homme6 (1993) air-water temperature linear regression model (SP model) increased for all 30 subbasins between the beginning of the study period in 2006 and the end in 2056; 29 of 30 increased by more than 2.16°C. The range of cumulative stream temperature change between these years was from 2.11°C to 4.16°C, with an average change of 3.08°C (Figure 10). Meduxnekeag (ME) was the only subbasin not to experience a projected stream temperature change greater than 2.16°C over the course of the study, with a total change of 2.11°C. 6 The SP model represents predominantly surface water streams, as stream temperature predictions using this model are much closer to air temperatures than those predicted by models developed for groundwater dominated streams, like the KEA model. 30 Stream Temperature (°C) Watershed Name 28 26 24 22 20 18 16 14 12 10 8 6 4 2 0 Buffalo-Whitewater, MN La Crosse-Pine, WI Lower Chippewa, WI Lower Wisconsin, WI Red Cedar, WI Wolf, WI Au Sable, MI Black, MI Dead-Kelsey, MI Escanaba, MI Thunder Bay, MI Aroostook, ME Lower Kennebec, ME Maine Costal, ME Meduxnekeag, ME Ausable, NY Chenango, NY Great Chazy-Saranac, NY Southern Long Island, NY Upper Genesee, NY Housatonic, CT Lower Connecticut, CT Quinnipiac, CT Saugatuck, CT Shetucket, CT Thames, CT Elk, WV Gauley, WV N. Branch Potomac, WV Lower Little Tennessee, NC 28 26 24 22 20 18 16 14 12 10 8 6 4 2 0 Buffalo-Whitewater, MN La Crosse-Pine, WI Lower Chippewa, WI Lower Wisconsin, WI Red Cedar, WI Wolf, WI Au Sable, MI Black, MI Dead-Kelsey, MI Escanaba, MI Thunder Bay, MI Aroostook, ME Lower Kennebec, ME Maine Costal, ME Meduxnekeag, ME Ausable, NY Chenango, NY Great Chazy-Saranac, NY Southern Long Island, NY Upper Genesee, NY Housatonic, CT Lower Connecticut, CT Quinnipiac, CT Saugatuck, CT Shetucket, CT Thames, CT Elk, WV Gauley, WV N. Branch Potomac, WV Lower Little Tennessee, NC Stream Temperature (°C) FIGURE 10. Stream temperatures projected for the years 2006 and 2056 for the 51 study streams using both the Stefan and Preud’homme (1993) (SP model) and Krider et al. (2013) (KEA model) air-stream stream temperature linear regression equations. Projected 2006 and 2056 Stream Temperatures: SP Model Stream Temperature 2006 (°C) Stream Temperature 2056 (°C) Projected 2006 and 2056 Stream Temperatures: KEA Model Stream Temperature 2006 (°C) Stream Temperature 2056 (°C) Watershed Name 31 Between 2006 and 2012, 26 subbasins increased in temperature and four decreased, however, none of these changes were greater than 2.16°C. The range of stream temperature change during this first study period was -0.87°C to 1.36°C, with an average change of 0.67°C. The first change in stream temperature greater than 2.16°C was recorded between the years 2012 and 2026, when the Lower Little Tennessee (NC) subbasin increased in temperature by 2.61°C. This second study period seemed to be characterized by a high number of stream cooling events, with 21subbasins cooling and only nine warming. The range of temperature change between 2012 and 2056 was -1.44°C to 1.36°C, with an average change of -0.05°C. Between 2026 and 2056 the subbasins experienced the greatest increase in stream temperature between study years. During this time, all 30 subbasins increased in temperature, with 21 of these changes being greater than 2.16°C. The range of temperature change was between 1.86°C and 3.32°C, with an average temperature change of 2.47°C. As a result of these changes in stream temperature, according to the SP model, 20 out of 30 subbasins containing streams currently listed as high quality brook trout habitat were given poorer habitat status designations at the end of the study in 2056 than at the beginning in 2006 (Figure 11). In essence, 20 out of 30 subbasins became warmer by more than 2.16°C and less conducive to brook trout growth and survival. Chronologically, changes in habitat status designation were rare at the beginning of the study and most prominently observed between the last years of the study (2026-2056), as a result of the high number of changes in stream temperature greater than 2.16°C during this period. As such, there were four changes in status designation during the 2006-2012 study period, one status change during the 2012-2006 study period, and 19 status changes during the 2026-2056 study period. 32 Figure 11. Change in thermal habitat status for brook trout between the years 2006 and 2056 predicted by the SP model. Thermal habitat status designations are described in Table 5. 33 Of the 20 subbasins that became less suitable for brook trout growth and survival between 2006 and 2056, eight subbasins initially supporting brook trout growth (status 2) in 2006 could no longer sustain brook trout growth by 2056 (status 3) and 12 subbasins with thermal conditions allowing for species survival in 2006 (status 3) increased in temperature beyond the brook trout’s upper thermal limit (status 4). Ten subbasins did not change status during this time period, although each increased in temperature. All ten remained at temperatures allowing for fish survival, but not growth (status 3) between the beginning and end of the study. The projected changes in stream temperature and thermal habitat status made by the SP model are listed in Table 6. On the other hand, as would be expected of a model developed to predict temperatures for groundwater dominated streams, stream temperatures projected using the Krider et al. (2013) linear regression model (KEA model) were cooler than those predicted by the SP model. Stream temperatures predicted for all 30 high quality subbasins increased, though not by more than the equation’s standard deviation between 2006 and 20567. The range of change between the beginning and end of the study was 0.94°C to 1.85°C, with an average change of 1.37°C (Figure 10). Since the same CCM projected air temperatures were used in calculating stream temperatures for both linear models (SP model and the KEA model), the same stream temperature fluctuations apparent in the SP model results were also present in the KEA results. Therefore, it is only pertinent to report the range and average temperature change and changes in thermal habitat status between study years for the KEA model. During the first study period, between 2006 and 2012, the changes in stream temperatures ranged from -0.39°C to 0.61°C, 7 When employing the KEA model predict stream temperatures, the model’s standard deviation used was 4.80°C. 34 Table 6. Average stream temperature (Average °C) for the warmest month (Month) projected by the three coupled climate models (CCMs) and the resulting thermal habitat status designation (Status) during the years 2006, 2012, 2026, and 2056 for each high quality HUC8. The air-water temperature conversion was conducted using Stefan and Preud’homme’s (1993) equation (SP). Projected Stream Temperatures (°C) and Thermal Habitat Status Designation (SP) Watershed Name Housatonic, CT Lower Connecticut, CT Quinnipiac, CT Saugatuck, CT Shetucket, CT Thames, CT Aroostook, ME Meduxnekeag, ME Lower Kennebec, ME Maine Costal, ME Dead-Kelsey, MI Escanaba, MI Black, MI Thunder Bay, MI Au Sable, MI Buffalo-Whitewater, MN Lower Little Tennessee, NC Ausable, NY Great Chazy-Saranac, NY Southern Long Island, NY Chenango, NY Upper Genesee, NY Wolf, WI La Crosse-Pine, WI Lower Chippewa, WI Red Cedar, WI Lower Wisconsin, WI North Branch Potomac, WV Gauley, WV Elk, WV Month 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 2006 SP Model 21.25 22.37 22.69 22.79 21.33 21.86 18.70 19.26 19.68 19.62 20.04 20.04 20.85 20.75 20.69 22.81 Status 3 3 3 3 3 3 2 2 2 2 2 2 3 3 3 3 Month 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 2012 SP Model 22.28 23.50 23.73 23.77 22.50 23.00 20.07 20.52 21.02 20.72 20.25 20.13 21.06 21.06 20.92 23.54 Status 3 3 3 3 3 3 2 3 3 3 2 2 3 3 3 3 Month 7 7 7 7 7 7 8 8 8 8 7 7 7 7 7 7 2026 SP Model 21.99 23.09 23.55 23.70 21.97 22.56 18.62 19.18 20.03 20.16 19.78 19.85 21.03 21.18 21.05 23.50 Status 3 3 3 3 3 3 2 2 2 2 2 2 3 3 3 3 Month 8 7 7 8 7 7 8 8 7 8 8 8 8 8 8 8 2056 SP Model 24.10 25.19 25.48 25.60 24.16 24.66 20.88 21.37 22.37 22.22 22.74 22.75 23.31 23.28 23.28 26.76 Status 3 3 4 4 3 3 3 3 3 3 3 3 3 3 3 4 8 7 7 7 7 7 7 7 7 7 7 23.60 18.75 19.41 23.30 20.90 20.78 21.88 23.00 22.39 22.05 22.78 3 2 2 3 3 3 3 3 3 3 3 7 7 7 7 7 7 7 7 7 7 7 22.73 19.96 20.62 24.28 21.53 20.85 22.47 23.73 23.21 22.86 23.47 3 2 3 3 3 3 3 3 3 3 3 7 7 7 7 7 7 7 7 7 7 7 25.34 19.20 19.89 24.23 21.34 21.26 22.20 23.87 22.81 22.35 23.87 4 2 2 3 3 3 3 3 3 3 3 7 8 8 8 8 8 8 8 8 8 8 27.63 21.74 22.45 26.27 23.91 23.85 25.13 26.94 26.03 25.66 26.94 4 3 3 4 3 3 3 4 4 4 4 7 8 7 22.00 21.20 22.61 3 3 3 7 7 7 21.92 21.06 22.56 3 3 3 7 7 7 23.35 22.64 24.12 3 3 3 8 8 8 25.21 25.15 26.68 4 4 4 35 with an average change of 0.30°C. Stream temperature change decreased during the second study period (2012-2026), cooling streams by -0.02°C on average. The range of change during this period was -0.64°C to 1.16°C. During the last study period (2026-2056) the greatest increases in stream temperature were observed, with temperature changes ranging from 0.83°C to 1.47°C and 1.10°C on average. It is important to note, however, than none of the changes in stream temperature were outside the equation’s standard deviation. When using the KEA model to predict stream temperatures, changes in stream temperature resulted in 15 changes in thermal habitat status between the beginning and the end of the study (2006-2056) (Figure 12). The first change in thermal habitat status, was again for the Lower Little Tennessee subbasin (NC) between 2012 and 2026, when it changed from optimal brook trout habitat (status 1) to habitat allowing for reduced growth rate (status 2). The remaining changes in habitat status occurred during the third study period between 2026 and 2056. All 15 reductions in thermal habitat status were a degradation from optimal habitat (status 1) to reduced growth conditions (status 2). The projected changes in stream temperature and thermal habitat status made by the KEA model are listed in Table 7. As a result of projected rising air temperatures, both the SP and KEA models predict reductions in thermal habitat quality for brook trout between 2006 and 2056. The SP model predicts that by 2056 40% of the high quality subbasins studied will be too warm to support brook trout populations during the warmest summer month, while brook trout inhabiting the remaining 60% of study subbasins will not be capable of growth during this period. The KEA model predicts that while 50% of studied subbasins will remain optimal thermal habitat for brook trout; thermal habitat quality for the other 50% of streams will be reduced. As such, I conclude that whether the high quality streams are predominantly surface water or groundwater 36 dominated, rising air temperatures related to climate change, overtime, will lead to substantial reductions in the quality of thermal habitat available to brook trout; ultimately resulting in extensive changes in their current distribution and productivity in the United States. Figure 12. Change in thermal habitat status for brook trout between the years 2006 and 2056 predicted by the KEA model. Thermal habitat status designations are described in Table 5. 37 Table 7. Average stream temperature (Average °C) for the warmest month (Month) projected by the three coupled climate models (CCMs) and the resulting thermal habitat status designation (Status) during the years 2006, 2012, 2026, and 2056 for each high quality HUC8. The air-water temperature conversion was conducted using Krider et al.’s (2013) equation (KEA). Projected Stream Temperatures (°C) and Thermal Habitat Status Designation (KEA) Watershed Name Housatonic, CT Lower Connecticut, CT Quinnipiac, CT Saugatuck, CT Shetucket, CT Thames, CT Aroostook, ME Meduxnekeag, ME Lower Kennebec, ME Maine Costal, ME Dead-Kelsey, MI Escanaba, MI Black, MI Thunder Bay, MI Au Sable, MI Buffalo-Whitewater, MN Lower Little Tennessee, NC Ausable, NY Great Chazy-Saranac, NY Southern Long Island, NY Chenango, NY Upper Genesee, NY Wolf, WI La Crosse-Pine, WI Lower Chippewa, WI Red Cedar, WI Lower Wisconsin, WI North Branch Potomac, WV Gauley, WV Elk, WV Month 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 2006 KEA Model 14.78 15.27 15.42 15.46 14.81 15.05 13.64 13.89 14.08 14.05 14.24 14.24 14.60 14.55 14.53 15.47 Status 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Month 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 2012 KEA Model 15.23 15.77 15.88 15.90 15.33 15.56 14.25 14.45 14.67 14.54 14.33 14.28 14.69 14.69 14.63 15.80 Status 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Month 7 7 7 7 7 7 8 8 8 8 7 7 7 7 7 7 2026 KEA Model 15.11 15.59 15.80 15.87 15.10 15.36 13.61 13.86 14.23 14.29 14.13 14.15 14.68 14.74 14.69 15.78 status 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Month 8 7 7 8 7 7 8 8 7 8 8 8 8 8 8 8 2056 KEA Model 16.04 16.53 16.65 16.71 16.07 16.29 14.61 14.83 15.27 15.21 15.44 15.44 15.69 15.68 15.68 17.22 Status 1 2 2 2 2 2 1 1 1 1 1 1 1 1 1 2 8 7 7 7 7 7 7 7 7 7 7 15.82 13.67 13.96 15.69 14.62 14.57 15.05 15.55 15.28 15.13 15.46 1 1 1 1 1 1 1 1 1 1 1 7 7 7 7 7 7 7 7 7 7 7 15.43 14.20 14.50 16.12 14.90 14.60 15.32 15.88 15.65 15.49 15.76 1 1 1 1 1 1 1 1 1 1 1 7 7 7 7 7 7 7 7 7 7 7 16.59 13.87 14.17 16.10 14.82 14.78 15.20 15.94 15.47 15.27 15.94 2 1 1 1 1 1 1 1 1 1 1 7 8 8 8 8 8 8 8 8 8 8 17.61 14.99 15.31 17.01 15.96 15.93 16.50 17.31 16.90 16.74 17.30 2 1 1 2 1 1 1 2 2 2 2 7 8 7 15.11 14.75 15.38 1 1 1 7 7 7 15.07 14.69 15.36 1 1 1 7 7 7 15.71 15.39 16.05 1 1 1 8 8 8 16.54 16.51 17.19 2 2 2 38 Stream Temperature Model Validation When I used the SP model to predict stream temperatures, of the 37 streams with field measured stream temperature available for the year 2006, the field measured temperatures of 16 streams (43%) were warmer than predicted (under-predicted) and 21 (57%) were cooler than predicted (over-predicted). The actual temperatures of the16 under-predicted streams ranged from 0.02°C to 5.30°C higher than projected, with an average under-prediction of 1.76°C. Of the 21 over-predicted streams, actual stream temperatures ranged from 0.04°C to 6.72°C cooler than predicted, with an average over-prediction of 2.39°C. Only 12 of the 37 streams (32%) were under or over-predicted, however, in accordance with the model’s standard deviation. The standard deviation of the SP model was 2.16°C. Therefore, I considered any actual stream temperature beyond +/- 2.16°C of the projection value to be under or over-predicted by this measure. Of these 12 streams, the field measured temperatures of four were warmer than the upper bound of the SP standard deviation by between 0.34°C and 3.14°C (average 1.29°C). The field measured temperatures of eight streams were cooler than predicted by between 0.92°C and 4.56°C (average 2.5°C) below the lower bound of the standard deviation (Figures 11 and 12). In 2012, like 2006, more streams were over predicted by model than were underpredicted. During this year, nine out of 22 streams were under-predicted by between 1.26°C and 4.26°C, with an average under-prediction of 2.59°C. The other thirteen streams were over predicted by between 0.55°C and 7.51°C, with an average over prediction of 3.01°C. Twelve of the 22 streams were over or under-predicted beyond the upper and lower bounds of the equation’s standard deviation in 2012. The four under-predicted streams were beyond the standard deviation by between 1.11°C and 2.10°C, with an average of 1.69°C beyond 39 the standard deviation. The eight over-predicted streams were between 0.15°C and 5.35°C beyond the standard deviation, with an average of 1.75°C beyond the lower limit of the standard deviation. Figures 13-16 depict validation charts for the SP model. The KEA model predicted much lower stream temperatures than the SP model. In 2006 all 37 streams were under predicted by the model by between 0.38°C and 10.87°C, with an average under-prediction of 5.88°C. Only 25 of these streams were cooler than the lower bound of the equation’s standard deviation (4.80°C), however. The 25 under-predicted streams were between 0.18°C and 6.07°C warmer than the upper limit of the standard deviation, with an average under-prediction of 5.88°C. Again, in 2012, all study streams with available data were under-predicted. The 22 streams with available field measured data in 2012 were between 0.60°C and 11.22°C warmer than predicted by the model. Of these 22 streams, 13 were under-predicted by between 0.61°C and 11.22°C beyond the standard deviation, with an average under-prediction of 5.87°C. Figures 17-20 depict validation charts for the KEA model. The field measured and projected stream temperatures (SP model and KEA model) for all sites in 2006 are summarized in Table 8 and those for 2012 are listed in Table 9. Individualized Air-Stream Temperature Linear Regression Models I created individualized air-stream temperature linear regression models (I-models) for eight streams, a subset of the 51 high quality brook trout streams selected for the study as a whole. Customized linear regression models best explain the relationship between air and stream temperatures for each individual stream, therefore they should more accurately predict stream temperatures than more generally applied models would. Thus, I created these I-models to compare with the SP model, to determine how consistent the results between the two models 40 Laurel, WV Bush, NY Sandy, ME Humaston, CT Jericho, CT Sutliffe, CT Good Hill, CT Bryant, CT Gilbert Bennett, CT Woods Pond, CT Morehouse, CT Cemetery, CT Riggs Street, CT Bunker Hill, CT E. Branch Ausable, NY Gulf, CT Mallory, CT Brooksvale, CT Saugatuck, CT Watermans, CT Ames, CT West, CT West Swamp, CT N. Branch Hamlin, CT Ballymahack, CT Mopang, ME Meduxnekeag, ME Salmon, ME Silver, MI Escanaba, MI Cady, WI Connors, WI Salmon Trout, MI Deep, NC Red, WI Embarrass, WI Au Sable, MI Stream Temperature (°C) FIGURE 13. Comparison between field measured (actual) stream temperature (°C) data and stream temperature projection projections made using the Stefan and Preud’homme d’homme (1993) (SP) model for the year 2006. Data is sorted by percent groundwater contribution to baseflow for or each stream from least to greatest. SP Model Validation and Baseflow Percentage: 2006 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 SP Prediction Actual Temperature Stream Name 41 Au Sable, MI Escanaba, MI Bryant, CT Bush, NY Connors, WI Good Hill, CT Jericho, CT Mallory, CT Meduxnekeag, ME Red, WI Silver, MI Sutliffe, CT Watermans, CT West, CT Woods Pond, CT Cemetery, CT E. Branch Ausable, NY Embarrass, WI Gilbert Bennett, CT Humaston, CT Riggs Street, CT Sandy, ME Brooksvale, CT Bunker Hill, CT Cady, WI Gulf, CT Laurel, WV Mopang, ME Morehouse, CT N. Branch Hamlin, CT Saugatuck, CT Salmon, ME Salmon Trout, MI West Swamp, CT Ames, CT Ballymahack, CT Stream Temperature (°C) FIGURE 14. Comparison between field measured (actual) stream temperature (°C) data and stream temperature projections made using the Stefan and Preud’homme d’homme (1993) (SP) model for the year 2006. Data is sorted by tthe he Strahler stream order for each gaged stream, from highest to lowest order. SP Model Validation and Stream Order: 2006 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 SP Prediction Actual Temperature Stream Name 42 FIGURE 15. Comparison between field measured (actual) stream temperature (°C) data and stream temperature projection projections made using the Stefan and Preud’homme d’homme (1993) (SP) model for the year 2012. Data is sorted by percent groundwater contribution to baseflow for each stream from least to greatest. Stream Temperature (°C) SP Model Validation and Baseflow Percentage: 2012 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 SP Prediction Actual Temperature Stream Name 43 FIGURE 16. Comparison between field measured (actual) stream temperature (°C) data and stream temperature projection projections made using the Stefan and Preud’homme d’homme (1993) (SP) model for the year 2012. Data is sorted by the Strahler stream order for each gaged stream, from highest to lowest order. Stream Temperature (°C) SP Model Validation and Stream Order: 2012 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 SW Prediction Actual Temperature Stream Name 44 Laurel, WV Bush, NY Sandy, ME Humaston, CT Jericho, CT Sutliffe, CT Good Hill, CT Bryant, CT Gilbert Bennett, CT Woods Pond, CT Morehouse, CT Cemetery, CT Riggs Street, CT Bunker Hill, CT E. Branch Ausable, NY Gulf, CT Mallory, CT Brooksvale, CT Saugatuck, CT Watermans, CT Ames, CT West, CT West Swamp, CT N. Branch Hamlin, CT Ballymahack, CT Mopang, ME Meduxnekeag, ME Salmon, ME Silver, MI Escanaba, MI Cady, WI Connors, WI Salmon Trout, MI Deep, NC Red, WI Embarrass, WI Au Sable, MI Stream Temperature (°C) FIGURE 17. Comparison between field measured (actual) stream temperature (°C) data and stream temperature projections made using the Krider et al. (2013) (KEA) model for the year 2006. Data is sorted by percent groundwater contribution to baseflow for each stream from least to greatest. KEA Model Validation and Baseflow Percentage: 2006 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 KEA Prediction Actual Temperature Stream Name 45 Au Sable, MI Escanaba, MI Bryant, CT Bush, NY Connors, WI Good Hill, CT Jericho, CT Mallory, CT Meduxnekeag, ME Red, WI Silver, MI Sutliffe, CT Watermans, CT West, CT Woods Pond, CT Cemetery, CT E. Branch Ausable, NY Embarrass, WI Gilbert Bennett, CT Humaston, CT Riggs Street, CT Sandy, ME Brooksvale, CT Bunker Hill, CT Cady, WI Gulf, CT Laurel, WV Mopang, ME Morehouse, CT N. Branch Hamlin, CT Saugatuck, CT Salmon, ME Salmon Trout, MI West Swamp, CT Ames, CT Ballymahack, CT Stream Temperature (°C) FIGURE 18. Comparison between field measured (actual) stream temperature (°C) data and stream temperature projections made using the Krider et al. (2013) (KEA) model for the year 2006. Data is sorted by the Strahler stream order for each gaged stre stream, from highest to lowest order. KEA Model Validation and Stream Order: 2006 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 KEA Prediction Actual Temperature Stream Name 46 FIGURE 19. Comparison between tween field measured (actual) stream temperature (°C) data and stream temperature projections made using the Krider et al. (2013) (KEA) model for the year 2012. Data is sorted by percent groundwater contribution to baseflow for each stream from least to greatest. Stream Temperature (°C) KEA Model Validation and Baseflow Percentage: 2012 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 KEA Prediction Actual Temperature Stream Name 47 FIGURE 20. Comparison between field measured (actual) stream temperature (°C) data and stream temperature projections made using the Krider et al. (2013) (KEA) model for the year 2012. Data is sorted by the Strahler stream order for each gaged stream, from highest to lowest order. Stream Temperature (C) KEA Model Validation and Stream Order: 2012 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 KEA Prediction Actual Temperature Stream Name 48 TABLE 8. Average projected stream temperatures for the years 2006 and 2012 using both the air-stream temperature conversion equations, Stefan and Preud’homme’s (1993) equation (Pro. SP) and et al. (2013) equation (Pro. KEA), compared with field measured stream temperatures for both years (Actual). The month (Month) projected and actual gage temperatures are provided for is also included. “x” represents no data. Actual Vs. Projected Stream Temperatures (°C) 2006 Stream Name Watershed Name Month Actual SP Model KEA Model Actual-SP Model Actual-KEA Model Good Hill Brook, CT Housatonic 7 21.27 21.25 14.78 0.02 6.49 Humaston Brook, CT Housatonic 7 22.15 21.25 14.78 0.90 7.37 Jericho Brook, CT Housatonic 7 23.27 21.25 14.78 2.02 8.49 Mallory Brook, CT Housatonic 7 23.08 21.25 14.78 1.83 8.30 Riggs Street Brook, CT Housatonic 7 17.39 21.25 14.78 -3.86 2.62 Sutliffe Brook, CT Housatonic 7 22.85 21.25 14.78 1.60 8.07 W. Brook, CT 7 23.03 21.25 14.78 1.79 8.26 7 x 22.37 15.27 x x W. Swamp Brook, CT Housatonic Lower Connecticut Lower Connecticut 7 18.95 22.37 15.27 -3.42 3.68 Brooksvale Stream, CT Quinnipiac 7 22.15 22.69 15.42 -0.54 6.73 Bunker Hill Brook, CT Quinnipiac 7 16.36 22.69 15.42 -6.33 0.94 Gulf Brook, CT Quinnipiac 7 23.62 22.69 15.42 0.94 8.21 N. B. Hamlin Brook, CT Quinnipiac 7 21.41 22.69 15.42 -1.27 6.00 Watermans Brook, CT Quinnipiac 6 18.73 20.54 14.46 -1.81 4.27 Bryant Brook, CT Saugatuck 7 21.87 22.79 15.46 -0.92 6.41 Cemetery Brook, CT Saugatuck 7 22.43 22.79 15.46 -0.36 6.97 Gilbert Bennett Brook, CT Saugatuck 7 22.75 22.79 15.46 -0.04 7.29 Morehouse Brook, CT Saugatuck 7 21.60 22.79 15.46 -1.19 6.14 Broad Swamp Brook, CT N. B. W. B. Saugatuck River, CT Saugatuck 7 21.55 22.79 15.46 -1.24 6.09 Woods Pond Brook, CT Saugatuck 7 22.04 22.79 15.46 -0.75 6.58 Ames Brook, CT Shetucket 7 21.84 21.33 14.81 0.52 7.03 Ballymahack Brook, CT Shetucket 7 22.56 21.33 14.81 1.24 7.75 Lake Pond Brook, CT Thames 7 x 21.86 15.05 x x Salmon Brook, ME Aroostook 7 20.35 18.70 13.64 1.65 6.71 Meduxnekeag River, ME 7 22.55 19.26 13.89 3.29 8.66 Sandy River, ME Meduxnekeag Lower Kennebec 8 16.02 19.10 13.82 -3.08 2.20 Mopang Stream, ME Maine Costal 7 24.92 19.62 14.05 5.30 10.87 Old Stream, ME Maine Costal 7 x 19.62 14.05 x x Pleasant River, ME Silver River, MI Maine Costal 7 x 19.62 14.05 x x Dead-Kelsey 8 18.74 18.88 13.76 -0.14 4.98 Salmon Trout River, MI Dead-Kelsey 7 14.88 20.04 14.24 -5.16 0.64 Escanaba River, MI Escanaba 6 19.09 16.40 12.62 2.69 6.47 49 Table 8 (cont’d). Black River, MI Comstock Creek, MI Black 7 x 20.85 14.60 x x Thunder Bay 7 x 20.75 14.55 x x Au Sable BuffaloWhitewater Lower Little Tennessee 8 17.72 19.58 14.04 -1.87 3.68 7 x 22.81 15.47 x x 8 26.10 23.60 15.82 2.49 10.27 8 17.12 17.91 13.29 -0.79 3.83 7 x 19.41 13.96 x x 7 x 23.30 15.69 x x 7 x 20.90 14.62 x x Bush Brook, NY Chenango Upper Genesee 7 19.07 20.78 14.57 -1.71 4.50 Embarrass River, WI Wolf 7 23.32 21.88 15.05 1.44 8.27 Red River, WI Wolf La CrossePine Lower Chippewa 7 22.28 21.88 15.05 0.40 7.23 7 x 23.00 15.55 x x 7 15.67 22.39 15.28 -6.72 0.38 Au Sable River, MI Snake Creek, MN Deep Creek, NC E. B. Ausable River, NY True Brook, NY Carmans River, NY Otselic River, NY Silver Creek, WI Cady Creek, WI Connors Creek, WI Ausable Great ChazySaranac Southern Long Island 7 16.60 22.05 15.13 -5.45 1.47 7 x 22.78 15.46 x x Johnnycake Run, WV Red Cedar Lower Wisconsin North Branch Potomac 7 x 22.00 15.11 x x N. Fork Cranberry River, WV Gauley 8 x 21.20 14.75 x x Laurel Fork, WV Elk 8 18.99 22.56 15.36 -3.57 3.63 Black Earth Creek, WI TABLE 9. Average projected stream temperatures for the years 2006 and 2012 using both the air-stream temperature conversion equations, Stefan and Preud’homme’s (1993) equation (Pro. SP) and et al. (2013) equation (Pro. KEA), compared with gage measured stream temperatures for both years (Actual). The month (Month) projected and actual gage temperatures are provided for is also included. “x” represents no data. Actual Vs. Projected Stream Temperatures (°C) 2012 Stream Name Watershed Name Month Actual SP Model KEA Model Actual-SP Model Actual-KEA Model Good Hill Brook, CT Housatonic 7 x 22.28 15.23 x x Humaston Brook, CT Housatonic 7 x 22.28 15.23 x x Jericho Brook, CT Housatonic 7 x 22.28 15.23 x x Mallory Brook, CT Housatonic 7 x 22.28 15.23 x x Riggs Street Brook, CT Housatonic 7 x 22.28 15.23 x x Sutliffe Brook, CT Housatonic 7 x 22.28 15.23 x x W. Brook, CT Housatonic Lower Connecticut 7 x 22.28 15.23 x x 7 26.77 23.50 15.77 3.28 11.00 Broad Swamp Brook, CT 50 Table 9 (cont’d). W. Swamp Brook, CT Lower Connecticut 7 x 23.50 15.77 x x Brooksvale Stream, CT Quinnipiac 7 x 23.73 15.88 x x Bunker Hill Brook, CT Quinnipiac 7 x 23.73 15.88 x x Gulf Brook, CT Quinnipiac 7 x 23.73 15.88 x x N. B. Hamlin Brook, CT Quinnipiac 7 x 23.73 15.88 x x Watermans Brook, CT Quinnipiac 7 x 23.73 15.88 x x Bryant Brook, CT Saugatuck 7 x 23.77 15.90 x x Cemetery Brook, CT Saugatuck 7 x 23.77 15.90 x x Gilbert Bennett Brook, CT Saugatuck 7 x 23.77 15.90 x x Morehouse Brook, CT Saugatuck 7 x 23.77 15.90 x x N. B. W. B. Saugatuck River, CT Saugatuck 7 x 23.77 15.90 x x Woods Pond Brook, CT Saugatuck 7 x 23.77 15.90 x x Ames Brook, CT Shetucket 7 x 22.50 15.33 x x Ballymahack Brook, CT Shetucket 7 x 22.50 15.33 x x Lake Pond Brook, CT Thames 7 26.77 23.00 15.56 3.77 11.22 Salmon Brook, ME Aroostook 7 x 20.07 14.25 x x Meduxnekeag River, ME 7 21.78 20.52 14.45 1.26 7.33 Sandy River, ME Meduxnekeag Lower Kennebec 7 x 21.02 x x x Mopang Stream, ME Maine Costal 7 x 20.72 14.54 x x Old Stream, ME Maine Costal 8 18.96 19.50 14.00 -0.55 4.95 Pleasant River, ME Silver River, MI Maine Costal 7 22.81 20.72 14.54 2.08 8.26 Dead-Kelsey 7 21.96 20.25 14.33 1.72 7.63 Salmon Trout River, MI Dead-Kelsey 7 15.90 20.25 14.33 -4.35 1.57 Escanaba River, MI Escanaba Black Watershed 7 24.39 20.13 14.28 4.26 10.11 7 15.30 21.06 14.69 -5.76 0.61 Comstock Creek, MI Thunder Bay 7 25.15 21.06 14.69 4.09 10.46 Au Sable River, MI Au Sable BuffaloWhitewater Lower Little Tennessee 8 17.58 19.39 13.95 -1.81 3.63 6 17.15 19.46 13.98 -2.31 3.17 7 24.15 22.73 15.43 1.42 8.72 7 x 19.96 14.20 x x 7 18.11 20.62 14.50 -2.51 3.62 7 16.77 24.28 16.12 -7.51 0.65 7 20.96 21.53 14.90 -0.58 6.05 Bush Brook, NY Chenango Upper Genesee 7 19.46 20.85 14.60 -1.38 4.87 Embarrass River, WI Wolf 7 x 22.47 15.32 x x Red River, WI Wolf La CrossePine 7 23.86 22.47 15.32 1.39 8.54 7 20.18 23.73 15.88 -3.55 4.30 Black River, MI Snake Creek, MN Deep Creek, NC E. B. Ausable River, NY True Brook, NY Carmans River, NY Otselic River, NY Silver Creek, WI Ausable Great ChazySaranac Southern Long Island 51 Table 9 (cont’d). Lower Chippewa Cady Creek, WI Connors Creek, WI 7 x 23.21 15.65 x x 7 x 22.86 15.49 x x 7 20.26 23.47 15.76 -3.20 4.50 Johnnycake Run, WV Red Cedar Lower Wisconsin North Branch Potomac 7 19.41 21.92 15.07 -2.51 4.34 N. Fork Cranberry River, WV Gauley 7 17.91 21.06 14.37 -3.15 3.54 Laurel Fork, WV Elk 7 x 22.56 14.37 x x Black Earth Creek, WI were and to determine whether or not using generally applied models is an effective way to predict stream temperatures. The slope values for the I-models ranged from 0.47 to 1.10 and yintercept values ranged from 0.84-11.61 (Table 10; Figures 21-28). TABLE 10. Individualized linear regression models for eight streams created using August monthly air temperature averages collected from NOAA’s National Climatic Data Center (http://www.ncdc.gov) and August monthly stream temperature averages from the USGS’s National Water Information System (http://waterdata.usgs.gov/nwis). All available air and stream temperature data from the year 2000 to 2014 was used. R2 values for each individualized linear regression model are also included in the table. Individualized Linear Regression Models Stream Name Broad Swamp Brook, CT Meduxnekeag River, ME Silver River, MI Escanaba River, MI Au Sable River, MI Deep Creek, NC Embarrass River, WI Red River, WI Linear Regression Models 1.10x+0.84 0.87x+4.04 0.94x+1.60 0.47x+11.61 0.63x+6.08 0.88x+2.85 0.53x+10.65 0.92x+2.58 R2 0.87 0.80 0.89 0.40 0.74 0.68 0.50 0.96 I compared the outputs of these I-models to field measured stream temperatures in 2006 and 2012, using air temperatures projected by the CCMs as the input values. Of the five streams with field measured stream temperature data in 2006, four were over predicted and one was under predicted by the models. Field measured temperatures for the over predicted streams were 52 FIGURE 21. Individualized air-stream stream temperature (°C) linear regression model for Broad Swamp Creek, Connecticut (CT). y = 1.0958x + 0.8383 R² = 0.8673 Broad Swamp Brook, CT Stream Temperature (°C) 29 27 25 23 21 19 17 15 15 17 19 21 23 Air Temperature (°C) 25 27 29 FIGURE 22. Individualized air-stream stream temperature (°C) linear regression model for Meduxnekeag River, Maine (ME). y = 0.874x + 4.0375 R² = 0.8009 Meduxnekeag River, ME Stream Temperature (°C) 29 27 25 23 21 19 17 15 15 17 19 21 23 Air Temperature (°C) 53 25 27 29 FIGURE 23. Individualized air air-stream stream temperature (°C) linear regression model for Silver River, Michigan (MI). Silver River, MI y = 0.9425x + 1.5979 R² = 0.8855 Stream Temperature (°C) 29 27 25 23 21 19 17 15 15 17 19 21 23 Air Temperature (°C) 25 27 29 FIGURE 24. Individualized air air-stream temperature (°C) linear regression model for Escanaba River, Michigan (MI). y = 0.4685x + 11.609 R² = 0.4014 Escanaba River, MI Stream Temperature (°C) 29 27 25 23 21 19 17 15 15 17 19 21 23 Air Temperature (°C) 54 25 27 29 FIGURE 25. Individualized air-stream stream temperature (°C) linear regression model for Au Sable River, Michigan (MI). Au Sable River, MI y = 0.6256x + 6.077 R² = 0.737 Stream Temperature (°C) 29 27 25 23 21 19 17 15 15 17 19 21 23 Air Temperature (°C) 25 27 29 FIGURE 26. Individualized air-stream stream temperature (°C) linear regression model for Deep Creek, North Carolina (NC). y = 0.8811x + 2.8519 R² = 0.6766 Deep Creek, NC Stream Temperature (°C) 29 27 25 23 21 19 17 15 15 17 19 21 23 Air Temperature (°C) 55 25 27 29 FIGURE 27. Individualized air-stream stream temperature (°C) linear regression model for Embarrass River, Wisconsin (WI). y = 0.5304x + 10.654 R² = 0.5021 Embarrass River, WI Stream Temperature (°C) 29 27 25 23 21 19 17 15 15 17 19 21 23 Air Temperature (°C) 25 27 29 Figure 28. Individualized air-stream stream temperature (°C) C) linear regression model for Red River, Wisconsin (WI). y = 0.9192x + 2.583 R² = 0.96313 Red River, WI Stream Temperature (°C) 29 27 25 23 21 19 17 15 15 17 19 21 23 Air Temperature (°C) 56 25 27 29 between 0.25°C and 0.87°C cooler than predicted. The over projected stream was warmer than predicted by 2.04°C. Overall, on average 2006 stream temperature predictions were 0.01°C cooler than predicted by the individualized models. Of the six streams with available stream temperature data in 2012, the temperatures of two streams were over predicted and the temperatures of four streams were under predicted. The two that were cooler than predicted were over predicted by 0.10°C and 0.49°C. Stream temperatures of the four under predicted streams ranged from 0.99°C to 2.73°C warmer than predicted. The average difference between field measured and individualized projections in 2012 was 0.97°C. In 2006 stream temperature status was consistent for three out of five streams between the status predicted by the individualized model and the status corresponding with the field measured stream temperature. One of the remaining streams was projected to be a status 3 stream (no growth), while the status corresponding with the actual stream temperature was a 4 (extirpation). It is important to note, however, that the predicted stream temperature was 2.04°C cooler than the field measured stream temperature and 1.24°C below the threshold of status 4. The other remaining stream was predicted to be a status 3 (no growth) by the individualized model, but the status corresponding to the field measured stream temperature was a status 2 (reduced growth). The predicted temperature was 0.87°C warmer than field measured stream temperature and 0.06°C above the threshold for status 3. In 2012, stream temperature was consistent for three out of six streams. Temperature for the remaining three streams were under predicted by the model, meaning the field measured stream temperatures were warmer than those predicted by the individualized models. Two streams predicted to be status 2 (reduced growth), were status 3 (no growth) according to field 57 recorded temperatures. The first of these two streams was 0.99°C warmer than predicted, and the prediction was 0.22°C below the threshold for status 3. The second stream was 2.73°C warmer than predicted, and the prediction was 0.82°C below the threshold for status 3. The third stream with a disparity between the predicted and field measured status was predicted to be a status 3 (no growth) stream, but field measured temperatures resulted in a status designation of 4, with an actual stream temperature 1.53°C warmer than predicted. The prediction was 0.59°C below the threshold for status 4. The stream temperatures project by the I-model projected and SP model are compared with field measured stream temperatures for 2006 and 2012 in Table 11. I also compared the stream temperature values produced by the individualized air-stream temperature linear regression models and the Stefan and Preud’homme (1993) linear regression model for beginning and end years of the study in 2006 and 2056 to determine how accurately the SP model predicted stream temperatures compared to the I-models. In 2006, the range of stream temperature difference (I-model output – SP-model output) was between -1.37°C and 3.12°C, with an average difference of 0.90°C. In 2056, the range of stream temperature difference between the two models ranged from -2.38°C and 4.01°C, with an average difference of 0.54°C. Importantly, only two streams in 2006 and two streams in 2056 had discrepancies between the status designations resulting from the individualized model and the SP model. In all four cases, the individualized models predicted a higher status designation than the SP model did. In 2006, two streams designated as status 3 by individualized models were designated as status 2 by the SP model. The individualized models predicted the streams to be 1.13°C and 0.87°C warmer than the SP model did. In both cases the SP model produced stream temperatures 1.31°C below the status 3 threshold. In 2056, two streams designated as status 4 (extirpation) by 58 their individualized models were designated as status 3 (no growth) by the SP model. The individualized models predicted the streams to be 4.01°C and 1.21°C warmer than the SP model did. The SP model produced stream temperatures 0.25°C and 0.17°C below the status 4 threshold. The stream temperatures projected for 2006 and 2056 using the I-models and SP model are listed in Table 12. Discussion Changing Stream Temperatures and Brook Trout Thermal Habitat Based on the results of this study, I predict an increase in stream temperature for all 30 studied subbasins (HUC8s) containing high quality brook trout streams between the years 2006 and 2056. While the magnitude of this change differs between the results produced by the SP model and the KEA model, both predict a general increase in stream water temperatures. The change in temperature between these years ranged from 2.11°C to 4.16°C, with an average change of 3.08°C according to the results of the SP model (Table 6; Figure 10) and ranged from 0.94°C to 1.85°C, with and average change of 1.37°C according to the results of the KEA model (Table 7; Figure 10). While none of the changes in stream temperature were outside of the equation’s standard deviation when using the KEA model, they were outside of this range for 29 of the 30 subbasins when using the SP model. These results indicate that rising air temperatures due to climate change will result in warmer stream water temperatures across the brook trout’s native U.S. range over the course of the next half-century. Consequently, increasing stream temperatures over this time period are projected to result in thermal habitat conditions that will be less conducive to brook trout growth and survival than they are currently for 20 of the 30 study subbasins according to the SP model and for 15 of 30 subbasins according to the KEA model. The SP model predicts thermal habitat quality for eight 59 Table 11. Comparison between stream temperatures and thermal habitat ratings projected by the individualized models (I-models), the Stephan and Preud’homme (1993) model (SP model), and field measured (Actual) temperature for 2006 and 2012. Individualized Model vs. Stefan and Preud'homme (1993) vs. Field Measured Stream Temperature (°C) Projections 2006 2012 Stream Name Broad Swamp Brook, CT Meduxnekeag River, ME Silver River, MI Escanaba River, MI Au Sable River, MI Deep Creek, NC Embarrass River, WI Red River, WI I Model x 19.82 19.19 x 18.21 24.06 x 20.56 Status x 2 2 x 2 3 x 3 SP Model x 18.43 18.96 x 19.58 23.60 x 20.19 Status x 2 2 x 2 3 x 2 Actual x 19.57 18.74 x 17.72 26.10 x 19.69 Status x 2 2 x 2 4 x 2 I Model 24.71 19.68 19.11 20.28 18.07 23.01 x x Status 3 2 2 2 2 3 x x SP Model 21.64 18.30 18.88 18.81 19.39 22.57 x x Status 3 2 2 2 2 3 x x Actual 26.24 22.41 19.01 21.27 17.58 24.15 x x Status 4 3 2 3 2 3 x x Table 12. Comparison between stream temperautres and thernmal habitat ratings projected by the individualized models (I-models) and the Stephan and Preud’homme (1993) model (SP model) for the years 2006 and 2056. Individualized Model vs. Stefan and Preud'homme (1993) Stream Temperature (°C) Projections 2006 2056 Stream Name I Model Status SP Model Status I model-SP Model I model Status SP Model Status I model-SP Model Broad Swamp Brook, CT 24.93 3 21.81 3 3.12 29.06 4 25.05 3 4.01 Meduxnekeag River, ME 19.82 2 18.43 2 1.39 22.80 3 21.37 3 1.44 Silver River, MI 19.19 2 18.96 2 0.24 23.34 3 22.74 3 0.60 Escanaba River, MI 20.39 2 19.01 2 1.37 22.42 3 22.75 3 -0.33 Au Sable River, MI 18.21 2 19.58 2 -1.37 20.90 3 23.28 3 -2.38 Deep Creek, NC 24.06 3 23.60 3 0.46 27.74 4 27.20 4 0.55 Embarrass River, WI 21.32 3 20.19 2 1.13 24.36 3 25.13 3 -0.77 Red River, WI 21.06 3 20.19 2 0.87 26.34 4 25.13 3 1.21 60 subbasins will degrade from supporting brook trout growth (status 2) in 2006 to conditions too warm for fish growth (status 3) by 2056. The 12 remaining streams predicted to experience changes in water temperature beyond the equation’s standard deviation between 2006 and 2056, will degrade from allowing brook trout survival without growth (status 3) to temperatures beyond the species’ upper thermal limits (status 4). In light of these results, I conclude that climate change will have a considerable impact on the quantity and distribution of thermal habitat available for brook trout in the United States. Other studies related to climate change and coldwater fishes in the U.S. support this conclusion. For example, Steen et al. (2010) produced two models to predict the future thermal habitat available to coldwater fishes in the midwestern state of Michigan. Both models included air temperature as a driver of stream temperature; one assumed a 3°C increase in air temperature by 2100 and the other assumed land cover change and a 5°C increase in air temperature over the same time period. Both models predict substantial reductions in the presence of coldwater fish species within Michigan’s Muskegon River watershed by the end of the 21st century and a 90% decrease in the probability of brook trout occurrence in one of the tributary streams studied (Steen et al., 2010). Additionally, Flebbe et al. (2006) predicted a 53% loss of brook trout habitat area along the southern end of the Appalachian Mountains by 2100 using data from climate projections by the more conservative Hadley Centre global circulation model and a 97% loss when assuming the projections of the more severe Canadian Centre global circulation model. The results from this study also suggest potential shifts in brook trout distribution and productivity, with fewer areas in the U.S. providing for their thermal habitat requirements by mid-century. Brook trout distribution is likely to contract in areas where summer stream 61 temperatures already result in poor or marginal habitat for this species. Near the southern limit of the brook trout’s native range, for example, current average summer air temperatures already result in water temperatures near the species’ upper thermal limit (Flebbe et al., 2006, Meisner, 1990a). Field recorded stream temperature for Deep Creek (NC), the southernmost stream in this study, provides an example of marginal summer habitat, which is likely to warm past the brook trout’s thermal tolerance in the near future. Deep Creek was consistently the warmest stream both in terms of field measured stream temperatures taken in 2006 and 2012 and in terms of stream temperature predictions across all study years. These potential distributional shifts will be driven by the lethal and sub-lethal effects of the increased metabolic costs associated with increased stream temperatures (Flebbe et al., 2006), including limited growth (Baldwin, 1957). In addition to the impact on growth and survival, I expect the brook trout’s reproductive ability will also be compromised at high temperatures because summer water temperatures exceeding 19°C for extended periods have been reported to hinder gametogenesis (Hokanson et al., 1973). According to the results of the SP model, average monthly stream temperature will exceed 19°C in all 30 subbasins during the warmest summer month by 2056. While the study suggests thermal habitat quality will decline in many portions of brook trout’s native range, conversely, thermal habitat could improve for brook trout populations in areas where their growth and reproduction are currently constrained by low temperatures (Meisner, 1990a, Meisner et al., 1988). Stream warming in these cooler regions of North America will increase the potential for brook trout to expand their range and increase productivity. Stream Temperature Model Validation 62 The results of the stream temperature model validation portion of this study suggest that the methodology used in this study is a reasonable way to predict current and future stream temperatures. The SP model produced results that were on average 0.60°C warmer than field measured temperatures in 2006 and 0.72°C warmer than field measured temperatures in 2012. The KEA model on the other hand, tended to predict stream temperatures to be cooler than field measured temperatures. Stream temperature predicted by the KEA model in 2006 were on average 5.88°C cooler than field measured stream temperature and were 5.87°C cooler than field measured temperature on average in 2012. The cooler temperatures predicted by the KEA model are not surprising considering that it was developed for streams receiving high levels of groundwater inputs (Krider et al., 2013). High levels of groundwater input tend to maintain streams at more constant temperatures throughout the year (Allan, 1995), making them less responsive to changes in air temperature, as I observed when comparing the results of the KEA model to the results of the SP model, which predicted stream temperatures that were much closer to air temperatures. While the SP model did over predict stream temperatures for some streams and under predict temperatures for others (Figures 11-14), the KEA model under predicted stream temperatures for all study streams in both 2006 and 2012 (Figures 15-18). This result, in conjunction with the SP model’s lower average discrepancy between field measured and predicted stream temperatures in both 2006 and 2012, indicates that the temperatures of my study streams were influenced more strongly by air temperatures than by groundwater inputs. The compelling exceptions to this conclusion were, Carmans River (NY) and Black River (MI), the two 1st order streams with baseflow consisting of ≥80% groundwater input (Figure 13, 14, 17, and 18). Stream temperatures for both sites were well predicted by the KEA model. The 63 stream temperature predicted for Carmans River in 2012 was under predicted by 0.65°C by the KEA model and over predicted by 7.51°C by the SP model. Similarly, in 2012 the KEA model under predicted the temperature of Black River by 0.61°C and the SP model over predicted its temperature by 5.76°C. Perhaps these two streams were predicted more accurately by the KEA model because they more closely resemble the streams found in the karst topography of southeastern Minnesota that the model was originally developed for. In any case, the SP model predicted stream temperatures better than the KEA model for the majority of my study sites. As such, I would encourage management agencies to choose to predict current and future stream temperatures using an air-water temperature linear regression model developed for surface water dominated streams. Since air temperature appears to be the dominant driver of stream temperature, I also suggest that management agencies prioritize strategies to mitigate and reduce air temperature directly above coldwater streams. Individualized Air-Water Temperature Linear Regression Models Stream temperature data is often difficult to obtain, since stream gages are often sporadically placed and may not collect data regularly over time. Additionally, maintaining stream gages is costly and lack of funding has recently led to reduced access to current gage recorded data (USGS, 2014). Air temperature data, on the other hand, are readily available and collected by a variety of institutions (weather stations, airports, etc.). The ease of access to air temperature data across the nation makes predicting stream temperatures from air temperatures using linear regression models attractive when actual stream temperature data are scarce. Additionally, the availability of future air temperature projections, like those produced by the three CCMs used in this study, makes the capacity to predict stream temperatures from air temperatures important. 64 A linear regression model developed by fitting a trend line to a plot of local historical air temperatures and concurrent stream temperature data would likely provide the most accurate means of predicting stream temperatures from air temperatures for a specific stream. The lack of available historical stream temperature data can be limiting in this regard, however, making it difficult to develop linear regression models for each stream of interest in large-scale management or research efforts. Even if the appropriate data were available, compiling air and stream temperature data, creating individualized linear regression models, and then customized future stream temperature predictions on a stream-by-stream basis could be very time consuming. Therefore, a linear regression model capable of predicting future temperatures for a wide range of streams would be a valuable tool for developing future brook trout management plans. Since I previously concluded the SP model to be more accurate in predicting temperatures for the streams included in this study, I chose to compare the outputs of this model to those produced by the individualized linear regression models. The purpose of this effort was to test the efficacy of the SP model for predicting stream temperatures compared to individualized linear regression models developed for each stream. I compared the values predicted by both models types for monthly average August stream temperatures during the years 2006 and 2056. In 2006 the range of temperature difference between the predictions from the two models was between -1.37°C and 3.12°C, with an average difference of 0.90°C. In 2056, the range of stream temperature difference between the two models ranged from -2.38°C and 4.01°C, with an average difference of 0.54°C. More biologically significant for brook trout than the differences in predicted stream temperatures between the two model types, are differences in stream temperature statuses 65 predicted by the models. In both 2006 and 2056, two of the eight streams had discrepancies in habitat status predictions between the two models. In all four cases the individualized models predicted a poorer habitat status designation than the SP model did, making the SP model generally more conservative than the individualized models. In 2006, two streams designated as status 3 by individualized models were designated as status 2 by the SP model. The individualized models predicted the streams to be 1.13°C and 0.87°C warmer than the SP model did. In both cases the SP model produced stream temperatures 1.31°C below the status 3 threshold, a value less than the standard deviation (2.16°C) associated with the SP model. In 2056, two streams designated as status 4 (extirpation) by their individualized models were designated as status 3 (no growth) by the SP model. The individualized models predicted the streams to be 4.01°C and 1.21°C warmer than the SP model did. The SP model produced stream temperatures 0.25°C and 0.17°C below the status 4 threshold, again values again less than the standard deviation of the SP model. Since the SP model did predict the same habitat statuses as the individualized models for six out of eight streams in both 2006 and 2056 and since in those cases when the two models did not predict the same status, the difference between the SP predicted value and the threshold of the status predicted by the individualized model was less than the standard deviation associated with the SP model. As such, the SP model appears to be an acceptable method of stream temperature prediction for those streams assessed. Factors influencing stream temperature While the SP and individualized models provide useful tools to estimate the potential effects of increasing air temperature on brook trout thermal habitat, air temperature is only one of a suite of potential factors influencing stream water temperatures. The variability between field 66 measured and SP model predicted stream temperatures in 2006 and 2012 supports this assertion. Field measured stream temperatures differed from the SP predicted values by more than 2.16°C (the equation’s standard deviation) for 32% of streams in 2006 and for 55% of streams in 2012, indicating the presence of additional factors influencing stream temperature. The R2 values of the individualized air-stream temperature linear regression equations also suggest that additional factors influence stream temperature. If stream temperatures were entirely determined by air temperature, then the R2 values of the equations would be equal to 1. Other key factors likely influencing stream temperatures include riparian zone shading and the magnitude of groundwater input. Vegetation providing overhead riparian zone shading, for example, can aid in cooling stream water by intercepting ultraviolet rays that would otherwise warm the stream surface on contact (Blann et al., 2002). Groundwater, which often provides streams with a relatively consistent source of water with temperatures generally within 1°C of the average annual air temperature throughout the year provides a thermal buffer, keeping streams warmer during the relatively cold air temperatures of the winter months and cooler during the relatively hot air temperatures of the summer months (Allan, 1995). Baseflow and Stream Order Since the SP model predicted temperature for the majority of streams more accurately than the KEA model, we can conclude that while groundwater likely plays a role in determining stream temperature, for the selected study streams air temperature exerted a greater influence on stream temperature than groundwater. Nevertheless, the role of groundwater in determining stream temperature is important and since for majority of streams, field measured stream temperatures were cooler than predicted by the by the SP model in both 2006 and 2012, I 67 hypothesized that the magnitude of groundwater input into the study streams influenced their temperatures. Percent groundwater contribution to baseflow ranged from 30.27% (Laurel Fork, WV) to 85.02% (Carmans River, NY). I divided the study streams into three categories, low (30.27≤%≤48.52%), medium (48.52<%≤66.77), and high baseflow (66.77<%≤85.02%) relative to the baseflow percentages of the study streams. Then, field measured and predicted stream temperature from both the KEA and SP models were categorized according to baseflow percentage (Figures 11, 13, 15, and 17). If groundwater input were to exert a strong influence on stream temperature, then surface water dominated streams, those with low groundwater inputs, should be better predicted by SP than those streams with high groundwater inputs. Conversely streams with high groundwater inputs should predicted better by the KEA model than the low baseflow streams. I was unable to discern any clear pattern relating groundwater input percentage with the accuracy of stream temperature predictions from either model. Streams from all three baseflow input categories, low, medium, and high baseflow were under predicted and over predicted by the SP model, showing no clear indication that the percent groundwater input played any role in the efficiency of the model’s predictive capacity in either 2006 or 2012 for study streams. The KEA model under predicted temperatures for all study streams in both 2006 and 2012. As such, it was not clear whether groundwater dominated streams (high groundwater input %) were predicted better by this model than surface water dominated streams (low groundwater input %). I also categorized the field measured a predicted stream temperatures by Strahler stream order (orders 1 through 6) because water temperature tends toward equilibrium with ambient air temperature as the stream departs from its source (Figures 12, 14, 16, and 18) (Hynes, 1970, 68 Ducharne, 2008). Therefore, higher order larger streams are typically warmer than smaller lower order streams. As with baseflow, I did not find any clear indication that the stream order played a role in predicting stream temperatures accurately using either the SP or KEA models. Therefore, I conclude, that for the streams included in this study, neither percent groundwater input nor stream order substantially influenced the predictive capacity of either the SP or KEA model. The one compelling exception to these findings, mentioned previously, was that the two streams in the high baseflow category (66.77<%≤85.02%) and lowest stream order category (stream order 1), were predicted well by the KEA model, indicating that when percent baseflow contribution is very high and stream order is very low, the combination of these two factors in combination are more influential than air temperature in determining stream temperature. Considerations for the future Air Temperature to Stream Temperature Conversion Alternative models for relating air temperature to water temperature should be considered for future studies, as linear regression models are not always the most accurate in capturing this relationship. For example, Morrill et al. (2005) found that the stream temperatures of 83% of the geographically diverse study sites assessed did not display a 1:1 relationship with air temperatures. Instead, they determined a non-linear S-shaped function to be a better descriptor of the air-stream temperature relationship (Morrill et al., 2005, Erickson and Stefan, 2000). This seems reasonable, given the fact that for some streams, the relationship between air and stream temperature becomes non-linear after air temperature has reached 25°C, a temperature regularly exceeded during the summer months in the states examined in my study; the reason for non- 69 linearity is likely related to evaporative cooling of the stream (Morrill et al., 2005, Erickson and Stefan, 2000). However, Morrill et al. (2005) did note that some streams exhibit a strong linear relationship between air and water temperature and that in these cases, using a non-linear model to estimate this relationship may offer little or no improvement (Morrill et al., 2005). Additionally, many other studies have observed streams with nearly 1:1 relationships between water and air temperature for weekly and monthly data making linear regression models an effective way to predict stream temperatures from air temperatures for some streams. For example, Pilgrim et al. (1998) found a near 1:1 relationship between air and stream temperature for 39 Minnesota streams and as did the paper which introduces the linear model used in this study, Stefan and Preud’homme (1993) also found a nearly linear relationship in 11 Mississippi basin streams. Given that the relationship between air and water temperature seems to be dependent upon the characteristics of the specific stream and surrounding conditions, the most effective way to predict stream temperatures from air temperature would be to investigate the linearity of the relationship between the two variables for each stream and develop a custom linear or nonlinear model based on the results. While this study is based on the premise that air temperature is one of the strongest predictors of stream temperature, the correlation found between low order streams with high levels of groundwater input and the general over prediction of stream temperature by the SP model, suggests that other factors, like groundwater input, also play an important role in influencing stream temperature. In addition to air temperature and groundwater inputs, other drivers of stream temperature include such things as percent riparian zone shading, runoff, ice 70 cover, and the presences of dams. Ideally, these additional drivers of stream temperature would be incorporated into a model to most accurately predict stream temperature. Climate Models As mentioned earlier in the methods section of this thesis, the emissions scenario used to project future air temperatures for this study, SRES A1B, assumes a moderate increase in atmospheric CO2 concentration (700 ppm by 2100) (Meehl et al., 2007). It is possible that CO2 emissions will follow a different emissions pathway, like the A2 pathway, which assumes CO2 concentrations of 820ppm by 2100 (Meehl et al., 2007), as a result of continued heavy reliance on fossil fuels to meet energy demands; or the B1 pathway, which assumes substantial reductions in emissions, predicting CO2 concentrations of 550ppm by 2100. If CO2 emissions continue along the lower emissions pathway (B1), then stream temperatures could be lower than predicted by this study, resulting in more favorable thermal habitat conditions for brook trout in the future. Conversely, if emissions proceed along the higher emissions pathway (A2), then stream temperatures could become warmer more quickly than predicted in this study, thus reducing thermal habitat for brook trout more rapidly than anticipated. Additionally, as the scientific community’s understanding of climate science grows, the latest GCMs and emissions scenarios, and downscaling methods should be used in further studies of the impact of climate change on brook trout habitat. Additional Effects of Climate Change on Thermal Habitat for Brook Trout In addition to direct effect that changes in air temperature have on stream temperature, climate change is likely to influence stream temperature indirectly as well. For example, climate change may influence groundwater and riparian zone shading, which are also drivers of stream temperature. 71 Since groundwater temperatures are ultimately driven by air temperatures, staying within 1°C of annual average air temperature (Allan, 1995), rising air temperatures should result in rising groundwater temperatures as well. The impact of climate change on groundwater temperature has not yet been studied in great depth, however, posing a challenge in projecting future water temperature for groundwater dominated streams (Kurylyk et al., 2013, Mayer, 2012). A study of the effects of air temperature and concurrent ground surface temperature on a small watershed catchment in east-central New Brunswick, Canada, did show that the temperature of shallow groundwater is responsive to seasonal changes in air temperature and the temperature of deeper groundwater is sensitive decadal climate change (Kurylyk et al., 2013). As such, the influence of changing groundwater temperature on stream temperature should be considered when modeling future stream temperature and the buffering effect of groundwater inputs on stream temperature should not be overestimated (Kurylyk et al., 2013). The riparian vegetation mitigating rising air temperatures by providing shading may also change as incidence of wildfires, droughts, insect, and pathogen outbreaks increase in response to climate change (Melillo et al., 2014), resulting in increasing stream temperatures and further reductions in thermal habitat status for brook trout. A study of the effect of changing vegetation regimes due to climate change on stream temperature in the southeastern United States, concluded that altered stream shading could result in increases in stream temperature of up to 7°C (Cooter and Cooter, 1990). Reductions in riparian vegetation could also lead to increases in the amount of surface runoff entering streams further contributing to stream warming (Hynes, 1970). In addition to the direct effects that rising stream temperatures are likely to have on brook trout, the effect that elevated stream temperatures will have on other members of the ecological 72 community will influence brook trout indirectly as well. One such concern is changing phenology as a result of climate change (Walther et al., 2002), which may lead to asynchrony between the lifecycle events of brook trout and other species in the aquatic environment. Changes in water temperature may lead to premature hatching of the eggs of fall spawning fish, like brook trout, leading to a mismatch between the pulse of prey species production and the onset of larval fish feeding (Shuter et al., 2012). Changes in the emergence of prey species may also prove problematic for brook trout. For example, the high altitude stream mayfly, Baetis bicaudatus, a prey species for trout, emerges as an adult insect earlier as stream temperatures warm (Harper and Peckarsky, 2006), which could result in a mismatch between the availability of mayflies as a prey resource and a critical period for energy consumption for brook trout. Harper and Peckarsky (2006) also note that the size and fecundity of mayflies emerging under increased water temperature conditions may be reduced, which likely means reduced quality of this food resource for trout even in the absence of phonological asynchrony. Management Implications In the face of increasing air temperatures due to global climate change, fisheries managers must identify the processes controlling stream temperature and prioritize efforts to mitigate the effects of rising air temperatures to maintain brook trout fisheries within their native range. I suggest that promoting high levels of riparian zone shading, reducing runoff, and protecting the sources of groundwater inputs into these coldwater systems would provide the most effective ways to maintain stream temperatures cool enough for brook trout to persist in United States. Riparian zone shading should be protected from removal and enhanced in areas where it has been lost to shield streams from direct exposure to ultraviolet radiation. This strategy to 73 reduce stream temperature will be of particular importance for streams receiving little or no groundwater input. Blann et al. (2002) found that suitable thermal habitat for brook trout could be expanded by introducing sufficient shading in previously unshaded stream reaches and indicated that wooded buffer zones produce the most shade, followed by successional vegetation buffers. Loss of riparian shading can also reduce the volume of cold water plumes in areas of high groundwater input, thus reducing important summertime thermal refugia for brook trout (Ebersole et al., 2003). In addition to reducing stream temperature by providing shading, streamside vegetation also reduces direct water runoff into streams (Hynes, 1970). Reducing runoff is critical for maintaining cool stream temperatures particularly in catchments with high levels of urbanization, because impervious surfaces, like pavement, increase runoff volume and warms water as it washes over these surfaces (Nelson and Palmer, 2007). Nelson and Palmer (2007), concluded that while spikes in stream temperature dissipated in a relatively short amount of time (3 hours), localized rainstorms could causes surges in temperature averaging 3.5°C. For streams receiving substantial inputs from groundwater aquifers, like those in this study receiving groundwater contributions to baseflow of ≥80%, protecting baseflow is critical. An important approach to maintaining baseflow levels high enough to mitigate increasing air temperatures is to carefully evaluate the effects of groundwater withdrawals for human use and consumption on stream temperature as the demand for this resource continues to increase over time because even slight reductions in baseflow can result in stream temperature increases (Waco and Taylor, 2010). In addition to directly safeguarding stream baseflow, managers must also be cognizant of the various landscape features promoting groundwater recharge, as these features ultimately 74 determine the amount and rate of groundwater input into streams. For example, streams located in areas with highly permeable, coarse soil types and low gradient topography (Winter, 1999, Stanford and Ward, 1993, Cey et al., 1998, Siitari et al., 2011), are likely to receive higher levels of groundwater input than streams in areas with steep terrain and more impermeable geology. As such, in low lying areas with highly permeable soils, such as those representative of large portions of the brook trout’s range in the Midwestern U.S., groundwater is an important factor influencing stream temperature (Siitari et al., 2011). The amount and rate of groundwater entering riparian systems has also been linked to land cover in addition to geologic structure. Notably, differing land cover types allow for distinctive rates of water percolation into soil, resulting in differing aquifer recharge rates. Grasslands, for example, allow for the highest rates of aquifer recharge followed in descending order by forested lands, croplands, and areas dominated by impervious surfaces, such as concrete (Waco and Taylor, 2010). Therefore, encouraging land cover types that promote higher rates of groundwater recharge and restricting the amount of area within a watershed dominated by impervious surfaces will be important in protecting groundwater quantity and stream baseflow in the future. Given the increasing temperatures predicted to result from global climate change, it is essential to protect and improve natural stream cooling mechanisms to help sustain brook trout thermal habitat and ultimately brook trout populations across their native U.S. range. Additionally, fisheries regulations must also change in response to rising air temperatures to protect brook trout populations from high angling pressure during thermally stressful events and by protecting the coldwater refugia these fish need to persist under adverse thermal regimes. Lastly, in areas where brook trout thermal habitats and thus populations are likely to be lost, we 75 must prepare to introduce stakeholders to new types of fishing opportunities, and ready ourselves and the public for new warmer water species assemblages as water temperatures become less conducive to stenothermic coldwater species (Lynch et al., 2010). As the brook trout’s range and productivity are impacted by rising air temperatures resulting from climate change, fisheries managers must prepare to address the concerns associated with the changing availability of this recreationally and economically valuable species. The role of coldwater fisheries managers in the future will become increasingly complicated, as the human population increases, land cover changes, competition for ground and surface water accelerates, and the challenges of increasing air temperatures threaten the brook trout’s existence across their native range. In order to cope with the complex issue of shifting brook trout range and productivity, resource managers must: identify and prioritize the protection of areas likely to maintain high quality brook trout habitat as air temperatures increase; safeguard and enhance the drivers of stream cooling; and appropriately adjust regulations in accordance with changes in brook trout productivity. 76 LITERATURE CITED 77 LITERATURE CITED Allan, D.J. (1995) Stream ecology: structure and function of running waters, Vol., Kluwer Academic Publishers, Dordrecht, The Netherlands. Baldwin, N.S. (1957) Food consumption and growth of brook trout at different temperatures. Transactions of the American Fisheries Society 86, 323-328. Beldade, R., Holbrook, S.J., Schmitt, R.J., Planes, S., Malone, D., Bernardi, G. (2012) Larger female fish contribute disproportionately more to self-replenishment. Proceedings of the Royal Society B-Biological Sciences 279, 2116-2121. [In English]. Blanchfield, P.J., Ridgway, M.S. (1997) Reproductive timing and use of redd sites by lakespawning brook trout (Salvelinus fontinalis). Canadian Journal of Fisheries and Aquatic Sciences 54, 747-756. [In English]. Blann, K., Nerbonne, J.F., Vondracek, B. (2002) Relationship of riparian buffer type to water temperature in the driftless area ecoregion of Minnesota. North American Journal of Fisheries Management 22, 441-451. Bowlby, J.N., Roff, J.C. (1986) Trout biomass and habitat relationships in southern Ontario streams Transactions of the American Fisheries Society 115, 503-514. [In English]. Brett, J.R. (1956) Some principles in the thermal requirements of fishes. Quarterly Review of Biology 31, 75-87. Cey, E.E., Rudolph, D.L., Parkin, G.W., Aravena, R. (1998) Quantifying groundwater discharge to a small perennial stream in southern Ontario, Canada. Journal of Hydrology 210, 2137. Cooter, E.J., Cooter, W.S. (1990) Impacts of greenhouse warming on water temperature and water quality in the southern United States. Climate Research 1, 1-12. Coutant, C.C. (1987) Thermal preference - When does an asset become a liability. Environmental Biology of Fishes 18, 161-172. Crisp, D.T., Howson, G. (1982) Effect of air-temperature upon mean water temperature in streams in the North Pennines and English Lake District. Freshwater Biology 12, 359367. Drake, M.T., Taylor, W.W. (1996) Influence of spring and summer water temperature on brook charr, Salvelinus fontinalis, growth and age structure in the Ford River, Michigan. Environmental Biology of Fishes 45, 41-51. 78 Ducharne, A. (2008) Importance of stream temperature to climate change impact on water quality. Hydrology and Earth System Sciences 12, 797-810. [In English]. Eastern Brook Trout Joint Venture. (2008) Conserving the Eastern Brook Trout: Action Strategies. Ebersole, J.L., Liss, W.J., Frissell, C.A. (2003) Cold water patches in warm streams: Physicochemical characteristics and the influence of shading. Journal of the American Water Resources Association 39, 355-368. [In English]. Erickson, T.R., Stefan, H.G. (2000) Linear air/water temperature correlations for streams during open water periods. Journal of Hydrologic Engineering 5, 317-321. [In English]. Flebbe, P.A., Roghair, L.D., Bruggink, J.L. (2006) Spatial Modeling to project southern Appalachian trout distribution in a warmer climate. Transactions of the American Fisheries Society 135, 1371-1382. Fry, F.E.J., Hart, J.S., Walker, K.F. (1946) Lethal temperature relations for a sample of young speckled trout, Salvelinus fontinalis Vol., The University of Toronto Press, Toronto. Harper, M.P., Peckarsky, B.L. (2006) Emergence cues of a mayfly in a high-altitude stream ecosystem: Potential response to climate change. Ecological Applications 16, 612-621. [In English]. Harris, A. (2010). United States Fish and Wildlife Service. Trout fishing in 2006: A demographic description and economic analysis. Helfman, G.S., Collette, B.B., Facey, D.E., Bowen, B.W. (2009) The Diversity of Fishes: Biology, Evolution, and Ecology, Vol., Blackwell Publishing, Chichester, UK. Henson, R. (2011) The Rough Guide to Climate Change: The Symptoms, The Science, The Solutions, Third edn (Rough Guides Reference, Vol., Rough Guides Ltd. Hislop, J.R.G. (1988) The influence of maternal length and age on the size and weight of the eggs and the relative fecundity of the haddock Melanogrammus-aeglefinus, in British waters. Journal of Fish Biology 32, 923-930. [In English]. Hokanson, K.E., McCormic, J.H., Jones, B.R., Tucker, J.H. (1973) Thermal requirements for maturation, spawning, and embryo survival of brook trout, Salvelinus-fontinalis. Journal of the Fisheries Research Board of Canada J 30, 975-984. Hokanson, K.E.F. (1977) Temperature requirements of some percids and adaptations to season temperature cycle. Journal of the Fisheries Research Board of Canada 34, 1524-1550. Houghton, J. (2009) Global Warming: The Complete Briefing, Fourth edn Vol., Cambridge University Press, Cambridge, U.K. 79 Hynes, H.B.N. (1970) The Ecology of Running Waters, Vol., University of Toronto Press. Karas, N. (1997) Brook Trout: A Thorough Look at North America's Great Native Trout — Its History, Biology, and Angling Possibilities. The Lyons Press, New York, New York U.S.A. Karl, T.R., Melillo, J.M., Peterson, T.C. (2009) Global Climate Change Impacts in the United States. Cambridge University Press. Krider, L.A., Magner, J.A., Perry, J., Vondracek, B., Ferrington, L.C. (2013) Air-water temperature relationships in the trout streams of southeastern Minnesota's carbonatesandstone landscape. Journal of the American Water Resources Association 49, 896-907. [In English]. Kurylyk, B.L., Bourque, C.P.A., MacQuarrie, K.T.B. (2013) Potential surface temperature and shallow groundwater temperature response to climate change: an example from a small forested catchment in east-central New Brunswick (Canada). Hydrology and Earth System Sciences 17, 2701-2716. [In English]. Lessard, J.L. (2000) Temperature effect of dams on coldwater fish and macroinvertebrate communities in Michigan. Master of Science, Michigan State University. Lynch, A.J., Taylor, W.W., Smith, K.D. (2010) The influence of changing climate on the ecology and management of selected Laurentian Great Lakes fisheries. Journal of Fish Biology 77, 1964-1982. MacCrimmon, H.R., Campbell, J.S. (1969) World distribution of brook trout, Salvelinus fontinalis. Journal of the Fisheries Research Board of Canada 26, 1699-1725. Magnuson, J.J., Destasio, B.T. (1997) Thermal niche of fishes and global warming. In: Global warming: implications for freshwater and marine fish. Society for experimental biology seminar series (Eds. C.M. Wood, D.G. McDonald), Cambridge University Press. Magnuson, J.J., Meisner, J.D., Hill, D.K. (1990) Potential changes in ther thermal habitat of Great-Lakes fish after global climate warming. Transactions of the American Fisheries Society 119, 254-264. Maurer, E.P., Brekke, L., Pruitt, T., Duffy, P.B. (2007) Fine-resolution Climate Projections Enhance Regional Climate Change Impact Studies. Eos Transactions, American Geophysical Union 88. Mayer, T.D. (2012) Controls of summer stream temperature in the Pacific Northwest. Journal of Hydrology 475, 323-335. [In English]. 80 McCormick, J.H., Hokanson, K.E.F., Jones, B.R. (1972) Effects of temperature on growth and survival of young brook trout, Salvelinus fontinalis. Fisheries Research Board of Canada 29, 1107-1112. Meehl, G.A., Covey, C., Delworth, T., et al. (2007) The WCRP CMIP3 multimodel dataset - A new era in climate change research. Bulletin of the American Meteorological Society 88, 1383-1394. Meisner, J.D. (1990a) Effect of climatic warming on the Southern margins of the native range of brook trout, Salvelinus fontinalis. Canadian Journal of Fisheries and Aquatic Sciences 47, 1065-1070. Meisner, J.D. (1990b) Potential loss of thermal habitat for brook trout, due to climatic warming, in 2 Southern Ontario streams. Transactions of the American Fisheries Society 119, 282291. Meisner, J.D., Rosenfeld, J.S., Regier, H.A. (1988) The role of groundwater in the impact of climate warming on stream salmonines. Fisheries 13, 2-8. Melillo, J.M., Richmond, T.T.C., Yohe, G.W. (2014) Climate change impacts on the United States: the third National Climate Assessment. 841. Michigan Department of Natural Resources (MI DNR). (2012) Brook Trout, Salvelinus fontinalis. http://www.michigan.gov/dnr/0,4570,7-153-10364_18958-96400--,00.html. Mohseni, O., Stefan, H.G. (1999) Stream temperature air temperature relationship: a physical interpretation. Journal of Hydrology 218, 128-141. Morrill, J.C., Bales, R.C., Conklin, M.H. (2005) Estimating stream temperature from air temperature: Implications for future water quality. Journal of Environmental Engineering-Asce 131, 139-146. [In English]. Nelson, K.C., Palmer, M.A. (2007) Stream temperature surges under urbanization and climate change: Data, models, and responses. Journal of the American Water Resources Association 43, 440-452. [In English]. Ozaki, N., Fukushima, T., Harasawa, H., Kojiri, T., Kawashima, K., Ono, M. (2003) Statistical analyses on the effects of air temperature fluctuations on river water qualities. Hydrological Processes 17, 2837-2853. Page, L.M., Burr, B.M. (1991) Freshwater Fishes, (Peterson Field Guides, Vol., Houghton Mifflin Company, New York, U.S.A. Pilgrim, J.M., Fang, X., Stefan, H.G. (1998) Stream temperature correlations with air temperatures in Minnesota: Implications for climate warming. Journal of the American Water Resources Association 34, 1109-1121. 81 Power, G. (1980) The Brook Charr, Salvelinus Fontinalis. In: Charrs: Salmonid Fishes of the Genus Salvelinus. (Ed. E.K. Balon), Dr. W. Junk Publishers, the Hague, pp. 141-203. Raleigh, R.F. (1982) Habitat suitability index models: Brook Trout. FWS/OBS-82/10.24. (Ed. F.a.W.S. U.S. Department of the Interior), p. 42. Shuter, B.J., Finstad, A.G., Helland, I.P., Zweimuller, I., Holker, F. (2012) The role of winter phenology in shaping the ecology of freshwater fish and their sensitivities to climate change. Aquatic Sciences 74, 637-657. [In English]. Shuter, B.J., Post, J.R. (1990) Climate, population viability, and the zoogeography of temperate fishes. Transactions of the American Fisheries Society 119, 314-336. Siitari, K.J., Taylor, W.W., Nelson, S.A.C., Weaver, K.E. (2011) The influence of land cover composition and groundwater on thermal habitat availability for brook charr (Salvelinus fontinalis) populations in the United States of America. Ecology of Freshwater Fish 20, 431-437. Stanford, J.A., Ward, J.V. (1993) An ecosystem perspecitve of alluvial rivers - connectivity and the hyporheic corridor. Journal of the North American Benthological Society 12, 48-60. Steen, P.J., Wiley, M.J., Schaeffer, J.S. (2010) Predicting Future Changes in Muskegon River Watershed Game Fish Distributions under Future Land Cover Alteration and Climate Change Scenarios. Transactions of the American Fisheries Society 139, 396-412. Stefan, H.G., Preud'homme, E.B. (1993) Stream temperature estimation from air temperature. Water Resources Bulletin: American Water Resources Association 29, 27-45. Strussmann, C.A., Saito, T., Takashima, F. (1998) Heat-induced germ cell deficiency in the teleosts Odontesthes bonariensis and Patagonina hatcheri. Comparative Biochemistry and Physiology a-Molecular and Integrative Physiology 119, 637-644. [In English]. United States Department of Agirculture, Natural Resources Conservation Servce Information about Hydrologic Unites and the Watershed Boundary Dataset. http://www.nrcs.usda.gov/wps/portal/nrcs/detail/national/water/watersheds/dataset/?cid= nrcs143_021616. United States (U.S.) Department of Interior, United States Fish and Wildlife Service, United States Department of Commerce Interior, United States Census Bureau. (2011) National Survey of Fishing, Hunting, and Wildlife-Associated Recreation. United States (U.S.) Geological Survey and United States (U.S.) Environmental Protection Agency. (2005) National Hydrography Dataset Plus - NHDPlus. United States Geological Survey (2014) USGS Threatened and Endangered Stations. http://streamstatsags.cr.usgs.gov/ThreatenedGages/ThreatenedGages.html. 82 Waco, K.E. (2009) The Influence of Groundwater withdrawal on available Brook Charr (Salvelinus fontinalis) Thermal Habitat in Twin and Chippewa Creeks, Osceola County, Michigan. Masters of Science, Michigan State University. Waco, K.E., Taylor, W.W. (2010) The influence of groundwater withdrawal and land use changes on brook charr (Salvelinus fontinalis) thermal habitat in two coldwater tributaries in Michigan, USA. Hydrobiologia 650, 101-116. Walther, G.R., Post, E., Convey, P., et al. (2002) Ecological responses to recent climate change. Nature 416, 389-395. [In English]. Wehrly, K.E., Wiley, M.J., Seelbach, P.W. (2003) Classifying regional variation in thermal regime based on stream fish community patterns. Transactions of the American Fisheries Society 132, 18-38. [In English]. Winter, T.C. (1999) Relation of streams, lakes, and wetlands to groundwater flow systems. Hydrogeology Journal 7, 28-45. 83