Evaluating Climate Change Impacts on Drinking Water Reservoirs and Wastewater System Resilience in Connecticut, U.S.A

University of Connecticut OpenCommons@UConn

Doctoral Dissertations University of Connecticut Graduate School

11-26-2019

Evaluating Climate Change Impacts on Drinking Water Reservoirs and Wastewater System Resilience in Connecticut, U.S.A.

Cristina Mullin University of Connecticut - Storrs, [email protected]

Follow this and additional works at: https://opencommons.uconn.edu/dissertations

Recommended Citation Mullin, Cristina, "Evaluating Climate Change Impacts on Drinking Water Reservoirs and Wastewater System Resilience in Connecticut, U.S.A." (2019). Doctoral Dissertations. 2353. https://opencommons.uconn.edu/dissertations/2353 Evaluating Climate Change Impacts on Drinking Water Reservoirs and Wastewater System Resilience in

Connecticut, U.S.A.

Cristina A. Mullin, PhD

University of Connecticut, 2019

Societies depend on the proper functioning and resilience of critical infrastructure systems including those for drinking water and wastewater, but these systems are vulnerable to natural and anthropogenic stressors. For example, wastewater systems are especially susceptible to extreme weather events while drinking water systems are vulnerable to diminished source water quality — both of which can disrupt proper functioning. Climate change and human development, economic problems, and pollution further challenge water system resilience. Despite the criticality of these systems and known vulnerabilities, little scholarship exists that aims to interrogate their resilience. This dissertation attempts to narrow this gap in the literature by: (1) investigating the sensitivity of reservoir water quality and wastewater system resilience to climate variability and change; (2) developing adaptation strategies to help improve wastewater system resilience; and, (3) advancing knowledge of historical and potential future reservoir water quality changes. This dissertation evaluates wastewater system resilience and reservoir water quality in Connecticut because climate induced increases in air temperature, extreme precipitation, and frequency of large storms stress water system functioning in the Northeast, U.S. and many other north temperate regions around the world.

Wastewater system resilience is assessed from a social sciences perspective using concepts of adaptive capacity and adaptive management. Results suggest human dimensions are important for building wastewater system resilience and that systems with managers who build and deploy diverse generic and specific adaptive capacities within an adaptive management framework are most resilient to storms. For drinking water systems, observations from six Connecticut reservoirs over sixteen years from

2003-2018 indicate: (1) thermal stability is increasing; (2) surface water temperature is increasing; (3) Cristina A. Mullin – University of Connecticut, 2019

bottom water temperature is decreasing; and, (4) surface and average water dissolved oxygen is increasing. Future projections suggest extreme high surface water temperatures and thermal stability will increase significantly by midcentury (2041-2070). Toxic cyanobacteria blooms may occur more often in the future under these extreme conditions, especially in reservoirs where eutrophication and harmful algal blooms are already a concern. This research aids in understanding, predicting, and informing appropriate strategies for managing water quality and wastewater systems under a warming climate.

Evaluating Climate Change Impacts on Drinking Water Reservoirs and Wastewater System Resilience in

Connecticut, U.S.A.

Cristina A. Mullin

B.S., University of Connecticut, Marine Sciences, 2015

M.S., University of Connecticut, Environmental Engineering, 2018

A Dissertation

Submitted in Partial Fulfillment of the

Requirements for the Degree of

Doctor of Philosophy

at the

University of Connecticut

2019

APPROVAL PAGE

Doctor of Philosophy Dissertation

Evaluating Climate Change Impacts on Drinking Water Reservoirs and Wastewater System Resilience in

Connecticut, U.S.A.

Presented by

Cristina A. Mullin, B.S., M.S.

Major Advisor ______

Dr. Christine Kirchhoff

Associate Advisor ______

Dr. Guiling Wang

Associate Advisor ______

Dr. Penny Vlahos

Associate Advisor ______

Dr. Marina Astitha

University of Connecticut

2019

Acknowledgements

In fall 2016, I was awarded an Environmental Engineering Graduate Assistance in Areas of National

Need (GAANN) Fellowship in water and policy, funded by the U.S. Department of Education, and began my pursuit of a PhD in Environmental Engineering at the University of Connecticut (UConn). I would like to thank the UConn Civil and Environmental Engineering Department and my advisor, Dr. Christine

Kirchhoff, for offering me this fellowship and granting me the opportunity to pursue the highest academic degree offered by the department. In addition to the GAANN fellowship, a municipal planning grant from the United States Department of Housing and Urban Development supported the research in Chapter 2. I would also like to thank important groups of people, without whom this dissertation would not have been possible: my major advisor, advisory committee, data providers, fellow graduate students, undergraduate research assistants, friends, and family.

I am indebted to my major advisor, Dr. Christine Kirchhoff, for her guidance and intellectual contributions which not only improved my dissertation research but contributed to my development as a scientist. I owe much gratitude to Dr. Kirchhoff and all of my advisory committee members, including Dr.

Guiling Wang, Dr. Penny Vlahos, and Dr. Marina Astitha, for their dedication to my success and for their professional guidance, valuable support, and beneficial critiques of my research and writing.

Motivation for the 3rd and 4th chapters of this dissertation came from conversations I had early on in my graduate journey with staff at South Central Connecticut Regional Water Authority, who were trying to understand how water quality and cyanobacteria productivity may be changing in their drinking water reservoirs. Specifically, I thank John Hudak and William Henley for their assistance, ideas, and for providing the reservoir data used in this dissertation. I also extend my sincere appreciation to our interview respondents whose insights were critical to the 2nd chapter and to Peter Watson for conducting the interviews.

iii

I am also very appreciative of my fellow graduate students, undergraduate research assistants, and friends for their support. I am grateful for mentorship from Michael Harris and Brendan Lenz, who selflessly gave their time to assist with the development of MATLAB codes and Excel Macros used for data analysis. Gratitude is also due to UConn Statistical Consulting team members Tim Moore and Renjie

Chen, for their assistance with the time series analysis in R. I would also like to thank two UConn,

Department of Civil and Environmental Engineering undergraduate student research assistants, Timothy

Cannata and Joshua Crittenden, who assisted with early data preprocessing, predictive model development, and editing and proofreading. Appreciation is also due to Berdakh Utemuratov, my labmate, for his motivation words and for providing both scholarly and mental support.

Finally, I’d like to thank my family for their mental, motivational, and financial support and for providing me with the early childhood experiences that ignited my scientific curiosity and fascination with water.

Table of Contents

1 Introduction …………………………………………………………………...…………….……….. 1

1.1 Motivation and Scope ...………...... ……………..………...... …... 2

2 Marshaling Adaptive Capacities Within an Adaptive Management Framework to Enhance the

Resiliency of Wastewater Systems ……………………………………...……………….…….…..... 4

2.1 Introduction and Literature Review ………………………………………………….…...... …..… 4

2.2 Data Collection and Analysis ……………………………………………….………..…...... …… 6

2.3 Results and Discussion …………………………………………...…………….…….….....…….. 7

3 Assessing the sensitivity of reservoir water quality to changes in air temperature and wind speed

in Connecticut, U.S.A. ……………...……………………...... ……………………….… 10

3.1 Introduction ………………………………………….…………………………….…………….. 10

3.2 Methods ………………………………………………………………………………….………. 13

3.3 Results ………………………………………………………………………………………...…. 18

3.4 Discussion ……………………………………………………………………………..………… 24

3.5 Conclusion …………………………………………………………………………….…………. 30

4 Future projections of lake temperature, thermal stability, and implications for cyanobacteria

blooms in Connecticut, U.S.A. reservoirs amidst a warming climate …………………..……….. 33

4.1 Introduction ………………………………………………………………………………..…….. 33

4.2 Methods ……………………………………………………………………………………….…. 37

4.3 Results and Discussion …………………………………………………………………….…….. 43

4.4 Conclusions …………………………………………………………………………….………... 57

5 Research Summary, Future Suggestions, and Concluding Remarks ….…………………..…….. 61

5.1 Research Summary ………………………………………………………………..……….…..… 61

5.2 Future Suggestions ………………………………………………………………………....……. 64

5.3 Concluding Remarks …………………………………………………………………………….. 65

6 Appendices ………………………………………………………………………….……..……...… 66

6.1 Chapter 3 - Supporting Material ……………………………...………………………………..…. 66

6.2 Chapter 4 - Supporting Material ……………………………………………………..……...……. 72

7 References …………………………….……………………………………………….…….…...…. 80

1 Introduction

Wastewater and drinking water systems provide services vital to human health and the health of environmental ecosystems. However, they are sensitive to climate variability and change including extreme weather, global warming, sea level rise; and anthropogenic stressors such as population growth, development, economic problems, pollution, and human decisions (Kundzewicz et al., 2008; Guikema et al., 2015; Arnell et al., 2015). Despite the importance of reservoir water quality, wastewater system functioning, and these known vulnerabilities, little scholarship exists that interrogates their resilience.

Resilience here is defined as the capacity of a water system to prepare for, cope with, recover from, and change to reduce vulnerability to stress, especially the impacts from extreme events and future climate change (IPCC, 2007; Francis & Bekera, 2014). This dissertation attempts to narrow this gap in the literature by: (1) investigating the sensitivity of reservoir water quality and wastewater system resilience to climate variability and change; (2) developing adaptation strategies to help improve wastewater system resilience; and, (3) advancing knowledge of historical water quality trends and methods for modelling the dynamic response of freshwater quality to future climate change.

The future of reservoir water quality and wastewater system resilience in Connecticut is examined because air temperature, extreme precipitation, and frequency of large storms are expected to increase faster in the Northeast than in any other region in the U.S., making water systems in this geographical area particularly vulnerable to climate change (Lynch et al, 2016; Karmalkar & Bradley, 2017; Seth et al.,

2019). Changes in the economy, development, aging infrastructure, and an uncertain regulatory environment further exacerbate these vulnerabilities. This dissertation addresses the following questions:

(1) What factors are important for improving wastewater system resilience to severe weather events and future climate change? (2) Has surface, average, or bottom water quality of select Connecticut reservoirs changed in recent decades? (3) Has air temperature and wind speed near lakes changed in recent decades and how are these changes affecting reservoir water quality? And, (4) How might future climate warming affect future water temperature, thermal stability, and cyanobacteria productivity by

1 midcentury? This dissertation aims to advance fundamental knowledge as well as provide useful information to guide water management and policy including informing appropriate strategies for building the resilience of vulnerable water and wastewater systems to climate change and anthropogenic stressors.

1.1 Motivation and Scope

Chapter 2 addresses the 1st research question. Given wastewater systems are vulnerable to severe weather events and climate change, there is a pressing need to advance understanding of what wastewater systems do to build resilience in practice, especially how human systems help build resiliency (or not).

Scholars have increasingly focused on understanding water, power, and transportation infrastructure resiliency, however much of the scientific focus is on how hard adaptation actions improve resiliency

(e.g., digging impoundments, raising equipment). A few prior studies suggest having high generic adaptive capacity (e.g., education, funding, knowledge, learning, information, skills, and leadership), high specific adaptive capacity (hard and soft adaptations associated with managing, reducing, and responding to climatic threats), and adaptive management is important for resilience (Folke et al., 2002; Yohe & Tol,

2002; Berkhout et al., 2006; Hess et al., 2012; Rudberg et al., 2012). However these concepts have not yet been applied to managing wastewater systems to improve resilience. Chapter 2 of this dissertation begins to fill this gap by assessing wastewater system resilience from a social sciences perspective using concepts of adaptive capacity and adaptive management.

Chapter 3 addresses research questions 2 and 3. Warming air temperatures (Karmalkar & Bradley,

2017), accelerating sea level rise (IPCC, 2007), changing wind patterns (Siver et al., 2018; Karnauskas et al., 2018), and more frequent extreme precipitation and storms (Horton et al., 2014; Wuebbles et al.,

2017), are linked with deteriorating surface water quality and water infrastructure (Kundzewicz et al.,

2008; Arnell et al., 2015). Understanding long-term changes in lake water quality is essential for protecting public health and the health of environmental ecosystems, and for enhancing resiliency of the water sector. Prior work suggests climate warming is deteriorating water quality by increasing water

2 temperature, thermal stability, and algal productivity in lakes. However, impacts of long-term changes in wind speed on lake water quality are less well known. Chapter 3 begins to address this gap by quantifying relationships between long-term changes in air temperature and wind speed and water quality trends from

2003-2018 in five Connecticut drinking water reservoirs. Water quality parameters evaluated in this chapter includes thermal stability and surface, average, and bottom water temperature, dissolved oxygen, specific conductivity, pH, and turbidity.

Chapter 4 addresses the 4th and final research question. Here, lake-specific empirical models are developed to assess possible future changes in water temperature and thermal stability in six Connecticut reservoirs. Historical (1971-2000) and future (2041-2070) changes are driven by air temperature projections from General Circulation Models (GCMs), under Representative Concentration Pathway

(RCP) 8.5. Additionally, observed relationships between water temperature, thermal stability, and cyanobacteria biomass are evaluated and used to understand how future changes in lake conditions may affect the occurrence of cyanobacteria blooms in drinking water reservoirs.

2 Marshaling Adaptive Capacities Within an Adaptive Management Framework to Enhance the

Resiliency of Wastewater Systems

This chapter contains a brief summary of the published manuscript: Mullin, C. A., & Kirchhoff, C. J. (2018). Marshaling Adaptive Capacities within an Adaptive Management Framework to Enhance the Resiliency of Wastewater Systems. Journal of the American Water Resources Association, 55(4), 906-919. https://doi.org/10.1111/1752-1688.12709

2.1 Introduction and Literature Review

Past storms including Alfred, Sandy, and Irene exposed the vulnerability of wastewater infrastructure and inflicted billions of dollars in damage in the Northeastern U.S. (Baylis et al., 2016). Wastewater treatment plants and associated collection systems are particularly vulnerable to storms because they tend to be in low lying areas and near water, where flooding may: (1) overload collection systems; (2) damage pump stations and electrical systems; (3) make the treatment plant inaccessible; and (4) put staff lives at risk. All of which hinder the ability of the system to treat wastewater to a sufficient level to protect public and environmental health. When infrastructure is flooded, and after water recedes, facilities may need rebuilding, processes may have to be restarted, etc. These recovery efforts are not instantaneous nor cheap. Given these vulnerabilities, there is a pressing need to advance understanding of what wastewater systems do to build resilience, especially how human systems help build resilience. Resilient infrastructure systems can reorganize, learn, and successfully adapt to stress or changing circumstances and maintain their desired state even after significant disruption (Folke et al., 2005; Pahl-Wostl, 2007;

Nelson et al., 2008; Pahl-Wostl, 2009). In this study, adaptive capacity and adaptive management are assessed as measures of wastewater system resiliency using data from interviews with wastewater system managers. This work contributes to filling an important gap in the literature by advancing understanding of the human dimensions of infrastructure resilience but also helps to advance resilience in the wastewater sector in practice.

Over the past several decades, scholars have increasingly focused on building water, power, and transportation infrastructure resiliency through hard adaptation actions (e.g., digging impoundments,

4 raising equipment) (Reed et al., 2009; Eisenack et al., 2011; Ouyang et al., 2012; Mugume et al., 2015;

Falco & Webb, 2015; Therrien et al., 2015; Bhamidipati, 2015; Mostafavi & Inman, 2016; Lin & Bie,

2016; Donovan & Work, 2017; Panteli & Mancarella, 2017; Shin et al., 2018). Hard adaptations can be temporary (i.e., adaptation actions implemented right before a storm hits aimed at increasing the wastewater system’s ability to cope with and reduce impacts from that storm) or permanent. Soft adaptations (e.g., making changes to address behavioral, social, economic, and governance factors) that also influence wastewater system performance and resilience are rarely considered (Juan-García et al.,

2017). Of the studies that do exist, most are modeling-based or are literature reviews that provide frameworks or guidance on what wastewater systems should do to increase resilience (Butler et al., 2017;

Schoen et al., 2015). To my knowledge, only one published empirical case study examines what systems are actually doing in practice to increase resilience (see Rudberg et al., 2012).

Building on prior scholarship, the concepts of generic and specific adaptive capacities (Folke et al.,

2002; Yohe & Tol, 2002; Berkhout et al., 2006; Hess et al., 2012; Rudberg et al., 2012), and adaptive management are used to assess the resilience of wastewater systems and how human dimensions influence wastewater system resilience. Having high adaptive capacity increases resilience by enhancing a system’s ability to anticipate disturbances and reduce impacts, to take advantage of opportunities or cope with consequences, to recover more quickly, and to make adaptive changes (IPCC, 2007). In addition to having high adaptive capacity, prior research suggests more adaptive (i.e., learning-based) management approaches which foster ongoing learning, proactivity, and transformation are better suited for sustaining resilience and reducing risk over time than traditional top down management (reactive, inflexible, and complacent), especially in response to climate change (Pahl-Wostl, 2007; Hess et al.,

2012; Kirchhoff & Dilling, 2016). Although application of adaptive management concepts has expanded in recent years, to my knowledge, adaptive management has not yet been applied to managing infrastructure systems to improve resilience. Here, the following research questions are investigated: (1)

How have wastewater systems been impacted by past storms and how do these impacts affect resilience?

(2) What kinds of generic adaptive capacities improve resilience? (3) What kinds of specific adaptive capacities are managers implementing and how do these capacities influence resilience? And, (4) How do adaptive capacity and adaptive management interrelate to foster resilience?

2.2 Data Collection and Analysis

Data from 31 focused qualitative interviews with wastewater system managers were used to understand impacts and responses from past storms and their implications for adapting to future climate change. The interviews addressed a range of topics and included questions about wastewater system characteristics (i.e., funding, size of system, location); the nature and severity of past storm impacts, what helped or hindered their emergency preparedness, response, and recovery; risk perceptions regarding climate change (i.e., future severe weather, sea level rise); the types of adaptive decisions and actions made (if any); what helped or hindered implementing adaptive decisions (e.g., availability of science, political support, leadership, adequate funding, see Moser & Ekstrom, 2010); and what factors have the strongest impact on resiliency. The interviewees were recruited from a random stratified (i.e., stratified by inland vs. coastal location, impacted vs. not impacted, and made changes vs. did not make changes) sample of 86 survey respondents (for more information on the survey, see Kirchhoff & Watson, 2019).

All interviewees are wastewater professionals comparable in age and education. The interview protocol was pilot tested with four wastewater managers not included in the final sample in July and August 2016, after which refinements were made to the protocol before initiating interviews for the study (for more information on the interview protocol, see Supporting Information in Mullin & Kirchhoff, 2018).

Interviews were recorded and transcribed.

Interview data was sorted into selective categories and themes (i.e., both predetermined and emergent) using a combination of both inductive (i.e., grounded theory) and deductive qualitative methodologies (Creswell, 2007). Qualitative data analysis software NVIVO 11 (QSR International) was used to code the interview transcripts. The following adaptive capacities were coded: (1) skilled staff (i.e.,

6 flexible, dependable, knowledgeable, well trained, and resourceful); (2) good asset management (i.e., proactive and ongoing maintenance, repair and replacement; monitoring and assessment; prevention of mechanical failures; and continual improvements); (3) good leadership (i.e., facilitates ongoing learning and change, proactivity, trust building, and useful connections); (4) sufficient funding (i.e., for both day- to-day and emergency operations); (5) soft adaptations: effective emergency preparedness, response and recovery; (6) hard adaptations: temporary or permanent adaptations to increase resiliency; and (7) adaptive management (i.e., continuous organizational learning, experimental adaptation, and transformation). 1-4 are generic adaptive capacities and 5-6 are specific adaptive capacities. Temporary storm resiliency interventions coded in category 6 include hard adaptation actions implemented right before a storm aimed at increasing the wastewater system’s ability to cope with and reduce impacts from that particular storm. These are typically removed after the event passes. Permanent resiliency interventions coded in category 6 are hard adaptations that remain in place and are intended to reduce potential impacts and help the wastewater system cope with consequences and/or recover more quickly from future storm events. The degree of historical impacts, the amount and diversity of adaptations, and wastewater system characteristics (i.e., funding availability, type and size of system, location) were considered and controlled for in the analysis of wastewater system resiliency.

2.3 Results and Discussion

Approximately half (i.e., 15 out of 31) of the wastewater system managers said their systems are located in a flood zone next to a river, tidal basin, or coastal area; and that their systems were significantly impacted by storm surge, tidal, coastal, or river flooding during past storms such as Alfred, Irene, and/or

Sandy. Results suggest the degree of historical impacts experienced by a wastewater system is positively correlated with the amount of adaptations made. Meaning that resilience strategies, learning, innovation, and change increase in response to experience with extreme weather events.

Results also suggest the most resilient wastewater systems are those with high adaptive capacities that employ an adaptive management approach and make ongoing adaptation investments over time. Greater

7 amounts of generic adaptive capacities (i.e., skilled staff and good leadership) help smooth both day‐to‐ day and emergency operations and provide a foundation for adaptive management. In turn, adaptive management helps managers: (1) build additional generic adaptive capacities; and, (2) develop and employ greater amounts and diversity of specific adaptive capacities (i.e., soft and/or hard adaptations).

Having a diversity of specific adaptive capacities is especially important for enhancing and sustaining resiliency. Adaptive management also enables managers to better understand their system's vulnerabilities, how those vulnerabilities change over time, and what specific actions may reduce those vulnerabilities. Finally, my work suggests wastewater system resilience critically depends on the capacities of the human systems for building resilience as much as or more so than relying only on physical infrastructure resilience.

Current literature on infrastructure resilience emphasizes structural resilience of the physical infrastructure, including the ability of the wastewater treatment processes to withstand higher or lower temperatures and flows, or the ability of wastewater system equipment and structures to withstand flooding (Juan-García et al., 2017). My findings suggest that resilience of the physical infrastructure is only one aspect of resilience, and that wastewater system resilience depends equally (or more so) on the capacity for resilience that the human systems that manage the physical system possess. That is, the most resilient wastewater systems are able to marshal, build, and deploy generic and specific adaptive capacities within an adaptive management framework to build resilience (see Fig. 1).

Fig. 1: An adaptive management framework provides a means to learn from disturbances and to marshal adaptive capacities to make informed adaptations to reduce vulnerability. Adaptive management in turn depends on the amount and diversity of underlying adaptive capacities (i.e., generic and specific) needed for resilience, which is the capacity to prepare, cope, and recover (originally published in Mullin & Kirchhoff, 2018).

Physical infrastructure resilience is an important contribution of overall wastewater system resilience, but it is embedded in and amplified by the adaptive capacities and adaptive decisions managers make as part of an adaptive management approach.

3 Assessing the sensitivity of reservoir water quality to changes in air temperature and wind

speed in Connecticut, U.S.A.

3.1 Introduction

Drinking water systems rely on surface water sources that are vulnerable to declining water quality driven by climate change and anthropogenic factors. For example, prior studies have shown climate warming is affecting large lakes around the world by increasing surface water temperature and promoting earlier onset and longer periods of thermal stratiﬁcation (Magnuson et al., 1997; King et al., 1998;

Dobiesz & Lester, 2009), but less is known about changes in smaller lakes. Anthropogenic inputs of nutrients into water resources also impact lakes by changing global biogeochemical cycles and increasing the global supply of nutrients (Blenckner, 2005; Descy et al., 2017). Together, these changes have been linked to greater biological productivity, toxic algal blooms, and higher extent and duration of anoxia during the summer months in lakes (Foley et al., 2011; Chapra et al., 2017; Tan et al., 2018). Finally, while higher air temperatures may reduce lake water quality, the impacts of long-term changes in wind speed on lake water quality are less well known, especially in shallow lakes which are likely to respond to changing wind speed more dramatically than larger lakes (Deng et al., 2018). This study begins to fill these gaps by quantifying observed changes in thermal stability, water temperature, dissolved oxygen, turbidity, pH, and specific conductivity in five southeastern Connecticut lakes using data from 2003-

2018. In addition to determining trends, long-term relationships between observed monthly water quality measurements and lagged wind speed and air temperature averages during the same time period are assessed for each study lake. To represent the effects of climate change, model-based historical (1971-

2000) and future (2041-2070) air temperature and wind speed corresponding to each lake is projected using GCMs under RCP 8.5.

Rising air temperatures affect lake water temperature and thermal stability. For example, Siver et al.

(2018) analyzed data from 1985-2015 in Candlewood Lake, Connecticut, and found both rising air temperatures and declining April-August average wind speed (31%) drove decreases in bottom water

10 temperature and stronger thermal stability during the summer. Rising air temperatures, declining wind speed, and stronger thermal stability are also linked to bottom water dissolved oxygen depletion

(Hutchinson, 1957; Foley et al., 2011). This likely occurs because strong thermal stability in summer isolates bottom water dissolved oxygen and temperature from access to atmospheric oxygen and heating.

Moreover, bottom water dissolved oxygen is used for microbial respiration (i.e., algal decomposition), and in eutrophic systems this typically leads to a reduction in summer bottom water dissolved oxygen

(Foley et al., 2011).

Long-term changes in turbidity and pH in lakes are not well understood, nor are the potential drivers of changes in these parameters. Turbidity measures water clarity based on the light scattering effects of suspended particulate matter in water; the higher the turbidity the less light is able to penetrate water.

Historically, soil erosion from arable land has been linked to higher turbidity in streams (Braskerud,

2001). Additionally, increasing wind speed is positively correlated with higher water turbidity due to sediment resuspension (Bever et al., 2018). Increasing algae in lakes leads to increases in turbidity as well

(Scheffer, 1998). For pH, natural factors are more likely to affect long-term changes than anthropogenic factors (Khatri & Tyagi, 2015). CO2 oversaturation, which reduces water pH levels and alkalinity, has been documented in surface waters (Stets et al., 2017). Moreover, soluble iron, phosphorus, and manganese are positively correlated with CO2 oversaturation and lower pH in lakes and soils (Kortelainen et al., 2013). CO2 oversaturation contributes to an imbalance in ΔCO2:ΔO2 stoichiometry in many rivers and streams though the level of imbalance depends on the buffering capacity of individual systems (Stets et al., 2017). In addition to CO2 oversaturation, anaerobic metabolism, acidification, and rooted plant respiration may contribute to this imbalance (Stets et al., 2017).

Specific conductivity (i.e., electrical conductivity) of freshwater is determined by the major ions calcium, sodium, and chloride and should remain relatively constant in lakes unless there is an input of dissolved ions, such as from agricultural or coal mining runoff or winter road salting activities (Granato &

Smith, 1999). Anthropogenic salinization of streams is well established (Kaushal et al., 2005; Cañedo-

Argüelles et al., 2016; Timpano et al., 2018; Kelly et al., 2019) while salinization of lakes is less well documented. Prior studies suggest surface coal mining (e.g., mountain top removal) (Cormier et al., 2013;

Timpano et al., 2018) and road salting activities during winter (Perera et al., 2013) are increasing the specific conductivity of northeastern U.S. streams. Specific conductivity has been found to increase as a function of the amount of impervious surface in watersheds (Kaushal et al., 2005), and Perera et al.

(2013) suggests that up to 40% of the chloride applied to roads may enter streams and lakes and lead to a gradual increase in specific conductivity. Salinization of freshwaters deteriorates drinking water quality

(Ramakrishna & Viraraghavan, 2005; Kaushal et al., 2005), and is an urgent ecological issue (Cañedo-

Argüelles et al., 2013) harming benthic macroinvertebrates, fish, salamander, and aquatic insects (Perera et al., 2013; Timpano et al., 2018).

Drinking water managers often take water temperature, dissolved oxygen, pH, specific conductivity, and turbidity depth profile measurements in reservoirs once per month or more frequently for monitoring purposes. Long-term time series including these parameters may provide valuable insights into how climate change or other factors may be affecting water quality. This study uses long-term reservoir water quality depth profiles provided by the South Central Connecticut Regional Water Authority to investigate the extent to which water quality in five small reservoirs is affected by long-term changes in air temperature and wind velocity over sixteen years from 2003-2018. This research focuses on Connecticut reservoirs as a case study because air temperature in the Northeast is increasing faster than any other region in the contiguous U.S. (Karmalkar & Bradley, 2017), with Connecticut showing significant increasing trends in annual temperature (+0.3 °F per decade since 1895) (Lynch et al, 2016; CTPCSAR,

2019). In this study, we aim to answer the following questions: (1) Has surface, average, or bottom water quality of select Connecticut reservoirs changed in recent decades? (2) How has air temperature and wind speed near lakes changed? And, (3) how do changing air temperature and wind speed affect reservoir water quality?

3.2 Methods

Reservoir depth profile data

The South Central Connecticut Regional Water Authority provided in situ water quality depth profile data for five reservoirs – Lake Whitney, Lake Dawson, Lake Watrous, Lake Gaillard, and Lake

Saltonstall. Data includes surface, average, and bottom water temperature (°C), dissolved oxygen (%

SAT), pH, specific conductivity (nS cm-1), turbidity (NTU), and total RTRM (unitless). In addition, we use vertical water temperature profile data to compute total RTRM (Birge & Juday, 1914; Kortmann et al., 1982) (Eq. 1).

Density of the Upper Layer −Density of the Lower Layer total RTRM = ∑푏표푡푡표푚 (1) 푠푢푟푓푎푐푒 Density at 5°C −Density at 4°C

Measurements for Lake Whitney, Lake Dawson, Lake Watrous, and Lake Gaillard were taken over a

16-year period from 2003-2018, while measurements in Lake Saltonstall were taken over a slightly shorter 14-year period from 2005-2018. Depth profile measurements were taken at one sampling point in each lake, once monthly from April-October, using a submersible multiparameter water quality probe every 1 meter, beginning at the reservoir surface and ending just above the reservoir bottom, except for

Lake Saltonstall. The max sampling depth in Lake Saltonstall was 12 m which is short of the average depth of 13.2 m. However, we estimate that the sampling point likely represents the lake well given the portion of the lake included up to 12 m represents approximately 91% of the lake volume.

Wind speed and air temperature data

Observed and statistically downscaled (i.e., 1/24th degree resolution, approximately 4km) daily air temperature and wind speed data over forty years from 1/1/1979-12/31/2018 is obtained from the gridMET dataset (publicly available at http://www.climatologylab.org/gridmet.html; Abatzoglou, 2013).

The gridMET data for each reservoir is taken from the grid cell where the reservoir is located. Daily mean air temperature is calculated by taking the average of the daily max and min air temperature. Daily mean

13 wind speed at 10m height used to assess long-term changes in wind speed and is calculated from the eastward and northward wind vectors, u (wind speed from the west) and v (wind speed from the south).

To represent the effects of climate change, model-based historical (1971-2000) and future (2041-

2070) max and min air temperature, and wind velocity from the west and east direction, corresponding to each lake are taken from MACAv2-METDATA dataset (publicly available at https://climate.northwestknowledge.net/MACA/data_csv.php; Abatzoglou & Brown, 2012). MACAv2-

METDATA is a statistically downscaled climate database produced by applying the Multivariate

Adaptive Constructed Analogs (MACA) approach (Abatzoglou & Brown, 2012) to 20 Coupled Model

Inter-Comparison Project Phase 5 (CMIP5) GCMs, using gridMET as the training data. RCP 8.5 was chosen for this analysis because greenhouse gas emissions have followed this scenario over the past decade (Sanford et al., 2014). HadGEM2-CC365 from the UK Met Office, CCSM4 from NCAR (U.S.), and GFDL-ESM2M from NOAA GFDL (U.S.) are used for analysis because they closely track available observations and capture a range of temperature sensitivity for Connecticut (Seth et al., 2019). Sensitivity is based on the expected time it will take for surface air temperature to reach a 2 °C increase relative to

2000 under RCP 8.5 based on the model projections (Miao et al., 2014). Models where it takes longer to reach the threshold 2 °C increase are considered less temperature sensitive. CCSM4 and GFDL-ESM2M have low-medium temperature sensitivity for Connecticut, while HadGEM2-CC365 has medium-high temperature sensitivity for Connecticut.

Evaluating the significance of long-term trends

Unique time series regression models are developed to assess changes in surface (upper 2m of the water column), average (total water column), and bottom (lower 2m of the water column) water quality trends in the study reservoirs; including water temperature, dissolved oxygen, pH, specific conductivity, and turbidity. Changes in total RTRM are also evaluated. The significance of trends depend on the noise of the data which depends on the variability (both natural and analytical), autocorrelation (seasonal and interannual cycles), and the length of the data period (Tiao et al., 1990, Weatherhead et al., 1998, Henson

14 et al., 2016; Staniec & Vlahos, 2017). Water quality trends are evaluated using an approach which isolates long-term trends from autocorrelation in data and noise due to natural variability. Prior work by Staniec and Vlahos (2017) found that less time is required to verify linear trends using three 4-month bins

(February-May, June-September, and October-January) compared to six 2-month bins, four 3-months, two 6-month bins, or one 12-month bin. Using 4-month binned averages helps to reduce the amount of noise in the time series associated with seasonal cycling (Staniec & Vlahos, 2017). This method requires equal temporal intervals between datapoints. Because only data from April-October is available, we use

October data as a proxy for the October-January bins, and April-May data as a proxy for the February-

May bins. We recognize that the October-January bin may have high uncertainty because October is a turnover month, meaning that water quality measurements taken in October may vary greatly depending on the exact sampling date and year, however this did appear to have greatly impacted our assessment of annual trends. We assess long-term trends using these proxies for the 4-month binned averages (February-

May, June-September, and October-January).

Building on this method further, we developed multiple regression linear time series models which can account for seasonal variation in the magnitude (slope) of trends. We fit two multiple regression models (varying intercepts and slopes) for each variable and conduct a Likelihood Ratio Test to select the most reasonable model (see: Eq. 2 and 3).

푌̂푖,휔(푡) = µ0퐷 + µ1퐹 + µ2퐴 + 휔푋(푡) + N(t) (2)

푌̂푖,휔(푡) = µ0퐷 + µ1퐹 + µ2퐴 + 휔0푋(푡) + 휔1푋(푡) + 휔2푋(푡) + 푁(푡) (3)

Here, 푦̂푖,,휔(푡) is the time series data (e.g., parameter at time t), 휔 is the magnitude (slope) of the trend,

푋(푡) is time, and 푁(푡) is the noise of the data which depends on the variability (both natural and analytical), autocorrelation (seasonal and interannual cycles), and the length of the data period (Tiao et al., 1990, Weatherhead et al., 1998, Henson et al., 2016; Staniec & Vlahos, 2017). In Eq. 2, the same

(average) slope is used for each seasonal bin but the October-January (µ0푑), February-May (µ1푓), and

June-September (µ2퐴) y-intercepts are allowed to vary. In Eq. 3, the y-intercepts and slope values (휔푖) are allowed to vary for each seasonal bin.

A Likelihood Ratio Test is conducted to compare the models. The Likelihood Ratio Test is based on the F test (Fotheringham et al., 2002, p. 94). If the test is significant (P ≤ 0.05), this suggests Eq. 3, the model with varying seasonal slopes, is the better model and should be used to assess trends for that parameter. If the test is not significant, then Eq. 2, the annual model, is the better model. A Student’s t- test is used to test for significance of the magnitude of the trend (Caldwell et al., 2014) consistent with prior studies (Decremer et al., 2014; Levine & Berliner, 1999). Significant trends are those with 95% confidence (P ≤ 0.05) or better (P ≤ 0.01 or P ≤ 0.001).

Evaluating observed relationships and projecting future change in wind speed and air temperature

The relationship between monthly surface, average, and bottom water quality data from April-

October for each reservoir and daily observed and statistically downscaled air temperature and wind speed from the gridMET dataset is assessed using linear regression. Specifically, the relationship between the 1-365 day antecedent air temperature and wind speed averages and each water quality parameter is evaluated to find the average that has the strongest linear relationship (i.e., highest R2) with the water quality parameter on the next day (see Fig. 2).

A) Observed Daily Air Temperature (T) and Monthly Total Relative Thermal Resistance to Mixing (RTRM) at Lake Saltonstall 40 Air Temperature 500 Total RTRM 30 400 20

C) 300 ° 10 200

Air ( T Air 0 Total RTRM Total

-10 100

-20 0 8/1/2004 4/28/2007 1/22/2010 10/18/2012 7/15/2015 4/10/2018 Time

R2 for Total RTRM Predictions vs. Number of Days B) Total RTRM vs. Best Fit Antecedent Air T Average for Used in Antecedent Air T Averages C) 1 Each Lake Saltonstall 500 Saltonstall Whitney 0.8 Whitney Watrous 400 Watrous Dawson Dawson 0.6 Gaillard 300 Gaillard

0.4 200 Total RTRM Total 0.2 100

and Air T T Average Air and 0 0

0 100 200 300 0 10 20 30 for Regression of Total RTRM Total of for Regression

2 Antecedent Air T Average (Number of Days Used in the

R Number of Days Used in Air T Average Average Determined by Highest R2 in Plot B) D) Observed Daily Wind Speed and Monthly Total RTRM at Lake Saltonstall Wind Speed 18 500 Total RTRM 16

) 400 1

- 14 12 300 10 200 8

6 100 RTRM Total

Wind Speed (m s (m Speed Wind 4 0 2 0 -100 8/1/2004 4/28/2007 1/22/2010 10/18/2012 7/15/2015 4/10/2018 Time E) R2 for Total RTRM Predictions vs. Number of Days F) Total RTRM vs. Best Fit Antecedent Wind Speed Used in Antecedent Wind Speed Averages Average for Each Lake 1 500 Saltonstall Saltonstall Whitney 0.8 400 Whitney Watrous Watrous Dawson 0.6 300 Dawson Gaillard Gaillard

0.4 200 Total RTRM Total 0.2 100 0

and Wind Speed Average Speed Wind and 0 for Regression of Total RTRM Total of Regression for

2 2 2.5 3.5 4.5 5.5 6.5

0 100 200 300 R Number of Days Used in Wind Speed Average Antecedent Wind Speed Average (Number of Days Used in the Average Determined by Highest R2 in Plot E) Fig. 2: Graphic showing methodology used to determine the relationship between air temperature (T) and wind speed with water quality parameters. Here, the relationship between air temperature (A-C) and wind speed (D-F) and total Relative Thermal Resistance to Mixing (RTRM) are shown as an example.

The information provided by R2 and the slope of the regression indicates the strength and sign (positive or negative) of the correlation.

Historical (1971-2000) and future (2041-2070) air temperature and wind speed are projected under

RCP 8.5 using downscaled climate data averaged from the 3 GCMs described earlier (CCSM4, GFDL-

ESM2M, and HadGEM2-CC365). Monthly average air temperature and wind speed across all lakes during the historical and future climate periods are compared and a 2-tailed t-test is conducted to compare the means for each month. Again, significant changes are those with 95% confidence (P ≤ 0.05) or better

(P ≤ 0.01 or P ≤ 0.001).

3.3 Results

Reservoir water quality trends & air temperature and wind induced impacts

Model coefficients, Likelihood Ratio Test results, and significance test results based on the Student’s t-test for each parameter and lake are included in Appendix 6.1, Supporting Material, Table 1. Results suggest the simpler linear trend model (Eq. 2), using the same slope for each bin, is appropriate in most cases. However the more complicated model (Eq. 3), using different slopes for each bin, is a better fit for surface, average, and bottom temperature trends in Lake Dawson, and for surface dissolved oxygen trends in Lake Saltonstall. For all other lakes and parameters, Eq. 2 is used to evaluate trends. Results from the regression between air temperature and wind speed with each water quality parameter including R2, the number of days used in the best fit air temperature and wind speed averages, slope, intercept, standard error, significance F, and Pearson Correlation Coefficients for each lake and parameter are reported in

Appendix 6.1, Supporting Material, Tables 2a and 2b.

Water temperature – Annual surface water temperature in Lake Whitney, Lake Watrous, Lake

Gaillard, and Lake Saltonstall is increasing by 0.05 to 0.11 °C yr-1 (total increase of approximately 0.77 to

1.69 °C from 2003-2018), though this trend is only significant in Lake Gaillard (P ≤ 0.01), where surface temperature is increasing by 0.113 °C yr-1 (total increase of 1.69 °C from 2003-2018). Annual average

18 water temperature is decreasing in Lake Whitney and Lake Watrous, and increasing in Lake Gaillard and

Lake Saltonstall, though none of these trends are significant. Annual bottom water temperature is decreasing in Lake Whitney, Lake Watrous, Lake Gaillard, and Lake Saltonstall by -0.05 to -0.10 °C yr-1

(total decrease of approximately -0.75 to -1.35 °C from 2003-2018), and the trend is significant in Lake

Gaillard and Lake Saltonstall (P ≤ 0.05), where bottom temperature is decreasing by -0.05 °C yr-1 and -

0.10 °C yr-1, respectively. In Lake Dawson, Eq. 3 is used to assess annual surface, average, and bottom temperature trends. February-May and June-September surface and average water temperature is increasing, though increases in February-May are occurring over 3 times faster compared to increases in

February-May, and only February-May trends are significant (P ≤ 0.01). February-May surface and average water temperature is increasing rapidly by 0.37 °C yr-1 and 0.27 °C yr-1 (total increase of 5.49 and

4.01 °C from 2003-2018), respectively. On the contrary, surface, average, and bottom temperature trends in January-October are all decreasing significantly by -0.202 °C yr-1, -0.239 °C yr-1, and -0.314 °C yr-1 (P

≤ 0.01), respectively. This is a total decrease of -3.03 °C, -3.58 °C, and -4.71 °C over sixteen years from

2003-2018.

Across lakes, surface, average, and bottom water temperature increase with persistent high air temperatures and decrease with persistent high wind speeds. Surface water temperature responds to changes in air temperature over the shortest period of time (8-21 days), followed by average water temperature (25-48 days), and bottom water temperature (69-233 days). Surface and average water temperature are also influenced by wind speed but over a longer period of time compared to air temperature. Surface water temperature responds to changes in average wind speed over 38-50 days while average water temperature responds to changes in average wind speed over 50-65 days. Both surface water temperature and average water temperature are highly correlated with antecedent mean air temperature averages and antecedent mean wind speed averages (R2 = 0.91 ± 0.03), however air temperature is the most influential parameter determining water temperature. The range of R2 values for bottom water temperature estimates using antecedent mean air temperature averages and antecedent mean

19 wind speed averages is high, ranging from 0.29-0.65, depending on the lake. This suggests other factors in addition to air temperature and wind speed may be more important for determining bottom water temperature in lakes than they are for determining surface or average temperature. On average, simple linear regression indicates that 36 ± 11 day antecedent air temperature averages can predict surface water temperature on a given day with an R2 of 0.91 ± 0.03. Similarly, 60 ± 6 day antecedent wind speed averages can predict average water temperature on a given day with an R2 of 0.79 ± 0.05.

Thermal stability – Annual total RTRM is increasing in all lakes by approximately 2.59 to 3.71 yr-1

(total increase of approximately 39 to 56), and increases are significant in 4 out of 5 lakes (all except Lake

Saltonstall). 8 ± 4 day antecedent air temperature averages are positively correlated with total RTRM with an R2 between 0.69-0.90 depending on the lake, indicating persistent high air temperature over a week or so leads to higher total RTRM. For wind speed, 25 ± 7 day antecedent wind speed averages are negatively correlated with total RTRM with an R2 between 0.38 to 0.72 across lakes, meaning persistent high wind speed over a month or so leads to lower total RTRM.

Dissolved oxygen – Lake Whitney, Lake Dawson, Lake Watrous, and Lake Gaillard are experiencing significant increases in annual surface dissolved oxygen of 1.44 to 1.74 % SAT yr-1 (P ≤ 0.05), and are becoming supersaturated with oxygen (i.e., greater than 100 % SAT) more often over time. In Lake

Saltonstall, February-May and October-January surface dissolved oxygen is increasing by 2.5 % SAT yr-1 and 1.2 % SAT yr-1, respectively (both significant, P ≤ 0.05), and June-September surface dissolved oxygen is increasing by 2.7 % SAT yr-1. Additionally, annual average dissolved oxygen is increasing in all five lakes by 1.46 % SAT yr-1, 1.24 % SAT yr-1, 0.72 % SAT yr-1, 0.85 % SAT yr-1, and 0.73 % SAT yr-1 in Lake Saltonstall, Lake Watrous, Lake Whitney, Lake Gaillard, and Lake Dawson, respectively (all significant with 95% confidence or higher). This is a total increase in dissolved oxygen of 15-35 % SAT at the surface and 11-19 % SAT on average, depending on the lake and time of year. Changes in annual bottom dissolved oxygen are not significant, suggesting a longer study period is needed to evaluate bottom dissolved oxygen trends.

Persistent high air temperature over 13-78 days leads to lower average dissolved oxygen across all study lakes (R2 between 0.33-0.73, depending on the lake). Similarly, persistent high air temperature over

11-106 days leads to lower bottom dissolved oxygen across lakes (R2 between 0.45-0.84, depending on the lake). On the other hand, persistent high wind speed over a longer period of time (39-127 days) leads to higher average and bottom dissolved oxygen with an R2 of 0.33-0.77 across all five lakes. Persistent average air temperature is a slightly better predictor of bottom water dissolved oxygen (R2 between 0.45-

0.84), compared to persistent average wind speed (R2 between 0.35-0.77). At the lake surface, these relationships between wind speed and air temperature are reversed in 4 out of 5 lakes, meaning that persistent higher air temperatures are associated with higher dissolved oxygen (R2 ≤ 0.02) and persistent high wind speeds are associated with lower dissolved oxygen at the surface (R2 between 0.04-0.28).

While these relationships are not as strong, this result suggests other factors in addition to air temperature and wind speed are very important for determining surface dissolved oxygen trends.

Turbidity – Annual surface turbidity is decreasing in all lakes by -0.46 to -0.19 NTU yr-1, and this trend is only significant in Lake Whitney and Lake Saltonstall (P ≤0.05), where annual surface turbidity is decreasing by -0.19 NTU yr-1 and -0.12 NTU yr-1, respectively. This equates to a total decrease in surface turbidity of -2.82 NTU and -1.56 NTU over sixteen years in Lake Whitney (2003-2018) and fourteen years (2005-2018) in Lake Saltonstall. Annual average turbidity is also decreasing in all lakes by -0.03

NTU yr-1 to -0.10 NTU yr-1, though this decrease is only significant in Lake Dawson and Lake Saltonstall

(P ≤0.05), where annual average turbidity is decreasing by -0.10 NTU yr-1 in both. This equates to a total decrease in average turbidity of -1.32 NTU and -1.34 NTU over sixteen years in Lake Gaillard (2003-

2018) and fourteen years (2005-2018) in Lake Saltonstall. Changes in bottom turbidity are more variable across lakes in that bottom turbidity is increasing in Lake Whitney and Lake Saltonstall, and decreasing in

Lake Dawson, Lake Watrous, and Lake Gaillard. However, bottom water turbidity trends are only significant in Lake Gaillard (P ≤0.05), where bottom turbidity is decreasing by -0.11 NTU yr-1. This equates to a total decrease in bottom turbidity of -1.64 NTU over sixteen years in Lake Gaillard.

The relationship between antecedent air temperature and wind speed averages and surface turbidity are not very strong, but highest in Lake Dawson and Lake Gaillard. In Lake Dawson and Lake Gaillard higher air temperature and lower wind speed over 3-4 days and 9-36 days respectively, corresponds with lower turbidity. In Lake Whitney, this relationship is opposite.

pH – Annual surface pH is decreasing by -0.01 to -0.03 yr-1 in all lakes (significant in Lake Dawson and Lake Saltonstall, P ≤ 0.01) except Lake Watrous, where annual surface pH is increasing by 0.008 yr-1

(not significant). In Lake Dawson and Lake Saltonstall, this equates to a total decrease in annual surface pH of -0.45 and -0.42 respectively over the 16-year study period. Annual average pH is decreasing in

Lake Whitney, Lake Dawson, and Lake Saltonstall by -0.14 to -0.27 yr-1 (only significant in Lake

Saltonstall, P ≤ 0.05). In Lake Saltonstall this equates to a total decrease in annual average pH of -0.36 over the 16-year study period. In contrast, annual average pH in Lake Watrous and Lake Gaillard is increasing (not significant). Annual bottom pH trends are similar to average pH trends in that bottom pH is decreasing in Lake Whitney, Lake Dawson, and Lake Saltonstall and increasing in Lake Watrous and

Lake Gaillard, however no bottom pH trends are significant.

Linear regression between pH and air temperature or wind speed averages are not as strong as they are for water temperature, total RTRM, or dissolved oxygen (all R2 ≤ 0.37). Nevertheless, Pearson

Correlation results suggest air temperature and wind speed have an equal and opposite effect on surface, average, and bottom water pH. Meaning persistent high air temperatures lead to lower surface, average, and bottom pH, while persistent high wind speeds lead to higher surface, average, and bottom pH across lakes. However this is not the case in Lake Whitney and Lake Saltonstall, where persistent high air temperature over 4-5 days leads to higher surface pH (Pearson Correlation Coefficient ≥ 0.40) and persistent high wind speed over 10-16 days leads to lower surface pH (Pearson Correlation Coefficient ≥

0.40) in Lake Whitney and Lake Saltonstall.

Specific Conductivity – Annual specific conductivity is increasing significantly in all lakes (P ≤

0.001) by 2.53 nS cm-1 yr-1 to 11.56 nS cm-1 yr-1 at the surface, 2.39 nS cm-1 yr-1 to 11.58 nS cm-1 yr-1 on

22 average, and 2.50 nS cm-1 yr-1 to 13.05 nS cm-1 yr-1 at the bottom. Changes in surface, average, and bottom specific conductivity are similar within a given lake but rates of change vary across lakes.

Average specific conductivity has increased by 173.40 nS cm-1 and 141.70 nS cm-1 total over the past sixteen years (2003-2018) in Lake Whitney and fourteen years (2005-2018) in Lake Saltonstall. The increase is less in Lake Dawson and Lake Watrous, where the average specific conductivity has increased by 83.52 cm-1 and 67.83 cm-1 over the last sixteen years (2003-2018), respectively. The increase is even less in Lake Gaillard, where average specific conductivity has increased by 35.90 cm-1 from 2003-2018.

In summary, changes in Lake Whitney and Lake Saltonstall are approximately twice that of Lake Dawson and Lake Watrous, and over 3 times that of Lake Gaillard. All changes in specific conductivity are significant with 95% confidence or higher.

There is a slight positive relationship between specific conductivity and antecedent air temperature over a long time period, meaning a higher 135-365 day antecedent air temperature average is correlated with higher surface, average, and bottom water specific conductivity (i.e., correlation between 0.28 and

0.57 depending on the lake). The relationship between antecedent wind speed averages and specific conductivity is not as clear across lakes.

Historical trends and future projections of air temperature and wind speed

Historical trends – The significance of historical air temperature and wind speed trends are assessed similarly to water quality trends. Results suggest the simpler linear trend model (Eq. 2), using the same slope for each bin, is appropriate for assessing historical air temperature and wind speed trends. Model coefficients, Likelihood Ratio Test results, and significance test results based on the Student’s t-test for air temperature and wind speed trends are also included in Appendix 6.1, Supporting Material, Table 1.

Air temperature at each reservoir location is increasing by 0.008 °C yr-1 to 0.017 °C yr-1 across lakes, though this increase is only significant at Lake Watrous and Lake Gaillard (P ≤ 0.05). This is a total increase of 0.54 °C and 0.66 °C over forty years at Lake Watrous and Lake Gaillard, respectively.

Furthermore, assuming a linear trend, this is a total increase of 0.11 °C to 0.25 °C over sixteen years

23 depending on the lake. Annual wind speed is increasing significantly by 0.005 m s-1 yr-1 to 0.006 m s-1 yr-1 at all lakes (P ≤ 0.001). This is a total increase of 0.19 to 0.30 m s-1 over forty years, or 0.07-0.08 m s-1 over sixteen years. A deeper look at the monthly scale suggests observed air temperature is increasing significantly year-round, while wind speed is increasing significantly in October across lakes.

Future climate model projections – Past and future climate model projections under RCP 8.5 (i.e., the average of HadGEM2-CC365, CCSM4, and GFDL-ESM2M) are used to assess future changes in air temperature and wind speed. Results from this assessment are included in Appendix 6.1, Supporting

Material, Table 3. According to the climate model projections, air temperature is increasing significantly during all months (P ≤ 0.001). Air temperature is expected to increase by 2.5 °C to 3.7 °C total over the

70-year period from 2041-2070 to 1971-2000, which is an increase of 0.036 °C yr-1 to 0.053 °C yr-1 depending on the lake. Wind speed changes are more variable throughout the year, with: (1) January,

April, May, June, October, November, and December wind speed decreasing; (2) February, March, July, and August wind speed increasing; and, (3) no changes in wind speed in September. On average annual wind speed is expected to decrease by -0.05 m s-1 total over the 70-year period from 2041-2070 to 1971-

2000 across lakes, which is a decrease of -0.0007 m s-1 yr-1. However only changes in April, May, June,

July, and October are significant (P ≤ 0.05), with April-June and October wind speed decreasing by -0.10 m s-1 to -0.22 m s-1 total over the 70-year period and July wind speed increasing by 0.09 m s-1 over the same period.

3.4 Discussion

Higher air temperatures are likely responsible for the increase in annual thermal stability (i.e., total

RTRM) and annual surface water temperature, especially during the spring and summer months; while increasing thermal stability associated with higher air temperatures likely combine to cause decreases in annual bottom water temperature. Persistent high air temperatures over a week or so lead to higher thermal stability across the study lakes, while persistent high wind speeds over approximately one month leads to lower thermal stability across all lakes. Also, water temperature increases with increasing average

24 air temperature over approximately one month and decreases with increasing average wind speed over approximately two months across all lakes. Air temperature impacts surface water temperature in the shortest amount of time compared to average and bottom water temperature. Air temperature impacts both water temperature and thermal stability over a shorter amount of time than wind speed, suggesting increases in air temperature may have more of an impact on thermal stability and surface water temperature than wind speed.

In 4 out of 5 lakes, seasonal differences in water temperature trends were not significant. However, in

Lake Dawson, February-May and June-September surface water temperature is increasing significantly, with February-May average water temperature increasing over 3 times faster than increases in June-

September. Also, in Lake Dawson average water temperature in October-January is decreasing significantly by about the same amount as February-May water temperature is increasing. These seasonal differences provide insights that may help explain long-term changes occurring in all of the study lakes.

This suggests spring and summer surface and average water temperature may be increasing fast as a result of higher air temperatures and resulting increases in thermal stability. Earlier thermal stability would lead to reduced lake mixing early in the season, which would separate cool winter bottom waters from warmer surface water temperatures early in the year. Therefore, a fast increase in surface water temperature and total RTRM in spring may explain the increase in annual surface water temperature and decrease in bottom water temperature observed in the study lakes. Future projections of air temperature suggest air temperatures will continue to increase and cause increases in surface water temperature and thermal stability in the future. Changes occurring during the fall months are less clear, though decreasing October-

January average water temperature in Lake Dawson suggests the fall turnover may actually be occurring earlier.

Impacts of future changes in wind speed on thermal stability and temperature are not clear due to discrepancies between observed trends and future projections of wind speed. Observed annual wind speed is increasing, however future projections of wind speed suggest wind speed will decline, especially in the spring months from April-May and in October. Regional variation in wind speed trends may explain

25 some of this inconsistency. For example, Siver et al. (2018) suggests April-August average wind speed declined 31% from 1985-2015 at Candlewood Lake, Connecticut which is located approximately 72 km northwest of the study lakes and further from the coastline. Therefore, it is possible changes in wind speed further south and closer to the coastline (i.e., where the study reservoirs are located) differ from inland changes in wind speed. On a much larger scale, the North Atlantic Oscillation (NAO) index (i.e., normalized sea-level pressure difference between the Azores and Iceland) quantifies large-scale atmospheric variability over the North Atlantic (Hurrell, 1995; Sarafanov, 2009), and interdecadal changes in the NAO index may impact westerly winds and storms, with stronger winds and storms occurring when the NAO index is high. The NAO index appears to have been relatively high over the study period from 2003-2018 (Hurrell & Phillips, 2018), and may have contributed to the higher observed wind speeds and dissolved oxygen. It is also possible the NAO index has a larger influence on lakes closer to the coast and may be partially responsible for regional inconsistencies. The future wind speed projections are consistent with Karnauskas et al. (2018) who found northern mid-latitude wind velocity may be declining as the climate warms due to large-scale reductions in air temperature differences between regions. However the future wind speed projections are likely uncertain due to the moderately complex land in the Northeast (e.g., high variation in elevation), and uncertainty in the vertical mixing strength of wind in the atmosphere (Hu et al., 2010; Frediani et al., 2016). Because of the potential influence of NAO and climate warming on wind speed, additional analysis and years of data are needed to determine weather future wind speeds will follow the historical trajectory (increase) or the climate model projections (decrease) in the future.

Theoretically, increases in wind speed throughout the year should combat increases in thermal stability and should lead to more uniform water temperature trends. Therefore it is unclear how the observed increases in wind speed may be influencing the observed changes water temperature and thermal stability. If the future projections of wind speed are accurate, then declining spring wind speed combined with higher air temperatures could lead to even faster rates of increase in total RTRM in the

26 study lakes in the future. Lower spring wind speed would lead to even higher summer thermal stability since changes in average wind speed over 1-2 months greatly affects thermal stability. Prior work by

Siver et al. (2018) supports the theory that bottom water temperatures may be declining due to decreases in spring wind speed and associated increases in thermal stability. Palmer et al. (2014) also found bottom water temperature is decreasing in Wisconsin lakes.

While this research found bottom water temperatures are declining due to higher air temperatures and resulting increases in thermal stability, other scholars are predicting future warming of bottom waters

(Butcher et al., 2015; Komatsu et al., 2007). For example, Butcher et al. (2015) suggests bottom water temperatures will increase about 33% based on a study of modulated thermal responses in 33 lakes across the U.S., while Komatsu et al. (2007) predicts bottom water temperature will increase in the Shimajigawa reservoir located in western Japan by 2.8 °C by the 2090s, compared to the 1990s. The contrasting vision of bottom water temperature may be because these studies evaluated annual trends but did not fully consider future changes in thermal stability. Alternatively lakes in different geographic locations and of varying shapes and sizes may be affected differently by these changes.

Results suggest annual surface and average dissolved oxygen is increasing significantly across all lakes possibly due to the higher wind speeds observed across the year. Generally, we found lower air temperatures and higher wind speeds are associated with higher average dissolved oxygen across the study lakes. Increased atmospheric aeration due to high wind speed may directly put oxygen and CO2 into the water (Schippers et al., 2004). This result is consistent with Palmer et al. (2014) who found average dissolved oxygen is increasing in Wisconsin lakes. However, changes in bottom water dissolved oxygen are inconclusive. Other studies found declining bottom dissolved oxygen levels in lakes due to rising air temperatures, declining wind speed, and stronger thermal stability (e.g., Hutchinson, 1957; Foley et al.,

2011; Schmidtko et al., 2017). Here, decreases in bottom water temperature are likely counteracting expected decreases in bottom dissolved oxygen. It is also possible bottom water respiration is decreasing due to reduced algal productivity as suggested by the observed decrease in annual turbidity.

While we would expect dissolved oxygen to decrease with increasing water temperature, we found the opposite effect at the lake surface where higher air temperatures are associated with higher surface dissolved oxygen. This may be occurring because biology may be a stronger factor affecting dissolved oxygen at the lake surface. Prior work suggested a productivity increase of 50% in eutrophic freshwater systems is expected with a 50% increase in atmospheric CO2 (Schippers et al., 2004). Supersaturation of dissolved oxygen in water can only occur when primary production exceeds respiration, therefore the observed increase in surface dissolved oxygen supersaturation during summer may be indicative of increasing biological productivity. Both increases in wind speed and increased biological productivity in combination with increasing surface water temperatures may be driving annual surface and average dissolved oxygen upward in these freshwater lakes. Although we did not find a significant bottom water dissolved oxygen trend, solubility of oxygen decreases with increasing conductivity, which means these reservoirs may be more at risk of hypoxic conditions in the bottom waters as specific conductivity continues to rise. Without long-term chlorophyll-a data, it is unclear whether the increase in dissolved oxygen is due to increases in wind speed alone or if changes in biological productivity also increase dissolved oxygen. Nevertheless, if these trends persist into the future, higher air temperature and higher wind speed may lead to even greater increases in surface and average dissolved oxygen. However future projections of wind speed suggest April, May, June, July, and October wind speed may decline. Therefore future changes in dissolved oxygen are uncertain because future decreases in wind speed may reduce or reverse this increasing trend over time.

Annual surface and average turbidity are decreasing in the study lakes indicating improved water clarity. On one hand, turbidity is decreasing most during the summer months suggesting higher total

RTRM and reduced lake mixing in summer may be contributing to reduced sediment resuspension and lower turbidity. On the other hand, improved water clarity may mean there is reduced algal growth. This would support the theory that dissolved oxygen is increasing mainly due to increasing wind speed rather than increasing biological productivity. However the study lakes are becoming supersaturated with

28 oxygen at the surface more often during the summer which counteracts this idea. Additional research is needed to understand long-term changes in biological activity occurring in the reservoirs.

Oxygen and CO2 oversaturation could also explain the reduction in pH observed in 4 out of 5 study reservoirs (Stets et al., 2017). Lower lake pH levels may create problems for water treatment such as an increase in soluble iron, phosphorus, and manganese. Changes in pH would be greater in individual lakes with lower buffering capacity and this capacity may continue to decrease over time as pH declines (Stets et al., 2017).

Specific conductivity is increasing in all of the study lakes. Lake Whitney and Lake Saltonstall have higher amounts of impervious land cover (roads, buildings, etc.) in their watershed, and these two reservoirs also have higher water specific conductivity and have experienced the highest increase in specific conductivity over time. High levels of impervious surface in watersheds where road salt is applied to roadways is associated with high salinization rates in streams and in underlying aquifers that provide input to lakes (Kaushal et al., 2005; Perera et al., 2013). While little prior research exists on elevated chloride levels in Connecticut surface waters, a Connecticut Transportation Institute report

(2015) mentions that sodium concentrations in raw intake water frequently exceed 28 mg L-1 in Lake

Whitney and Mianus Reservoir in Connecticut. Prolonged chloride concentrations in the hundreds of mg

L-1 have also been observed in heavily urbanized areas in Toronto and Baltimore (Health Canada, 2001;

Kaushal et al., 2005). Kohlraush’s Law and the total capacity of each lake was used to determine how much salt would have needed to be applied to cause the observed increase in average specific conductivity (35.9-173.7 nS cm-1). If attributed entirely to addition of NaCl, the change in average specific conductivity in the study reservoirs could have been caused by an increase of 80-200 mg of NaCl per liter of water, depending on the lake. Theoretically 113-2994 pounds of dissolved road salt added to each of the lakes could account for this increase in average water specific conductivity observed. With approximately 200 pounds of salt and sand applied per lane mile, and an estimated 280,000-560,000 tons total used on Connecticut roads during snow and ice conditions (Connecticut Transportation Institute,

2015), it is very likely winter road salt is causing the observed increase in specific conductivity. Kaushal

29 et al. (2005) found that if salinity continues to increase at the present rate, many surface waters in the northeastern U.S. will not be suitable for human consumption or freshwater life by the end of next century.

3.6 Conclusion

In this study, long-term changes in thermal stability and surface, average, and bottom water temperature, dissolved oxygen, specific conductivity, pH, and turbidity are evaluated from 2003-2018 in four Connecticut drinking water reservoirs and from 2005-2018 in a fifth reservoir. In addition, the importance of changes in wind speed and air temperature on each water quality parameter is assessed over the same time period for each lake. Finally, future changes in air temperature and wind speed are projected out to midcentury (2041-2070) using General Circulation Model averages under RCP 8.5 to provide insights into possible future changes.

Surface water temperatures are increasing, bottom water temperatures are decreasing, and thermal stability is increasing in the study lakes mainly as a result of increasing air temperatures. Surface water temperature may be increasing faster in February-May, while average water temperature may be decreasing in October-January, suggesting spring stratification and the fall turnover may both be occurring earlier in the year. A faster rise in surface water temperature and thermal stability in spring may explain the increase in annual surface water temperature and decrease in annual bottom water temperature observed in most of the study lakes.

There is a strong relationship between persistent high air temperatures occurring over 1-2 weeks and increases in thermal stability on the following day. Consequently, it is likely rising air temperature and subsequent increases in thermal stability are driving surface water temperature up and bottom water temperature down. These changes are not likely driven by the observed increases in annual wind speed because higher wind speeds over 2-5 weeks are linked to lower thermal stability. Future projections suggest air temperature will continue to increase while spring wind speed is expected to decline by midcentury. If these future projections are accurate, then the decline in spring wind speed combined with

30 higher air temperatures will likely lead to even greater increases in surface water temperature, decreases in bottom water temperature, and increases in thermal stability in the study lakes in the future.

Annual surface and average dissolved oxygen levels are significantly increasing in all five study reservoirs, with surface dissolved oxygen increasing at a faster rate than average dissolved oxygen. The observed increase in surface and average dissolved oxygen in the reservoirs is likely due to a combination of increased atmospheric aeration due to higher wind speeds and higher biological productivity associated with higher air temperatures. All lakes are becoming supersaturated with oxygen more often in the summer which suggests biology may be a significant factor affecting dissolved oxygen at the lake surface during the summer. Future changes in dissolved oxygen may continue on the current trajectory or may be offset by projected future decreases in spring wind speed.

Changes in specific conductivity, pH, and turbidity are not greatly affected by climate, yet some of these parameters are changing over time for other reasons. Annual surface, average, and bottom specific conductivity is increasing significantly in all five lakes, and the two reservoirs with the highest percentage of impervious surfaces in their watershed are experiencing the greatest increase. These increases are consistent with what would be expected if increases are driven by road salt application and subsequent runoff and shallow aquifer inputs into lakes. Gradual salinization of reservoirs due to winter road salt application may have profound impacts on drinking water quality and treatment over time including the mobilization of iron, phosphorus, manganese, radium, and radon. The additional financial costs and potential human health impacts of road salt application are not currently accounted for, nor are the impacts on biology. Surface pH is decreasing in 4 out of 5 lakes possibly as a result of CO2 oversaturation. Changes in pH are greater in individual lakes with lower buffering capacity and this capacity may continue to decrease over time as pH declines, possibly leading to faster reduction in pH in some lakes in the future. In addition, surface and average turbidity is decreasing across all five lakes, possibly as a result of higher annual thermal stability and associated decreased vertical lake mixing and sediment resuspension or reduction in biological activity.

Overall, this research quantifies historical trends and provides insights into future changes in reservoir water quality in Connecticut. Whole lake dynamics are not fully represented by discrete monthly vertical profile samples, suggesting that additional information would be gained by future studies observing changes in the water quality parameters at higher temporal and spatial resolution. Missing winter data made it difficult to parse out seasonal changes in quality, therefore it is especially important for future studies to include full winter data if possible. Despite these limitations, many of the observed annual water quality trends have high certainty (i.e., 95% confidence). The observed changes in quality may have large impacts on reservoir water quality and treatment in the future. Studies should attempt to understand the costs and benefits of the changes brought to light and their associated impacts on drinking water treatment processes. For instance, future studies should aim to draw a more direct link between road salt usage, higher sodium and chloride levels in drinking water reservoirs, and the economic impacts of associated decreases in raw water quality. Findings may be generalizable to other freshwater reservoirs in

Connecticut, the Northeast U.S., and north temperate regions broadly.

4 Future projections of lake temperature, thermal stability, and implications for cyanobacteria

blooms in Connecticut, U.S.A. reservoirs amidst a warming climate

4.1 Introduction

Global warming induced changes in lake surface water temperature and thermal stability threaten the quality of freshwater resources around the world (Adrian et al., 1995; Magnuson et al., 1997; Arhonditsis et al., 2004), including reservoirs used for drinking water supply (Delpla et al., 2011). Studies of historical trends have shown lake surface water temperatures are warming, thermal stability is increasing, and lakes are becoming stratified earlier in the year (see Chapter 3). Warmer surface water temperatures and stronger thermal stability are also linked to higher biological growth (Downing & Duarte, 2009;

Descy et al., 2017), and cyanobacteria blooms (Paerl & Huisman, 2009; Wagner & Adrian, 2009; Rigosi et al., 2014; Persaud et al., 2015). Scholars have begun to examine climate change impacts on surface water temperature and algal productivity, especially in moderate sized and large temperate lakes (King et al., 1998; Dobiesz & Lester, 2009; Piccolroaz, 2016; Deng et al., 2018). Yet few studies evaluate impacts on small temperate lakes, and to our knowledge no prior studies have assessed future changes in lake thermal stability using air temperature projections from state-of-the-art GCMs. This study addresses this gap in understanding by: (1) developing empirical models for predicting daily surface, average, and bottom water temperature, and thermal stability using daily air temperature; (2) evaluating future changes in water temperature and thermal stability driven with statistically downscaled historical (1971-2000) and future (2041-2070) air temperature projections from GCMs; and, (3) providing insights into how climate induced changes in water temperature and thermal stability may affect the prevalence of cyanobacteria blooms in the future.

Climate impacts on lakes vary depending on geographic location (Livingstone & Dokulil, 2001), catchment features including lake shoreline and watershed characteristics (Kratz et al., 1997), lake morphometry (Magnuson et al., 1990), and lake history (Blenckner, 2005). North temperate lakes are potentially more vulnerable to climate warming than lakes in other geographic locations, because

33 increases in air temperature over past centuries have been greatest in northern latitudes (Giorgi et al.,

2005). However, prior research indicates climate warming induced impacts on thermal stability may vary depending on differences in lake morphology (i.e., surface area, surface shape, underwater form and depth, and shoreline irregularity), even in lakes that are located in the same geographic area (Fee et al.,

1996; Blenckner, 2005). Biological responses are complex and are more likely to vary across lakes in the same geographical area. Lake history may also affect how lakes respond to climate forcings (Blenckner,

2005). For example, historical land use in the catchment area can contribute nutrients to lakes and promote harmful algal blooms. Blooms may continue for many years even if nutrient loads have declined via internal nutrient recycling from sediments (Ahlgren, 1976). Lakes affected by historical nutrient loading may become eutrophic under climate warming due to internal release of phosphorus from sediments (Hamilton et al., 2002). Finally, lakes impacted by other human disturbances such as acidification (e.g., from road salt application) may respond differently to climate induced changes than lakes with a different history (Blenckner, 2005).

A study of 235 globally distributed lakes from 1985-2009 found lake surface water temperature is warming by approximately 0.34 °C per decade on average (O’Reilly et al., 2015). Summer surface water temperatures in large north temperature lakes (e.g., the Laurentian Great Lakes) are increasing at a faster rate compared to the global average (Schneider & Hook, 2010; O’Reilly et al., 2015). Winter surface water temperatures are also increasing and leading to more frequent ice-free days from February-April in small, north temperate lakes (Mishra et al., 2011). Prior studies on historical changes in lakes also indicate lakes are becoming thermally stratified earlier in the year, and that stratified conditions are lasting longer.

For example, Siver et al. (2018) assessed changes in thermal stability in a Connecticut lake between 1985-

2015 and found that the strength of summer thermal stability is increasing. Elsewhere, stratified conditions are emerging 28 ± 6 days earlier in Lake Blelham, Tarn, England (Foley et al., 2011), 16 days earlier in Lake Washington, U.S. (Winder & Schindler, 2004), and 3 weeks earlier in Lake Heiligensee,

Germany (Adrian et al., 1995). Prior research based on 26 lakes around the world from 1970-2010

34 suggests that lake morphometry (average depth, surface area, and volume) and average lake temperature affect lake thermal stability, and deeper lakes with higher average water temperatures are more likely to experience large changes in thermal stability over time (Kraemer et al., 2015).

Warm, calm, lake conditions during summer months are ideal for cyanobacteria growth (Elliott et al.,

2006; Jöhnk et al., 2008; Paerl & Huisman, 2008), and cyanotoxin production in freshwater lakes (Huber et al., 2012; Taranu et al., 2012; Bosse et al., 2019). Prior studies found warm water temperatures coupled with persistent thermal stability during the summer months correlated with cyanobacteria bloom formation and maximum biomass in two moderately sized temperate lakes [i.e., Brandy Lake, Canada

(Persaud et al, 2015); and Müggelsee Lake, Germany (Wagner & Adrian, 2009)]. Recent research also suggests cyanotoxins in lakes are increasing with rising regional temperatures due to associated increases in the dominance of toxin-producing cyanobacteria (Pilon et al., 2019). Studies suggest surface water temperatures above 25 °C may trigger cyanobacteria blooms if nutrient levels are also high (Yasarer &

Sturm, 2016). However, optimal growth rates for toxin producing Microcystis aeruginosa are between

28–29 °C (Thomas & Litchman, 2016), and production of microcystin-LR is highest at approximately 26

°C (Crettaz et al., 2017). Another study of multiple cyanobacteria species found slightly higher temperatures (30.6 ± 2.3 °C) were associated with optimal growth rates on average (Nalley et al., 2018), which for north temperate regions may only occur during the summer (Robarts & Zohary, 1987).

Cyanotoxins may be harmful to humans and animals if ingested, inhaled, or contacted directly by skin

(Zanchett & Oliveira-Filho, 2013; Trevino-Garrison et al., 2015). Consequently, scientists and health organizations around the world are concerned with protecting the health of the public, animals (including pets), and lake ecosystems from cyanotoxins. But protecting people and animals from cyanotoxins is not straightforward. For example, not all cyanobacteria species produce toxins, and different species produce toxins with variable health affects (Hudon et al., 2016). Research also suggests that cyanotoxins can be present (or absent) among low to high density algal blooms (Subbiah et al., 2019). This means that cyanobacteria may produce toxins exceeding the threshold levels for drinking water at low, moderate or

35 high cell counts (WHO, 2003; O’Keeffe, 2019). A further complication is that algal species and toxin data are often not available and that current toxicological data is insufficient to set specific guidelines for all known cyanotoxins (Ibelings et al., 2014; Farrer et al., 2015), despite at least 46 strains of cyanobacteria with known toxic effects in vertebrates (Sivonen & Jones, 1999).

At this time, typically the World Health Organization (WHO, 2003) guidelines that relate cyanobacteria biomass (cells mL-1) to toxin risk are used to inform drinking water guidelines and regulations. According to their guidelines, if Microcystis-dominated algal blooms are present, then a density of 100,000 cells mL-l and 20,000 cells mL-l of cyanobacteria biomass is equivalent to approximately 20 ppb and 2-4 ppb of Microcystin-LR, respectively (Pilotto et al., 1997; Sivonen & Jones,

1999). In addition, the U.S. Environmental Protection Agency suggests a toxin guideline of 14 ppm to prevent adverse human health effects from human exposure to microcystin-LR (EPA 1989, 1997, 2003).

Based on these prior studies, the Massachusetts Department of Public Health (MDPH) estimated a cell count of 70,000 cells mL-1 would correspond to a toxin level of approximately 14 ppb of toxin in

Microcystis-dominated algal blooms (MDPH, 2014). While these guidelines were developed based on exposure to microcystin-LR, they can be used to provide a conservative risk level estimate for any cyanotoxin (Graham et al., 2009; Ibelings et al., 2014; MDPH, 2014).

Herein, we address the following research questions: (1) What is the relationship between air temperature, water temperature, and thermal stability in Connecticut reservoirs? (2) How may climate warming affect water temperature and thermal stability by midcentury? And, (3) How might future changes in lake water temperature and thermal stability affect the occurrence of cyanobacteria blooms?

To answer these questions, we use data from six small, temperate reservoirs in Connecticut to evaluate observed relationships between air temperature, water temperature, and thermal stability over a 16-year period from 2003-2018 and water temperature, thermal stability, and cyanobacteria blooms over a 8-year period from 2011-2018. We use downscaled, GCM-projected climate forcing to assess potential future changes in water temperature and thermal stability. Finally, we examine how future increase of water

36 temperature and thermal stability may affect the occurrence of high-risk cyanobacteria blooms in the study reservoirs. We focus on Connecticut reservoirs as a case study because: (1) Connecticut shows significant increasing trends in annual air temperature (0.54 °C per decade) since 1895 (Seth et al., 2019;

Lynch et al., 2016); (2) Connecticut is situated in the Northeast where air temperatures are increasing faster than any other region in the contiguous U.S. (Karmalkar & Bradley, 2017); and, (3) studies show the Northeast U.S. will likely experience the largest increases in harmful cyanobacteria bloom occurrence due to both increasing water temperature and nutrient loading (Chapra et al., 2017).

4.2 Methods

Case study reservoirs

The six Southeastern Connecticut reservoirs included in this study – Lake Whitney, Lake Dawson,

Lake Glen, Lake Watrous, Lake Gaillard, and Lake Saltonstall – are small to moderately sized with varying watershed land cover. All are man-made reservoirs except for Lake Saltonstall. 4 out of 6 reservoirs are in highly forested areas (Glen, Watrous, Dawson, Gaillard), one is in a more highly urbanized area (Whitney), and one is in a mixed suburban and forested area (Saltonstall). Figure 3 shows the location of the study lakes and surrounding land cover while Table 1 summarizes watershed characteristics, morphometric and physical lake characteristics.

Fig. 3: Location of study reservoirs and watersheds with color-coded land cover. Watersheds are outlined in red and open water is highlighted light blue. Sampling locations in each reservoir are marked with a black circle. Publicly available United States Geological Survey Gap national land cover data derived from 2011 imagery (USGS, 2011) and watershed boundary shapefiles were used to estimate watershed land cover for each reservoir. The ‘Developed and Other Human Use’ land cover (highlighted yellow) also includes agricultural land, developed vegetation, and recently disturbed or modified land. The ‘Forest and Natural’ land cover (highlighted green), also includes woodland, introduced and semi natural vegetation, and nonvascular and sparse vascular rock vegetation. For more information on the land cover definitions see, USGS (2011). Latitudes and longitudes were recorded in decimal degrees in North American Datum of 1983. This map was created using ArcMap version 10.4.1 (ESRI Inc., Redlands, CA).

Table 1: Relevant reservoir characteristics and percent watershed land cover types are presented. Total capacity is the total reservoir volume at the spillway elevation. Developed watershed land cover includes agricultural land, developed vegetation, and recently disturbed or modified land. Natural land cover includes forest, woodland, introduced and semi natural vegetation, and nonvascular and sparse vascular rock vegetation. For more information on the watershed land cover definitions see, USGS (2011). Lake Whitney Dawson Glen Watrous Gaillard Saltonstall Average 4.7 4.7 5.6 5.8 13.1 13.2 Depth (m) Max Depth 9 11 13 13 29 36 (m) Max Sampling 8 (site A), 12 9 11 9 12 22 Depth (m) (site B) Total Capacity 3,353,873 1,332,464 594,309 2,460,517 59,203,812 20,819,755 (m3) Surface Area 720,341 284,494 105,218 424,920 4,512,249 1,578,275 (m2) Watershed 95,121,459 2,091,258 4,333,622 8,832,604 20,603,862 6,682,491 Area (m2) Residence 0.06 ± 0.01 1.06 ± 0.30 0.23 ± 0.06 0.46 ± 0.13 5.02 ± 0.97 6.03 ± 1.69 Time1 (yrs.) Watershed 49% developed, 13% developed, 14% developed, 14% developed, 6% developed, 21% developed, Land Cover 50% natural, 72% natural, 84% natural, 81% natural, 72% natural, 54% natural, (%) 1% water 15% water 2% water 5% water 22% water 25% water 1Residence times were computed by dividing the total capacity of the reservoir by the annual inflow, assuming reservoirs are filled to their max capacity at the spillway elevation. We estimate annual inflow assuming 50% of the average annual precipitation over the watershed each year from 1998 to 2016 is runoff/streamflow into the reservoir. Annual precipitation is calculated from observed daily and monthly precipitation data from gauges throughout the region supplied by the South Central Connecticut Regional Water Authority. Finally, we assume all stream diversions are open.

Data and statistical analyses

The South Central Connecticut Regional Water Authority provided long-term, in situ depth profile data including water temperature and phycocyanin fluorescence (an indicator of cyanobacteria biomass) for the six study reservoirs. Profile measurements were taken monthly from April to October each year using a submersible multiparameter water quality sonde. All measurements were taken at one sample location in each reservoir except for Lake Saltonstall, which has two sampling sites (Fig. 3).

Measurements begin at the water surface (0m) and end 1m above the reservoir bottom. Water temperature measurements were taken over a 16-year period from 2003-2018 in all but Lake Saltonstall B where measurements were taken over a slightly shorter, 14-year period from 2005-2018. Cyanobacteria biomass data is available for an 8-year period from 2011-2018, with 2011 corresponding to the year the submersible probe for measuring phycocyanin fluorescence was added. The device makes reliable estimates for cyanobacteria biomass (R2 = 0.73-0.83) in vivo in units of cells mL-1 (Izydorczyk et al.,

2005; Brient et al., 2008; Janex-Habibi et al., 2011; Zamyadi et al., 2012), compared to the International

Organization for Standardization method (R2 = 0.97, p < 0.05, between methods) (Gregor and Maršálek,

2004). Cyanobacteria biomass estimates are generally insensitive to interference from chlorophyll-a, turbidity, and dissolved organics, but humic substances (solid organic compounds) or very dense algal blooms can cause overestimation (Gregor & Maršálek, 2004). A few high bottom water cyanobacteria measurements greater than 4 standard deviations away from the mean were discovered during data cleaning. These measurements, likely caused by sediment resuspension when the sonde was bounced off the lake bottom before measurement (Gregor & Maršálek, 2004), were removed before analysis.

Here, we use a combination of laboratory data indicating species information, and cyanobacteria biomass cell counts to consider risks posed by any common, toxic, blooming cyanobacteria species.

Specifically, total cyanobacteria biomass consisting of less than 20,000 cells mL-1 is considered low-risk, between 20,000-70,000 cells mL-1 is considered moderate-risk, and over 70,000 cells mL-1 is considered high-risk. We compute the depth integrated total cyanobacteria biomass, integrated every 1m beginning at the surface and ending 1m above the reservoir bottom, to assess general risk posed by high amounts of cyanobacteria biomass in reservoirs. We do this to provide a more conservative risk level estimate because cyanobacteria blooms may occur at any water depth. For this analysis, we define surface water temperature as the integrated water temperature from the lake surface down 2m (i.e., average temperature of depths ~0, ~1.0, and ~2.0m). We define bottom water temperature as the bottom 2m of the depth profile. Average water temperature is the average temperature integrated over the whole depth profile.

A well-established measure of lake thermal stability (Birge, 1910; Birge & Juday, 1914; Kortmann et al., 1982), Relative Thermal Resistance to Mixing (RTRM), was used to quantify thermal stratification in the study lakes. RTRM is a dimensionless value which quantifies the strength of density differences between layers of the water column based on the vertical temperature gradient, which is a function of thermal stability in the water column (Eq. 4).

Density of the Upper Layer −Density of the Lower Layer RTRM = (4) Density at 5°C −Density at 4°C

Total RTRM indicates the total water column stability (additive), while max RTRM identifies the location and intensity of the thermocline (steepest density gradient).

Statistical analyses are used to identify environmental conditions (surface water temperature, average water temperature, bottom water temperature, total RTRM, and max RTRM values) that correlate with total cyanobacteria biomass over the observed period in each of the six study lakes. Laboratory sample data from 2014-2018 compliment the cyanobacteria biomass data and indicate which species of toxin- producing cyanobacteria were present and dominating in the study reservoirs during past blooms. Species information is included here even though we lack toxin data, because potentially any amount of cyanobacteria can be a risk for toxins if toxin producing cyanobacteria are present, and any amount of toxin can be a health risk (Zanchett & Oliveira-Filho, 2013).

Sixteen years (2003-2018) of downscaled, daily max and min air temperature data corresponding to the grid cell where each reservoir is located is taken from the gridMET dataset (publicly available at http://www.climatologylab.org/gridmet.html; Abatzoglou, 2013). The gridMET data has a spatial resolution of 1/24th degree (approximately 4km). Model-based historical and future air temperature corresponding to each lake are taken from MACAv2-METDATA (Abatzoglou & Brown, 2012).

MACAv2-METDATA is a statistically downscaled climate database produced by applying the

Multivariate Adaptive Constructed Analogs (MACA) approach (Abatzoglou & Brown, 2012) to 20

Coupled Model Inter-Comparison Project Phase 5 (CMIP5) GCMs, using gridMET as the training data.

RCP 8.5 was chosen for this analysis because greenhouse gas emissions over the past decade have followed the high radiative forcing scenario (Sanford et al., 2014). 3 out of the 20 CMIP5 GCMs were chosen for further analysis including: (1) HadGEM2-CC365 from the UK Met Office; (2) CCSM4 from

NCAR (U.S.); and, (3) GFDL-ESM2M from NOAA GFDL (U.S.). These 3 GCMs were chosen because they closely track available observations and capture a range of temperature sensitivity for Connecticut

(Seth et al., 2019). Sensitivity is based on the expected time it will take for surface air temperature to reach a 2 °C increase relative to 2000 under RCP 8.5 based on the model projections (Miao et al., 2014).

Models where it takes longer to reach the threshold 2 °C increase are considered less temperature sensitive. CCSM4 and GFDL-ESM2M have low-medium temperature sensitivity for Connecticut, while

HadGEM2-CC365 has medium-high temperature sensitivity for Connecticut.

Model development and validation

Empirical linear and 2nd order polynomial regression models based on observed relationships between air temperature averages and surface water temperature, total RTRM, and max RTRM over the study period were developed to predict daily water temperature, total RTRM, and max RTRM using air temperature alone. All monthly water temperature measurements available over this time period are included in the model development, and although there are a few missing months, this did not affect our analysis. This is consistent with Piccolroaz (2016), who used the empirical relationship between daily air temperature and inconsistent monthly surface water temperature observations to predict daily surface water temperature. While winter data from November-March is not available, we tested the effect of including representative winter month data for Lake Gaillard and found no significant change in the model results with winter data included. However, we did find lakes respond differently to changes in climate and must be assessed individually (Blenckner, 2005; O’Reilly et al., 2015). Thus, lake-specific empirical models were developed to predict water temperature and RTRM at each reservoir at the daily timescale. Models were developed through data driven calibration of monthly water temperature from the observational dataset with daily observed air temperature data from the gridMET dataset. The 1-365 day air temperature maximum (tmax), minimum (tmin), and average (tavg) from the days before, were evaluated to determine the best predictor of surface water temperature on the preceding day.

Ultimately, we derived a linear regression equation for each lake where: tw is water temperature, ta,i is average air temperature, i quantifies the number of days averaged for the prediction, and A and B are constants (Eq. 5).

tw = 퐵 × ta,i + 퐴 (5)

Unique, best fit models for each lake were also developed to predict daily total and max RTRM from surface water temperature. These empirical models are based on monthly observed relationships between surface water temperature, total RTRM, and max RTRM over the study period at each reservoir. We

nd derived a 2 order polynomial equation for each lake where: tw = surface water temperature; r = total or max RTRM, depending on which is being predicted; and A, B, and C are constants (Eq. 6).

2 r = C × tw + B × tw + A (6)

The best predictive models for surface water temperature, average water temperature, bottom water temperature, total RTRM, and max RTRM for each lake were selected based on the root mean square error (RMSE) and mean bias (MB) leave-one-out cross validation estimates and R2. Model accuracy was also evaluated by comparing measurements to predicted values (residual = measured - predicted).

4.3 Results and Discussion

Relationships between air temperature, water temperature, and thermal stability in small, north temperate reservoirs

Using sixteen years of observed daily average air temperature and monthly water temperature for calibration, unique, optimized, empirical models were developed for each reservoir to predict daily surface, average, and bottom water temperature, and daily total and max RTRM (Table 2).

Table 2: Surface water temperature (T), average water T, bottom water T, total Relative Thermal Resistance to Mixing (RTRM), and max RTRM models derived for each lake are presented along with their respective R2, leave-one-out cross validation root mean square error (RMSE), and mean bias (MB) estimates. Here, tw = water temperature, rt = total RTRM, rm = max RTRM, and tavg,i = average air T, with i quantifying the number of days averaged for the prediction. Lake Model Whitney Dawson Glen Watrous Gaillard Saltonstall A Saltonstall B

Air temperature to tw = 0.94 × tw = 1.06 × tw = 0.96 × tw = 1.07 × tw = 1.10 × tw = 1.08 × tw = 1.09 × Surface Water T tavg,9 + 2.99 tavg,8 + 1.72 tavg,15 + 3.02 tavg,15 + 1.97 tavg,19 + 1.20 tavg,13 + 1.16 tavg,13 + 0.89 R2 0.95 0.95 0.90 0.95 0.96 0.95 0.95 RMSE 1.2 1.3 1.6 1.3 1.3 1.4 1.4 MB -0.002 -0.004 -0.0043 -0.002 -0.001 -0.003 -0.004 Air T to Average tw = 0.54 × tw = 0.73 × tw = 0.84 × tw = 0.71 × tw = 0.52 × tw = 0.91 × tw = 0.60 × Water T tavg,26 + 6.11 tavg,25 + 4.37 tavg,43 + 2.15 tavg,47 + 4.61 tavg,48 + 4.99 tavg,25 + 2.13 tavg,36 + 4.66 R2 0.90 0.89 0.92 0.87 0.93 0.95 0.94 RMSE 1.0 1.5 1.5 1.7 0.9 1.1 0.9 MB -0.002 -0.002 0.001 3.74E-04 -0.003 -0.003 -0.002 Air T to Bottom tw = -0.01 × y = 0.002 × tw = 0.02 × tw = 0.01 × tw = -0.02 × tw = 0.02 × tw = 0.01 × 2 2 2 2 2 2 2 Water T tavg,81 + 0.36 × tavg,69 + 0.33 × tavg,66 + 0.19 × tavg,85 + 0.21 × tavg,67 + 0.58 × tavg,67 - 0.03 × tavg,233 -0.02 ×

tavg,81 + 7.45 tavg,69 + 7.24 tavg,66 + 4.06 tavg,85 + 6.47 tavg,67 + 4.30 tavg,67 + 8.40 tavg,233 + 7.59 R2 0.44 0.48 0.84 0.66 0.53 0.79 0.42 RMSE 1.4 2.8 2.3 2.6 1.0 2.1 1.2 MB -0.006 -0.004 0.002 -0.005 -0.001 -0.005 2.69E-04 2 2 2 2 2 2 2 Surface Water T to rt = 0.89 × tw rt = 0.98 × tw rt = 0.43 × tw rt = 0.95 × tw rt = 0.78 × tw rt = 1.25 × tw - rt = 0.75 × tw Total RTRM - 9.56 × tw + - 16.39 × tw + - 2.66 × tw + - 16.59 × tw + - 5.64 × tw + 29.28 × tw + - 4.10 × tw - 17.49 73.88 15.82 88.65 1.69 183.56 17.03 R2 0.97 0.82 0.43 0.74 0.99 0.70 0.99 RMSE 19.4 59.5 79.2 68 12.2 71.5 14.9 MB 0.03 -0.11 -0.009 -0.05 0.02 0.003 0.03 2 2 2 2 2 2 2 Surface Water T to rm = 0.15 × tw rm = 0.22 × tw rm = 0.21 × tw rm = 0.45 × tw rm = 0.02 × tw rm = 0.69 × tw rm = 0.18 × tw Max RTRM + 0.66 × tw - - 1.60 × tw + - 2.98 × tw + - 9.25 × tw + + 5.36 × tw - - 16.33 × tw + + 4.34 × tw - 16.09 1.01 19.09 54.19 41.91 101.46 59.50 R2 0.84 0.61 0.41 0.60 0.73 0.62 0.72 RMSE 14.7 32.2 30.4 38.7 25.3 45.8 43.2 MB 0.01 -0.02 -0.01 -0.05 0.04 -0.06 -0.05 Depending on the lake, the 8-19 day, 25-48 day, and 66-233 day average air temperature from the days or

weeks before the prediction were used to predict surface, average, and bottom water temperature

respectively on the subsequent day. Surface water temperature in Lake Whitney and Lake Dawson, the

two shallowest lakes with the smallest average depth, are most affected by average air temperatures from

the week before (8-9 days), while the other four lakes are most affected by average air temperatures from

a little over two weeks before (13-19 days). This variation supports previous findings that climate

warming may have different impacts on lake water temperature depending on lake morphology, even for

those located in the same geographical area (Fee et al., 1996; Blenckner, 2005).

Overall, the average air temperature to surface and average water temperature empirical models

provide good performance for temperature predictions across reservoirs (RMSE between 0.9-1.7 °C)

(Table 2). There is a strong correlation between predicted and measured surface water temperature values

and the residuals are evenly spaced, indicating high model accuracy (Appendix 6.2, Supporting Material,

Figs. 1 and 2). In Lake Whitney, Lake Dawson, and Lake Saltonstall A, the 25-26 day average air

44 temperature was most highly correlated with average water temperature on the preceding day, while the

36-48 day average air temperature was more highly correlated with average water temperature in Lake

Glen, Lake Watrous, Lake Gaillard, and Lake Saltonstall B. Again, it is not surprising that the two smallest reservoirs are more affected by more recent air temperatures (25-26 days) compared to the larger lakes (36-48 days); however, it is somewhat surprising that the two sampling locations in Lake Saltonstall respond so differently. Site A corresponds with a shallow location closer to the shoreline in Lake

Saltonstall, while site B corresponds with a deeper location closer to the center of the lake.

The range of bottom water temperatures experienced by the study lakes varies greatly, and lakes which experienced a higher range of bottom water temperatures have higher model accuracy for bottom water temperature predictions (Appendix 6.2, Supporting Material, Fig. 3). There is no clear relationship between the range of bottom water temperatures experienced by each lake and average lake depth. Lake

Whitney, Lake Saltonstall B, and Lake Gaillard experience only low water temperatures over the study period, ranging from approximately 2-15 °C. Lake Glen, Lake Saltonstall A, Lake Dawson, and Lake

Watrous experienced higher temperatures during the study period, ranging from approximately 2-25 °C.

Lake Dawson and Lake Watrous also have lower R2 values relative to Lake Glen and Saltonstall A, likely due to slight underprediction of bottom water temperatures at higher temperatures (above approximately

15 °C) in these lakes. In Lake Whitney, Lake Dawson, Lake Glen, Lake Gaillard, and Lake Saltonstall A, the 66-81 day average air temperature was most highly correlated with bottom water temperature on the preceding day, while the 233 day average air temperature was most highly correlated with bottom water temperature in Lake Saltonstall B. Differences between the two sampling locations in Lake Saltonstall may be explained again by their proximity to the lake shoreline and max depth at the sampling location; site A is located closer to the lake shoreline and in a shallower location compared to Lake Saltonstall site

B, therefore it experienced higher max bottom water temperatures. The max depth at site B is approximately 36m (Table 1), which is much deeper than all of the other reservoirs likely because Lake

Saltonstall is the only natural lake. The sample max depth at site B is only 12m, however it is clear that

45 the bottom water temperatures in Lake Saltonstall (below at least 12m) remain cool even during the summer, and are most affected by average air temperatures over half a year (233 days).

The total and max RTRM models provide good performance for some reservoirs, but not all (R2 between 0.41-0.99). For Lake Whitney, Lake Gaillard, and Lake Saltonstall B, there is a strong correlation between predicted and measured total RTRM values and the residuals are evenly spaced, indicating high model accuracy (Appendix 6.2, Supporting Material, Fig. 4). Total RTRM is directly correlated with surface water temperatures on any given day in these 3 lakes. These 3 lakes are also the lakes which experienced only cool bottom water temperatures (2-15 °C) over the study period, indicating there is also a clear relationship between higher lake thermal stability and cooler bottom water temperatures in Lake Whitney, Lake Gaillard and Lake Saltonstall B. The total RTRM model for Lake

Saltonstall A, Lake Watrous, and Lake Dawson have slightly lower R2 values due to the higher RMSE of predictions for these two lakes (Table 2). Lake Glen’s surface water temperature to total RTRM model is the least accurate, mostly because it slightly underpredicts total RTRM at higher RTRM values (above

200). Higher variation and inaccuracy of total RTRM predictions corresponds to higher variation and inaccuracy of bottom water temperature predictions and vice versa in all of the study lakes.

Lake Whitney and Lake Gaillard, followed closely by Lake Saltonstall B, have the strongest relationship between lake surface water temperature on a given day and max RTRM. However, the range of max RTRM values experienced by the study lakes varies and lakes which experienced a higher range of max RTRM values have lower model accuracy for max RTRM predictions (Appendix 6.2, Supporting

Material, Fig. 5). This is likely due to underprediction of max RTRM at high max RTRM measurements

(above 100).

Future projections of water temperature and thermal stability

Water temperature and RTRM are projected to be significantly warmer across the year and at the

monthly timescale at midcentury, compared to the historical period (Table 3, Fig. 4).

Table 3: 30-year historical (1971-2000) and future (2041-2070) mean monthly water temperature (T) and total and max Relative Thermal Resistance to Mixing (RTRM) projections averaged over all six reservoirs are presented. A 2-sided paired t-test is used to evaluate the significance of long-term changes in monthly and annual means over the 70-year period. Projected changes in water T and total RTRM) are based on the General Circulation Model averages under Representative Concentration Pathway 8.5. Significant differences in the means for historical and future are indicated with asterisks on the total change. Standard deviations across lakes for each month are reported. Surface Water T (°C) Average Water T (°C) Bottom Water T (°C) Total RTRM Max RTRM Calendar Time Mean Change Mean Change Mean Change Mean Change Mean Change Month Period Jan. Historical 0.3 ± 1.2 4.0 ± 1.0 7.3 ± 0.4 63.4 ± 16.1 10.2 ± 4.0 ***3.6 ***2.4 ***1.0 ***-30.5 ***-5.9 Future 3.9 ± 1.5 6.4 ± 1.1 8.3 ± 0.5 32.9 ± 10.3 4.3 ± 2.0 Feb. Historical 1.1 ± 1.3 3.2 ± 0.9 6.5 ± 0.2 53.4 ± 15.3 7.9 ± 3.6 ***2.5 ***2.0 ***0.9 ***-19.5 ***-3.5 Future 3.6 ± 1.7 5.3 ± 1.1 7.5 ± 0.4 33.9 ± 11 4.4 ± 2.1 Mar. Historical 5.2 ± 1.2 5.3 ± 1.1 6.6 ± 0.2 25.4 ± 4.5 3.2 ± 1.2 ***2.5 ***1.7 ***0.8 **-3.3 **1.6 Future 7.7 ± 1.3 7.0 ± 1.2 7.4 ± 0.4 22.1 ± 3.3 4.8 ± 2.0 Apr. Historical 10.4 ± 0.6 8.5 ± 1.2 7.4 ± 0.4 30.9 ± 4.9 10.6 ± 2.2 ***11.7 ***3.0 ***1.9 ***0.8 ***26.5 Future 13.4 ± 1.2 10.4 ± 1.4 8.2 ± 0.5 57.4 ± 14.9 22.3 ± 6.0 May Historical 16.4 ± 0.8 12.4 ± 1.2 8.8 ± 0.5 96.3 ± 13.8 38.1 ± 5.3 ***2.8 ***1.9 ***0.9 ***52.2 ***19.8 Future 19.2 ± 0.9 14.4 ± 1.3 9.6 ± 0.6 148.5 ± 20.9 57.9 ± 7.7 Jun. Historical 21.7 ± 0.7 16.1 ±1.1 10.5 ± 0.5 205.6 ± 19.4 79.3 ± 7.1 ***2.7 ***1.8 ***1.0 ***78.3 ***28.3 Future 24.4 ± 0.9 17.9 ± 1.2 11.4 ± 0.6 283.9 ± 28 107.6 ± 10.1 Jul. Historical 25.3 ± 0.6 19.0 ± 0.8 12.1 ± 0.5 309.8 ± 19 117.1 ± 6.8 ***3.3 ***2.1 ***1.1 ***115.3 ***40.9 Future 28.6 ± 1.0 21.1 ± 1.0 13.2 ± 0.6 425.1 ± 37.4 158 ± 13.2 Aug. Historical 25.7 ± 0.6 20.1 ± 0.4 13.3 ± 0.3 319.3 ± 19.5 120.6 ± 7.0 ***3.6 ***2.4 ***1.3 ***130.3 ***46 Future 29.3 ± 0.9 22.5 ± 0.7 14.6 ± 0.5 449.6 ± 36 166.6 ± 12.6 Sep. Historical 21.8 ± 0.7 18.5 ± 0.9 13.3 ± 0.3 209.2 ± 18.3 80.8 ± 6.7 ***3.5 ***2.4 ***1.4 ***102.4 ***36.9 Future 25.3 ± 1.2 20.9 ± 1.1 14.7 ± 0.4 311.6 ± 34.1 117.7 ± 12.2 Oct. Historical 15.6 ± 0.7 14.8 ± 1.3 12.2 ± 0.5 85 ± 10.6 34 ± 4.1 ***3.2 ***2.2 ***1.3 ***57.8 ***22.1 Future 18.8 ± 1.1 17.1 ± 1.5 13.5 ± 0.6 142.8 ± 21.7 56.1 ± 8.1 Nov. Historical 9.6 ± 0.9 10.7 ± 1.2 10.4 ± 0.5 26.7 ± 4.5 8.6 ± 2.4 ***3.1 ***2.1 ***1.2 ***24.5 ***11.3 Future 12.7 ± 1.3 12.8 ± 1.4 11.6 ± 0.7 51.2 ± 13.2 19.9 ± 5.6 Dec. Historical 4.2 ± 1.2 7.0 ± 1.2 8.7 ± 0.5 30.3 ± 7.9 3.7 ± 1.5 ***3.4 ***2.2 ***1.1 ***-6.8 **1.6 Future 7.6 ± 1.3 9.2 ± 1.3 9.8 ± 0.6 23.5 ± 3.9 5.3 ± 2.0 Annual Historical 13.2 ± 9.1 11.7 ± 5.9 9.8 ± 2.5 121.9 ± 119.6 43.1 ± 53.4 ***3.1 ***2.1 ***1.1 ***44.4 ***17.8 Future 16.3 ± 9.2 13.8 ± 6.0 10.9 ± 2.7 166.3 ± 167.6 60.9 ± 67.7 *p≤0.05, **p≤0.01, ***p≤0.001. The standard errors reported here is are one standard deviation uncertainties calculated for the 30-year monthly means.

Fig. 4: Histogram of the annual distribution of temperature and Relative Thermal Resistance to Mixing (RTRM) (i.e., the frequency of days with certain surface water temperature and RTRM ranges) during the reference period (1971-2000) and midcentury period (2041-2070). The data shown here is a multi-model (i.e., HadGEM2-CC365, CCSM4, GFDL-ESM2M) ensemble of water temperature and thermal stability under Representative Concentration Pathway 8.5.

Mean annual total RTRM and max RTRM are expected to increase from April-November across lakes by midcentury, with the most significant increases in July-September (Table 3) while thermal stability is projected to occur two-four weeks earlier. These findings are consistent with Siver et al. (2018), who found that the strength of summer thermal stability is increasing in Candlewood Lake, Connecticut and with Adrian et al. (1995), Winder & Schindler (2004), and Foley et al. (2011) who found earlier onset of thermal stability in their study lakes. In addition to occurring earlier, results show thermal stability is projected to last two-four weeks longer by midcentury. Fig. 4 shows that the distribution of temperature and RTRM is moving towards the right, meaning that cooler temperature and lower RTRM values are decreasing over time while annual water temperature and RTRM is increasing.

On average across reservoirs, mean annual surface, average, and bottom water temperatures are expected to increase by approximately 0.44 °C, 0.30 °C, and 0.16 °C per decade from now until 2070

(Table 3). We expect lake surface water temperatures in small Connecticut reservoirs to increase at a faster rate than the global lake average (i.e., 0.34 °C per decade). This is consistent with other north temperate lakes including all of the Laurentian Great Lakes which are all also warming faster than the global lake average (Schneider & Hook, 2010; O’Reilly et al., 2015). Here, surface water temperature is increasing at a slightly slower rate than near-surface air temperatures in Connecticut which are increasing

0.54 °C per decade (Seth et al., 2019; Lynch et al., 2016). Significant increases in surface water temperature occur year-round, with an average annual increase of 3.1 °C expected on average across lakes by midcentury. This rate of increase in surface water temperature by midcentury is slightly higher than increases expected for larger north temperate U.S. lakes. For example, Piccolroaz et al. (2018) used climate forcing from GCMs to simulate lake surface temperature warming during the summer months up to 2.9 °C under RCP 8.5 at Lake Tahoe by the end of the 21st century.

While most prior studies focus only on climate warming of surface water temperatures, because our data included vertical lake profile measurements, we were able to examine average and bottom water temperatures. In addition to surface water warming, we found that climate warming is increasing average and bottom water temperatures too, though at a slower rate. Mishra et al. (2011) found increasing air temperatures are increasing the number of ice-free days in north temperate lakes by 0.2-2.0 days per decade, depending on the lake. Our results indicate significant increases in surface, average, and bottom water temperature year-round, with a concomitant decrease in total and max RTRM in January and

February likely indicating lakes will be more well mixed in the winter. Ice free days will likely decrease in the study reservoirs as well with both rising winter water temperatures and better lake mixing, especially in January and February.

Correlating future changes in lake water temperature and thermal stability with the occurrence of cyanobacteria blooms

Toxin producing cyanobacteria species Aphanizomenon and Anabaena, common in temperate regions, did appear in the laboratory samples as a dominate species in all of the study reservoirs at some

49 point from 2014-2017, while Microcystis only dominated in one laboratory sample taken during the same time period (i.e., low risk < 20,000 cell mL-1). Microcystis, Aphanizomenon, and Anabaena can produce microcystin-LR (Chen et al., 2007; Lyon-Colbert et al., 2018). Aphanizomenon is also known to produce significant levels of cylindrospermopsin and saxitoxin (Lyon-Colbert et al., 2018), algal toxins that can also be harmful to human health (Schmale et al., 2019). Subbiah et al. (2019) detected higher amounts of these specific algal toxins in a large southwest U.S. reservoir (2015-2017) during times when water temperatures were highest. Despite the presence of these algal species no cyanotoxins were measured in detectable amounts in the reservoirs (Personal Communication with the South Central Connecticut

Regional Water Authority).

Generally, higher lake surface water temperatures, higher total RTRM values, and lower bottom water temperatures were moderately correlated with higher amounts of cyanobacteria biomass across lakes (Appendix 6.2, Supporting Material, Table 1). A two-sample t-test assuming unequal variances indicated there is no significant difference in bottom water temperature across the low, moderate and high-risk bins (Appendix 6.2, Supporting Material, Table 2). However, there is a significant difference in mean surface water temperature, mean average water temperature, and mean total RTRM between high and low risk bins, and high and moderate risk bins. There is only a significant difference in max RTRM between the high and low risk bins. High-risk cyanobacteria blooms typically occurred when surface water temperature and total RTRM was highest during the year, which typically co-occurs at different times from July-September.

From 2011-2018 across all lakes, high-risk cyanobacteria blooms, exceeding 70,000 cells mL-1, occurred in 4.54% of samples (Appendix 6.2, Supporting Material, Table 3). 15 out of the 20 high-risk blooms occurred in Lake Whitney, 2 in Lake Saltonstall, and 3 in Lake Watrous. Furthermore, Lake

Watrous, Lake Whitney, and Lake Saltonstall each experienced high or moderate-risk cyanobacteria blooms during 3 ± 1 out of the 7 months sampled per year on average. The other 3 reservoirs, Lake

Gaillard, Lake Dawson, and Lake Glen, experienced moderate risk cyanobacteria blooms for 1 ± 0.2

50 months out of the 7 months sampled per year on average. Generally, high-risk blooms occurred when surface water temperatures were 25-26 °C in Lake Whitney and Lake Saltonstall, and 18 °C in Lake

Watrous. High-risk blooms in Lake Watrous correspond with significantly lower surface water temperatures compared to the other two reservoirs, likely due to artificial aeration that occurs in Lake

Watrous and not the other two lakes. Lake Watrous had an aeration system installed in 2003, which was replaced by a Layer Aeration system in 2016. The original system mixed the whole water column, while the new system uses a detailed heat budget based on total and max RTRM to inform the depth-discrete mixing treatment, which strategically redistributes heat in the water column and manipulates the location of the thermocline. The high-risk blooms in Lake Watrous occurred in 2001 (2) and 2016 (1), before the new Layer Aeration system was installed in 2016. The uniformity in surface (18.1 °C), average (17.8 °C), and bottom (17.2 °C) water temperature in Lake Watrous on average during the 3 blooms indicates the original layer aeration system was used but not able to prevent high-risk blooms from occurring. Average water temperature in all 3 lakes was 18.2 ± 1.4 °C when the high-risk blooms occurred, indicating average water temperature may be a better indicator of high-risk blooms across lakes than lake surface water temperature.

The higher amounts of human development (e.g., agriculture and impermeable surfaces) in Lake

Whitney’s and Lake Saltonstall’s watersheds are associated with higher amounts of algal nutrients (i.e., nitrogen and phosphorus) entering these lakes from runoff (Lacher et al., 2019). This historical land use in the watershed area provided nutrients to these lakes and may explain why Lake Saltonstall and Lake

Whitney are historically more susceptible to high-risk cyanobacteria blooms (Ahlgren, 1976). Beyond catchment characteristics and watershed land cover, lake morphology may play a role here, especially surface area to average depth ratios and irregular shorelines (Blenckner, 2005). Lake Saltonstall, Lake

Whitney, and Lake Watrous have long, jagged lake shorelines relative to the other study lakes, and as a result these reservoirs have more shallow areas relative to other lakes, which would allow these areas to mix more easily and warm more quickly (Mason et al., 2016). We found evidence of this occurring when

51 we compared the two sampling sites in Lake Saltonstall. Site A is located closer to the lake shoreline and in a shallower location compared to Lake Saltonstall site B. As a result, site A warmed more easily and is more well mixed than site A, as indicated by higher max bottom water temperatures and more variable total RTRM responses to changes in air temperature. At site B, the bottom water temperatures in Lake

Saltonstall (below at least 12m) remain cool even during the summer, indicating strong thermal stability at this location. Furthermore, Lake Whitney may be most susceptible to cyanobacteria blooms compared to the other study reservoirs because is the shallowest reservoir but has the largest and most developed watershed area. Lake Whitney is also prone to internal nutrient loading from sediments which will likely be amplified by climate warming (Ahlgren, 1976). Lake Whitney also has the fastest residence time which may contribute to bloom susceptibility.

Figure 5 includes the observed relationships between water temperature, thermal stability, and cyanobacteria biomass in the six study reservoirs from 2011-2018.

Fig. 5: Relationships between water temperature, thermal stability, and cyanobacteria biomass in six Connecticut reservoirs from 2011-2018. The amounts of cyanobacteria biomass associated with low, moderate, and high-risk shown here are adapted from the World Health Organization (2003) and the Massachusetts Department of Public Health (2019) guidelines which are based on toxin risk. Mean surface water temperature (23.7 ± 3.1), mean average water temperature (18.2 ± 1.4), mean bottom water temperature (11.8 ± 3.1), mean total RTRM (282 ± 131), and mean max RTRM (90 ± 63) correlating with high-risk levels of cyanobacteria biomass across lakes were calculated, along with their standard deviations, and are included here. The gray bands do not indicate a strength of relationship but are included to help visualize the associations between each parameter and high-risk levels of cyanobacteria biomass (Appendix 6.2, Supporting Material, Table 3). Specific temperature and RTRM ranges appear to be associated with high-risk levels of cyanobacteria biomass. Specifically, we found surface, average, and bottom water temperatures between 20.6-26.8 °C,

16.8-19.6 °C, and 8.7-14.9 °C, respectively, total RTRM between 151-413, and max RTRM between 27-

153, corresponded with high-risk blooms across lakes (cyanobacteria biomass > 70,000 cells mL-1). Some of these ranges are large due to the high standard deviations on the mean values that are associated with

53 high-risk blooms for each parameter, which has to do with high variation across lakes, especially in total and max RTRM values.

The co-occurrence of specific temperatures and RTRM ranges appears to be what is important for determining the magnitude of a cyanobacteria bloom, not just exceeding a certain water temperature or

RTRM value. Therefore, the co-occurrence of certain water temperature and thermal stability ranges may be more important for predicting high-risk blooms across lakes than each parameter on its own. For example, bottom water temperatures must remain below approximately 11.8 ± 3.1 °C, while at the same time average water temperature exceeds 18.2 ± 1.4 °C, and surface water temperature exceeds 23.7 ± 3.1

°C. In thinking about future bloom potential, if there is a narrow or longer window of time when these temperatures are occurring, then blooms may increase or decrease depending on varying vertical temperature changes. In the next section, these ranges are used to assess how future changes in water temperature and RTRM might affect future cyanobacteria blooms.

Changes in specific water temperature and thermal stability ranges that may affect cyanobacteria blooms

Here, we evaluate future changes in the annual frequency of days where water temperatures are within a certain range (i.e., surface, average, and bottom water temperatures between 20.6-26.8 °C, 16.8-

19.6 °C, and 8.7-14.9 °C, respectively), or above a certain threshold (e.g., 15 °C, 20 °C, 25 °C, and 30

°C). In addition, we evaluate future changes in the annual frequency of days where total RTRM is between 151-413, over 413, and over 475; and max RTRM is between 27-153, over 153, and over 200.

These specific ranges were chosen because they are generally associated with higher amounts of cyanobacteria biomass across lakes (Fig. 5). Mean annual frequency of days where water temperature and

RTRM values are within a certain range, or above a certain threshold during the historical period compared to the future period are reported in Supporting Material, Table 4. At the lake specific level, there will be significant changes in the future water temperature, total RTRM, and max RTRM ranges that corresponded with high-risk blooms historically. Generally, future projections indicate surface, average, and bottom water temperatures in these specific ranges are decreasing in most lakes. This is especially

54 true for surface water temperatures between 20.6-26.8 °C. Changes in average and bottom temperature vary more at the lake-specific level. However, in most cases decreases in these ranges go hand and hand with increases in even higher temperature and RTRM thresholds.

Future projections indicate surface water temperatures between 20.6-26.8 °C will decrease by 26 ± 8 days yr-1, from 94 ± 3 days yr-1 across lakes during the reference period to 68 ± 10 days in the future, with little variability across lakes. However, surface waters temperatures above 25 °C and 30 °C are expected to increase by 48 ± 7 and 18 ± 11 days yr-1, respectively, across lakes. Therefore, milder temperatures are decreasing while extreme high surface water temperatures are expected to increase. However, extreme surface water temperatures above 30 °C are beyond the range of max temperatures observed over the study period, and evaporative cooling at these higher temperatures may affect the linear relationship between future air temperatures and future surface water temperatures. Nevertheless, if realized, these extreme surface temperatures will have profound implications for water quality.

Average water temperatures are also increasing. All lakes except Lake Gaillard indicate average water temperatures between 16.8-19.6 °C will decrease by 14 ± 10 days by midcentury. In Lake Gaillard they are expected to increase by 70 days yr-1. However, the five lakes which are expected to experience a decrease in days when average water temperatures are between 16.8-19.6 °C are all also expected to experience an increase of 38 ± 18 days yr-1 where average water temperatures exceed a higher average temperature of 20 °C. None of the lakes experienced average water temperatures above 25 °C historically, and none are expected to experience significant increases in average water temperatures above this threshold in the future except for Lake Saltonstall A, which is expected to experience an additional 58 days yr-1 when average water temperature exceeds 25 °C by midcentury. Lake Saltonstall A is located close of the shore, therefore average water temperatures in near-shore areas may be warming at a faster rate than areas closer to the center of the lake. This finding is consistent with Mason et al. (2016) who found surface water temperatures in near-shore areas in the Laurentian Great Lakes are warming more quickly than more central areas.

Bottom water temperatures between 8.7-14.9 °C are expected to decrease, remain about the same, or increase depending on the lake, though the lakes that are expected to experience decreases in these temperatures are the same lakes that are likely to experience large increases in bottom water temperatures exceeding 15 °C or higher. Specifically, days when bottom water temperatures are between 8.7-14.9 °C are expected to decrease in Lake Glen and Lake Watrous, remain about the same in Lake Dawson, and increase in Lake Whitney and Lake Saltonstall. In Lake Glen and Lake Watrous, the two lakes where temperatures between 8.7-14.9 °C are expected to decrease, temperatures above 15 °C are expected to increase by 38 and 51 days yr-1 respectively. Lake Whitney, Lake Gaillard, and Lake Saltonstall B did not experience bottom water temperatures above 15 °C during the historical period are at not expected to in the future. Also, Lake Gaillard did not experience any days where bottom water temperatures were between 8.7-14.9 °C in the past and is not expected to in the future. However the bottom water temperature model for Lake Gaillard, Lake Whitney, and Lake Saltonstall B slightly underpredicts bottom water temperatures at higher temperatures.

At the lake specific level, days when total RTRM measurements are between 151-413 are expected to increase some lakes and decrease in others, though in lakes where measurements in this range are decreasing the number of days where total RTRM measurements will exceed 413 are expected to increase. Lake Saltonstall B is the only exception here, where both days when total RTRM measurements are between 151-413 and days above 413 are both projected to decrease. Specifically, total RTRM between 151-413 is expected to increase by 8-48 days in Lake Glen, Lake Watrous, Lake, Gaillard, and

Lake Saltonstall A, and decrease by 8-60 days in Lake Whitney, Lake Dawson, and Lake Saltonstall B.

However, in Lake Whitney and Lake Dawson, days where total RTRM exceeds 413 and 475 are expected to increase by 83-108 days and 51-69 days yr-1, respectively, indicating a significant increase in total

RTRM in these lakes. Lake Gaillard and Lake Saltonstall A are also expected to experience an increase in days (i.e., 51 and 18 days yr-1, respectively) where total RTRM exceeds 413. Lake Gaillard is also expected to experience very high total RTRM above 475 for an additional 11 days yr-1 by midcentury.

Max RTRM is also expected to increase significantly in all lakes by midcentury. Though days with max RTRM measurements between 27-15, which corresponded with high-risk amount of cyanobacteria biomass historically, are decreasing in most lakes except for Lake Saltonstall A and B. Lake Saltonstall B did not experience any days where max RTRM is above 153 or 200 historically and is not expected to in the future. However in the other five lakes and in Lake Saltonstall A, days where max RTRM is above

153 and 200 are expected to increase by 45 ± 14 yr-1 and 36 ± 18 days yr-1, respectively across lakes.

4.4 Conclusions

In this study, we project water temperature and thermal stability out to midcentury in six southeastern Connecticut lakes to assess long-term changes in lake water quality, including possible changes in cyanobacteria bloom risks. Lake specific, empirical models were derived to predict daily surface water temperature (R2 between 0.90-0.96), average water temperature (R2 between 0.87-0.95), bottom water temperature (R2 between 0.42-0.84), total RTRM (R2 between 0.43-0.99), and max RTRM

(R2 between 0.41-0.84). These models were derived using sixteen years of daily air temperature, and monthly in situ depth profile water temperature measurements. Future changes in water temperature and thermal stability are driven with statistically downscaled historical (1971-2000) and future (2041-2070) air temperature projections from 3 GCMs (i.e., HadGEM2-CC365, CCSM4, GFDL-ESM2M).

When calibrating the models, we found climate warming affects surface water temperatures differently depending on lake morphology, even those located in the same geographical area. For example, surface water temperatures in very shallow lakes (i.e., average depth approximately 5m) are most affected by average air temperatures averaged over 1 week before, while slightly deeper lakes

(average depth approximately 6m or higher) are most affected by average air temperatures averaged over

2-3 weeks before. Average lake temperature and bottom water temperature respond to changes in air temperature over a longer period of time (i.e., 30 days or more) compared to surface water temperatures.

Bottom water temperatures are affected by past average air temperatures over the longest period of time

(i.e., approximately 2-6 months). The proximity of the sample location to the lake shoreline and the max depth at the sampling location impacted average and bottom water temperature responses to past average air temperatures. While there is also no clear relationship between the range of bottom water temperatures experienced by each lake and average lake depth, we did find there is a strong relationship between high thermal stability and cooler bottom water temperatures. Specifically, the 3 lakes which experienced only cool water temperatures historically, also experienced strong thermal stability responses to past average air temperatures. Lower RMSE of total RTRM predictions corresponded to lower RMSE of bottom water temperature predictions and vice versa in all of the study lakes.

Across lakes, model results suggest that increasing air temperatures will lead to significant increases in monthly surface water temperature, average water temperature, and bottom water temperature by midcentury across the year. Surface water temperatures are increasing fastest (0.44 °C per decade), followed by average water temperatures (0.30 °C per decade), and then bottom water temperatures (0.16

°C per decade). Expected increases in surface water temperature per decade are higher than the global average and are also slightly higher than increases expected for large north temperate lakes. This indicates small lakes in Connecticut are warming faster than the global average, and faster than much larger lakes nearby. However, expected increases in surface water temperature are slightly less than expected increases in regional air temperature. Additionally, total and max RTRM is increasing from April-

November, with the most significant decreases in lake mixing expected from July-September. Thermal stability is projected to begin 2-4 weeks earlier and last 2-4 weeks longer by midcentury. Moreover, ice free days will likely decrease in the study reservoirs during the winter as a result of increases in water temperature and lake mixing during January and February

Historically, high-risk cyanobacteria blooms occurred in 3 out of 6 of the study lakes during the

July-September months. Higher lake surface water temperatures, higher total RTRM values, and lower bottom water temperatures were moderately correlated with higher amounts of cyanobacteria biomass.

These 3 lakes have low surface area to perimeter ratios and more irregularly shaped shorelines relative to

58 the other study lakes. Our results suggest shallow, near-shore areas warm at a faster rate, therefore we speculate lake surface and shoreline shape may be important for promoting cyanobacteria blooms. Future studies focused on understanding how differences in lake shorelines, specifically perimeter to surface area ratios, affect lake vulnerability to high-risk cyanobacteria blooms would be beneficial. Lake history may also play a role in lake vulnerability to cyanobacteria blooms. 2 out of 3 lakes that experience high-risk cyanobacteria blooms historically are located in areas with high amounts of human development in their watersheds. Higher development in a lake watershed is associated with higher amounts of algal nutrients in lakes, therefore reducing nutrient inputs into these reservoirs may be especially important for mitigating the frequency and severity of future blooms in vulnerable reservoirs.

Generally, specific water temperature and RTRM ranges that corresponded with high-risk amounts of cyanobacterial biomass historically are expected to decrease in the future, while more extreme temperatures and RTRM values are expected to increase. Therefore, it is unclear how future changes in water temperature and thermal stability will affect the occurrence of high-risk cyanobacteria blooms in the study lakes. We anticipate that future increases in surface water temperature and thermal stability will increase biological growth and shift phytoplankton community composition to favor cyanobacteria species more often over time in reservoirs where cyanobacteria blooms are already a concern, provided nutrient levels are sufficient to support growth. We also suspect that lakes which historically experience moderate risk blooms may be more susceptible to high-risk blooms in the future.

Somewhat concerning, Aphanizomenon, Anabaena, and Microcystis appeared in the laboratory samples as dominate species in all of the study reservoirs at some point. All have the ability to produce microcystin-LR, and Aphanizomenon is known to produce significant levels of cylindrospermopsin and saxitoxin. These toxins are harmful to human health and are expected to increase with rising regional air temperatures. Given these species dominate at times in these Connecticut reservoirs, and likely other north temperate reservoirs, it would be beneficial for future studies to evaluate what specific water temperatures are associated with production of cylindrospermopsin and saxitoxin. Also, increased

59 monitoring of these cyanotoxins in addition to microcystin may be warranted, especially in Connecticut and in other north temperate regions where air temperatures are increasing faster than in other parts of the country. It is important to keep in mind that species of cyanobacteria that dominated historically may not be the same species that dominate in the future. As water temperatures begin to exceed historical temperature ranges, different cyanobacteria species, including those that grow best at more extreme temperatures may dominate more often in the future.

Beyond cyanobacteria blooms, future increases in water temperature and thermal stability may lead to other water quality issues. For example, strong thermal stability is linked to hypoxia, which is problematic for drinking water management as increased frequency or severity of hypoxic conditions may enhance sediment release of manganese and phosphorus (algal nutrients), which are difficult to remove in water treatment. More work is needed to better understand and characterize risks posed by increased water temperature and thermal stability to lake ecosystems, drinking water sources, and recreational lake activities as the climate continues to change. Understanding water temperature and thermal stability changes on a local scale is vital for informing future water management for water utilities, including scenario planning and adaptation strategies (e.g., artificial mixing, or adding and connecting water sources and infrastructure). This study may provide useful information for understanding and informing strategies for managing drinking water reservoirs under a warming climate.

5 Research Summary, Future Suggestions, and Concluding Remarks

5.1 Research Summary

Societies depend on the proper functioning and resilience of critical infrastructure systems including those for drinking water and wastewater, yet these systems are vulnerable to extreme weather, pollution, population growth, and human decisions. To date, little scholarship exists that interrogates how these stressors combine to impact reservoir water quality, wastewater system resilience, and water resource management broadly. To bridge this gap, this dissertation addressed the following questions: (1) What factors are important for improving wastewater system resilience to severe weather events and future climate change? (2) Has surface, average, or bottom water quality of select Connecticut reservoirs changed in recent decades? (3) Has air temperature and wind speed near lakes changed in recent decades and how are these changes affecting reservoir water quality? And, (4) How might future climate warming affect future water temperature, thermal stability, and cyanobacteria productivity by midcentury? The future of reservoir water quality and wastewater system resilience in Connecticut is evaluated because air temperature, extreme precipitation, and frequency of large storms are increasing in the Northeast.

Chapter 2 addresses the 1st research question. Approximately half of the wastewater system managers interviewed stated their systems are located areas vulnerable to flooding (i.e., next to a river, tidal basin, or coastal area). Most of these systems were significantly impacted by storm surge and/or tidal, coastal, or river flooding during past storms (e.g., Winter Storm Alfred, Hurricane Irene, and

Superstorm Sandy). Interview results suggest that systems more impacted by past storms tend to be more resilient to future storms because these past experiences drive learning, innovation, and adaptation.

Results also suggest the most resilient wastewater systems are those with high adaptive capacities that employ an adaptive management approach to make ongoing adaptation investments over time. Greater amounts of generic adaptive capacities (i.e., skilled staff and good leadership) help smooth both day‐to‐ day and emergency operations and provide a foundation for adaptive management. In turn, adaptive

61 management helps managers both build more generic adaptive capacities, and develop and employ greater amounts and diversity of specific adaptive capacities (i.e., soft and/or hard adaptations) that are especially important for enhancing and sustaining resiliency. Adaptive management also enables managers to better understand their system's vulnerabilities, how those vulnerabilities change over time, and what specific actions may reduce those vulnerabilities. Finally, my work suggests wastewater system resilience critically depends on the capacities of the human systems for building resilience as much as or more so than relying only on physical infrastructure resilience. This research contributes to filling an important gap in the literature by advancing our understanding of the human dimensions of infrastructure resilience and has practical implications for advancing resilience in the wastewater sector.

Chapter 3 addresses the 2nd and 3rd research questions. Long-term changes in thermal stability, and surface, average, and bottom water temperature, dissolved oxygen, turbidity, pH, and specific conductivity and are evaluated from 2003-2018 in five Connecticut drinking water reservoirs. Changes in wind speed and air temperature evaluated over the same period, and future air temperature and wind speed are assessed using GCMs under RCP 8.5. Relationships between air temperature, wind speed, and each water quality parameter are evaluated to determine if climate change may be the main driver of water quality changes. Prior literature was used to inform other potential drivers of change (i.e., both natural and anthropogenic) for parameters that were not largely impacted by changing wind speed and air temperature. Results suggest higher air temperatures are driving higher surface water temperatures, cooler bottom water temperatures, and stronger thermal stability in the study reservoirs. Persistent high air temperatures over 1-2 weeks are correlated strongly with high thermal stability, while persistent high wind speed over 2-5 weeks is correlated with low thermal stability. Significant increases in surface and average dissolved oxygen were also observed and are likely caused by a combination of increasing annual wind speed and higher biological productivity associated with higher air temperatures. Results suggest changes in specific conductivity, pH, and turbidity are not greatly associated with changes in air temperature and wind speed. However, decreasing surface and average turbidity across lakes may be a result of higher annual thermal stability and associated decreases in vertical lake mixing and sediment

62 resuspension. Annual specific conductivity is increasing significantly across lakes, likely as a result of road salting during the winter months and subsequent runoff from impervious surfaces in the lake watersheds into nearby streams or into the reservoirs directly. Surface pH is decreasing in most study lakes, however a longer study period is needed to confirm this trend and understand potential drivers of pH change. Increases in surface water temperature, thermal stability, and specific conductivity pose the highest threat to reservoir water quality, while decreases in bottom water temperature, increases in dissolved oxygen, and decreases in turbidity may have a positive impact.

Chapter 4 addresses the final research question. Here, lake-specific empirical models were developed and used to predict future (2041-2070) water temperature and thermal stability in six Connecticut reservoirs using air temperature projections from GCMs under RCP 8.5. In addition, observed relationships between the occurrence of high-risk cyanobacteria blooms, defined as those exceeding

70,000 cells mL-1, and corresponding values of water temperature and RTRM (i.e., thermal stability) were established. High-risk amounts of cyanobacteria biomass historically occurred in 3 out of 6 reservoirs and corresponded with specific high water temperature and RTRM ranges. These observed relationships were used to evaluate how future changes in lake conditions may affect the occurrence of future cyanobacteria blooms. On average across lakes, future projections suggest that surface, average, and bottom water temperature may continue to increase by 0.44°C, 0.30 °C, and 0.16 °C per decade, respectively.

Moreover, future projections indicate the most significant increases in thermal stability may occur from

July-September. Conversely, lake mixing during January and February is expected to decrease as water temperatures warm, which may increase the number of ice-free days in winter. Small reservoirs in

Connecticut are warming faster than global lakes on average, but it is unclear how future changes in water temperature and thermal stability will affect the occurrence of high-risk cyanobacteria blooms. Lake conditions that corresponded with high-risk blooms historically (e.g. surface water temperature of 23.7 ±

23.1, average water temperature of 18.2 ± 1.4, total RTRM of 282 ± 131) will likely decrease in the future, while days with more extreme high water temperatures and high RTRM are expected to increase.

Higher extreme surface water temperatures may promote the growth of toxic cyanobacteria species which tend to prefer more extreme temperatures, especially in reservoirs where cyanobacteria blooms are already a concern.

5.2 Future research suggestions

Future studies should further test and refine this research to advance water management and to develop solutions to address the resiliency and water quality concerns discussed herein. For wastewater systems, the adaptive management framework proposed in Chapter 2 may help systems develop solutions to increase resiliency and reduce impacts from severe weather events. But to be useful, this framework must be applied and rigorously tested among a broader sample of wastewater systems and across infrastructure domains. For drinking water systems, key findings from chapters 3 and 4 may provide useful knowledge and inform appropriate actions that water managers and decision makers can take to improve resilience and ensure water systems can provide safe, clean drinking water under a changing climate. Many conventional drinking water treatment systems cannot remove cyanotoxins and upgrades are expensive, therefore smart solutions are needed to increase water system resilience to increasing toxic cyanobacteria blooms in Connecticut and other north temperate regions broadly. Increased water quality monitoring may help to determine lake specific solutions. For instance, aeration systems that provide depth discrete vertical mixing, carbon filtration systems to remove toxins, riparian buffering systems to reduce nutrient inputs, and physical removal of legacy nutrients from sediments, and policies to reduce nutrient inputs from agriculture and wastewater are all potential management solutions that may help to reduce lake vulnerability to cyanobacteria blooms and other water quality concerns.

While findings from chapters 2-4 may be generalizable to other drinking and wastewater systems in north temperate regions, more work is needed to broaden the geographic study range and further improve the generalizability of this research. Finally, future research should attempt to build on the frameworks and tools developed and further enhance our understanding of climate change impacts on drinking water reservoirs and wastewater systems.

5.3 Concluding remarks

This dissertation assesses climate change impacts on wastewater system resilience and drinking water quality. Findings may aid in understanding, predicting, and informing appropriate strategies for managing these systems under a changing climate. Chapter 2 contributes to filling an important gap in the literature by advancing our understanding of the human dimensions of infrastructure resilience, while

Chapters 3 and 4 enhance understanding of climate change impacts on small drinking water reservoirs, including how future increase of air temperature may affect midcentury water temperature, thermal stability, and cyanobacteria blooms in small reservoirs. Beyond advancing science, Chapter 2 proposes an adaptive management framework that may help to enhance wastewater system adaptation and resilience in practice. Furthermore, Chapters 3 and 4 produce novel statistical time series and empirical prediction models to investigate the dynamic responses of freshwater quality to a variable and changing climate.

Finally, this dissertation interrogates drinking water reservoir and wastewater system resilience with the goal of providing useful information to help water managers ensure high quality, reliable, drinking water supplies, and undisrupted wastewater treatment, even in a changing climate. Ongoing relationships with state, regional, and local drinking water managers, climate change experts, and decision makers will be leveraged to broadly disseminate study findings and tools that may have applications for state, regional, and local drinking water planning and management.

6 Appendices

6.1 Chapter 3 - Supporting Material

Supporting Material, Table 1: Linear trend analysis results for each lake and parameter. If the Likelihood Ratio Test is significant (P ≤ 0.05, 95 % confidence), then the seasonal model (Eq. 3) is used, and in all other cases Model 1 (Eq. 2) is used. Linear trends for Lake Whitney, Dawson, Watrous, and Gaillard are evaluated from 2003-2018, while linear trends for Lake Saltonstall are assessed over a slightly shorter period from 2005-2018. Air temperature and wind speed is assessed over a longer period from January 1979 to December 2018. The total change noted is calculated from the annual slope and is dependent on the total number of years of data available for each lake. The significance of linear trends for each lake and for each parameter are determined using a Student’s t-test, where if P (>|t|) is ≤ 0.05, then the trend is determined significant with 95% confidence. The model standard error is included. Likelihood Water Quality Time Ratio Test Slope Standard Total Parameter Lake Period (P-value) Eq. (change yr-1) Error t -value P (>|t|) change Whitney 2003-2018 0.92 2 0.090 0.068 1.32 0.19 1.44 Dawson 2003-2018 0.001 3 (Feb.-May) 0.37 0.11 3.29 **0.002 5.92 Surface Water 3 (Jun.-Sep.) 0.11 0.15 -1.69 0.099 1.76 Temperature 3 (Oct.-Jan.) -0.20 0.15 -3.77 ***5.0e-4 -3.2 (°C) Watrous 2003-2018 0.09 2 0.051 0.069 0.74 0.464 0.816 Gaillard 2003-2018 0.85 2 0.11 0.042 2.66 *0.011 1.76 Saltonstall 2005-2018 0.79 2 0.054 0.071 0.76 0.451 0.756 Whitney 2003-2018 0.59 2 -0.0017 0.051 -0.03 0.97 -0.0272 Dawson 2003-2018 0.003 3 (Feb.-May) 0.27 0.11 2.41 *0.020 4.32 Average Water 3 (Jun.-Sep.) 0.012 0.15 -1.71 0.095 0.192 Temperature 3 (Oct.-Jan.) -0.24 0.15 -3.38 **0.002 -3.84 (°C) Watrous 2003-2018 0.35 2 -0.041 0.063 -0.65 0.52 -0.656 Gaillard 2003-2018 0.73 2 0.028 0.027 1.03 0.31 0.448 Saltonstall 2005-2018 0.85 2 0.024 0.047 0.51 0.61 0.336 Whitney 2003-2018 0.41 2 -0.090 0.047 -1.92 0.062 -1.44 Dawson 2003-2018 0.02 3 (Feb.-May) 0.21 0.15 1.48 0.15 3.36 Bottom Water 3 (Jun.-Sep.) -0.049 0.20 -1.35 0.19 -0.784 Temperature 3 (Oct.-Jan.) -0.31 0.20 -2.69 **0.010 -4.96 (°C) Watrous 2003-2018 0.53 2 -0.087 0.071 -1.22 0.23 -1.392 Gaillard 2003-2018 0.80 2 -0.050 0.025 -1.97 *0.054 -0.8 Saltonstall 2005-2018 0.30 2 -0.10 0.035 -2.89 **0.006 -1.4 Whitney 2003-2018 0.99 2 2.59 1.42 1.82 0.076 41.44 Total Relative Dawson 2003-2018 0.79 2 3.71 1.48 2.50 *0.016 59.36 Thermal Watrous 2003-2018 0.09 2 2.84 1.29 2.20 *0.033 45.44 Resistance to Mixing (RTRM) Gaillard 2003-2018 0.82 2 3.13 0.88 3.54 ***0.001 50.08 Saltonstall 2005-2018 0.69 2 1.69 1.39 1.22 0.23 23.66 Whitney 2003-2018 0.27 2 1.52 0.39 3.88 ***4.0e-4 24.32 Dawson 2003-2018 0.29 2 1.59 0.22 7.09 ***1.1e-8 25.44 Surface Water Watrous 2003-2018 0.69 2 1.74 0.23 7.58 ***1.9e-9 27.84 Dissolved Gaillard 2003-2018 0.81 2 1.44 0.20 7.36 ***4.0e-9 23.04 Oxygen (% SAT) Saltonstall 2005-2018 0.05 3 (Feb.-May) 2.52 0.50 5.00 ***1.5e-5 35.28 3 (Jun.-Sep.) 2.72 0.71 0.28 0.78 38.08 3 (Oct.-Jan.) 1.17 0.71 -1.89 0.067 16.38 Whitney 2003-2018 0.22 2 0.72 0.33 2.21 *0.033 11.52 Average Water Dawson 2003-2018 0.21 2 0.73 0.26 2.84 **0.007 11.68 Dissolved Watrous 2003-2018 0.54 2 1.24 0.20 6.26 ***1.5e-7 19.84 Oxygen (% SAT) Gaillard 2003-2018 0.23 2 0.85 0.21 4.08 ***1.9e-4 13.6 Saltonstall 2005-2018 0.11 2 1.46 0.29 5.08 ***1.1e-5 20.44 Whitney 2003-2018 0.37 2 -0.11 0.53 -0.21 0.84 -1.76 Bottom Water Dawson 2003-2018 0.49 2 -0.29 0.56 -0.51 0.61 -4.64 Dissolved Watrous 2003-2018 0.36 2 0.73 0.44 1.68 0.10 11.68 Oxygen (% SAT) Gaillard 2003-2018 0.39 2 0.31 0.32 0.98 0.33 4.96 Saltonstall 2005-2018 0.09 2 0.053 0.47 0.11 0.91 0.742 Whitney 2003-2018 0.52 2 -0.013 0.014 -0.97 0.34 -0.208 Surface pH Dawson 2003-2018 0.50 2 -0.030 0.0099 -3.05 **0.004 -0.48 Watrous 2003-2018 0.25 2 0.0084 0.012 0.69 0.49 0.1344

Gaillard 2003-2018 0.66 2 -0.011 0.013 -0.88 0.38 -0.176 Saltonstall 2005-2018 0.41 2 -0.032 0.012 -2.78 **0.0083 -0.448 Whitney 2003-2018 0.45 2 -0.014 0.014 -0.97 0.34 -0.224 Dawson 2003-2018 0.41 2 -0.018 0.012 -1.48 0.15 -0.288 Average pH Watrous 2003-2018 0.11 2 0.0092 0.012 0.74 0.47 0.1472 Gaillard 2003-2018 0.82 2 0.0037 0.012 0.30 0.76 0.0592 Saltonstall 2005-2018 0.48 2 -0.027 0.013 -2.04 *0.048 -0.378 Whitney 2003-2018 0.49 2 -0.032 0.026 -1.23 0.22 -0.512 Dawson 2003-2018 0.58 2 -0.014 0.019 -0.71 0.48 -0.224 Bottom pH Watrous 2003-2018 0.23 2 0.0054 0.015 0.36 0.72 0.0864 Gaillard 2003-2018 0.83 2 0.0053 0.016 0.34 0.74 0.0848 Saltonstall 2005-2018 0.39 2 -0.034 0.020 -1.73 0.093 -0.476 Whitney 2003-2018 0.85 2 -0.19 0.083 -2.27 *0.028 -3.04 Dawson 2003-2018 0.39 2 -0.066 0.036 -1.82 0.075 -1.056 Surface Turbidity (NTU) Watrous 2003-2018 0.43 2 -0.046 0.058 -0.79 0.44 -0.736 Gaillard 2003-2018 0.61 2 -0.059 0.039 -1.52 0.14 -0.944 Saltonstall 2005-2018 0.70 2 -0.12 0.045 -2.68 *0.011 -1.68 Whitney 2003-2018 0.66 2 -0.10 0.085 -1.23 0.23 -1.6 Dawson 2003-2018 0.22 2 -0.045 0.033 -1.34 0.19 -0.72 Average Turbidity (NTU) Watrous 2003-2018 0.32 2 -0.029 0.051 -0.56 0.58 -0.464 Gaillard 2003-2018 0.87 2 -0.088 0.037 -2.36 *0.023 -1.408 Saltonstall 2005-2018 0.62 2 -0.10 0.042 -2.45 *0.019 -1.4 Whitney 2003-2018 0.51 2 0.062 0.19 0.33 0.74 0.992 Dawson 2003-2018 0.14 2 -0.025 0.037 -0.67 0.51 -0.4 Watrous 2003-2018 0.33 2 -0.025 0.047 -0.53 0.60 -0.4 Bottom Turbidity (NTU) Gaillard 2003-2018 0.86 2 -0.11 0.042 -2.59 *0.013 -1.76 Saltonstall 2005-2018 0.02 3 (Feb.-May) -0.22 0.083 -2.65 *0.012 -3.08 3 (Jun.-Sep.) -0.20 0.12 0.18 0.856 -2.8 3 (Oct.-Jan.) 0.065 0.12 2.43 *0.020 0.91 Whitney 2003-2018 0.08 2 11.56 1.12 10.31 ***3.4e-13 184.96 Surface Specific Dawson 2003-2018 0.64 2 5.39 0.43 12.40 ***1.3e-15 86.24 Conductivity Watrous 2003-2018 0.89 2 4.59 0.37 12.51 ***6.4e-16 73.44 -1 (nS cm ) Gaillard 2003-2018 0.93 2 2.53 0.19 13.61 ***<2e-16 40.48 Saltonstall 2005-2018 0.96 2 10.90 0.67 16.38 ***<2e-16 152.6 Whitney 2003-2018 0.47 2 11.58 1.52 7.63 ***1.6e-9 185.28 Average Specific Dawson 2003-2018 0.63 2 5.57 0.44 12.55 ***8.5e-16 89.12 Conductivity Watrous 2003-2018 0.97 2 4.52 0.38 11.93 ***3.1e-15 72.32 -1 (nS cm ) Gaillard 2003-2018 0.74 2 2.39 0.18 12.96 ***<2e-16 38.24 Saltonstall 2005-2018 0.88 2 10.90 0.66 16.59 ***<2e-16 152.6 Whitney 2003-2018 0.77 2 13.045 3.19 4.08 ***1.9e-4 208.72 Bottom Specific Dawson 2003-2018 0.63 2 5.97 0.51 11.81 ***6.2e-15 95.52 Conductivity Watrous 2003-2018 1.00 2 4.57 0.69 6.61 ***4.7e-8 73.12 (nS cm-1) Gaillard 2003-2018 0.31 2 2.50 0.22 11.59 ***8.2e-15 40 Saltonstall 2005-2018 0.78 2 10.91 0.68 16.02 ***<2e-16 152.74 Whitney 1979-2018 0.79 2 0.0123 0.0068 1.82 0.07 0.492 Dawson 1979-2018 0.78 2 0.0129 0.0067 1.92 0.06 0.516 Air Temperature Watrous 1979-2018 0.76 2 0.0138 0.0067 2.06 *0.04 0.552 (°C) Gaillard 1979-2018 0.71 2 0.0169 0.0068 2.50 *0.01 0.676 Saltonstall 1979-2018 0.69 2 0.0076 0.0069 1.10 0.27 0.304 Whitney 1979-2018 0.22 2 0.0055 0.0019 2.99 **0.003 0.22 Dawson 1979-2018 0.25 2 0.0048 0.0018 2.71 **0.008 0.192 Wind Speed Watrous 1979-2018 0.24 2 0.0048 0.0018 2.70 **0.008 0.192 (m s-1) Gaillard 1979-2018 0.24 2 0.0048 0.0018 2.70 **0.008 0.192 Saltonstall 1979-2018 0.23 2 0.0054 0.0019 2.88 **0.005 0.216 Significance codes: P ≤ ***0.001, P ≤ **0.01, P ≤ *0.05

Supporting Material, Table 2a: Observed correlation (R2) between the 2-365 day lagged mean air temperature averages and water quality parameter predictions on the subsequent day in five Connecticut lakes. Air Temperature Water Quality Standard Pearson Parameter Lake R2 Days Slope Intercept Error Significance F Correlation Whitney (n=113) 0.94 8 0.93 3.10 1.3 ***<0.001 0.97 Dawson (n=118) 0.95 8 1.04 1.94 1.3 ***<0.001 0.97 Watrous (n=130) 0.95 14 1.05 2.03 1.3 ***<0.001 0.97 Surface Water Temperature Gaillard (n=164) 0.96 21 1.06 1.79 1.3 ***<0.001 0.98 (°C) Saltonstall (n=104) 0.95 13 1.07 1.26 1.4 ***<0.001 0.97 Whitney (n=112) 0.9 26 0.54 6.11 1.0 ***<0.001 0.95 Dawson (n=118) 0.89 25 0.73 4.38 1.5 ***<0.001 0.94 Watrous (n=129) 0.87 47 0.71 4.61 1.7 ***<0.001 0.93 Average Water Temperature Gaillard (n=163) 0.93 48 0.52 5.00 0.9 ***<0.001 0.96 (°C) Saltonstall (n=103) 0.94 36 0.6 4.66 0.9 ***<0.001 0.97 Whitney (n=113) 0.42 94 0.16 8.46 1.4 ***<0.001 0.65 Dawson (n=118) 0.48 69 0.39 6.96 2.8 ***<0.001 0.69 Watrous (n=128) 0.65 97 0.51 5.63 2.3 ***<0.001 0.81 Bottom Water Temperature Gaillard (n=163) 0.29 102 0.11 6.78 1.2 ***<0.001 0.54 (°C) Saltonstall (n=104) 0.41 233 0.23 6.36 1.2 ***<0.001 0.64 Whitney (n=113) 0.90 6 21.45 -181.89 41 <0.001 0.95 Dawson (n=116) 0.77 7 21.71 -201.35 65 <0.001 0.88 Watrous (n=130) 0.69 5 19.6 -163.45 74 <0.001 0.83 Gaillard (n=155) 0.92 15 24.53 -196.02 39 <0.001 0.96 Total RTRM Saltonstall (n=104) 0.94 9 24.64 -208.94 36 <0.001 0.97 Whitney (n=113) 0.23 3 1.88 60.27 19.83 ***<0.001 0.48 Dawson (n=118) 0.16 365 4.94 38.33 9.80 ***<0.001 0.40 Surface Water Watrous (n=130) 0.15 365 5.57 34.89 11.06 ***<0.001 0.39 Dissolved Oxygen (% Gaillard (n=160) 0.18 1 0.67 81.93 9.24 ***<0.001 0.42 SAT) Saltonstall (n=105) 0.17 4 1.12 79.08 14.56 ***<0.001 0.41 Whitney (n=113) 0.60 47 -2.27 93.57 11.16 ***<0.001 -0.77 Dawson (n=118) 0.54 13 -2.12 106.99 10.93 ***<0.001 -0.73 Average Water Watrous (n=130) 0.33 18 -1.30 97.40 10.27 ***<0.001 -0.57 Dissolved Oxygen (% Gaillard (n=160) 0.73 78 -2.10 100.75 8.45 ***<0.001 -0.85 SAT) Saltonstall (n=105) 0.50 52 -1.98 103.9 11.82 ***<0.001 -0.71 Whitney (n=113) 0.74 30 -3.89 90.02 12.99 ***<0.001 -0.86 Dawson (n=118) 0.7 11 -4.85 128.05 17.89 ***<0.001 -0.84 Bottom Water Watrous (n=128) 0.45 18 -2.93 104.81 17.86 ***<0.001 -0.67 Dissolved Oxygen (% Gaillard (n=159) 0.74 106 -3.48 97.87 14.43 ***<0.001 -0.86 SAT) Saltonstall (n=105) 0.84 46 -4.73 115.87 11.8 ***<0.001 -0.92 Whitney (n=113) 0.17 4 0.04 7.24 0.51 ***<0.001 0.41 Dawson (n=118) 0.07 211 -0.02 7.87 0.38 **0.003 -0.26 Watrous (n=130) 0.07 194 -0.02 7.81 0.44 **0.003 -0.26 Gaillard (n=164) 0.02 153 -0.01 7.96 0.51 *0.045 -0.14 Surface pH Saltonstall (n=105) 0.37 5 0.05 7.76 0.40 ***<0.001 0.61 Whitney (n=113) 0.04 5 0.02 7.26 0.52 *0.02 0.2 Dawson (n=118) 0.05 5 -0.02 7.63 0.40 *0.02 -0.22 Watrous (n=130) 0.07 60 -0.02 7.45 0.41 **0.003 -0.26 Gaillard (n=164) 0.18 49 -0.03 7.71 0.43 ***<0.001 -0.42 Average pH Saltonstall (n=105) 0.13 296 -0.08 9.1 0.39 ***<0.001 -0.36 Whitney (n=113) 0.02 94 0.02 7.18 0.93 0.18 0.14 Dawson (n=118) 0.06 5 -0.03 7.64 0.67 **0.007 -0.24 Watrous (n=128) 0.06 10 -0.03 7.29 0.57 **0.005 -0.24 Gaillard (n=163) 0.21 46 -0.04 7.49 0.52 ***<0.001 -0.46 Bottom pH Saltonstall (n=105) 0.04 365 -0.17 9.75 0.63 *0.04 -0.2 Whitney (n=113) 0.07 91 0.17 1.4 4.34 **0.004 0.26

Dawson (n=118) 0.15 4 -0.07 2.04 1.02 ***<0.001 -0.39 Watrous (n=130) 0.1 7 -0.08 2.51 1.40 ***<0.001 -0.32 Surface Turbidity Gaillard (n=160) 0.14 9 -0.08 2.27 1.22 <0.001 -0.37 (NTU) Saltonstall (n=105) 0.03 220 -0.07 2.87 1.87 0.09 -0.17 Whitney (n=113) 0.14 110 0.16 2.29 2.80 ***<0.001 0.37 Dawson (n=118) 0.02 5 -0.03 1.58 1.07 0.11 -0.14 Watrous (n=130) 0.11 211 0.10 0.75 1.47 ***<0.001 0.33 Average Turbidity Gaillard (n=160) 0.08 9 -0.06 2.1 1.17 ***<0.001 -0.28 (NTU) Saltonstall (n=105) 0.04 281 -0.11 3.53 1.28 *0.05 -0.2 Whitney (n=113) 0.1 104 0.25 2.96 5.24 ***<0.001 0.32 Dawson (n=118) 0.05 110 0.05 0.89 1.55 **0.01 0.22 Watrous (n=128) 0.22 64 0.20 -0.44 2.46 ***<0.001 0.47 Bottom Turbidity Gaillard (n=159) 0.06 204 0.08 0.86 1.91 **0.003 0.24 (NTU) Saltonstall (n=105) 0.02 101 0.04 2.59 2.15 0.18 0.14 Whitney (n=113) 0.14 365 0.028 -0.05 0.06 ***<0.001 0.37 Dawson (n=118) 0.13 365 0.013 -0.01 0.03 <0.001 0.36 Surface Watrous (n=130) 0.08 365 0.0086 0.02 0.02 <0.001 0.28 Specific Conductivity Gaillard (n=164) 0.31 365 0.0088 -0.02 0.01 ***<0.001 0.56 (mS cm-1) Saltonstall (n=105) 0.07 361 0.020 0.07 0.05 **0.006 0.26 Whitney (n=113) 0.11 151 0.0036 0.23 0.07 ***<0.001 0.33 Dawson (n=118) 0.13 365 0.0129 -0.01 0.03 <0.001 0.36 Average Watrous (n=130) 0.09 365 0.0088 0.02 0.02 ***<0.001 0.30 Specific Conductivity Gaillard (n=164) 0.31 365 0.0084 -0.02 0.01 ***<0.001 0.56 (mS cm-1) Saltonstall (n=105) 0.07 365 0.0166 0.07 0.05 **0.005 0.26 Whitney (n=113) 0.10 135 0.0059 0.24 0.13 ***<0.001 0.32 Dawson (n=118) 0.16 150 0.0020 0.11 0.03 ***<0.001 0.40 Watrous (n=128) 0.12 148 0.0012 0.10 0.02 ***<0.001 0.35 Bottom Specific Conductivity Gaillard (n=163) 0.32 365 0.0087 -0.02 0.01 ***<0.001 0.57 (mS cm-1) Saltonstall (n=105) 0.07 365 0.0163 0.08 0.05 ***<0.001 0.26 Significance codes: P ≤ ***0.001, P ≤ **0.01, P ≤ *0.05

Supporting Material, Table 2b: Observed correlation (R2) between the 2-365 day lagged mean wind speed averages and water quality parameter predictions on the subsequent day in five Connecticut lakes. Wind Speed Water Quality Standard Significance Pearson Parameter Lake R2 Days Slope Intercept Error F Correlation Surface Water Whitney (n=113) 0.74 43 -6.91 48.41 2.71 ***<0.001 -0.86 Temperature (°C) Dawson (n=118) 0.68 38 -8.05 -52.74 3.41 ***<0.001 -0.82 Watrous (n=130) 0.67 38 -8.09 52.64 3.37 ***<0.001 -0.82 Gaillard (n=164) 0.78 43 -8.66 54.94 3.11 ***<0.001 -0.88 Saltonstall (n=104) 0.78 50 -7.95 53.64 2.87 ***<0.001 -0.88 Average Water Whitney (n=112) 0.77 50 -4.2 33.4 1.52 ***<0.001 -0.88 Temperature (°C) Dawson (n=118) 0.73 60 -5.9 41.1 2.31 ***<0.001 -0.85 Watrous (n=129) 0.75 59 -6.3 41.9 2.31 ***<0.001 -0.87 Gaillard (n=163) 0.85 65 -4.8 33.5 1.35 ***<0.001 -0.92 Saltonstall (n=103) 0.84 65 -4.9 36.1 1.38 ***<0.001 -0.92 Bottom Water Whitney (n=113) 0.4 127 -1.52 17.80 1.38 ***<0.001 -0.63 Temperature (°C) Dawson (n=118) 0.44 101 -3.78 29.20 2.86 ***<0.001 -0.66 Watrous (n=128) 0.62 101 -5.19 34.95 2.79 ***<0.001 -0.79 Gaillard (n=163) 0.34 140 -1.26 13.87 1.15 ***<0.001 -0.58 Saltonstall (n=104) 0.41 233 -2.03 18.83 1.17 ***<0.001 -0.64 Total RTRM Whitney (n=113) 0.66 32 -149.12 819.37 73.4 <0.001 -0.81 Dawson (n=116) 0.41 17 -136.24 725.75 102.6 <0.001 -0.64 Watrous (n=130) 0.38 19 -135.73 708.40 105 <0.001 -0.62 Gaillard (n=155) 0.69 32 -169.21 908.68 80.5 <0.001 -0.83 Saltonstall (n=104) 0.72 25 -166.95 931.62 75.5 <0.001 -0.85 Whitney (n=113) 0.28 7 -13.11 148.10 19.2 ***<0.001 -0.53 Dawson (n=118) 0.04 362 -15.66 164.21 10.45 *0.02 -0.20

Surface Water Watrous (n=130) 0.05 363 -18.26 178.87 11.67 **0.01 -0.22 Dissolved Oxygen (% Gaillard (n=160) 0.24 248 13.26 30.23 8.93 ***<0.001 0.49 SAT) Saltonstall (n=105) 0.15 7 -6.65 126.97 14.73 ***<0.001 -0.39 Average Water Whitney (n=113) 0.56 82 18.68 -25.72 11.76 ***<0.001 0.75 Dissolved Oxygen (% Dawson (n=118) 0.44 39 17.21 0.34 12.11 ***<0.001 0.66 SAT) Watrous (n=130) 0.33 41 12.01 26.85 10.2 ***<0.001 0.57 Gaillard (n=160) 0.72 109 21.09 -22.15 8.53 ***<0.001 0.85 Saltonstall (n=105) 0.47 66 17.13 -2.97 12.17 ***<0.001 0.69 Bottom Water Whitney (n=113) 0.69 50 32.3 -113.35 14.22 ***<0.001 0.83 Dissolved Oxygen Dawson (n=118) 0.5 39 37.2 -107.11 14.82 ***<0.001 0.71 (%SAT) Watrous (n=128) 0.35 41 23.7 -40.42 19.41 ***<0.001 0.59 Gaillard (n=159) 0.73 127 37.1 -113.83 14.82 ***<0.001 0.85 Saltonstall (n=105) 0.77 65 39.6 -134.68 14.27 ***<0.001 0.88 Surface pH Whitney (n=113) 0.15 16 -0.28 9.14 0.52 ***<0.001 -0.39 Dawson (n=118) 0.09 238 0.32 6.16 0.38 ***<0.001 0.30 Watrous (n=130) 0.06 193 0.21 6.66 0.44 **0.003 0.24 Gaillard (n=164) 0.04 298 0.46 5.64 0.51 **0.015 0.20 Saltonstall (n=105) 0.37 10 -0.34 10.12 0.39 ***<0.001 -0.61 Average pH Whitney (n=113) 0.05 345 0.72 4.07 0.52 *0.02 0.22 Dawson (n=118) 0.09 2 0.10 6.96 0.4 ***<0.001 0.30 Watrous (n=130) 0.08 65 0.19 6.41 0.4 ***0.001 0.28 Gaillard (n=164) 0.22 55 0.33 5.84 0.42 ***<0.001 0.47 Saltonstall (n=105) 0.13 291 0.58 5.37 0.39 ***<0.001 0.36 Bottom pH Whitney (n=113) 0.03 1 0.1 7.03 0.92 0.08 0.17 Dawson (n=118) 0.07 2 0.14 6.56 0.66 **0.004 0.26 Watrous (n=128) 0.05 40 0.21 5.99 0.57 **0.01 0.22 Gaillard (n=163) 0.25 53 0.44 4.99 0.5 ***<0.001 0.50 Saltonstall (n=105) 0.05 358 0.91 3.3 0.62 *0.02 0.22 Surface Turbidity Whitney (n=113) 0.09 118 -1.78 11.99 4.31 **0.002 -0.30 (NTU) Dawson (n=118) 0.19 3 0.42 -0.93 0.99 ***<0.001 0.44 Watrous (n=130) 0.07 30 0.64 -1.51 1.42 **0.003 0.26 Gaillard (n=160) 0.12 36 0.66 -1.84 1.71 ***<0.001 0.35 Saltonstall (n=105) 0.04 365 -2.40 13.80 1.86 *0.04 -0.20 Average Turbidity Whitney (n=113) 0.17 124 -1.66 12.07 2.77 ***<0.001 -0.41 (NTU) Dawson (n=118) 0.1 3 0.29 -0.1 1.03 ***<0.001 0.32 Watrous (n=130) 0.1 179 -0.80 5.42 1.48 ***<0.001 -0.32 Gaillard (n=160) 0.05 292 -1.19 6.76 1.19 **0.004 -0.22 Saltonstall (n=105) 0.03 286 0.75 -1.3 1.29 0.1 0.17 Bottom Turbidity Whitney (n=113) 0.1 109 -2.41 17.26 5.25 <0.001 -0.32 (NTU) Dawson (n=118) 0.08 112 -0.68 4.51 1.52 0.001 -0.28 Watrous (n=128) 0.21 69 -1.94 10.73 2.47 <0.001 -0.46 Gaillard (n=159) 0.1 218 -1.30 7.79 1.86 <0.001 -0.32 Saltonstall (n=105) 0.02 365 1.91 -6.06 2.15 0.15 0.14 Surface Specific Whitney (n=113) 0.05 6 -0.02 0.3212 0.0642 *0.02 -0.22 Conductivity (mS cm- Dawson (n=118) 0.07 246 -0.02 0.2323 0.029 **0.003 -0.26 1) Watrous (n=130) 0.06 316 -0.03 0.2659 0.0239 **0.004 -0.24 Gaillard (n=164) 0.01 8 0 0.0672 0.013 0.18 0.10 Saltonstall (n=105) 0.01 4 0.01 0.2383 0.0474 0.24 0.10 Average Specific Whitney (n=113) 0.1 142 -0.03 0.4246 0.07 ***<0.001 -0.32 Conductivity (mS cm- Dawson (n=118) 0.09 214 -0.02 0.2187 0.03 ***0.001 -0.30 1) Watrous (n=130) 0.08 246 -0.02 0.1959 0.02 ***0.001 -0.28 Gaillard (n=164) 0.02 8 0 0.0657 0.01 0.09 0.14 Saltonstall (n=105) 0.01 2 0 0.2480 0.05 0.33 0.10 Bottom Specific Whitney (n=113) 0.13 141 -0.06 ***0.61 0.12 ***<0.001 -0.36 Conductivity (mS cm- Dawson (n=118) 0.11 175 -0.02 ***0.22 0.03 ***<0.001 -0.33 1 ) Watrous (n=128) 0.11 246 -0.02 ***0.22 0.02 ***<0.001 -0.33 Gaillard (n=163) 0.02 8 0.002 ***0.06 0.01 0.06 0.14 Saltonstall (n=105) 0.01 360 -0.03 **0.41 0.05 0.32 -0.10 Significance codes: P ≤ ***0.001, P ≤ **0.01, P ≤ *0.05

Supporting Material, Table 3: Historical and future average mean air temperature and wind speed averaged over five Connecticut lakes. The historical and future 30-year averages are the average monthly air temperature and wind speed projections from 3 General Circulation Models (HadGEM2-CC365, CCSM4, and GFDL-ESM2M) under Representative Concentration Pathway 8.5. Future 30- Total Historical Historical 30- year Avg. Change 30-year Future 30- Total year Avg. Mean Air over Avg. Mean year Avg. Change Mean Air temperatur the 70- Wind Speed Mean Wind over the temperature e (°C) Change year (m s-1) Speed Change 70-year P, 2- (°C) (2041- yr-1 period P, 2-tailed t- (1971- (m s-1) yr-1 period tailed t- Month (1971-2000) 2070) (°C) (°C) test 2000) (2041-2070) (m s-1) (m s-1) test Jan. -1.9 ± 3.1 1.8 ± 2.8 0.053 3.7 <0.001 *** 5.66 ± 1.25 5.57 ± 1.21 -0.0013 -0.09 0.27 Feb. -0.1 ± 2.6 2.5 ± 3.0 0.037 2.6 <0.001 *** 5.47 ± 1.21 5.60 ± 1.28 0.0019 0.13 0.13 Mar. 4.0 ± 2.6 6.5 ± 2.7 0.036 2.5 <0.001 *** 5.31 ± 1.23 5.32 ± 1.27 0.0001 0.01 0.89 Apr. 9.4 ± 2.4 12.3 ± 2.6 0.041 2.9 <0.001 *** 4.66 ± 0.99 4.53 ± 0.94 -0.0019 -0.13 0.01 ** May. 15.1 ± 2.3 17.7 ± 2.3 0.037 2.6 <0.001 *** 4.02 ± 0.85 3.85 ± 0.80 -0.0024 -0.17 0.00 *** Jun. 20.0 ± 2.1 22.8 ± 2.2 0.040 2.8 <0.001 *** 3.66 ± 0.73 3.56 ± 0.67 -0.0014 -0.10 0.04 * Jul. 23.0 ± 1.6 26.4 ± 1.8 0.049 3.4 <0.001 *** 3.43 ± 0.63 3.52 ± 0.63 0.0013 0.09 0.05 * Aug. 22.4 ± 1.7 26.0 ± 1.8 0.051 3.6 <0.001 *** 3.41 ± 0.65 3.43 ± 0.65 0.0003 0.02 0.63 Sep. 18.2 ± 2.2 21.6 ± 2.4 0.049 3.4 <0.001 *** 3.86 ± 0.83 3.86 ± 0.85 0 0 0.92 Oct. 11.9 ± 2.6 15.0 ± 2.9 0.044 3.1 <0.001 *** 4.70 ± 1.10 4.48 ± 1.10 -0.0031 -0.22 0.00 *** Nov. 6.4 ± 2.6 9.5 ± 2.9 0.044 3.1 <0.001 *** 5.26 ± 1.26 5.25 ± 1.29 -0.0001 -0.01 0.95 Dec. 1.2 ± 2.8 4.8 ± 2.8 0.051 3.6 <0.001 *** 5.67 ± 1.28 5.59 ± 1.24 -0.0011 -0.08 0.28 Annual 10.9 ± 8.9 14.0 ± 9.0 0.044 3.1 <0.001 *** 4.59 ± 1.35 4.54 ± 1.35 -0.0007 -0.05 0.05 * Significance codes: P ≤ ***0.001, P ≤ **0.01, P ≤ *0.05

6.2 – Supporting Material

Supporting Material, Figure 1: Air temperature to surface water temperature predictive model accuracy for each lake. Residuals (residual = measured - predicted) are the leave-one-out cross validation estimates. There is a strong correlation between predicted and measured surface water temperature values. The residuals are evenly spaced and fall close to the R2=1 line, indicating high model accuracy.

Supporting Material, Figure 2: Lagged air temperature to average water temperature predictive model accuracy for each lake. Residuals (residual = measured - predicted) are the leave-one-out cross validation estimates. There is a strong correlation between predicted and measured average water temperature values. The residuals are evenly spaced and fall close to the R2=1 line, indicating high model accuracy.

Supporting Material, Figure 3: Lagged air temperature to bottom water temperature empirical model accuracy for each lake. Residuals (residual = measured - predicted) are the leave-one-out cross validation estimates. The range of bottom water temperatures experience by each lake vary greatly.

Supporting Material Figure 4: Surface water temperature to total RTRM empirical model accuracy for each lake. Residuals (residual = measured - predicted) are the leave-one-out cross validation estimates. There is a strong correlation between predicted and measured total RTRM values in Lake Whitney, Lake Gaillard, and Lake Saltonstall B. For these 3 lakes, the residuals are evenly spaced and fall close to the R2=1 line, indicating high model accuracy. For Lake Saltonstall A and Lake Watrous, the residuals are evenly spaced but fall further away from the R2=1 line, which is consistent with the slightly lower R2 value for these lake models (Table 2). Lake Glen’s surface water temperature to total RTRM model is the least accurate, mostly because it slightly under predicts total RTRM at high RTRM values (above 200).

Supporting Material Figure 5: Surface water temperature to max RTRM predictive model accuracy for each lake. Residuals (residual = measured - predicted) are the leave-one-out cross validation estimates.

Supporting Material, Table 1: Correlation between low, moderate, and high-risk levels of cyanobacterial biomass, water temperature (T), Relative Thermal Resistance to Mixing measurements (RTRM). Mean Surface Mean Average Mean Bottom Mean Total Mean Max Mean Total Cyanobacteria Biomass (cells mL-1) Water T Water T Water T RTRM RTRM All available data (n=438) 0.14 0.15 0.01 0.15 0.10 Low Risk ≤ 20,000 (n=309) 0.01 0.08 0.06 -0.02 0.05 Moderate Risk = 20,000-70,000 (n=109) 0.14 0.16 0.04 0.11 0.12 High Risk ≥ 70,000 (n=20) 0.37 0.14 -0.44 0.38 0.26

Supporting Material, Table 2: Results from two-sample t-tests assuming unequal variances for water temperature and total Relative Thermal Resistance to Mixing (RTRM) ranges associated with low, moderate and high-risk blooms. Total cyanobacteria biomass consisting of less than 20,000 cells mL-1 is considered low-risk, between 20,000-70,000 cells mL-1 is considered moderate-risk, and over 70,000 cells mL-1 is considered high-risk. Parameter Low vs. High Risk Bins Moderate vs. High Risk Bins Low vs. Moderate Risk Bins Surface Water T Low High Moderate High Low Moderate Mean 19.7 23.7 19.3 23.7 19.7 19.3 Variance 37.0 8.0 28.7 8.0 37.0 28.7 Observations 311 20 110 20 311 110 Hypothesized Mean Difference 0 0 0 df 32 48 215

76 t Stat -5.55 -5.35 0.56 P(T<=t) one-tail 1.99E-06 1.23E-06 0.290 t Critical one-tail 1.69 1.68 1.65 P(T<=t) two-tail 3.99E-06 2.46E-06 0.570 t Critical two-tail 2.04 2.01 1.97 Average Water T Low High Moderate High Low Moderate Mean 15.5 18.2 16.0 18.2 15.5 16.0 Variance 21.5 1.8 16.8 1.8 21.5 16.8 Observations 311 20 110 20 311 110 Hypothesized Mean Difference 0 0 0 df 58 93 214 t Stat -6.95 -4.53 -1.12 P(T<=t) one-tail 1.80E-09 8.60E-06 0.132 t Critical one-tail 1.67 1.66 1.65 P(T<=t) two-tail 3.60E-09 1.72E-05 0.263 t Critical two-tail 2.00 1.99 1.97 Bottom Water T Low High Moderate High Low Moderate Mean 11.5 11.8 11.7 11.8 11.5 11.7 Variance 21.1 7.7 18.9 7.7 21.1 18.9 Observations 311 20 110 20 311 110 Hypothesized Mean Difference 0 0 0 df 26 38 201 t Stat -0.46 -0.15 -0.40 P(T<=t) one-tail 0.325 0.441 0.344 t Critical one-tail 1.71 1.69 1.65 P(T<=t) two-tail 0.649 0.881 0.687 t Critical two-tail 2.06 2.02 1.97 Total RTRM Low High Moderate High Low Moderate Mean 179 282 166 282 179 166 Variance 18058 13659 16218 13659 18058 16218 Observations 311 20 110 20 311 110 Hypothesized Mean Difference 0 0 0 df 22 28 201 t Stat -3.79 -4.02 0.89 P(T<=t) one-tail 5.06E-04 1.99E-04 0.187 t Critical one-tail 1.72 1.70 1.65 P(T<=t) two-tail 0.001 3.98E-04 0.374 t Critical two-tail 2.07 2.05 1.97 Max RTRM Low High Moderate High Low Moderate Mean 69 90 72 90 69 72 Variance 3569 2521 3871 2521 3569 3871 Observations 311 20 110 20 311 110 Hypothesized Mean Difference 0 0 0 df 23 31 185 t Stat -1.81 -1.40 -0.51 P(T<=t) one-tail 0.042 0.086 0.306 t Critical one-tail 1.71 1.70 1.65 P(T<=t) two-tail 0.083 0.171 0.612 t Critical two-tail 2.07 2.04 1.97 Total Cyanobacteria Biomass (cells mL-1) Low High Moderate High Low Moderate Mean 10541 123228 33647 123228 10541 33647 Variance 25516449 2464073702 132966946 2464073702 25516449 132966946 Observations 311 20 110 20 311 110 Hypothesized Mean Difference 0 0 0 df 19 19 124 t Stat -10.15 -8.03 -20.34 P(T<=t) one-tail 2.07E-09 7.91E-08 1.29E-41 t Critical one-tail 1.73 1.73 1.66 P(T<=t) two-tail 4.15E-09 1.58E-07 2.59E-41 t Critical two-tail 2.09 2.09 1.98

Supporting Material, Table 3: Water temperature and thermal stability values correlating with risk levels from exposure to cyanobacteria. The percentage of samples for each lake are categorized by low, moderate, or high health risk according to the Massachusetts Department of Health guidelines for human recreational exposure to cyanobacteria biomass in recreational lakes based on World Health Organization risk and EPA (2003) papers which associate cell counts with toxin risks and probability of adverse health effects (Sivonen & Jones, 1999). The specific values included in this table are the mean values for each parameter and associated standard deviations, that corresponding with low, moderate, and high-risk cyanobacteria blooms. This table does not indicate significance. Low Risk Moderate Risk High Risk Lake Measure (means) <20,000 cells mL-1 20,000-70,000 cells mL-1 >70,000 cells mL-1 Whitney (n=55) Percentage of Samples 36.36% 36.36% 27.27% Total Cyanobacteria 8554 ± 2458 38189 ± 6874 118497 ± 39413 Surface Water T 14.6 ± 2.5 20.1 ± 1.3 24.5 ± 1.6 Average Water T 11.9 ± 2.2 16.7 ± 1.2 18.1 ± 0.7 Bottom Water T 9.0 ± 2.4 11.9 ± 1.3 10.8 ± 0.8 Total RTRM 92 ± 42 189 ± 31 324 ± 52 Max RTRM 28 ± 13 68 ± 14 88 ± 23 Dawson (n=59) Percentage of Samples 89.83% 10.17% - Total Cyanobacteria 9423 ± 2180 26254 ± 7182 - Surface Water T 20.9 ± 2.7 23.4 ± 2.6 - Average Water T 17.6 2.9 18.9 ± 0.4 - Bottom Water T 14.3 3.7 14.0 ± 4.1 - Total RTRM 174 ± 50 235 ± 158 - Max RTRM 73 ± 37 80 ± 52 - Glen (n=61) Percentage of Samples 85.25% 14.75% - Total Cyanobacteria 9176 ± 3089 26254 ± 1620 - Surface Water T 19.5 ± 1.3 19.0 ± 2.1 - Average Water T 16.0 ± 1.9 17.9 ± 2.5 - Bottom Water T 13.5 ± 2.7 17.0 ± 3.0 - Total RTRM 128 ± 40 52 ± 24 - Max RTRM 49 ± 15 25 ± 16 - Watrous (n=68) Percentage of Samples 73.53% 22.06% 4.41% Total Cyanobacteria 11614 ± 1878 31903 ± 6455 78122 ± 223 Surface Water T 21.5 ± 2.3 19.7 ± 1.5 18.1 ± 0.8 Average Water T 16.4 ± 1.9 17.2 2.8 17.8 ± 0.4 Bottom Water T 12.5 ± 2.3 15.1 4.2 17.2 ± 0.2 Total RTRM 218 ± 55 123 ± 66 32 ± 35 Max RTRM 86 ± 34 54 ± 24 13 ± 16 Gaillard (n=86) Percentage of Samples 93.02% 6.98% - Total Cyanobacteria 11834 ± 2306 28109 ± 8906 - Surface Water T 19.3 ± 1.8 12.0 ± 2.3 - Average Water T 13.8 ± 0.7 8.7 ± 0.7 - Bottom Water T 8.6 ± 0.8 6.5 ± 0.3 - Total RTRM 208 ± 46 67 ± 43 - Max RTRM 69 ± 11 19 ± 12 - Saltonstall A (n=55) Percentage of Samples 63.64% 34.55% 1.82% Total Cyanobacteria 11752 ± 3337 35233 ± 4174 100136 Surface Water T 19.0 ± 3.5 18.2 ± 3.4 25.9 Average Water T 17.7 ± 3.2 16.5 ± 3.3 22.2 Bottom Water T 14.9 ± 3.2 13.9 ± 3.2 15.9 Total RTRM 113 ± 67 118 ± 39 292 Max RTRM 70 ± 45 56 ± 27 199 Saltonstall B (n=54) Percentage of Samples 37.04% 61.11% 1.85% Total Cyanobacteria 15672 ± 3855 35627 ± 8967 108833 Surface Water T 16.6 ± 6.2 19.1 ± 2.5 26.2 Average Water T 13.1 ± 4.0 14.8 ± 1.9 17.9 Bottom Water T 8.6 ± 2.5 8.7 ± 1.1 8.7 Total RTRM 155 ± 137 195 ± 59 409 Max RTRM 70 ± 64 88 ± 43 203 Across Lakes (n=438) Percentage of Samples 70.52% 24.94% 4.54% Total Cyanobacteria (y) 10,541 ± 3350 33,647 ± 7,615 123,228 ± 34,548 Surface Water T 19.7 ± 3.7 19.3 ± 3.2 23.7 ± 3.1 Average Water T 15.5 ± 3.2 16.0 ± 3.3 18.2 ± 1.4 Bottom Water T 11.5 ± 3.6 11.7 ± 4.0 11.8 ± 3.1 Total RTRM 179 ± 77 166 ± 81 282 ± 131 Max RTRM 69 ± 37 72 ± 36 90 ± 63

Supporting Material, Table 4: Projected change in the number of days per year where water temperature (T) and Relative Thermal Resistance to Mixing (RTRM) values are within a certain range, or above a certain threshold. The change reported is the difference between the average number of days per year during the historical period (1971-2000) and the average number of days during the future period (2041-2070). Positive numbers indicate an increase in the number of days per year in the future and vice versa. The standard deviation of the change across lakes is reported. Average Parameter Whitney Dawson Glen Watrous Gaillard Saltonstall A Saltonstall B Across Lakes Range Time Period (days yr-1) (days yr-1) (days yr-1) (days yr-1) (days yr-1) (days yr-1) (days yr-1) (days yr-1)

1971-2000 99 93 95 96 95 91 90 94 ± 3

20.6-26.8 2041-2070 79 64 84 64 62 61 60 68 ± 10 (°C)

Change -20 -29 -11 -32 -33 -30 -30 -26 ± 8 T 1971-2000 26 48 16 46 48 56 56 43 ± 15 ≥ 25 2041-2070 81 93 77 93 93 97 97 90 ± 8 Change 54 45 60 47 45 41 41 48 ± 7 1971-2000 0 0 0 0 0 0 0 0 ± 0 ≥ 30 2041-2070 4 22 2 20 21 28 29 18 ± 11 Surface Water Surface Change 4 22 2 20 21 28 29 18 ± 11

1971-2000 87 49 48 57 15 34 84 53 ± 26

16.8-19.6 2041-2070 64 40 37 44 85 33 57 51 ± 18

(°C)

T Change -23 -9 -10 -13 70 -1 -27 -2 ± 33

1971-2000 0 57 50 40 0 92 0 34 ± 36 ≥ 20 2041-2070 44 100 96 90 0 124 48 72 ± 43 Change 44 43 46 50 0 32 48 38 ± 18 1971-2000 0 0 0 0 0 0 0 0 ± 0 ≥ 25 2041-2070 0 2 3 0 0 58 0 9 ± 22 Average Water Average Change 0 2 3 0 0 57 0 9 ± 21

1971-2000 264 189 90 156 0 160 193 150 ± 84

8.7-14.9 2041-2070 313 189 82 140 0 171 262 165 ± 105 (°C)

Change 49 1 -8 -16 0 11 68 15 ± 32 T 1971-2000 0 59 80 48 0 95 0 40 ± 41 ≥ 15 2041-2070 0 101 118 99 0 130 0 64 ± 61 Change 0 42 38 51 0 34 0 24 ± 23 1971-2000 0 0 0 0 0 0 0 0 ± 0 ≥ 20 2041-2070 0 0 53 0 0 53 0 15 ± 26 Bottom Water Bottom Change 0 0 53 0 0 53 0 15 ± 26 1971-2000 219 118 188 170 164 151 231 177 ± 39 151-413 2041-2070 159 110 201 217 172 177 180 174 ± 34 Change -60 -8 13 48 8 27 -51 -3 ± 40 1971-2000 32 70 0 0 0 0 1 15 ± 27 ≥ 413 2041-2070 83 108 0 2 51 18 0 37 ± 44 Change 51 38 0 2 51 18 -1 23 ± 24

Total RTRM Total 1971-2000 1 8 0 0 0 0 0 1 ± 3 ≥ 475 2041-2070 51 69 0 0 11 0 0 19 ± 29 Change 50 61 0 0 11 0 0 17 ± 27 1971-2000 129 111 124 124 112 89 113 115 ± 13 27-153 2041-2070 97 97 106 91 103 97 125 102 ± 11 Change -32 -14 -18 -33 -9 9 12 -12 ± 18 1971-2000 0 40 9 1 0 0 0 7 ± 15 ≥ 153 2041-2070 35 88 46 73 34 41 0 46 ± 29 Change 35 48 37 72 34 41 0 38 ± 21

Max RTRM Max 1971-2000 0 17 1 0 0 0 0 3 ± 6 ≥ 200 2041-2070 23 76 29 58 19 27 0 33 ± 26 Change 23 59 28 58 19 27 0 31 ± 21

7 References

Abatzoglou, J. T. (2013). Development of Gridded Surface Meteorological Data for Ecological Applications and Modelling. International Journal of Climatology, 33(1), 121-131. https://doi.org/10.1002/joc.3413

Abatzoglou, J. T., & Brown, T. J. (2012). A Comparison of Statistical Downscaling Methods Suited for Wildfire Applications. International Journal of Climatology, 32(5), 772-780. https://doi.org/10.1002/joc.2312

Adrian, R., Deneke, R., Mischke, U., Stellmacher, R., & Lederer, P. (1995). A Long-Term Study of the Heiligensee (1975-1992). Evidence for Effects of Climatic Change on the Dynamics of Eutrophied Lake Ecosystems. Archiv für Hydrobiologie, 133(1), 315-337.

Ahlgren, I. (1976). Role of Sediments in the Process of Recovery of a Eutrophicated Lake. Interactions between Sediments and Fresh Water, https://doi.org/10.1007/978-94-011-9802-8_55

Arhonditsis, G. B., Brett, M. T., Degasperi, C. L., & Schindler, D. E. (2004). Effects of Climatic Variability on the Thermal Properties of Lake Washington. Limnology and Oceanography, 49(1), 256-270. https://doi.org/10.4319/lo.2004.49.1.0256

Arnell, N. W., Halliday S.J., Battarbee, R. W., Skeffington, R. A., & Wade, A. J. (2015). The Implications of Climate Change for the Water Environment in England. Progress in Physical Geography, 39(1), 93-120. https://doi.org/10.1177/0309133314560369

Baylis, J., Edmonds, A.J., & Grayson, M. E. (2016). NIAC Water Sector Resilience Final Report and Recommendations. Washington, DC: National Infrastructure Advisory Council

Berkhout, F., Hertin, J., & Gann, D. M. (2006). Learning to Adapt: Organisational Adaptation to Climate Change Impacts. Climatic Change, 78(1), 135–56. https://doi.org/10.1007/s10584-006-9089-3

Bever, A. J., MacWilliams, M. L., & Fullerton, D. K. (2018). Influence of an Observed Decadal Decline in Wind Speed on Turbidity in the San Francisco Estuary. Estuaries and Coasts, 41(7), 1943-1967. https://doi.org/10.1007/s12237-018-0403-x

Bhamidipati, S. (2015). Simulation Framework for Asset Management in Climate-Change Adaptation of Transportation Infrastructure. Transportation Research Procedia, 8,17–28. https://doi.org/10.1016/j.trpro.2015.06.038

Birge, E. A. (1910). An Unregarded Factor in Lake Temperatures. Transactions of the Wisconsin Academy of Sciences, Arts and Letters, 16, 989–1004.

Birge, E. A., & Juday, C. (1914). A Limnological Study of the Finger Lakes of New York. Bureau of Fisheries, 32, 526–610.

Blenckner, T. (2005). A Conceptual Model of Climate-Related Effects on Lake Ecosystems. Hydrobiologia, 533(1), 1-14. https://doi.org/10.1007/s10750-004-1463-4

Bosse, K. R., Sayers M. J., Shuchman, R. A., Fahnenstiel, G. L., Ruberg, S. A., Fanslow, D. L., et al. (2019). Spatial-Temporal Variability of in Situ Cyanobacteria Vertical Structure in Western Lake Erie: Implications for Remote Sensing Observations. Journal of Great Lakes Research, 45(3), 480- 489. https://doi.org/10.1016/j.jglr.2019.02.003

Boyd, E., & Juhola, S. (2015). Adaptive Climate Change Governance for Urban Resilience. Urban Studies, 52(7), 1234–64. https://doi.org/10.1177/0042098014527483

Braskerud, B. C. (2001). The Influence of Vegetation on Sedimentation and Resuspension of Soil Particles in Small Constructed Wetlands. Journal of Environmental Quality, 30(4), 1447-1457. https://doi.org/10.2134/jeq2001.3041447x

Brient, L., Lengronne, M., Bertrand, E., Rolland, D., Sipel, A., Steinmann, D., et. al. (2008). A Phycocyanin Probe as a Tool for Monitoring Cyanobacteria in Freshwater Bodies. Journal of Environmental Monitoring, 10(2), 248-255. https://doi.org/10.1039/b714238b

Brown, S., Versace, V. L., Laurenson, L., Ierodiaconou, D., Fawcett, J., & Salzman, S. (2012). Assessment of Spatiotemporal Varying Relationships Between Rainfall, Land Cover and Surface Water Area Using Geographically Weighted Regression. Environmental Modeling and Assessment, 17(3), 241-254. https://doi.org/10.1007/s10666-011-9289-8

Bruine De Bruin, W., Parker, A. M., & Fischhoff, B. (2007). Individual Differences in Adult Decision- Making Competence. Journal of Personality and Social Psychology, 92(5), 938–56. https://doi.org/10.1037/0022-3514.92.5.938

Butcher, J. B., Nover, D., Johnson, T. E., & Clark., C. M. (2015). Sensitivity of Lake Thermal and Mixing Dynamics to Climate Change. Climatic Change, 129(1), 295-305. https://doi.org/10.1007/s10584-015-1326-1

Butler, D., Ward, S., Sweetapple, C., Astaraie-Imani, M., Diao, K., Farmani, R., & Fu, G. (2017). Reliable, Resilient and Sustainable Water Management: The Safe & SuRe Approach. Global Challenges, 1(1), 63–77. https://doi.org/10.1002/gch2.1010

Caldwell, P. M., Bretherton, C. S., Zelinka, M. D, Klein, S. A., Santer, B. D., & Sanderson, B. M. (2014). Statistical Significance of Climate Sensitivity Predictors Obtained by Data Mining. Geophysical Research Letters, 41(5), 1803-1808. https://doi.org/10.1002/2014GL059205

Camacho, A. E. (2009). Adapting Governance to Climate Change: Managing Uncertainty through a Learning Infrastructure. Emory Law Journal, 59, 4–77. https://doi.org/10.2139/ssrn.1352693

Cañedo-Argüelles, M., Hawkins, C. P., Kefford, B. J., Schafer, R. B., Dyack, B. J., Brucet, S., Buchwalter, D. (2016). Saving Freshwater from Salts. Science, 351(6276) 914-916. https://doi.org/10.1126/science.aad3488.

Cañedo-Argüelles, M., Kefford, B. J., Piscart, C., Prat, N., Schäfer, R. B., & Schulz, C. (2013). Salinisation of Rivers: An Urgent Ecological Issue. Environmental Pollution, 173, 157-167. https://doi.org/10.1016/j.envpol.2012.10.011.

Carvalho, L., Mcdonald, C., de Hoyos, C., Mischke, U., Phillips, G., Borics, G., et al. (2013). Sustaining Recreational Quality of European Lakes: Minimizing the Health Risks from Algal Blooms through Phosphorus Control. Journal of Applied Ecology, 10(2), 248-255, https://doi.org/10.1111/1365- 2664.12059

Chapra, S. C., Boehlert, B., Fant. C, Bierman, V. J., Henderson, J., Mills, D., et al. (2017). Climate Change Impacts on Harmful Algal Blooms in U.S. Freshwaters: A Screening-Level Assessment. Environmental Science & Technology, 51(16), 8933-8943. https://doi.org/10.1021/acs.est.7b01498

Chen, H., Burke, J. M., Dinsmore, W. P., Prepas, E. E., & Fedorak, P. M. (2007). First Assessment of Cyanobacterial Blooms and Microcystin-LR in the Canadian Portion of Lake of the Woods. Lake and Reservoir Management, 23, 169-178. https://doi.org/10.1080/07438140709353920

Connecticut Transportation Institute, UConn: Mahoney, J., Jackson, E., Larsen, D., Vadas, T., Wille, K., & Zinke, S. (2015). Winter Highway Maintenance Operations: Connecticut. Connecticut Academy of Science and Engineering. Retrieved from http://www.ctcase.org/reports/WinterHighway2015/winter-highway-2015.pdf

Cormier, S.M., Wilkes, S.P., & Zheng, L. (2013). Relationship of Land Use and Elevated Ionic Strength in Appalachian Watersheds. Environmental Toxicology and Chemistry, 32(2), 296-303. https://doi.org/10.1002/etc.2055

Couture, R. M., De Wit, H. A., Tominaga, K., Kiuru, P., & Markelov, I. (2015). Oxygen Dynamics in a Boreal Lake Responds to Long-Term Changes in Climate, Ice Phenology, and DOC Inputs. Journal of Geophysical Research: Biogeosciences, 120(11), 2441-2456. https://doi.org/10.1002/2015JG003065

Creswell, J. W. (2007). Qualitative Inquiry and Research Design: Choosing Among Five Approaches (2nd ed.), Thousand Oaks, CA: Sage Publications.

Crettaz, M. M. C., Lorena, R., Jorge, O. A., Sandro, G., Daniela, S., Dario, A., & Leda, G. (2017). Mathematical Modeling of Microcystis Aeruginosa Growth and Microcystin-LR Production in Culture Media at Different Temperatures. Harmful Algae, 67, 13-25. https://doi.org/10.1016/j.hal.2017.05.006

CTPCSAR. (2019). Connecticut Physical Climate Science Assessment Report. Groton, CT: Connecticut Institute for Resilience & Climate Adaptation

Decremer, D., Chung, C. E., Ekman, A. M. L., & Brandefelt, J. (2014). Which Significance Test Performs the Best in Climate Simulations? Tellus, Series A: Dynamic Meteorology and Oceanography, 66(1). https://doi.org/10.3402/tellusa.v66.23139

Delpla, I., Baures, E., Jung, A. V., Clement, M., & Thomas, O. (2011). Issues of Drinking Water Quality of Small Scale Water Services towards Climate Change. Water Science and Technology, 63, 227- 232. https://doi.org/10.2166/wst.2011.038

Deng, J., Zhang, W., Qin, B., Zhang, Y., Paerl, H. W., & Salmaso, N. (2018). Effects of Climatically- Modulated Changes in Solar Radiation and Wind Speed on Spring Phytoplankton Community Dynamics in Lake Taihu, China. PLoS ONE, 13(10), e0205260. https://doi.org/10.1371/journal.pone.0205260

Descy, J. P., Darchambeau, F., Lambert, T., Stoyneva-Gaertner, M. P., Bouillon, S., & Borges, A. V. (2017). Phytoplankton Dynamics in the Congo River. Freshwater Biology, 62(1), 87-101. https://doi.org/10.1111/fwb.12851

Dobiesz, N. E., & Lester, N. P. (2009). Changes in Mid-Summer Water Temperature and Clarity across the Great Lakes between 1968 and 2002. Journal of Great Lakes Research, 35(3), 371-384. https://doi.org/10.1016/j.jglr.2009.05.002

Donovan, B, & Work, D. B. (2017). Empirically Quantifying City-Scale Transportation System Resilience to Extreme Events. Transportation Research Part C: Emerging Technologies, 79, 333– 346. https://doi.org/10.1016/j.trc.2017.03.002

Downing, J. A., & Duarte, C. M. (2009). Abundance and Size Distribution of Lakes, Ponds and Impoundments. Encyclopedia of Inland Waters (pp. 469–478). Cambridge, MA: Academic Press

Eisenack, K., Stecker, R., Reckien, D., & Hoffmann, E. (2011). Adaptation to Climate Change in the Transport Sector: A Review of Actions and Actors. Mitigation and Adaptation Strategies for Global Change, 17(5), 451–469. https://doi.org/10.1007/s11027-011-9336-4

Elliott, J. A., Jones, I. D., & Thackeray, S. J. (2006). Testing the Sensitivity of Phytoplankton Communities to Changes in Water Temperature and Nutrient Load, in a Temperate Lake. Hydrobiologia, 559(1), 401-411. https://doi.org/10.1007/s10750-005-1233-y

Engle, N. L., & Lemos, M. C. (2010). Unpacking Governance: Building Adaptive Capacity to Climate Change of River Basins in Brazil. Global Environmental Change, 20(1), 4–13. https://doi.org/10.1016/j.gloenvcha.2009.07.001

EPA. (1989). Risk Assessment Guidance for Superfund. Volume 1: Human Health Evaluation Manual (Part A). Washington, DC: Environmental Protection Agency, Office of Emergency and Remedial Response

EPA. (1997). Exposure Factors Handbook. Washington, DC: Environmental Protection Agency, Office of Research and Development

EPA. (2003). User’s Manual- Swimmer Exposure Assessment Model (SWIMODEL) Version 3.0. Washington, DC: Environmental Protection Agency, Office of Pesticides Programs, Antimicrobials Division

EPA. (2019). Health and Ecological Effects. Washington. DC: Environmental Protection Agency. Retrieved from https://www.epa.gov/nutrient-policy-data/health-and-ecological-effects#what1

Falco, G. J., & Webb, W. R. (2015). Water Microgrids: The Future of Water Infrastructure Resilience.” Procedia Engineering, 118, 50–57. https://doi.org/10.1016/j.proeng.2015.08.403

Farrer, D., Counter, M., Hillwig, R., & Cude, C. (2015). Health-Based Cyanotoxin Guideline Values Allow for Cyanotoxin-Based Monitoring and Efficient Public Health Response to Cyanobacterial Blooms. Toxins, 7(2), 457-477. https://doi.org/10.3390/toxins7020457

Fee, E. J., Hecky, R. E., Kasian, S. E. M., & Cruikshank, D. R. (1996). Effects of Lake Size, Water Clarity, and Climatic Variability on Mixing Depths in Canadian Shield Lakes. Limnology and Oceanography, 41(5), 912-920. https://doi.org/10.4319/lo.1996.41.5.0912

Foley, B., Jones, I. D., Maberly, S. C., & Rippey, B. (2011). Long-Term Changes in Oxygen Depletion in a Small Temperate Lake: Effects of Climate Change and Eutrophication. Freshwater Biology, 57(2), 278-289. https://doi.org/10.1111/j.1365-2427.2011.02662.x

Folke, C., Carpenter, S., Elmqvist, T., Gunderson, L., Holling, C. S., & Walker, B. (2002). Resilience and Sustainable Development: Building Adaptive Capacity in a World of Transformations. Ambio, 31(5), 437–40. https://doi.org/10.1639/0044-7447(2002)031[0437:RASDBA]2.0.CO;2

Folke, C., Hahn, T., Olsson, P, & Norberg, J. (2005). Adaptive Governance of Social-Ecological Systems. Annual Review of Environment & Resources, 30(1), 441–73. https://doi.org/10.1146/annurev.energy.30.050504.144511

Ford, J. D., Berrang-Ford, L., & Paterson, J. (2011). A Systematic Review of Observed Climate Change Adaptation in Developed Nations. Climatic Change, 106(2), 327-336 https://doi.org/10.1007/s10584-011-0045-5

Fotheringham, A. S., Brunsdon, C., & Charlton, M. (2002). Geographically Weighted Regression: The Analysis of Spatially Varying Relationships. Hoboken, NJ: John Wiley. Francis, R., & Bekera, B. (2014). A Metric and Frameworks for Resilience Analysis of Engineered and Infrastructure Systems. Reliability Engineering and System Safety, 121, 90-103. https://doi.org/10.1016/j.ress.2013.07.004

Frediani, M. E. B., Hacker, J. P., Anagnostou, E. N., & Hopson, T. (2016). Evaluation of PBL Parameterizations for Modeling Surface Wind Speed during Storms in the Northeast United States. Weather and Forecasting, 31(5), 1511-1528. https://doi.org/10.1175/WAF-D-15-0139.1

Giorgi, F., & Bi, X. Q. (2005). Updated Regional Precipitation and Temperature Changes for the 21st Century from Ensembles of Recent AOGCM Simulations. Geophysical Research Letters, 32(21), L21715. https://doi.org/10.1029/2005GL024288

Graham, J. L., & Jones, J. R. (2009). Microcystin in Missouri Reservoirs. Lake and Reservoir Management, 25(3), 253-263. https://doi.org/10.1080/07438140903143239

Granato, G. E., & Smith, K. P. (1999). Estimating Concentrations of Road-Salt Constituents in Highway- Runoff from Measurements of Specific Conductance. U.S. Geological Survey Water-Resources Investigations Report 99-4077. https://doi.org/10.3133/wri994077

Gregor, J., & Maršálek, B. (2004). Freshwater Phytoplankton Quantification by Chlorophyll a: A Comparative Study of in Vitro, in Vivo and in Situ Methods. Water Research, 38(3), 517-522. https://doi.org/10.1016/j.watres.2003.10.033

Guikema, S., McLay, L., & Lambert, J. H. (2015). Infrastructure Systems, Risk Analysis, and Resilience- Research Gaps and Opportunities. Risk Analysis. 35(4), 560-561. https://doi.org/10.1111/risa.12416

Hamilton, D. P., Spillman, C. M., Prescott, K. L., Kratz, T. K., & Magnuson, J. J. (2002). Effects of Atmospheric Nutrient Inputs and Climate Change on the Trophic Status of Crystal Lake, Wisconsin. SIL Proceedings (1922-2010), 28(1), 467-470. https://doi.org/10.1080/03680770.2001.11902625

Health Canada. (2001). Priority Substance List Assessment Report: Road Salt. Environment Canada, Minister of Public Works and Government Services.

Henson, S. A., Beaulieu, C., & Lampitt, R. (2016). Observing Climate Change Trends in Ocean Biogeochemistry: When and Where. Global Change Biology, 22(4), 1561-1571. https://doi.org/10.1111/gcb.13152

Hess, J. J., McDowell, J. Z., & Luber, G. (2012). Integrating Climate Change Adaptation into Public Health Practice: Using Adaptive Management to Increase Adaptive Capacity and Build Resilience. Environmental Health Perspectives, 120(2), 171-179. https://doi.org/10.1289/ehp.1103515

Holling, C. S. (1996). Engineering Resilience versus Ecological Resilience. Engineering Within Ecological Constraints, (pp. 31–44). Washington, DC: The National Academies Press. https://doi.org/10.17226/4919

Horton, R., Yohe, G., Easterling, W., Kates, R., Ruth, M., Sussman, E., et al. (2014). Ch. 16: Northeast. Climate Change Impacts in the United States: The Third National Climate Assessment. In U.S. Global Change Research Program (pp. 1–24). https://doi.org/10.7930/J0SF2T3P

Hu, X. M., Nielsen-Gammon, J. W., & Fuqing Z. (2010). Evaluation of Three Planetary Boundary Layer Schemes in the WRF Model. Journal of Applied Meteorology and Climatology, 49, 1831-1844. https://doi.org/10.1175/2010JAMC2432.1

Huber, V., Wagner, C., Gerten, D., & Adrian, R. (2012). To Bloom or Not to Bloom: Contrasting Responses of Cyanobacteria to Recent Heat Waves Explained by Critical Thresholds of Abiotic Drivers. Oecologia. 169(1), 245-256. https://doi.org/10.1007/s00442-011-2186-7

Hudon, C. P., Gagnon, S., Poirier Larabie, C., Gagnon, A., Lajeunesse, M. L., & Quilliam, M. A. (2016). Spatial and Temporal Variations of a Saxitoxin Analogue (LWTX-1) in Lyngbya Wollei (Cyanobacteria) Mats in the St. Lawrence River (Québec, Canada). Harmful Algae, 57, 69-77. https://doi.org/10.1016/j.hal.2016.06.001

Hurrel, J. W. (1995). Decadal trends in the North Atlantic Oscillation: regional temperatures and precipitation. Science, 269(5224), 676-679.

Hurrell, J. (1995). Decadal Trends in the North Atlantic Oscillation: Regional Temperatures and Precipitation. Science, 269, 676-679. https://doi.org/10.1126/science.269.5224.676

Hurrell, J., & Phillips, A. (2018). The Climate Data Guide: Hurrell North Atlantic Oscillation (NAO) Index (station-based). National Center for Atmospheric Research. Last modified 04 Aug 2018. Retrieved from https://climatedataguide.ucar.edu/climate-data/hurrell-north-atlantic-oscillation-nao- index-station-based

Hutchinson, G. E. (1957). A Treatise on Limnology. (Vol. 1). New York, NY: John Wiley.

Ibelings, B. W., Backer, L. C., Kardinaal, W. E. A., & Chorus, I. (2014). Current Approaches to Cyanotoxin Risk Assessment and Risk Management around the Globe. Harmful Algae, 40, 63-74. https://doi.org/10.1016/j.hal.2014.10.002

IPCC. (2007). Climate Change 2007: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. https://doi.org/10.2134/jeq2008.0015br

Izydorczyk, K., Tarczynska, M., Jurczak, T., Mrowczynski, J., & Zalewski, M. (2005). Measurement of Phycocyanin Fluorescence as an Online Early Warning System for Cyanobacteria in Reservoir Intake Water. Environmental Toxicology, 20(4), 425-430. https://doi.org/10.1002/tox.20128

Janex-Habibi, M. L., Baudin, I., Bruchet, A., Esperanza, M., Courtois, S., Decottignies, V., et al. (2011). Managing the Risk of Emerging Algae Toxins in Drinking Water: Development of a Targeted Analytical Approach. Water Practice and Technology, 5(4), wpt2010079. https://doi.org/10.2166/wpt.2010.079

Jöhnk, K. D., Huisman, J., Sharples, J., Sommeijer, B., Visser, P. M., & Stroom, J. M. (2008). Summer Heatwaves Promote Blooms of Harmful Cyanobacteria. Global Change Biology, 14(3), 495-512. https://doi.org/10.1111/j.1365-2486.2007.01510.x

Juan-García, P., Butler, D., Comas, J., Darch, G., Sweetapple, C., Thornton, A., & Corominas, L. (2017). Resilience Theory Incorporated into Urban Wastewater Systems Management. State of the Art. Water Research, 115(15), 149-161. https://doi.org/10.1016/j.watres.2017.02.047

Karmalkar, A. V., & Bradley, R. S. (2017). Consequences of Global Warming of 1.5 °C and 2 °C for Regional Temperature and Precipitation Changes in the Contiguous United States. PLoS ONE, 12(1), e0168697. https://doi.org/10.1371/journal.pone.0168697

Karnauskas, K. B., Lundquist, J. K., & Zhang, L. (2018). Southward Shift of the Global Wind Energy Resource under High Carbon Dioxide Emissions. Nature Geoscience, 11, 38-43 https://doi.org/10.1038/s41561-017-0029-9

Kaushal, S. S., Groffman, P. M., Likens, G. E., Belt, K. T., Stack, W. P., Kelly, V. R., Band, L. E., & Fisher, G. T. (2005). Increased Salinization of Fresh Water in the Northeastern United States. Proceedings of the National Academy of Sciences of the United States of America, 102(38), 13517- 13520. https://doi.org/10.1073/pnas.0506414102

Kelly, V. R., Findlay, S. E., Hamilton, S. K., Lovett, G. M., & Weathers, K. C. (2019). Seasonal and Long-Term Dynamics in Stream Water Sodium Chloride Concentrations and the Effectiveness of Road Salt Best Management Practices. Water, Air, and Soil Pollution, 230(1), 13. https://doi.org/10.1007/s11270-018-4060-2

Khatri, N., & Tyagi, S. (2015). Influences of Natural and Anthropogenic Factors on Surface and Groundwater Quality in Rural and Urban Areas. Frontiers in Life Science. 8(1), 23-39. https://doi.org/10.1080/21553769.2014.933716

King, J. R., Shuter, B. J., & Zimmerman, A. P. (1998). Erratum: The Response of the Thermal Stratification of South Bay (Lake Huron) to Climatic Variability. Canadian Journal of Fisheries and Aquatic Science, 55(8), 1999. https://doi.org/10.1139/f98-097

Kirchhoff, C. J., & Dilling, L. (2016). The Role of U.S. States in Facilitating Effective Water Governance under Stress and Change. Water Resources Research, 52(4), 2951-2964. https://doi.org/10.1002/2015WR018431

Kirchhoff, C. J., & Watson, P. (2019). Are Wastewater Systems Adapting to Climate Change? Journal of the American Water Resources Association (JAWRA), 55(4), 869-880. https://doi.org/https://doi.org/10.1111/1752-1688.12748

Komatsu, E., Fukushima, T. & Harasawa, H. (2007). A modeling approach to forecast the effect of long- term climate change on lake water quality. Ecological Modelling, 209(2-4), 351-366. https://doi:10.1016/j.ecolmodel.2007.07.021

Kortelainen, P., Rantakari, M., Pajunen, H., Mattsson, T., Juutinen, S., Larmola, T., Alm, J., Silvola, J., & Martikainen, P. J. (2013). Carbon Evasion/Accumulation Ratio in Boreal Lakes Is Linked to Nitrogen. Global Biogeochemical Cycles, 27(2), 363-374. https://doi.org/10.1002/gbc.20036.

Kortmann, R. W., Henry, D. D., Kuether, A., & Kaufman, S. (1982). Epilimnetic Nutrient Loading by Metalimnetic Erosion and Resultant Algal Responses in Lake Waramaug, Connecticut. Hydrobiologia, 91, 501-510. https://doi.org/10.1007/PL00020039

Kraemer, B.M., Anneville, O., Chandra, S., Dix, M., Kuusisto, E., Livingstone, D. M., et al. (2015). Morphometry and Average Temperature Affect Lake Stratification Responses to Climate Change. Geophysical Research Letters, 42(12), 4981-4988. https://doi.org/10.1002/2015GL064097

Kratz, T. K., Webster, K.E., Bowser, C.J., Magnuson, J. J., & Benson, B. J. (1997). The Influence of Landscape Position on Lakes in Northern Wisconsin, Freshwater Biology. 37(1), 209-217. https://doi.org/10.1046/j.1365-2427.1997.00149.x

Kundzewicz, Z. W., L. J. Mata, N. W. Arnell, P. Döll, B. Jimenez, K. Miller, T. Oki, Z. Şen, and I. Shiklomanov. (2008). The Implications of Projected Climate Change for Freshwater Resources and Their Management. Hydrological Sciences Journal. 53(1), 3-10. https://doi.org/10.1623/hysj.53.1.3

Lacher, I. L., Ahmadisharaf, E., Fergus, C., Akre, T., Mcshea, W. J., Benham, B. L., & Kline, K. S. (2019). Scale-Dependent Impacts of Urban and Agricultural Land Use on Nutrients, Sediment, and Runoff. Science of The Total Environment, 652, 611-622. https://doi.org/10.1016/j.scitotenv.2018.09.370

Levine, R. A., & Berliner, L. M. (1999). Statistical Principles for Climate Change Studies. Journal of Climate, 12, 564-574. https://doi.org/10.1175/1520-0442(1999)012<0564:SPFCCS>2.0.CO;2

Lin, Y., & Bie, Z. (2016). Study on the Resilience of the Integrated Energy System. Energy Procedia, 103, 171–76. https://doi.org/10.1016/j.egypro.2016.11.268

Livingstone, D. M. (2003). Impact of Secular Climate Change on the Thermal Structure of a Large Temperate Central European Lake. Climatic Change, 57(1), 205-225. https://doi.org/10.1023/A:1022119503144

Livingstone, D. M., & Dokulil, M. T. (2001). Eighty Years of Spatially Coherent Austrian Lake Surface Temperatures and Their Relationship to Regional Air Temperature and the North Atlantic

Oscillation. Limnology and Oceanography, 46(5), 1220-1227. https://doi.org/10.4319/lo.2001.46.5.1220

López-Abbate, M. C., Molinero, J., Perillo, G. M. E., Barría, d. C., Pettigrosso, R. E., Guinder, V. A., et al. (2019). Long-term changes on estuarine ciliates linked with modifications on wind patterns and water turbidity. Marine Environmental Research, 144, 46-55. https://doi:10.1016/j.marenvres.2018.12.001

Løvsletten, O., & Rypdal. M. (2016). Statistics of Regional Surface Temperatures after 1900: Long- Range versus Short-Range Dependence and Significance of Warming Trends. Journal of Climate, 29, 4057-4068. https://doi.org/10.1175/JCLI-D-15-0437.1

Lynch, C., Seth, A., & Thibeault, J. (2016). Recent and Projected Annual Cycles of Temperature and Precipitation in the Northeast United States from CMIP5. Journal of Climate, 29, 347-365. https://doi.org/10.1175/JCLI-D-14-00781.1

Lyon-Colbert, A., Su, S., & Cude, C. (2018). A Systematic Literature Review for Evidence of Aphanizomenon Flos-Aquae Toxigenicity in Recreational Waters and Toxicity of Dietary Supplements: 2000–2017. Toxins, 10(7), 254. https://doi.org/10.3390/toxins10070254

Magee, M. R., Wu, C. H., Robertson, D. M., Lathrop, R. C., & Hamilton, D. P. (2016). Trends and Abrupt Changes in 104 Years of Ice Cover and Water Temperature in a Dimictic Lake in Response to Air Temperature, Wind Speed, and Water Clarity Drivers. Hydrology and Earth System Sciences, 20(5), 1681-1702. https://doi.org/10.5194/hess-20-1681-2016

Magnuson, J. J., Meisner, J. D., & Hill, D. K. (1990). Potential Changes in the Thermal Habitat of Great Lakes Fish after Global Climate Warming. Transactions of the American Fisheries Society, 119(2), 254-264. https://doi.org/10.1577/1548-8659(1990)119<0254:pcitth>2.3.co;2

Magnuson, J. J., Webster, K. E., Assel, R. A., Bowser, C. J., Dillon, P. J., Eaton, J. G., et al. (1997). Potential Effects of Climate Changes on Aquatic Systems: Laurentian Great Lakes and Precambrian Shield Region. Hydrological Processes, 11(8), 825-871.https://doi.org/10.1002/(SICI)1099- 1085(19970630)11:8<825::AID-HYP509>3.0.CO;2-G

Mason, L. A., Riseng, C. M., Gronewold, A. D., Rutherford, E. S., Wang, J., Clites, A., et al. (2016). Fine-Scale Spatial Variation in Ice Cover and Surface Temperature Trends across the Surface of the Laurentian Great Lakes. Climatic Change, 138(1), 71-83. https://doi.org/10.1007/s10584-016-1721- 2

May, B., & Plummer, R. (2011). Accommodating the challenges of climate change adaptation and governance in conventional risk management: Adaptive collaborative risk management (ACRM). Ecology & Society, 16(1), 47. https://doi:10.5751/ES-03924-160147

MDPH. (2014). MDPH Guidelines for Cyanobacteria in Freshwater Recreational Water Bodies in Massachusetts. Boston, MA: Massachusetts Department of Public Health, Bureau of Environmental Health Retrieved from http://archives.lib.state.ma.us/handle/2452/615560

Miao, C., Duan, Q., Sun, Q., Huang, Y., Kong, D., Yang, T., et al. (2014). Assessment of CMIP5 Climate Models and Projected Temperature Changes over Northern Eurasia. Environmental Research Letters, 9(5), 055007. https://doi.org/10.1088/1748-9326/9/5/055007

Mishra, V., Cherkauer, K. A., Bowling, L. C., & Huber, M. (2011). Lake Ice Phenology of Small Lakes: Impacts of Climate Variability in the Great Lakes Region. Global and Planetary Change, 76(3), 166-185. https://doi.org/10.1016/j.gloplacha.2011.01.004

Moore, T. S., Churnside, J. H., Sullivan, J. M., Twardowski, M. S., Nayak, A. R., McFarland, M. N., et al. (2019). Vertical Distributions of Blooming Cyanobacteria Populations in a Freshwater Lake from LIDAR Observations. Remote Sensing of Environment, 225, 347-367. https://doi.org/10.1016/j.rse.2019.02.025

Moser, S. C., & Ekstrom, J. A. (2010). A Framework to Diagnose Barriers to Climate Change Adaptation. Proceedings of the National Academy of Sciences, 107(51), 22026–22031. https://doi.org/10.1073/pnas.1007887107

Mostafavi, A., & Inman, A. (2016). Exploratory Analysis of the Pathway towards Operationalizing Resilience in Transportation Infrastructure Management. Built Environment Project and Asset Management, 6(1), 106–18. https://doi.org/10.1108/BEPAM-03-2015-0011

Mugume, S. N., Gomez, D. E., Fu, G., Farmani, R., & Butler, D. (2015). A Global Analysis Approach for Investigating Structural Resilience in Urban Drainage Systems. Water Research, 81, 15–26. https://doi.org/10.1016/j.watres.2015.05.030

Nalley, J. O., O’Donnell, D. R., & Litchman, E. (2018). Temperature Effects on Growth Rates and Fatty Acid Content in Freshwater Algae and Cyanobacteria. Algal Research, 35, 500-507. https://doi.org/10.1016/j.algal.2018.09.018

Nelson, R., Howden, M., & Smith, M. S. (2008). Using Adaptive Governance to Rethink the Way Science Supports Australian Drought Policy. Environmental Science and Policy, 11(7), 588–601. https://doi.org/10.1016/j.envsci.2008.06.005

O’Keeffe, J. (2019). Cyanobacteria and Drinking Water – Occurrence Risks Management and Knowledge Gaps for Public Health EN. National Collaborating Centre for Environmental Health

O’Reilly, C. M., Sharma, S., Gray, D. K., Hampton, S. E., Read, J. S., Rowley, R. J., et al. (2015). Rapid and Highly Variable Warming of Lake Surface Waters around the Globe. Geophysical Research Letters. 42(24), 10773-10781. https://doi.org/10.1002/2015GL066235

Ouyang, M., Dueñas-Osorio, L., & Min, X. (2012). A Three-Stage Resilience Analysis Framework for Urban Infrastructure Systems. Structural Safety, 36–37, 23–31. https://doi.org/10.1016/j.strusafe.2011.12.004

Paerl, H. W., & Huisman, J. (2008). Climate: Blooms like It Hot. Science, 320(5872), 57-58. https://doi.org/10.1126/science.1155398

Paerl, H. W., & Huisman, J. (2009). Climate Change: A Catalyst for Global Expansion of Harmful Cyanobacterial Blooms. Environmental Microbiology Reports, 1(1), 27-37. https://doi.org/10.1111/j.1758-2229.2008.00004.x

Pahl-Wostl, C. (2007). Transitions towards Adaptive Management of Water Facing Climate and Global Change. In Water Resources Management, 21(1), 49–62. https://doi.org/10.1007/s11269-006-9040- 4

Pahl-Wostl, C. (2009). A Conceptual Framework for Analysing Adaptive Capacity and Multi-Level Learning Processes in Resource Governance Regimes. Global Environmental Change, 19(3): 354– 65. https://doi.org/10.1016/j.gloenvcha.2009.06.001

Palmer, M. E., Yan, N. D. & Somers, K. M. (2014). Climate change drives coherent trends in physics and oxygen content in North American lakes. Climatic Change, 124(1-2), 285-299. https://doi:10.1007/s10584-014-1085-4

Panteli, M., & Mancarella, P. (2017). Modeling and Evaluating the Resilience of Critical Electrical Power Infrastructure to Extreme Weather Events. IEEE Systems Journal, 11(3), 1733–42. https://doi.org/10.1109/JSYST.2015.2389272

Perera, N., Gharabaghi, B., & Howard, K. (2013). Groundwater Chloride Response in the Highland Creek Watershed Due to Road Salt Application: A Re-Assessment after 20years. Journal of Hydrology, 479, 159-168. https://doi.org/10.1016/j.jhydrol.2012.11.057

Persaud, A. D., Paterson, A. M., Dillon, P. J., Winter, J. G., Palmer, M., & Somers, K. M. (2015). Forecasting Cyanobacteria Dominance in Canadian Temperate Lakes. Journal of Environmental Management, 151, 343-352. https://doi.org/10.1016/j.jenvman.2015.01.009

Piccolroaz, S. (2016). Prediction of Lake Surface Temperature Using the Air2water Model: Guidelines, Challenges, and Future Perspectives. Advances in Oceanography and Limnology, 7(1), 36-50. https://doi.org/10.4081/aiol.2016.5791

Piccolroaz, S., Healey, N. C., Lenters, J. D., Schladow, S. G., Hook, S. J., Sahoo, G. B., & Toffolon, M. (2018). On the Predictability of Lake Surface Temperature Using Air Temperature in a Changing Climate: A Case Study for Lake Tahoe (U.S.A.). Limnology and Oceanography, 63(1), 243-261. https://doi.org/10.1002/lno.10626

Pilon, S., Zastepa, A., Taranu, Z. E., Gregory-Eaves, I., Racine, M., Blais, J. M., et al. (2019). Contrasting Histories of Microcystin-Producing Cyanobacteria in Two Temperate Lakes as Inferred from Quantitative Sediment DNA Analyses. Lake and Reservoir Management, 35(1), 102-117. https://doi.org/10.1080/10402381.2018.1549625

Pilotto, L. S., Douglas, R. M., Burch, M. D., Cameron, S., Beers, M., Rouch, G. J., Robinson, P., et al. (1997). Health Effects of Exposure to Cyanobacteria (Blue-Green Algae) during Recreational Water-Related Activities. Australian and New Zealand Journal of Public Health. 21(6), 562-566. https://doi.org/10.1111/j.1467-842X.1997.tb01755.x

Ramakrishna, D. M., & Viraraghavan, T. (2005). Environmental Impact of Chemical Deicers - A Review. Water, Air, and Soil Pollution, 166(1), 49-63. https://doi.org/10.1007/s11270-005-8265-9

Reed, D. A., Kapur, K. C., & Christie, R. C. (2009). Methodology for Assessing the Resilience of Networked Infrastructure. IEEE Systems Journal, 3(2), 174–80. https://doi.org/10.1109/JSYST.2009.2017396

Rigosi, A, Carey, C. C., Ibelings, B. W., & Brookes, J. D. (2014). The Interaction between Climate Warming and Eutrophication to Promote Cyanobacteria Is Dependent on Trophic State and Varies among Taxa. Limnology and Oceanography, 59(1), 99-114. https://doi.org/10.4319/lo.2014.59.1.0099

Rijke, J., Farrelly, M., Brown, R., & Zevenbergen, C. (2013). Configuring Transformative Governance to Enhance Resilient Urban Water Systems. Environmental Science and Policy, 25, 62–72. https://doi.org/10.1016/j.envsci.2012.09.012

Robarts, R. D., & Zohary, T. (1987). Temperature Effects on Photosynthetic Capacity, Respiration, and Growth Rates of Bloom‐forming Cyanobacteria. New Zealand Journal of Marine and Freshwater Research, 21(3), 391-399. https://doi.org/10.1080/00288330.1987.9516235

Rudberg, P. M., Wallgren, O., & Swartling, Å. G. (2012). Beyond Generic Adaptive Capacity: Exploring the Adaptation Space of the Water Supply and Wastewater Sector of the Stockholm Region, Sweden. Climatic Change, 114(3), 707–721. https://doi.org/10.1007/s10584-012-0453-1

Sanford, T., Frumhoff, P. C., Luers, A., & Gulledge, J. (2014). The Climate Policy Narrative for a Dangerously Warming World. Nature Climate Change, 4, 164-166. https://doi.org/10.1038/nclimate2148

Sarafanov, A. (2009). On the Effect of the North Atlantic Oscillation on Temperature and Salinity of the Subpolar North Atlantic Intermediate and Deep Waters. ICES Journal of Marine Science, 66(7),1448-1454. https://doi.org/10.1093/icesjms/fsp094.

Scheffer, M. (1998). Ecology of Shallow Lakes. DeAngelis DL, Manly BFJ, editors. First edition 1998: Springer, 378.

Schippers, P., Lürling, M. & Scheffer, M. (2004). Increase of atmospheric CO2 promotes phytoplankton productivity. Ecology Letters, 7(6), 446-451. https://doi:10.1111/j.1461-0248.2004.00597.x

Schmale, D .G., Ault, A. P., Saad, W., Scott, D. T., & Westrick, J. A. (2019). Perspectives on Harmful Algal Blooms (HABs) and the Cyberbiosecurity of Freshwater Systems. Frontiers in Bioengineering and Biotechnology, 7. https://doi.org/10.3389/fbioe.2019.00128

Schmidtko, S., Stramma, L., & Visbeck, M. (2017). Decline in Global Oceanic Oxygen Content during the Past Five Decades. Nature, 542, 335-339. https://doi.org/10.1038/nature21399

Schneider, P., & Hook, S. J. (2010). Space Observations of Inland Water Bodies Show Rapid Surface Warming since 1985. Geophysical Research Letters, 37(22). https://doi.org/10.1029/2010GL045059

Schoen, M., Hawkins, T., Xue, X., Ma, C., Garland, J., & Ashbolt, N. J. (2015). Technologic Resilience Assessment of Coastal Community Water and Wastewater Service Options. Sustainability of Water Quality and Ecology, 6, 75–87. https://doi.org/10.1016/j.swaqe.2015.05.001

Seth, A., Wang, G., Kirchhoff, C., Lombardo, K., Stephenson, S., Anyah, R., & Wu , J. (2019). Connecticut Physical Climate Assessment Report (PCSAR): Observed trends and projections of temperature and precipitation. Groton, CT: Connecticut Institute for Resilience and Climate Adaptation. Retrieved from https://circa.uconn.edu/wp- content/uploads/sites/1618/2019/08/CTPCSAR-Aug2019.pdf

Shin, S., Lee, S., Judi, D. R., Parvania, M., Goharian, E., McPherson, T., & Burian, S. J. (2018). A Systematic Review of Quantitative Resilience Measures for Water Infrastructure Systems. Water (Switzerland) 10 (2). https://doi.org/10.3390/w10020164

Siver, P. A., Marsicano, L. J., Lott, A. M., Wagener, S., & Morris, N. (2018). Wind Induced Impacts on Hypolimnetic Temperature and Thermal Structure of Candlewood Lake (Connecticut, U.S.A.) from 1985–2015. Geo: Geography and Environment, 5(2), e00056. https://doi.org/10.1002/geo2.56

Sivonen, K., & Jones, G. (1999). Cyanobacterial Toxins. In I. Chorus, J. Bartram (Eds), Toxic cyanobacteria in water: A guide to their public health consequences, monitoring and management (pp. 41–112). London, UK: E & FN Spon

Smith, V. H. (1983). Low Nitrogen to Phosphorus Ratios Favor Dominance by Blue-Green Algae in Lake Phytoplankton. Science, 221(4611), 669-671. https://doi.org/10.1126/science.221.4611.669

Staniec, A., & Vlahos, P. (2017). Timescales for Determining Temperature and Dissolved Oxygen Trends in the Long Island Sound (LIS) Estuary. Continental Shelf Research, 151, 1-7. https://doi.org/10.1016/j.csr.2017.09.013

Steffen, M. M., Davis, T. W., McKay, R. M. L., Bullerjahn, G. S., Krausfeldt, L. E., Stough, J. M. A., et al. (2017). Ecophysiological Examination of the Lake Erie Microcystis Bloom in 2014: Linkages between Biology and the Water Supply Shutdown of Toledo, OH. Environmental Science & Technology, 51(12), 6745-6755. https://doi.org/10.1021/acs.est.7b00856

Stets, E. G., Butman, D., McDonald, C. P., Stackpoole, S. M., DeGrandpre, M. D., & Striegl, R. G. (2017). Carbonate Buffering and Metabolic Controls on Carbon Dioxide in Rivers. Global Biogeochemical Cycles, 31(4), 663-677. https://doi.org/10.1002/2016GB005578

Subbiah, S., Karnjanapiboonwong, A., Maul, J. D., Wang, D., & Anderson, T. A. (2019). Monitoring Cyanobacterial Toxins in a Large Reservoir: Relationships with Water Quality Parameters. PeerJ, 7, e7305. https://doi.org/10.7717/peerj.7305

Tan, Z., Yao, H. & Zhuang, Q. (2018). A Small Temperate Lake in the 21st Century: Dynamics of Water Temperature, Ice Phenology, Dissolved Oxygen, and Chlorophyll A. Water Resources Research, 54(7), 4681-4699. https://doi.org/10.1029/2017WR022334

Taranu, Z. E., Zurawell, R. W., Pick, F., & Gregory-Eaves, I. (2012). Predicting Cyanobacterial Dynamics in the Face of Global Change: The Importance of Scale and Environmental Context. Global Change Biology, 18(12), 3477-3490. https://doi.org/10.1111/gcb.12015

Therrien, M. C., Beauregard, S., & Valiquette-L’Heureux, A. (2015). Iterative Factors Favoring Collaboration for Interorganizational Resilience: The Case of the Greater Montréal Transportation Infrastructure. International Journal of Disaster Risk Science, 6(1), 75–86. https://doi.org/10.1007/s13753-015-0044-7

Thomas, M. K., & Litchman, E. (2016). Effects of Temperature and Nitrogen Availability on the Growth of Invasive and Native Cyanobacteria. Hydrobiologia, 763(1), 357-369. https://doi.org/10.1007/s10750-015-2390-2

Tiao, G. C., Reinsel, G. C., Xu, D., Pedrick, J. H., Zhu, X., Miller A. J., et al. (1990). Effects of Autocorrelation and Temporal Sampling Schemes on Estimates of Trend and Spatial Correlation. Journal of Geophysical Research, 95(D12), 20507-20517. https://doi.org/10.1029/jd095id12p20507

Timpano, A. J., Zipper, C. E., Soucek, D. J., & Schoenholtz, S. H. (2018). Seasonal Pattern of Anthropogenic Salinization in Temperate Forested Headwater Streams. Water Research, 133, 8-18. https://doi.org/10.1016/j.watres.2018.01.012

Trevino-Garrison, I., Dement, J., Ahmed, F. S., Haines-Lieber, P., Langer, T., Ménager, H., et al. (2015). Human Illnesses and Animal Deaths Associated with Freshwater Harmful Algal Blooms—Kansas. Toxins, 7(2), 353-366. https://doi.org/10.3390/toxins7020353

USGS. (2016). GAP/LANDFIRE National Terrestrial Ecosystems 2011. Gap Analysis Program, U.S. Geological Survey, https://doi.org/10.5066/F7ZS2TM0

Vystavna, Y., Hejzlar, J., & Kopáček, J. (2017). Long-Term Trends of Phosphorus Concentrations in an Artificial Lake: Socioeconomic and Climate Drivers. PLoS ONE, 12(10), e0186917. https://doi.org/10.1371/journal.pone.0186917

Wagner, C., & Adrian, R. (2009). Cyanobacteria Dominance: Quantifying the Effects of Climate Change. Limnology & Oceanography, 54(6), 2460-2468. https://doi.org/10.4319/lo.2009.54.6_part_2.2460

Weatherhead, E. C., Reinsel, G. C., Tiao, G. C., Meng, X. L., Choi, D., Cheang, W. K., et al. (1998). Factors Affecting the Detection of Trends: Statistical Considerations and Applications to Environmental Data. Journal of Geophysical Research Atmospheres, 103(D14), 17149-17161 https://doi.org/10.1029/98JD00995

WHO. (2003). Algae and Cyanobacteria in Freshwater. In: Guidelines for Safe Recreational Water Environments. Vol.1 - Coastal and Fresh Waters. Geneva, Switzerland: World Health Organization Publishing.

Winder, M., & Schindler, D. E. (2004). Climatic Effects on the Phenology of Lake Processes. Global Change Biology, 10(11), 1844-1856. https://doi.org/10.1111/j.1365-2486.2004.00849.x

Wuebbles, D.J., Fahey, D. W., Hibbard, K. A., Deangelo, B., Doherty, S., Hayhoe, K., et al. (Eds.) (2017). Executive Summary. In Climate Science Special Report: Fourth National Climate Assessment, Volume I (pp. 12-34). Washington, DC: Global Change Research Program.

Yasarer, L. M. W., & Sturm, B. S. M. (2016). Potential Impacts of Climate Change on Reservoir Services and Management Approaches. Lake and Reservoir Management, 32(1), 13-26. https://doi.org/10.1080/10402381.2015.1107665

Yohe, G., & Tol, R. S. J. (2002). Indicators for Social and Economic Coping Capacity—moving toward a Working Definition of Adaptive Capacity. Global Environmental Change, 12(1), 25–40. https://doi.org/10.1016/S0959-3780(01)00026-7

Zamyadi, A., McQuaid, N., Prévost, M., & Dorner, S. (2012). Monitoring of Potentially Toxic Cyanobacteria Using an Online Multi-Probe in Drinking Water Sources. Journal of Environmental Monitoring, 14, 579-588. https://doi.org/10.1039/c1em10819k

Zanchett, G., & Oliveira-Filho, E. C. (2013). Cyanobacteria and Cyanotoxins: From Impacts on Aquatic Ecosystems and Human Health to Anticarcinogenic Effects. Toxins, 5(10), 1896-1917. https://doi.org/10.3390/toxins5101896