Science of the Total Environment 521–522 (2015) 11–18

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Simulating the effect of climate change on stream temperature in the Trout Lake Watershed, Wisconsin William R. Selbig U.S. Geological Survey — Wisconsin Water Science Center, 8505 Research Way, Middleton, WI, USA

H I G H L I G H T S • • • •

A stream temperature model was calibrated for three streams in northern Wisconsin. The effect of climate change on stream temperature was simulated in each stream. Annual average stream temperature was projected to rise from 1 to 3 °C by 2100. Forecasts of stream temperature exceeded optimal ranges for brook trout.

a r t i c l e

i n f o

Article history: Received 29 January 2015 Received in revised form 17 March 2015 Accepted 18 March 2015 Available online 28 March 2015 Editor: F. Riget Keywords: Climate change Stream temperature Model Brook trout Thermal toxicity Wisconsin

a b s t r a c t The potential for increases in stream temperature across many spatial and temporal scales as a result of climate change can pose a difficult challenge for environmental managers, especially when addressing thermal requirements for sensitive aquatic species. This study evaluates simulated changes to the thermal regime of three northern Wisconsin streams in response to a projected changing climate using a modeling framework and considers implications of thermal stresses to the fish community. The Stream Network Temperature Model (SNTEMP) was used in combination with a coupled groundwater and surface water flow model to assess forecasts in climate from six global circulation models and three emission scenarios. Model results suggest that annual average stream temperature will steadily increase approximately 1.1 to 3.2 °C (varying by stream) by the year 2100 with differences in magnitude between emission scenarios. Daily mean stream temperature during the months of July and August, a period when cold-water fish communities are most sensitive, showed excursions from optimal temperatures with increased frequency compared to current conditions. Projections of daily mean stream temperature, in some cases, were no longer in the range necessary to sustain a cold water fishery. Published by Elsevier B.V.

1. Introduction Thermal regimes in stream ecosystems are fundamentally important to fish and other aquatic organisms, especially those with low tolerances for extended periods above or below optimal thresholds (Isaak et al., 2012). In many cases, changes to the thermal regime can be attributed to anthropogenic drivers such as removal of riparian vegetation, thermal effluent from powerplants, water storage in reservoirs, and urbanization. While these alterations to flow and temperature are deleterious to receiving waters, they are generally restricted in spatial scale, allowing for aquatic communities to shift towards more suitable habitat (Wenger et al., 2011). These opportunities for relocation are lessened as stresses to the ecosystem become more pervasive. Climate change is one such stress that can strongly dictate the distribution and abundance of individual species because changes in air temperature, atmospheric radiation, and the timing and magnitude of precipitation patterns can E-mail address: [email protected].

http://dx.doi.org/10.1016/j.scitotenv.2015.03.072 0048-9697/Published by Elsevier B.V.

affect entire ecosystems and river networks. While previous studies have considered the potential effect of climate change on the distribution of fish in North America, the majority have been relatively coarse in scale that focus on broad landscapes or large spatial catchments greater than 500 km2 (Lyons et al., 2010). While these studies yield important insights, they represent a small fraction of total stream habitat available. Other studies are more regional, focusing on changes in stream temperature in relation to elevation or latitude. For example, Null et al. (2013) showed an overall reduction in viable coldwater habitat in California's Sierra Nevada, shifting more towards higher elevations. Relatively few studies have emphasized the variation in thermal conditions in a smaller geographical context, or individual streams (Lyons et al., 2010; Steen et al., 2010; Williams et al., 2009). A better understanding of smallscale thermal response to the potential warming effects of climate change will better direct the managerial decision-making process. The potential for increases in stream temperature across many spatial and temporal scales poses a challenge for environmental managers. Streams that may currently be suitable as a recreational cold-water

12

W.R. Selbig / Science of the Total Environment 521–522 (2015) 11–18

sport fishery could become increasingly fragmented as fish find refuge in less impacted areas. Furthermore, as climate change advances, aquatic communities already constrained by warmer stream temperatures could result in net losses of habitat or a generalized shift in fish assemblages towards more tolerant species (Mohseni et al., 1998; Wenger et al., 2011; Isaak et al., 2012). Increases in stream temperature as a result of climate change could also result in changes to water quality, such as dissolved oxygen (Ficklin et al., 2013). An important step in understanding the importance of climate change on sensitive aquatic resources is describing rates at which suitable metrics might change given variations in projected climate forcings. Given the uncertainty associated with long range projections, the exact consequences of a warming climate on stream temperature could vary depending on several factors, such as precipitation patterns and seasonality of temperature shifts (Nelson and Palmer, 2007). Several studies have characterized the response in stream temperature to changing climate variables either through examination of historical trends [for example Isaak et al., 2012] or by forecasting through model simulation. A variety of stream temperature models are available, ranging from complex advection/dispersion models (Stefan and Sinokrot, 1993) to more simplistic linear regression models that incorporate one or more climate variables, such as air temperature (Stefan and Preud'homme, 1993; Nelson and Palmer, 2007). However, these models often ignore or approximate complicated interactions between groundwater and surface water systems neglecting important feedback loops with other dynamic hydrologic processes such as evapotranspiration, soil-zone flow, and surface runoff (Ficklin et al., 2013; Hunt et al., 2013). For example, a decreasing snowmelt contribution to streamflow from less snowfall will reduce the amount of cold-water inputs resulting in warmer winter, spring, and early summer stream temperatures (Ficklin et al., 2013). Thus, forecasts of stream temperature should account for both the atmospheric and coupled groundwater/surface water hydrologic system responses to climate change. The goal of this study was to simulate the historic and potential future stream temperatures in three select streams in the Trout Lake watershed in Vilas County, Wisconsin. While there have been many other studies examining the response of stream temperature to climate change, most make use of regional models that are too coarse for meaningful management options, or use surrogates, such as air temperature, as predictors of stream temperature and do not include the dynamic hydrological processes at the local scale. This study made use of a coupled surface water and groundwater model called GSFLOW. GSFLOW is an integration of the U.S. Geological Survey's Precipitation-Runoff Modeling System (PRMS) and MODFLOW. The objectives for the model described in this paper included forecasts of the effects of climate-change scenarios on streamflow and stream temperature. Therefore, streamflow results from the coupled GSFLOW model were linked to the Stream-Network TEMPerature model (SNTEMP) (Bartholow, 1991). This approach allows propagation of potential temperature changes in the atmosphere to coldwater streams and informs questions related to projected impacts on stream ecology and the potential risk to fish communities. The following is a partial digest of a previously published U.S. Geological Survey Scientific Investigations Report. Full details of the study can be found in Hunt et al. (2013). 1.1. Site description The Trout Lake watershed (105 km2) is in Vilas County, Wisconsin, USA (Fig. 1). The basin is comprised of many small lakes with watersheds that are completely forested with a mixture of coniferous and deciduous species. The poorly drained glacial landscape has resulted in numerous wetland areas, ranging from bogs to fens (Hunt et al., 2006). Lakes are well connected to the groundwater system and many lakes are flow-through lakes with respect to groundwater. Streamflow in the area is dominated by groundwater which can account for over 80% of total streamflow; however, surface runoff can be appreciable during spring snowmelt (Gebert et al., 2009). Annual precipitation

averages about 81.5 cm/years (National Climatic Data Center, 2014); average groundwater recharge is estimated to be 27 cm/years (Hunt et al., 1996), and has been estimated to range from about 15 to 50 cm/years at local areas within the basin (Dripps et al., 2006). Mean monthly temperatures range from − 18 to −7 °C in January to + 12 to +25 °C in July (National Climatic Data Center, 2014). Three mainstem and associated tributaries to Trout Lake were selected for measurement and simulation of stream temperature: North Creek, Stevenson Creek, and Upper Allequash Creek (Fig. 1). Mean baseflow at the mouth ranged between 0.09 m3/s at Stevenson Ck. and 0.13 m3/s at Allequash Ck. (Hunt et al., 2006). Channels are widest and more diffuse near the headwaters as each stream becomes integrated with a larger wetland system. Channel widths range from approximately 30 m near the headwaters of both Upper Allequash and Stevenson Ck. to less than 3 m near the mouth of Upper Allequash. Topography surrounding each stream consists primarily of wetland and forested lowlands. 2. Methods 2.1. Description of temperature model The instream water temperature model SNTEMP, developed and supported by the U.S. Fish and Wildlife Service, was selected to predict stream temperatures in the Trout Lake stream network. A modified version of SNTEMP called TRPA Stream Temperature for Windows (http://trpafishbiologists.com/sindex.html) provided a graphical user interface to simplify data entry and export. SNTEMP is a steady-state, one-dimensional heat transport model that predicts daily mean and maximum temperatures as a function of stream distance and environmental heat flux (Bartholow, 1991). A heat-transport equation describes the downstream movement of heat energy in the water and actual exchange of heat energy between the water and its surrounding physical environment (Theurer et al., 1984). Net heat flux is calculated by parameter inputs describing the meteorological, hydrological, stream geometry, and shade setting for a network of stream segments that define North, Stevenson, and Upper Allequash Creeks. Each stream was discretized into two or more segments. Each segment is considered constant and represents uniform width, groundwater accretion rates, and relatively homogeneous topographic and riparian vegetation conditions. As such, major transitions in any of these categories would support creation of a new stream segment. Final versions of conceptual models created for North, Stevenson, and Upper Allequash Creek, resulted in 7, 3, and 4 stream segments, respectively. Each stream segment requires a physical description of stream geometry, hydrology, and shading variables. Meteorological variables, on the other hand, are considered more global in nature and were applied to all stream segments equally. SNTEMP assumes that all input data, including meteorological and hydrological variables, can be represented by 24-hour averages (Bartholow, 1991). 2.1.1. Data collection and field measurements Data used to calibrate SNTEMP came from a variety of sources. Meteorological data came from publicly available historical datasets for the modeled region, hydrologic data were provided by previously calibrated hydrologic models, and field measurements of stream geometry and riparian shading were done by using methods described in Bartholow (1989). Some parameters, such as dust coefficients, ground reflectivity, and Manning's n values were not measurable and were supplemented by published data sources. 2.1.2. Hydrology Hydrologic data consists of stream discharge and water temperatures. SNTEMP requires both upstream discharge and temperature data for each modeled stream segment. For calibration, daily mean discharge data was based on a coupled groundwater-surface water flow model called GSFLOW (Markstrom et al., 2008). GSFLOW is an

W.R. Selbig / Science of the Total Environment 521–522 (2015) 11–18

13

Fig. 1. Location of North Creek, Stevenson Creek, and Upper Allequash Creek in the Trout Lake Watershed, Wisconsin.

integration of the U.S. Geological Survey (USGS) Precipitation-Runoff Modeling System, PRMS (Leavesley et al., 1983, 2005), with versions of the USGS Modular Groundwater Flow Model MODFLOW-2005 (Harbaugh, 2005) and MODFLOW-NWT (Niswonger et al., 2011). Full details of the construction and calibration of both groundwater and surface water models for the Trout Lake watershed can be found in Hunt et al. (2013). Use of the calibrated GSFLOW model provided discharge values for the upstream node of each stream segment. Differences in discharge between segments were assumed to be from additions in groundwater through lateral accretion. This allowed the model to adequately describe changes in stream temperature with changing flow rates as well as other stream characteristics such as geometry or shading. Headwater segments of North, Stevenson, and Upper Allequash Creeks were assumed to have zero flow and thus did not require discharge for model execution. SNTEMP allows the user to assume a zero flow headwater which disregards any associated temperature data because there cannot be any water temperature if there is not any water discharge. All discharge gained from the zero flow headwater to the next stream segment was considered to be lateral accretion of groundwater.

Mean daily groundwater temperatures were based on measured values from thermocouples placed approximately 15 cm underneath the streambed at select locations in North, Stevenson, and Allequash Creeks (Hunt et al., 2006). Because the location of each sensor was within a gaining section of the stream, resulting temperatures were considered a reasonable estimate for groundwater prior to entering the stream. Although measured at a single location within each stream, the measured groundwater temperatures were applied equally across each stream segment in the model. 2.1.3. Geometry, shading, and meteorology Stream geometry data consists of the network layout of the main stem and all tributaries. Stream width was measured in the field where possible and estimated by use of spatially rectified aerial photography for areas that were inaccessible. The width of the stream segment represents an average and was assumed to remain constant for all values of flow. Manning's n values were initially estimated based on reported ranges for natural channels by Gupta (1989) and were varied during calibration of the model. The thermal gradient, the rate of thermal input from the streambed to the water, in each stream reach was kept at the default

14

W.R. Selbig / Science of the Total Environment 521–522 (2015) 11–18

Table 1 Global circulation models used to simulate future climate conditionsa. Model ID

Organization

CGCM3.1 (T47), 2005

Canadian Centre for Climate Modeling and Analysis, Canada CNRM-CM3, 2004 Centre National de Recherches Meteorologiques, France CSIRO-MK3.0, 2001 Commonwealth Scientific and Industrial Research Organization (CSIRO) Atmospheric Research, Australia GFDL-CM2.0, 2005 National Oceanic and Atmospheric Administration (NOAA) Geophysical Fluid Dynamics Laboratory (GFDL) MIROC3.2 (medres), 2004 Center for Climate Systems Research, National Institute for Environmental Studies and Frontier Research Center for Global Change, Japan MRI-CGCM2.3.3, 2003 Meteorological Research Institute, Japan a

during the warmest summer months of July and August (National Climatic Data Center, 2014). Calibration of stream temperature during winter months were of less concern since many of the lakes and streams in the modeled area are completely or partially frozen. Although SNTEMP was used to predict daily mean stream temperature from the headwater to the mouth of each stream, calibration targets were located only near the mouth. Measured values used for comparison to predicted values were based on daily mean stream temperatures collected by Hunt et al. (2006). Although all SNTEMP input variables were initially considered calibration parameters, the greatest change to simulated daily mean stream temperatures was effected by adjusting air temperature, streamflow, groundwater discharge, and

Reproduced from Hunt et al. (2013).

value of 1.65 (unitless). Elevation, latitude, longitude, and river kilometer locations were acquired through a global positioning system (GPS). Stream azimuth was determined by using USGS 1:24,000-scale topographic maps. Values used to describe stream geometry parameters for each stream reach are available in the Supplementary online material. Stream shading parameters including, topographic angle, vegetation offset, crown width, shade density, and riparian-vegetation height were measured at several locations on the left and right banks of each stream segment then averaged to provide a single value for each streamsegment bank. In some cases, shading parameters were estimated based on aerial photography. Values used to describe shading parameters for each stream reach are available in the Supplementary online material. Meteorological data came from both measured and published sources. SNTEMP uses only one set of meteorological data that is applied to all stream segments. A weather station at the Noble F. Lee Municipal airport in Woodruff, WI (http://lter.limnology.wisc.edu/dataset/northtemperate-lakes-lter-meteorological-data-woodruff-airport) was used for all daily mean air temperature, humidity, and wind speed data. Cloud cover was estimated by using a calculated percent possible sunshine for Woodruff, Wisconsin (http://aa.usno.navy.mil). Dust coefficients and ground reflectivity were taken from published values described in Tennessee Valley Authority (Theurer et al., 1984). A ground reflectivity value of 0.08 was used for all stream segments to represent leaf and needle forest (Theurer et al., 1984). Although solar radiation data was available, better agreement between predicted and observed stream temperatures was achieved by using model calculated values. Other meteorological values, including evaporation factors and the Bowen ratio, made use of default values prescribed by Theurer et al. (1984). 2.1.4. Calibration The time period April through September 2002 was selected for model calibration due to having the longest continuous water temperature and discharge datasets at multiple locations throughout North, Stevenson, and Upper Allequash Creeks during the GSFLOW model calibration period. This period also corresponds to when habitat and aquatic organisms are vulnerable to increases in stream temperature

Table 2 Select emission scenarios used to simulate future climate conditionsa. Scenario Description A1B

A2

B1

a

Rapid economic growth, global population peaking in mid-century and declining thereafter, and introduction of new and efficient technologies with a balance across all energy sources Very heterogeneous world with self-reliance and preservation of local identities with continuously increasing population growth, and slow regional economic growth and technological change Convergent world with population change as described in the A1 scenarios with rapid changes towards a service and information economy with clean and resource-efficient technologies

Reproduced from Hunt et al. (2013).

Fig. 2. Simulated and observed daily mean stream temperatures at the calibration points in A) North Creek, B) Stevenson Creek, and C) Upper Allequash Creek, April through September 2002.

W.R. Selbig / Science of the Total Environment 521–522 (2015) 11–18 Table 3 Summary statistics for predicted and observed differences in daily mean stream temperature at calibration points in SNTEMP, April–September, 2002. Stream

R2

Min error

Max error

Mean error

% of projections N 1 °C

Stevenson North Allequash

0.93 0.95 0.95

b0.5° b0.5° b0.5°

−5.9 −4.7 −4.2

−1.4 −0.6 −0.6

66% 44% 49%

groundwater temperature through lateral accretion. Of these four variables, air temperature was a measured value, thus considered relatively well known and not adjusted during calibration. Similarly, although streamflow was not directly measured at all locations, it was considered well constrained by the calibrated flows simulated by the coupled GSFLOW model. All other SNTEMP input variables were subject to trial-and-error adjustment until minimal differences were achieved between simulated and measured calibration locations. Calibration was considered complete when three criteria were met: (1) high correlation between simulated and observed daily mean stream temperatures, (2) minimal difference between average simulated and average observed stream temperatures, and (3) minimal difference between individual daily mean simulated and observed stream temperatures (Hunt et al., 2013).

2.2. Climate change scenarios A range of future climatic conditions were based on precipitation and air temperature output from six General Circulation Models (GCMs) detailed in the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Special Report on Emission Scenarios (2007) (Table 1). For each GCM, three emission scenarios representing a range of growth in population, technological advances, and the economy were considered (Table 2). Data produced by these regional GCMs were then downscaled to the Trout Lake area by the Wisconsin Initiative on Climate Change Impacts (WICCI) for improved resolution (2011). More details on GCMs and emission scenarios can be found in Hunt et al. (2013). Each GCM and emission scenario was run to simulate a 140 year period (1961–2100); however, the first 20 years were discarded to ensure sufficient time for the models to equilibrate between groundwater and

15

surface water dynamics. Temporally downscaled GCM projections of daily mean air temperature were imported directly into SNTEMP. GCM outputs of air temperature and precipitation were used by GSFLOW as a means to simulate future changes in stream discharge and solar radiation which were then imported into SNTEMP. Future estimates of groundwater temperature were based on a regression between measured air temperature and shallow groundwater temperature during the calibration period of April–September, 2002. Results of the regression showed strong correlation with R2 values of 0.86, 0.89, and 0.89 for North, Stevenson, and Upper Allequash Creek, respectively. The regression was then used to project daily mean groundwater temperature given changes to daily mean air temperature. Wind speed, relative humidity, and percent possible sunshine reflected 30-year monthly normal (1981–2020) from the nearest available weather station in Green Bay, Wisconsin (National Climatic Data Center, 2015). These values were repeated for all simulated years. All other SNTEMP parameters retained their calibrated values as described in Hunt et al. (2013). 3. Results and discussion 3.1. SNTEMP Over the six month calibration period, all three models were able to simulate changes in stream temperature associated with seasonal fluctuations in air temperature, stream discharge, and other climatic drivers, although the magnitude of change varied compared to measured values (Fig. 2). This is supported by the high correlation coefficients between predicted and observed values (Table 3). The model was also able to replicate relational differences between modeled streams observed by Hunt et al. (2006) in that Upper Allequash Creek was coldest followed by North Creek, then Stevenson Creek. When averaged across the six month calibration period, all three models generally underestimated stream temperature with negative mean error ranging from −0.6 to −1.4 °C (Table 3). One reason could be the association of these three streams with wetlands. Hunt et al. (2006) identified the presence of underlying peat sediments in the modeled area which are capable of having relatively high insulation capacity. Additionally, water color in the wetland streams is generally dark brown due to increased tannins derived from the peat. Houser (2006) has shown darker

Fig. 3. Difference between predicted and observed daily mean stream temperature outside and within a 1 °C boundary at North Creek, Stevenson Creek, and Upper Allequash Creek, April through September 2002.

16

W.R. Selbig / Science of the Total Environment 521–522 (2015) 11–18

water to become warmer as it absorbs more solar radiation. These insulating properties would not be well represented in the model. When compared by day, SNTEMP was less able to model the system dynamics as evidenced by the percentage of days where differences exceeded +/− 1 °C (Table 3). Excursions from a 1° difference are greatest in the late spring and early summer months; likely a result of rapid fluctuations in air temperature over short time periods often experienced in spring (Fig. 3). These large swings in air temperature are less pronounced in July through September. As such, differences between predicted and observed values generally remained within 1 °C. The efficiency of transferring effects of rapid air-temperature change to stream temperature is expected to be only approximately simulated, given the relatively coarse stream-reach discretization and piecewiseconstant daily time step of the model; that is, SNTEMP is a successive steady-state model in which one assumes that all input data, including meteorological and hydrological variables, can be represented by 24hour averages. This assumption is often appropriate for large river systems that tend to exhibit gradually varying temperatures over time, thereby containing sufficient heat capacity to mitigate large shortterm variations in air temperature. However, the study streams simulated here are small, wetland-dominated streams. Overall, the models were considered acceptable given a similar response to seasonal fluctuations in climate as well as mean errors that were much smaller than the N 20 °C range simulated. 3.2. Climate change Based on a 10-year moving average, annual stream temperature gradually increased over the 100 year simulation period for all six GCMs and three emission scenarios (Fig. 4). This was expected given that all GCM models showed substantial increases in both maximum and minimum air temperature, parameters to which stream temperature is highly sensitive. Projections of warmer air temperatures result in more precipitation coming in the form of rainfall, particularly during periods that are normally dominated by snowfall (Hunt et al., 2013). In response, average annual groundwater recharge becomes less changed over the 100-year simulation period, a reflection of the increased rainfall during winter months — a period of plant senescence. Subsequently, increases in groundwater recharge resulting from increased winter rainfall are less available for soil-zone evapotranspiration; the net effect on annual streamflow is not as large as other changes in the basin. Therefore, increases in stream temperature are likely more influenced by a steadily increasing atmospheric temperature than modest changes in streamflow. Additional information on basin-scale changes to hydrologic flows and storage can be found in Hunt et al. (2013). Similar to calibration outcomes, Stevenson Creek remained the warmest of all three modeled streams with gains in annual average stream temperatures ranging from 1.7 to 3.2 °C while North Creek and Upper Allequash Creek showed relatively smaller gains ranging from 1.4 to 2.9 and 1.1 to 2.2 °C, respectively (Fig. 4). Gains in annual average stream temperature were generally steady over the 100-year simulation with the largest gains represented by the A2 emission scenario and the least represented by B1. This pattern was repeated for each modeled stream with the MIROC3.2 (medres) GCM (A2 emission) resulting in the largest increase in stream temperature over time and the CSIROMK3.0 GCM (B1 emission) the least. The rate of increase varied slightly from 0.2 to 0.3 °C every 10 years in both North Creek and Stevenson Creek and 0.1 to 0.2 °C every 10 years in Upper Allequash Creek. Relative increases in stream temperature tend to flatten for the A1B and B1 emission scenarios over the last 20 years while the A2 scenario continued to steadily increase. This pattern is similar to simulated minimum and maximum air temperatures for each emission scenario in the six GCMs and is likely a reflection of the stagnation in population growth over the last half of the century that are projected in scenarios A1B and B1. However, as projections expand farther into the future, variability becomes larger and more prone to increased uncertainty.

Fig. 4. Climate change simulations for three emission scenarios showing annual average stream temperature for A) North Creek, B) Stevenson Creek, and C) Upper Allequash Creek.

3.3. Thermal toxicity North, Stevenson and Upper Allequash Creeks currently have been designated as a Class I or Class II trout stream in which conditions are sufficient to support natural reproduction of wild trout species (Wisconsin Department of Natural Resources, 2015). Quantiles of daily mean stream temperatures representing the last 5 years (2095– 2099) of GCM emission simulation periods were created for North Creek and Stevenson Creek (Table 4) as a way to estimate how often optimal temperatures needed to support a cold water fishery are

W.R. Selbig / Science of the Total Environment 521–522 (2015) 11–18 Table 4 Quantile estimation of daily mean stream temperature in North Creek and Stevenson Creek for three GCM emission scenarios compared to measured values. Upper Allequash Creek did not have sufficient measured data for quantile estimation. Values represent only the months of July and August for the simulated period of 2095–2099, and measured period of 2013–2014. All values are in °C. 75%

90% 95%

99%

14.3 16.2 17.6 18.3

Percentile 10% 25% 50% North Creek 15.4 16.3 17.5 16.6 17.4 18.2 17.9 18.6 19.2 19.0 19.9 20.8

19.0 19.0 19.7 21.6

21.2 19.5 20.3 22.3

22.1 19.9 20.5 22.5

23.0 20.4 21.0 23.6

15.3 17.8 19.5 19.9

Stevenson Creek 16.3 17.5 18.7 18.3 19.3 20.2 19.8 21.0 21.6 20.7 22.1 22.8

20.4 21.2 22.1 23.7

22.4 22.1 22.7 24.5

23.7 22.1 22.9 25.0

25.7 22.7 23.4 25.8

1%

5%

Measured B1 A1B A2

12.4 15.2 16.6 17.6

Measured B1 A1B A2

13.9 16.7 18.3 18.4

––

Upper Allequash Creek Measured –– –– –– –– –– –– –– –– –– B1 13.9 14.7 15.1 15.8 16.3 16.8 17.5 17.6 18.1 A1B 15.1 15.8 16.2 16.8 17.3 17.8 18.2 18.4 18.7 A2 16.0 16.6 17.0 17.8 18.5 19.2 19.7 20.0 20.6

17

A1B emission scenarios and only slight excursions above the optimal range in the A2 scenario. As evident by the increased frequency of excursions above the optimal range for brook trout, a persistent increase in daily mean stream temperature as a result of climate change affects the diversity of fish species found in northern Wisconsin streams. Similar conclusions were made by Lyons et al. (2010) when assessing the impact of climate change on the distribution of 50 different species of fish across Wisconsin. Using GIS-based models, Lyons et al. (2010) found even small increases in air and water temperatures will have major effects on the distribution of fish. Changes to fish species would be most dramatic in the cold headwater streams of northern Wisconsin where cold- and cool-water fishes would decline and eventually disappear with only a small subset of warm-water fishes to replace them (Lyons et al., 2010). Use of these models provides valuable insights into the challenges environmental managers face when trying to maintain a cold water fishery.

4. Conclusion

a

a

Temperature threshold classifications based on Creaser [1930].

Temperature threshold classifications based on Creaser (1930).

exceeded. Each distribution represents only the months of July and August, the two hottest months of the year based on the 30 year normal for measured air temperature near Trout Lake (National Climatic Data Center, 2014). Similar quantiles for measured daily mean stream temperature (2013 and 2014) were estimated for comparison to current conditions. Insufficient measured data were available to create a similar distribution for Upper Allequash Creek. Summer stream temperature is the most important single factor influencing distribution and production of some cold water fishes (MacCrimmon and Campbell, 1969). Observations of self-sustaining brown trout (Salmo trutta) populations have reported optimal temperature ranges between 12 and 18 °C with a maximum threshold of approximately 22 °C (Bell, 2006). Similarly, brook trout (Salvelinus fontinalis), a species that has historically been recorded in North, Stevenson, and Upper Allequash Creeks [Wisconsin Department of Natural Resources, personal communication, January 2015], are most stable when temperatures do not exceed 19 °C (Creaser, 1930). Each cell in Table 4 was shaded to represent increasing stress to brook trout as a function of stream temperature. Each shade represents the optimal, critical, and lethal stream temperatures generally tolerated by brook trout. The incidence of fish mortality may not necessarily occur for temperatures described as lethal in Table 4 since the degree of thermal stress outside of these tolerance zones is a function of the exposure duration and rate at which temperature changes (Jonsson and Jonsson, 2009); therefore, Table 4 should be used as guidelines and not strict tolerances. The optimal range of stream temperature is currently met in North Creek at least 50% of the time with relatively few excursions above critical thresholds. Optimal temperatures are exceeded with increasing frequency for the simulation period with greater occurrence in the A2 emission scenario. Stevenson Creek, the warmest of the three modeled streams, had measured values exceed optimal conditions at greater frequency than North Creek but still remained below lethal thresholds for brook trout 75% of the time. Similar to North Creek, all three GCM emission scenarios exceeded optimal temperatures with increasing frequency. Projections for both the A1B and A2 emission scenarios indicate that daily mean stream temperatures in Stevenson Creek are unlikely to be cool enough for a healthy trout population, therefore making it difficult to support its current designation as a Class II trout stream. Upper Allequash Creek showed resiliency to climate warming with stream temperatures remaining optimal for brook trout in both the B1 and

A one-dimensional stream temperature model was developed to assess how forecasted changes to climate can affect the thermal regime of three streams in the Trout Lake Watershed in northern Wisconsin. The Stream Network Temperature Model (SNTEMP) was able to predict daily mean stream temperature values with reasonable accuracy compared to measured values over a six month calibration period. Despite violations in the assumptions of the model's predictive ability (mean error of − 1.4 for Stevenson Creek), the high correlation coefficients (R2 ≥ 0.93) suggested appropriate use for long range forecasting. The calibrated model was then used to simulate the response of daily mean stream temperature to measured values of riparian shading and stream geometry, forecasts of stream discharge and groundwater temperature from a coupled groundwater and surface-water flow model (GSFLOW), and changes to air temperature and solar radiation from six global circulation models (GCMs) and three emission scenarios. Results showed steady increases in annual average stream temperature over the 100 year simulation period (2000–2100) for all GCMs and emission scenarios. The A2 and B1 emission scenarios resulted in the largest and smallest gains in stream temperature, respectively. When compared to current values, daily mean stream temperatures over the last 5 years of the 100 year simulation period exceeded optimal temperatures for brook trout with increased frequency in both North and Stevenson Creeks during the warmest months of the year (July and August). Two of the three GCM emission scenarios for Stevenson Creek suggest that future daily mean stream temperatures may no longer be suitable to sustain a cold-water fishery. This information can be used by watershed managers to help make important decisions on future protection or mitigation efforts in the Trout Lake Watershed.

Acknowledgments Much of the stream temperature model description, data collection effort, and calibration were reproduced from Hunt et al. (2013). Full details of calibration and simulation of the coupled groundwater and surface water flow model are also contained in Hunt et al. (2013). Data used to generate tables and figures are available in the Supplementary online material. The author would like to thank John Risley and Steven Markstrom for their helpful comments. The U.S. Geological Survey Climate and Land Use Change Research and Development Program and the Water, Energy and Biogeochemical Budgets Project provided financial support necessary to complete this paper. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

18

W.R. Selbig / Science of the Total Environment 521–522 (2015) 11–18

Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.scitotenv.2015.03.072. References Bartholow, J.M., 1989. Stream temperature investigations—field and analytical methods. U.S. Fish and Wildlife Service Instream Flow Information Paper No. 13. Biological Report 89(17) (139 pp.). Bartholow, J.M., 1991. A modeling assessment of the thermal regime for an urban sport fishery. Environ. Manag. 15 (6), 833–845. http://dx.doi.org/10.1007/bf02394821. Bell, J.M., 2006. The Assessment of Thermal Impacts on Habitat Selection, Growth, Reproduction, and Mortality in Brown Trout (Salmo trutta L): A Review of the Literature. Applied Ecological Services (AES Project 05–0206, 23 pp.). Creaser, C.W., 1930. Relative importance of hydrogen-ion concentration, temperature, dissolved oxygen, and carbon dioxide tension, on habitat selection by brook-trout. Ecology 11 (2), 246–262. Dripps, W.R., Hunt, R.J., Anderson, M.P., 2006. Estimating recharge rates with analytic element models and parameter estimation. Ground Water 44 (1), 47–55. http://dx.doi. org/10.1111/j.1745-6584.2005.00115.x. Ficklin, D.L., Stewart, I.T., Maurer, E.P., 2013. Effects of climate change on stream temperature, dissolved oxygen, and sediment concentration in the Sierra Nevada in California. Water Resour. Res. 49 (5), 2765–2782. http://dx.doi.org/10.1002/wrcr.20248. Gebert, W.A., Walker, J.F., Hunt, R.J., 2009. Use of streamflow data to estimate average annual recharge in Wisconsin, 1970–99. U.S. Geological Survey Fact Sheet FS-20093092 (4 pp.). Gupta, R.S., 1989. Hydrology and Hydraulic Systems. Prentice Hall, Englewood Cliffs, N.J. (739 pp., (Reissued 1995, Prospect Heights, Ill., Waveland Press)). Harbaugh, A.W., 2005. MODFLOW-2005, the U.S. Geological Survey modular groundwater model — the ground-water flow process. U.S. Geological Survey Techniques and Methods 6-A16. Houser, J.N., 2006. Water color affects the stratification, surface temperature, heat content, and mean epilimnetic irradiance of small lakes. Can. J. Fish. Aquat. Sci. 63 (11), 2447–2455. http://dx.doi.org/10.1139/f06-131. Hunt, R.J., Krabbenhoft, D.P., Anderson, M.P., 1996. Groundwater inflow measurements in wetland systems. Water Resour. Res. 32 (3), 495–507. http://dx.doi.org/10.1029/ 95wr03724. Hunt, R.J., Strand, M., Walker, J.F., 2006. Measuring groundwater-surface water interaction and its effect on wetland stream benthic productivity, Trout Lake watershed, northern Wisconsin, USA. J. Hydrol. (320), 370–384 http://dx.doi.org/10.1016/j. jhydrol.2005.07.02. Hunt, R.J., Walker, J.F., Selbig, W.R., Westenbroek, S.M., Regan, R.S., 2013. Simulation of climate-change effects on streamflow, lake water budgets, and stream temperature using GSFLOW and SNTEMP, Trout Lake Watershed, Wisconsin. U.S. Geological Survey Scientific Investigations Report 2013–5159 http://dx.doi.org/10.3133/ sir20135159 (118 pp.). Intergovernmental Panel on Climate Change, 2007. Summary for policymakers. Climate Change 2007—The Physical Science Basis, Contributions of Working Group 1 to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge and New York (18 pp.). Isaak, D.J., Wollrab, S., Horan, D., Chandler, G., 2012. Climate change effects on stream and river temperatures across the northwest U.S. from 1980–2009 and implications for salmonid fishes. Clim. Chang. 113, 499–524. http://dx.doi.org/10.1007/ s10584-011-0326-z. Jonsson, B., Jonsson, N., 2009. A review of the likely effects of climate change on anadromous Atlantic salmon Salmo salar and brown trout Salmo trutta, with particular reference to water temperature and flow. J. Fish Biol. 75, 2381–2447. http://dx.doi.org/10. 1111/j.1095-8649.2009.02380.x.

Leavesley, G.H., Lichty, R.W., Troutman, B.M., Saindon, L.G., 1983. Precipitation-runoff modeling system—user's manual. U.S. Geological Survey Water-Resources Investigation Report 83–4238 (207 pp.). Leavesley, G.H., Markstrom, S.L., Viger, R.J., Hay, L.E., 2005. USGS Modular Modeling System (MMS)—Precipitation-Runoff Modeling System (PRMS) MMS-PRMS. In: Singh, V., Frevert, D. (Eds.), Watershed Models. CRC Press, Boca Raton, Fla., pp. 159–177. Lyons, J., Stewart, J.S., Mitro, M., 2010. Predicted effects of climate warming on the distribution of 50 stream fishes in Wisconsin, U.S.A. J. Fish Biol. 77, 1867–1898. http://dx. doi.org/10.1111/j.1095-8649.2010.02763.x. MacCrimmon, H.R., Campbell, J.C., 1969. World distribution of brook trout, Salvelinus fontinalis. J. Fish. Res. Board Can. 26, 1699–1725. Markstrom, S.L., Niswonger, R.G., Regan, R.S., Prudic, D.E., Barlow, P.M., 2008. GSFLOW—Coupled Ground-Water and Surface-Water Flow Model based on the integration of the Precipitation-Runoff Modeling System (PRMS) and the Modular Ground-Water Flow Model (MODFLOW-2005). U.S. Geological Survey Techniques and Methods (book 6, chap. D1, 240 pp.). Mohseni, O., Stefan, H.G., Erickson, T.R., 1998. A nonlinear regression model for weekly stream temperatures. Water Resour. Res. 34 (10), 2685–2692. National Climatic Data Center, 2014. Station Normals of Temperature, Precipitation, and Heating and Cooling Degree Days (1981–2010), Minocqua, WI, USA. http://ncdc. noaa.gov (accessed on 12/19/2014). National Climatic Data Center, 2015. Local Climatological Data Tables. http://www.ncdc. noaa.gov/nespls/olstore.prodspecific?prodnum=C00095-PUB-A0001#TABLES (accessed on 2/25/2015). Nelson, K.C., Palmer, M.A., 2007. Stream temperature surges under urbanization and climate change: data, models, and responses. J. Am. Water Resour. Assoc. 43 (2), 440–452. http://dx.doi.org/10.1111/j.1752-1688.2007.00034.x. Niswonger, R.G., Panday, S., Ibaraki, M., 2011. MODFLOW-NWT, a Newton formulation for MODFLOW-2005. U.S. Geological Survey Techniques and Methods (book 6, chap. A37, 44 pp.). Null, S.E., Viers, J.H., Deas, M.L., Tanaka, S.K., Mount, J.F., 2013. Stream temperature sensitivity to climate warming in California's Sierra Nevada: impacts to coldwater habitat. Clim. Chang. 116, 149–170. http://dx.doi.org/10.1007/s10584-012-0459-8. Steen, P.J., Wiley, M.J., Schaeffer, J.S., 2010. Predicting future changes in Muskegon River watershed game fish distributions under future land cover alteration and climate change scenarios. Trans. Am. Fish. Soc. 139, 396–412. Stefan, H.G., Preud'homme, E.B., 1993. Stream temperature estimation from air temperature. Water Resour. Res. 29, 27–45. http://dx.doi.org/10.1111/j.1752-1688.1993. tb01502.x. Stefan, H.G., Sinokrot, B.A., 1993. Projected global climate change impact on water temperatures in five north central U.S. streams. Climate Change 24, 353–381. http://dx. doi.org/10.1007/bf01091855. Theurer, F.D., Voos, K.A., Miller, W.J., 1984. Instream water temperature model. Instream Flow Information Paper 16. U.S. Fish and Wildlife Service (FWS/OBS-84/15). Wenger, S.J., Isaak, D.J., Luce, C.H., Neville, H.M., Fausch, K.D., Dunham, J.B., Dauwalter, D.C., Young, M.K., Elsner, M.M., Rieman, B.E., Hamlet, A.F., Williams, J.E., 2011. Flow regime, temperature, and biotic interactions drive differential declines of trout species under climate change. Proc. Natl. Acad. Sci. U. S. A. 108 (34), 14175–14180. http://dx.doi.org/10.1073/pnas.1103097108. Williams, J.E., Haak, A.L., Neville, H.M., Colyer, W.T., 2009. Potential consequences of climate change to persistence of cutthroat trout populations. N. Am. J. Fish Manag. 29, 533–548. Wisconsin Department of Natural Resources, 2015. Trout Stream Maps for Vilas County, Wisconsin. http://dnr.wi.gov/topic/fishing/documents/trout/trout_maps/Vilas_ color_landscape.pdf (accessed on 01/06/2015). Wisconsin Initiative on Climate Change Impacts, 2011. Wisconsin's Changing Climate—Impacts and Adaptation. Nelson Institute for Environmental Studies, University of Wisconsin—Madison and the Wisconsin Department of Natural Resources, Madison, Wis. (217 pp.).