Final Report: Alterations of Water Availability, Water Quality and Fish Habitats in Cold Regions by Climate Change

EPA Grant Number: R824801
Title: Alterations of Water Availability, Water Quality and Fish Habitats in Cold Regions by Climate Change
Investigators: Stefan, Heinz G.
Institution: University of Minnesota , St. Anthony Falls Laboratory
EPA Project Officer: Chung, Serena
Project Period: October 1, 1995 through September 30, 1998
Project Amount: $300,000
RFA: Regional Hydrologic Vulnerability to Global Climate Change (1995) RFA Text |  Recipients Lists
Research Category: Global Climate Change , Ecological Indicators/Assessment/Restoration , Water , Climate Change

Objective:

The main objective of the research was to develop and apply computational simulation methods, which link hydrology, water quality and fish habitats in lakes to climate conditions. Projected climate warming will have a particularly strong impact on ecosystems and aquatic resources in cold regions. Cold regions were defined as regions where water bodies develop an ice cover. Their migration to higher latitudes (or altitudes) under climate warming provided the regional focus of the research.

Summary/Accomplishments (Outputs/Outcomes):

Lake water temperatures and dissolved oxygen were emphasized in the first part of the study. Ice covers were included. In the second part, fish habitat criteria in terms of water temperature and dissolved oxygen were applied to project what fish species could survive and grow well in different types of lakes at different locations under a 2xCO2 climate scenario.

Methodology. A rational and systematic method of analysis was developed. The simulations of lake assemblages use three types of databases: lake hydrometric data, regional meteorological data, and fish response to water quality data. Water quality was simulated by deterministic, process-oriented, unsteady models capable of representing daily water quality changes in a diversity of lakes over long periods of time. Criteria for fish response to water quality were used to determine habitat and fish productivity.

This methodology was first developed and applied for the open water season with emphasis on the upper thermal tolerance limits of fishes and lower limits of dissolved oxygen tolerance. Its applicability was demonstrated, and the climate impact on coldwater, coolwater, and warmwater fishes was projected using the State of Minnesota as a test region. In this research project, the methology was extended to winter simulations (i.e., lower temperatures and ice-cover), and to the entire contiguous United States. Investigations were made for 27 lake types placed at 209 locations around the contiguous U.S. Parameters for survival and good growth of fish were evaluated for each lake type at each location under both past and projected future 2xCO2 climate. Simulation results are presented in the form of tables and maps.

Accomplishments and Research Results. The results project substantial changes in response to projected global warming, for example, shifts in cold-water fish habitats, the potential for invasion of warmwater fishes into these habitats, changes in ice conditions, etc.

Through modeling, a better understanding of how different types of lakes respond to climate and climate changes has been developed. Critically important environmental variables have been put into focus. The results are not limited to information in cold seasons (e.g., thickness of ice) covers and winterkill potential. Water temperature, dissolved oxygen and fish habitat under open water (summer) conditions have also been assessed. The projected changes of water quality, ice-cover and fish habitat parameters with projected climate change have been quantified within the accuracy of the models. The results can be of use to legislators, resource managers/planners, and the public at large. The results can also serve as input to economic assessments.

The accomplishments and research results can be summarized under four subtopics: ice-covers, water availability, water quality, and fish habitat.

1. Ice-cover Information

Information pertaining to climate and climate change effects on lake ice-covers was generated by two substantially different methods: (1) deterministic modeling of physical ice growth and decay on lakes using multi-year daily weather data as input, and (2) collection/ assembly and analysis of historical data on lake ice covers. In support of the first method, field studies were conducted to measure real-time temperature, snow, ice, and radiation characteristics. The results of ice-cover studies re summarized in six St. Anthony Falls Laboratory, University of Minnesota, project reports and eight journal articles. Highlights are as follows:

a) Ice and snow cover model. An ice and snow cover model for freshwater lakes associated with a deterministic, one-dimensional water temperature model has been developed. The lake parameters required as model input are lake surface area (As), maximum depth (Hmax), and Secchi depth as a measure of light attenuation and trophic state. The model is driven by daily weather data and operates year-round over multiple years. The model has been validated with extensive data. Standard errors between simulated and measured values are 0.12 m for maximum ice thicknesses on lakes, 0.07 m for snow covers and less than 6 days for ice formation and decay dates. The model has been applied to simulate effects of climate change on winter ice and snow covers of 27 lake types first under Minnesota climate conditions and then for 209 locations in the U.S. The projected climate changes due to a doubling of atmospheric CO2 were obtained from the output of the Canadian Climate Center Global Circulation Model (CCC GCM) and the Goddard Institute of Space Studies at Columbia University (GISS GCM). To illustrate the effect of project climate change on ice cover characteristics, simulations were made with inputs of past climate conditions (1961-79) and with the projected (2xCO2) climate scenarios.

b) Supporting field studies. Several supporting field studies had to be conducted to justify and validate relationships and parameter values used in the ice cover model formulation and development.

  • Solar radiation is a most important component of the heat budget of an ice cover, especially in spring during the ice-melt period. Surface reflectance/albedo controls how much of the incident solar radiation is retained by an ice cover. Surface reflectance varies roughly with color and texture of the ice/snow cover on a lake. Irradiance measurements were taken over a freshwater lake in Minnesota for 3 months during the winter of 1996-1997. The mean albedo of new snow was measured as 0.83 ± 0.028, while the mean ice albedo was measured as 0.38 ± 0.033. The period from December 17, 1996 to February 17, 1997, was marked as the nonmelt, or high albedo, season when albedo decayed at an average rate of 0.02 per day. During the melt season the albedo decay rate varied from 0.10 to 0.20 per day. Two albedo models were developed for the entire winter season; they use separate equations for the nonmelt and melt periods. The mean absolute error between values observed on the lake surface and values predicted by the model was 0.023 for the first model and 0.029 for the second model. Albedo predictions from both new models and three existing models were also compared to 11 years of observed surface albedo data collected over land. The new model predictions for the land data were as good, if not better, than those by the three existing models.

  • Transmission and attenuation of radiation in the snow and ice covers of a lake. Continuous radiation measurements above and below the snow/ice layer of a lake were made and analyzed to provide attenuation coefficients for solar radiation in snow and ice. Attenuation coefficients in snow were on the order of 30 to 60 m-1. Attenuation coefficient in ice were between 2 and 4 m-1.

  • Heat conduction and thermal conductivities of lake ice and snow covers. A stationary thermister array in a 10 m deep lake in Minnesota was used to measure temperature profiles throughout an ice cover, in the snow and air above, and the water below, continuously throughout a season. An average thermal conductivity for lake ice was calculated to be 2.18 Wm-1°C-1.

  • Ice growth and decay. The temperature time series records also provided detailed information on the thickness of the ice cover and rates of growth and decay of the ice cover on its upper (white ice) and lower (black ice) sides.

  • Density current intrusions. Evidence of density currents (submerged buoyancy driven flows) was obtained during this study. Elevated salinity in spring runoff is primarily responsible for the density flows observed.

c) Compilation and analysis of historical field data on ice covers

  • Records on the dates of ice cover formation, dates of ice-out and ice thickness from selected lakes in Colorado, Maine, Minnesota, New York, and Wisconsin were assembled. The data were correlated with lake and climate characteristics by multi-variate linear regression.

  • Lake ice characteristics of 10 large lakes in Minnesota were related to climate parameters, geographic location, lake surface area, and depth by multiple linear regressions. The regression equations were developed because they are much easier to use than a deterministic, unsteady simulation model that requires time series of weather data as input. Some of the regression equations were employed to hindcast the ice-in dates, ice-out dates, ice cover duration, and maximum ice thicknesses for several freshwater lakes in the United States and one in Canada. The hindcast results were compared with field data from the same lakes. The standard errors between observed and predicted ice-in dates, ice-out dates, ice cover duration, and maximum ice thicknesses for the lakes tested are 4 to 7 days, and 0.05 to 0.06 m, respectively, depending on the equations used. Ice-in date is slightly better predicted than ice-out date and ice cover duration. Because the lakes used to verify the regression equations are in Wisconsin, Maine, and Ontario, the relationships are considered applicable at least to lake in the north-central United States and in southern Ontario.

d) Projected climate change effects on ice covers. Under past climate (1961-79), the total duration of ice cover on Minnesota lakes is about 140 days in the southern half and about 160 days in the northern half. Under 2xCO2 climate, the periods are projected to be shortened to about 85 days and 120 days, respectively. Variations in lake morphometry cause variation of up to 6 days in the mean value of the ice-in date. Climate change is projected to delay the ice-in date by 2 to 3 weeks. Trophic state and geometry of a lake have little effect on ice-in or ice-out date, but latitude is important. Ice-out presently varies by about 4 to 5 weeks from the southern to the northern border of Minnesota. After climate change, ice-out dates are projected to be about 6 to 7 weeks earlier in the south, and 4 to 5 weeks earlier in the north. This is a very substantial change. Annual standard deviations from average dates are presently about 1 week and will increase to 2 weeks. Under a 2xCO2 climatic scenario, maximum ice thickness is projected to be reduced by about 50 percent. These changes would endanger snowmobiles and fishermen because of the reduced bearing capacity of lake ice.

To simulate effects of projected climate change on ice covers of small lakes in the entire northern contiguous U.S., a process-based ice/snow cover model, associated with a deterministic, one-dimensional year-round water temperature model, was run on a continental scale. The dependence of ice cover characteristics on latitude, elevation, and lake characteristics was obtained by making simulations for 27 lake types at 209 locations across the contiguous U.S. Lake surface averages ranged from 0.2 to 10 km2 and maximum depths from 4 m to 24 m. Weather records for the period 1961-1979 were used to represent past climate conditions. The projected climate changes due to a doubling of atmospheric CO2 were obtained from the output of the Canadian Climate Center Global Circulation Model.

Over the contiguous U.S., it is projected that there will be a substantial decrease in locations where an ice cover forms on freshwater lakes. Under the 2xCO2 climate scenario lake ice covers are projected to form every year only in eastern Montana, North Dakota, Minnesota, Wisconsin, Vermont, New Hampshire, and Maine. Under past climate (1961-1979), the duration (cumulative days) of ice cover has been on the average up to 165 days near the northern border of the contiguous U.S. (International Falls, MN). This duration of ice covers is projected to be shortened by 45 days in the coldest parts of the contiguous U.S. and by a maximum of 89 days in warmer areas (Rock Springs, WY), after climate warming. Climate change is projected to delay ice formation by up to 40 days and to advance ice-out by up to 67 days in Michigan; the mean value is 37 days for the country. Under a 2xCO2 climatic scenario, maximum ice thicknesses are projected to be on average reduced by up to 0.44 m (near Sault Ste. Marie, MI) due to climate warming; the mean reduction is 0.21 m for the contiguous U.S. These changes would eliminate fish winterkill problems in shallow lakes, but may endanger snowmobiles and fishermen because of reduced bearing capacity of lake ice.

2. Water Availability Information

The water budget of lakes controls lake levels. Its main components are surface water inflow, precipitation, evaporation, groundwater inflow and outflow, and surface outflow. Surface water runoff information requires field measurements or watershed runoff models. For climate change effect studies runoff models (deterministic or stochastic or parametric) or other methods must be used. In a study separate from this one we have developed and applied a parametric watershed runoff model particularly suitable for water budget and climate change effect studies. We have also applied a modified version of the USDA/Agricultural Research Service's Soil and Water Assessment Tool (SWAT) model to the same watersheds. It was found that landuse has a stronger effect on watershed runoff than projected climate change. Moderate changes in quantity, intensity and seasonal distribution of precipitation projected by GCMs were found to alter model prediction of runoff significantly especially in southern watersheds where only a small fraction of annual precipitation turns into runoff.

Evaporation can affect a lake's water surface elevation and hence its depth and volume as well. On lakes with outflow, a control structure is, however, often used to control and maintain lake levels. Cumulative evaporative water loss from a lake surface was calculated from the cumulative evaporative heat loss in the lake heat budget. Lake surface area and maximum depth as well as trophic status have little influence on evaporative water losses. Therefore, mean annual cumulative evaporative water losses over the simulation period (1961-1979) were averaged for 27 lake types. Due to climate warming, mean annual evaporative water losses from lakes are projected to increase. Near Austin, Texas, the annual evaporative water loss would increase from 1.5 m (present) to 1.9 m (2xCO2), and near Duluth, Minnesota, from 0.6 m (present) to 0.8 m (2xCO2). Mean annual evaporative water losses are projected to be as high as 2.8 m at Las Vegas, Nevada, where air temperature is high and humidity is low; for 209 locations in the contiguous U.S., the mean value is 1.2 m and the standard deviation is 0.4 m. Over the 18-year simulation period, standard deviations of mean annual cumulative evaporative water losses are higher at southern latitudes (0.10 to 0.16 m) than at northern latitudes (0.08 m) under both present and 2xCO2 climate conditions. Simulated changes of annual evaporative water losses due to projected climate change are statistically significant above the 0.1 m value and are largest (up to 0.4 m) in the north-central part of the U.S. where projected mean annual air temperature increases are also largest. Large increases in evaporation are projected to occur in a wide belt with its centerline stretching from Montana through Nebraska and Illinois to Virginia. The smallest increases are projected near the Pacific coast (0.0 m near Olympia, Washington). The mean increase in evaporation for 209 locations in the contiguous U.S. is projected to be 0.17 m with a standard deviation of 0.10 m.

3. Water Quality Information

In a previous related study, the SWAT was chosen from among available models as the most appropriate tool for simulating the effects of climate change on streamflow and water quality in rural watersheds. It simulates the hydrology of a rural watershed on a daily basis and most of its input parameters are physical quantities that can be determined without calibrating to measured streamflow and water quality data. Its algorithms are designed to simulate the response of all of the components of a watershed, including its vegetation, to changes in land cover, land management practices, climate, and atmospheric CO2 concentrations.

In this study the emphasis was on lake water quality, particularly water temperature and dissolved oxygen. These two water quality parameters were selected because they represent two most influential constraints for fish habitat, and they can be more readily simulated than other parameters. After developing and applying a year round water temperature model, including ice cover formation, to Minnesota, simulations were extended to the entire contiguous U.S. The model development made use of the field studies on ice covers described earlier. It also used a new submodel on heat exchange between the water and the sediments of a lake. That process is of minor significance in summer, but becomes important under an ice cover. The oxygen dynamics under an ice cover were described by a separate submodel.

a) Water temperature simulation model for lakes and its application to Minnesota. The numerical simulation model for daily water temperature profiles in a lake solves the one-dimensional, unsteady heat transfer equation. Solar radiation adsorption in the water column is the main contributor to the heat source term. Heat exchange between the lake and the atmosphere is treated as a source/sink term for the topmost water layer of a lake during the open water season. It includes short-wave solar radiation, long-wave atmospheric solar radiation, evaporation and convection through the water surface, and backradiation from the water surface. The MINLAKE96 model includes also heat exchange between each water layer and its adjoining sediments. The equation is solved numerically using an implicit finite difference scheme and a Gaussian elimination method using a timestep of one day and a layer thickness of one meter.

In cold regions, the model simulates ice thicknesses and sediment temperature profiles (heat conduction equation) first, then determines the heat source/sink term, and finally solves the heat transfer equation to obtain water temperature profiles below the lake ice cover. The model uses a stacked layer system, the layers consisting of lake sediments, water, ice cover and snow cover. Snow thickness is determined from snow accumulation, followed by compaction and melting of snow by surface heat input (convection, rainfall, solar radiation) and melting within the snow layer due to internal adsorption of short wave radiation, and transformation of wetted snow to ice when cracks in the ice cover allow water to spill onto the ice surface. In the model, ice growth occurs from the ice-water interface downward and from the ice surface upward. Ice decay occurs at the snow/ice interface, ice-water interface, and within the ice layer.

The snow and ice thickness models were originally developed by Gu and Stefan in 1990 but have been used with some modifications. To make projections of ice cover characteristics for lakes, a new algorithm, which replaces the previous empirical and lake size dependent criteria for the date of ice formation, was developed and incorporated in the model. The new algorithm uses a full heat budget equation to estimate surface cooling quantifies the effect of forced convective (wind) mixing and includes the latent heat removed by ice formation. The algorithm has a fine (0.02 m) spatial resolution near the water surface where temperature gradients before freeze-over are the greatest. Predicted freeze-over dates were compared with observations in nine Minnesota lakes for multiple (1 to 36) years. The difference between the simulated and observed ice formation dates was less than 6 days for all lakes studied.

The year-round results of daily water temperature simulations obtained with the model were tested against extensive water temperature measurements at different water depths (over 5,000 data points) for a total of 48 'lake-years'. Average standard error between simulated and measured water temperatures year round was 1.4°C with a range from 0.92 to 1.69°C, and 0.5°C for the ice-cover season. Water temperature errors are mostly the result of small errors in the prediction of the depth of the thermocline or surface mixed layer.

The temperature features which are typical of the open water season of temperate zone lakes have previously been reported. Noteworthy are the following additional points in the new simulation results:

  • Lake water temperatures in temperate zone lakes reach their lowest values immediately before ice cover formation. (The uppermost 1 m layer is an exception.)

  • Lake water temperatures gradually rise during the winter. Increases are on the order of 1 to 2°C over the winter. The necessary heat is supplied from the sediments with some additional radiation input especially in late spring.

  • Water temperatures near the lake bottom can become slightly larger than 4°C.

  • Lake surface area has a small effect on the winter temperature regime.

  • Differences of water temperature between the water surface and the lake bottom are smaller in oligotrophic lakes than in eutrophic lakes because greater Secchi depth is associated with lower attenuation coefficients and hence deeper penetration of solar radiation into a lake.

  • The presence of an ice cover, although thinning and increasingly transparent in spring, delays the heating of the water to 4°C isothermal conditions. Hence the year-round model shows the spring overturn a few days later than the previous open water model. The water temperatures reach, however, identical levels within two to four weeks so that the temperatures thereafter, especially the high midsummer temperatures, are hardly affected.

  • The inclusion of sediment heat transfer for all layers does lower some of the previously predicted temperature maxima. Reductions are on the order of 1°C.

b) Projected climate warming effects on Minnesota lakes. To simulate projected future water temperature regimes in response to a doubling of atmospheric CO2, monthly adjustments of weather variables were applied to the historical weather data base.

Project effects of climate warming are:

  • Surface water temperature maxima are 3 to 4°C higher.

  • Water temperature stratification is stronger.

  • The ice cover period is shorter.

  • Bottom temperatures are projected to become warmer by 0.2 - 0.6°C, probably because of greater heat storage in the lake sediments during the summer.

  • Lake temperature characteristics that are less affected by climate change are:
    • Winter water temperatures seem unchanged.
    • Water temperature minima still occur at the end of fall turnover, shortly before ice cover formation.
    • Heating of water temperatures in winter is still apparent.
    • Variation of water temperature with depth are strongly dependent on lake morphometry (lake surface area and maximum depth).

The GISS and CCC climate projection models differ in their forecast, and hence the effects on water temperatures also differ somewhat. Compared to the CCC model, the GISS model input to the lake water quality model gives slightly delayed ice formation dates but the same length of ice cover periods. In order to identify the sensitivity of lake temperature characteristics to climate and climate change more clearly, a large number of simulation results had to be processed further. As a first step the water temperature profiles simulated for each day over multiple years were averaged. From these average profiles, several water temperature characteristics were extracted, e.g., the maximum and minimum water temperatures near the surface and in the deepest portion of a lake, and the lengths of the periods of stratification and of ice cover. These selected characteristics (parameters) have ecological significance, e.g., for the survival and growth of fishes.

c) Lake water temperature simulations for the contiguous U.S. Before the model was applied to continental-scale temperature simulations, the model formulation and input requirements were reviewed to examine if significant geographic adjustments were required. The processes which control water temperatures in a lake include heat exchange: (1) through the water surface, (2) between sediment and lake water, and (3) in the water itself through turbulent diffusion. Internal sources of heat are absorbed radiation and latent heat of freezing or melting of ice/snow covers. It was concluded that there is no major geographic constraint in these physical process submodel.

Extending the previous water temperature studies of Minnesota Lakes, this research project was concerned with climate change effects on water temperatures of lakes, especially small lakes, with surface areas up to 10 km2 and depths up to 24 m in the entire contiguous United States. The deterministic (process-oriented), one-dimensional water temperature model was applied to 27 generic lake types at 209 locations in the contiguous U.S. The lake parameters required as model input were surface area (As), maximum depth (HMAX), and Secchi depth as a measure of radiation attenuation and trophic state. The model was driven by daily weather data and operates year-round over multiple years. Daily weather records from 209 stations in the contiguous U.S. for the period 1961-1979 were used to represent present climate conditions. The projected climate changes due to a doubling of atmospheric CO2 were obtained from the output of the Canadian Climate Center General Circulation Model (CCC GCM). To illustrate the effect of projected climate change on lake water temperature characteristics, separate maps were prepared for values simulated with present climate conditions (1961-1979), and with the projected 2xCO2 climate scenario as input, and for the difference between values simulated for projected and present climate conditions, respectively.

The model input data were local weather data. They were taken as representative of the weather station locations, without interpolation. The model output obtained with these local input values was spatially interpolated. Because 209 weather stations were used for the entire continent, the spatial interpolations were not excessive. In mountainous areas, the weather stations cover only the low elevations. Alpine conditions were excluded from this study.

The 2xCO2 climate scenario was constructed from present weather data and increments simulated at grid points with about 273 km to 378 km spacing in longitude and 417 km spacing in latitude. No mesoscale interpolation between grid point and weather station locations was made following an EPA protocol.

The model of lake temperatures used is constructed from heat transfer equations. Coefficients used in the equations for turbulent vertical diffusion coefficients were derived from lake data in different regions and are hence applicable on a continental scale. The lake surface areas are assumed to be more or less circular, and real lakes will deviate from this assumption. Thus, wind sheltering may be over- or underestimated for non-circular lakes. The winter portion of the model describes snow depths and ice albedos by the best available equations derived from field data. These relationships are empirical.

Previous studies of lake temperature regimes under 2xCO2 climate scenarios had looked at individual, specific lakes. Herein generic lakes with different characteristics and in many different regions of the U.S. were investigated. This type of information has not previously been generated. Results of previous studies of individual lakes seem to fit with the results of this generic study.

5.2°C, ecological impacts such as shifts in species distributions, e.g., for fishes, are most likely. Parameters used to characterize water temperatures and thermal stratification in a lake include maximum and minimum surface temperature, maximum and minimum lake bottom temperature, maximum surface to bottom temperature difference, open water stratification ratio, and number of days with hypolimnetic temperature less than 8°C. The maximum surface water temperatures ranged from 19.6 to 32.8°C, and 22.9 to 35.6°C, over 209 locations investigated in the contiguous U.S. under present and a 2xCO2 climate conditions, respectively. Alpine conditions were not included. Geographic latitude has a strong influence on maximum surface water temperatures (13°C differences from the northern to south border in the contiguous U.S.). Climate change is projected to cause an increase in maximum surface water temperatures by up to 5.2°C. Deep lakes (24 m) have slightly higher (2 to 3°C) minimum surface water temperatures than shallow lakes (4 m) under a 2xCO2 climate scenario. Up to 5.1°C increase in minimum surface or bottom water temperatures (in southern latitude lakes) is projected.
The maximum difference between surface and bottom water temperatures in stratified lakes is projected to increase by 1 to 2°C with a local maximum of 3.2°C due to climate warming. The changes point towards a longer stratification season. This may cause serious ecological problems, e.g., more frequent bottom dissolved oxygen depletion and consequences for fish survival.

Duration of bottom water temperatures of less than 8°C depends significantly on lake geometry and geographic location (8°C water temperature is an average limit for feeding of warm-water fish species). This duration ranges anywhere from 0 (shallow, large lakes in southern latitudes) to 365 days (deep, small stratified lakes in northern latitudes) under both present and projected climate conditions. Climate change is projected to reduce this duration. This may provide more growth space or time for warm-water fish species in stratified lakes.

Increases of surface and bottom water temperatures during the summer can also pose serious problems for some fish species, e.g. occurrence of lethal temperatures or temperatures above the good growth limit. Increases of surface and bottom water temperatures in lakes of cold regions delay ice formation and shorten the ice cover period.

d) Dissolved oxygen model for lakes and its application to Minnesota. Dissolved oxygen is crucial for lake ecosystems. A deterministic, one-dimensional model was developed and used to simulate the effect of different climate scenarios on dissolved oxygen conditions in lakes. The numerical simulation model for daily dissolved oxygen profiles in a lake solves the one-dimensional, unsteady transport equation for dissolved oxygen (DO) as a function of depth (z) and time (t). The equation includes vertical turbulent diffusion, sedimentary oxygen demand, oxygen production by photosynthesis, and the first order decay of BOD and plant respiration. In the one-dimensional DO model, the lake simulated is divided into well-mixed horizontal layers from the water surface to the lake bottom. The oxygen transfer through the water surface (reaeration ) during the open water season is used as an oxygen source or sink term in the topmost water (surface) layer of the lake. Sedimentary oxygen demand is treated as a source/sink term for each layer, since each layer is in contact with sediments. Secondary effects of atmospheric CO2 increase on photosynthesis, e.g. through pH changes are not included in the model.

For the DO simulation in a lake during the winter ice cover period modifications include (a) zero gas exchange between the atmosphere and the water body; (b) oxygen consumption by plant respiration is very small and is not presented as a separate sink, (c) water column oxygen demand, BOD, by detrital and other matter, is set constant regardless of trophic status of a lake; and (d) sedimentary oxygen demand is made dependent on trophic state. During the winter months when irradiance is naturally low photosynthesis in ice and snow-covered lakes is predominantly light-limited. In the model Chla is specified by a mean annual value. The model contains a function that calculates typical seasonal chlorophyll cycles. DO concentrations were simulated after water temperature and snow/ice covers had been simulated.

The year-round results of daily DO simulations obtained with the model were tested against extensive DO measurements in Minnesota lakes for a total of 48 ?lake years'. Average standard error between year-found simulations and measurements for 5,378 data pairs for dissolved oxygen concentrations was 1.94 mg/liter. Errors related largely to the difficulty in predicting thin snow cover thicknesses which control photosynthetic radiation available under the ice cover.

Specific lake parameters such as minimum DO concentrations at the lake surface or the lake bottom, duration of anoxia and percentage of anoxic lake volumes were extracted. These DO characteristics were plotted in a coordinate system with a lake geometry ratio (As0.25/HMAX) on one axis and Secchi depth on the other. The lake geometry ratio measures the effect of stratification. To illustrate the effect of projected climate change, separate graphs were presented for values simulated with past climate data (1961-1979) and with projected 2xCO2 climate as input.

Oxygen dynamics and oxygen depletion in winter are strongly related to trophic state of a lake. Oligotrophic shallow lakes have few DO problems whereas eutrophic shallow lakes suffer from DO depletion in winter under identical weather conditions. Projected DO concentrations for a 2xCO2 climate scenario show the virtual disappearance of winter anoxia, even in shallow lakes because the ice cover period is considerably shortened.

Bottom water anoxia can and does occur both in summer and in winter with climate warming. Summer bottom anoxia is projected to become longer in stratified deep lakes, but shorter in medium deep and shallow lakes. The effect of trophic state on anoxia remains pronounced regardless of climate. Oligotrophic lakes tend to have higher bottom DO concentrations than eutrophic lakes, primarily because of the lower sediment oxygen demand, and higher photosynthetic rates, especially in shallow lakes. In some lake types, mostly polymictic oligotrophic lakes, anoxia never occurs. These lakes experience sufficient vertical mixing so that oxygen rich water is frequently in contact with lake sediments. The longest duration of anoxia is estimated for eutrophic lakes which typically have high sediment oxygen demand, and higher water column stability than oligotrophic lakes. Climate change can lengthen (by up to 30 days) or shorten (by up to 75 days) the total period of hypolimnetic anoxia. Eutrophic, polymictic lakes appear to benefit the most from climate change.

A significant breakpoint in lake behavior often occurs at (As0.25/HMAX) values around
5 m.-0.5 Lakes with geometry ratios lower than 4 tend to be seasonally stratified (dimictic), while those above 6 tend to be polymictic. Between 4 and 6 seems to be a transition. Plotting lake characteristics in a coordinate system of lake geometry ratio and Secchi depth allows an overview of the behavior of many different lakes under the same climate setting. Since the model must specify a number of process coefficients and rates only the average behavior can be captured, and individual lakes will deviate from this ?mean.' Nevertheless, getting some quantitative information on the ?average' behavior of many different lakes is considered to be of value.

e) Lake dissolved oxygen simulations for the contiguous U.S. Each winter, hundreds of ice-covered, shallow lakes in the northern U.S. are aerated to prevent winterkill, the death of fish due to oxygen depletion under the ice. How will the projected climate warming influence winterkill and the need to artificially aerate lakes? To answer this question the deterministic, one-dimensional year-round water quality model, which simulates daily DO profiles and associated water temperatures as well as ice/snow covers on lakes, was applied.

Because the main objective of the investigation was to identify trends and changes in response to climate conditions, and not extreme values which may occur over the simulation period, the simulated daily dissolved oxygen profiles were averaged over the 18-year simulation period (Results for the first year, 1961, were excluded to avoid transients from the initial conditions). The results are therefore long-term averages and can not be compared to instantaneous measurements in a lake.

Simulated dissolved oxygen (mg/liter) varies significantly with season and depth. In a large, shallow, eutrophic lake the dissolved oxygen concentrations are more or less uniformly distributed with depth during the open-water season. Due to strong wind mixing, bottom DO concentrations in this shallow lake are typically high enough (>4 mg/liter) in the summer to support fish under both past and projected 2xCO2 climate conditions. During the ice cover period and under past climate conditions the same lakes become, however, totally anoxic for several days. Under the projected 2xCO2 climate scenario, anoxic conditions in the winter disappear, because the ice cover period is shortened by about 40 days due to climate warming and reduced snowcover allows for stronger photosynthetic oxygen production under the ice-cover during the 2xCO2 climate scenario.

Parameters used to characterize winter DO constraints on fishes and possible winterkill include duration of ice covers, minimum under-ice and lake bottom DO concentrations, maximum percentages of lake volume where DO < 0.1 mg/liter or 3 mg/liter, and total number of days (duration) where under-ice DO is less than 0.1 mg/liter or 3 mg/liter. A DO concentration of 3 mg/liter was selected based on typical DO requirements for cold and cool water fish species.

Under past (1962-1979) climate conditions, lake anoxia reaching from the lake bottom to the ice-water interface, i.e. 100 percent of lake volume, currently affects many northern shallow eutrophic lakes. Duration of under-ice anoxia can last from 7 to 43 days in northern shallow, eutrophic lakes of the contiguous U.S. These lakes are candidates for winterkill if artificial aeration is not implemented. Oligotrophic shallow lakes do not appear to have problems with fish winterkill. Under the 2xCO2 climate scenario, winterkill resulting from winter anoxia is projected to disappear from northern shallow lakes because the maximum percentage of lake volume with anoxic conditions is only up to 9 percent and under-ice DO will not reach anoxic conditions. When DO concentrations are less than 3 mg/liter, fish species may still experience stress. Under past climate conditions, the duration of DO < 3 mg/liter conditions under the ice cover is predicted to last up to 80 days; this low DO condition affects especially shallow eutrophic lakes in northern latitudes of the contiguous U.S. Under the projected 2xCO2 climate conditions, the simulations indicate that still up to 80 percent of a shallow lake's volume can have DO concentrations of less than 3 mg/liter in winter. Low lake bottom DO conditions push fish upward into colder water temperatures (0 to 2oC) below the ice cover. This may cause some mortality of fish species through osmoregulatory dysfunction.

4. Fish Habitat Information

Freshwater fish habitat is constrained by several physical and biological parameters that relate to shelter, reproduction, water quality, food supply, and human interference. Water temperature, channel geometry, and streamflow are important to fish habitat in streams. In lakes, water temperature and DO concentrations are two of the most significant water quality parameters affecting survival and growth of fishes.

Climate warming would alter water temperature and DO characteristics in lakes. These changes are in turn expected to have a profound effect on indigenous fish populations. Water temperatures and DO concentrations in Minnesota lakes under a projected 2xCO2 GISS (Goddard Institute for Space Stud) climate scenario were previously simulated for the open-water season only, and the results were used to estimate potential future fish habitats under a 2xCO2 scenario.

Since lakes in cold regions have ice covers over several months, low DO concentrations in winter can also significantly influence presence or survival of fish assemblages. Dissolved oxygen concentrations diminish over time in ice-covered lakes due to sedimentary oxygen demand and water column oxygen demand, while an ice cover prevents oxygen replenishment (reaeration) from the atmosphere. With a snow/ice cover on a lake, irradiance of the water is strongly attenuated, and photosynthetic oxygen production is correspondingly reduced. In shallow eutrophic lakes, DO concentrations can therefore drop to low values (0.2-2.0 mg/liter) resulting in fish death near the end of an extended ice cover period. The death of fish due to oxygen depletion under ice (winterkill) is a significant fisheries management problem in shallow lakes, both in the north-central United States and Canada. For example hundreds of lakes are annually threatened by winterkill in the state of Minnesota; nearly 200 aeration permits have been issued in every winter since 1988 to overcome the problem.

In this research project previous fish habitat estimates in Minnesota lakes were extended from open-water (summer) conditions to ice-cover (winter) conditions. Lakes in Minnesota were studied because the winter period is of substantial length, and lake water quality and fish data were available for model validation. Fish habitat was determined for both the ice-cover period and the open-water season from daily water temperature and DO profiles, which were simulated by a process-oriented, one-dimensional year-round water quality model. The model was run over a continuous 19-year period under past and projected 2xCO2 climate scenarios. Estimated fish habitat for past climate was validated against fish observations in 3002 Minnesota lakes.

The research examined fish habitat in ?generic' lake types classified by lake geometry (surface area, maximum depth) and trophic state measured by Secchi depth. A lake classification was developed from the Minnesota Lake Database. Lake surface areas (As) chosen were 0.2, 1.7 and 10.0 km2 for small, medium and large lakes, respectively. Maximum depths (Hmax) chosen were 4, 13 and 24 m for shallow, medium-depth and deep lakes, respectively. With these numbers one obtains 9 lake types ranging from relatively large and shallow lakes to relatively small and deep lakes. The likelihood of a strong or weak stratification in a lake can be related to the lake geometry ratio As0.25/Hmax. The above 9 types of lakes cover geometry ratios from 0.9 to 14.1. Polymictic lakes have the highest geometric ratio, while strongly stratified (dimictic or monomictic) lakes have the lowest geometry ratio. The transition occurs between 4 and 6. The lake bathymetry (the shape of the lake basin) was characterized by a "typical" function fitted to the data for 122 Minnesota lakes.

Secchi depth, a common limnological measure of lake transparency, was used to represent both trophic state and radiation attenuation of a lake. The trophic state expresses primary productivity (photosynthesis of plants) and ranges from oligotrophy (nutrient-poor, biologically unproductive) to eutrophy (nutrient-rich, productive). The radiation attenuation of a lake is used to quantify how much solar energy reaching the water surface can penetrate through a water column to heat water and to support photosynthesis of aquatic plants. Secchi depths are inversely proportional to radiation attenuation of a lake. Secchi depths of 1.2, 2.5, and 4.5 m were selected for eutrophic, mesotrophic, and oligotrophic lakes, respectively. Therefore the 27 ?generic' lake types were characterized by a 3x3x3 matrix consisting of (a) three different lake surface areas, (b) three lake maximum depths and (c) three Secchi depths.

a) Temperature and DO criteria for fish habitat. Fish habitat has several inter-dependent physical and biological characteristics. The presence and survival of fish in a lake is related to accessibility, suitable ecological conditions, human interference, and resistance to episodic natural events. Lake water temperature and DO concentrations are considered two of the most significant water quality parameters affecting survival and growth of fishes. In this study suitable fish habitat in the water column over time was determined by having water temperature and DO within suitable ranges. Water temperature criteria for three fish guilds (cold-water, cool-water, and warmwater) were developed from laboratory and field data by Eaton. The guild designations for 29 fish species were suggested by Hokanson. Three temperature levels: the lower good-growth temperature limit (LGGT), the upper good-growth temperature limit (UGGT), and the lethal temperature threshold (LT) were developed. Fish experience optimum growth when water temperature falls between the lower and upper good-growth temperature limits. The LGGT was the mean temperature between zero net growth and maximum growth, and the UGGT is the U.S. EPA upper temperature criterion for growth [optimum temperature plus 1/3 (ultimate incipient lethal temperature minus optimum temperature)]. The upper lethal temperature threshold is water temperature that fish cannot be acclimated to without causing death. In essence, the LT was determined from field data that relate maximum annual water temperature to geographic distribution.

The availability of suitable DO concentrations can also control the presence of freshwater fish species and fish populations in lakes. Fish guilds have DO concentration requirements, below which mortality is more likely to occur or growth is impaired. DO criteria used were obtained from available EPA data. Appropriate DO survival criteria values are 2.5 mg/l for warmwater species and 3.0 mg/l for cool- and cold-water species. DO criteria were developed for fish species during the open water season ? the summer period. For ice-covered lakes, DO criteria could be set at lower values, but data from previous studies are still inadequate to establish specific dissolved oxygen tolerance limits for fish winterkill. The sensitivity of fish habitat during the winter to three DO survival limits was therefore studied.

b) Fish survival in Minnesota lakes. Fish thermal and DO survival and growth criteria were applied to the simulated 18-year average daily water temperature and DO profiles. During the summer, low DO concentrations near the lake bottom push fish upward, and warm water temperatures near the water surface push fish downward in search of optimum conditions for growth and survival. When isolines of lethal temperature (LT) and limiting DO for a fish species intersect each other, summer uninhabitable space develops over the entire depth of a stratified lake. Summer kill for fish species in a lake is expected to occur if non-survival conditions last more than seven days. Three habitats can be identified:

(1) Uninhabitable space where temperature is above or DO is below the survival criteria.
(2) Good-growth habitat if temperature is between the upper and lower good-growth limits and DO is above the survival limit.
(3) Restricted growth habitat if temperature is above the upper growth temperature but below the survival limit of temperature, or if temperature is below the lower good-growth limit, and if DO is above the survival limit.

When thermal stratification starts earlier in spring under a 2xCO2 CCC scenario and sedimentary and water column oxygen demands increase with water temperature, dissolved oxygen concentrations start to decrease earlier and decrease more rapidly over time. Lake volumes where DO is less than the survival limit for fish are therefore greater under a 2xCO2 scenario in seasonally stratified lakes.

Lake types that provide fish habitat during both winter and summer under present climate conditions but are projected to become uninhabitable in summer under a 2xCO2 CCC climate scenario are identified by the simulation results.

Under climate warming it is projected that fish winterkill will disappear from shallow lakes in Minnesota, but non-survival conditions are projected to develop during the summer in mesotrophic and oligotrophic shallow lakes for cold-water fish species (due to higher water temperature), but not for cool- and warmwater fish species. In a previous study, with a projected GISS climate scenario and an open-water water quality model, there were sixteen lake types, including all shallow lakes and some of the medium-depth and deep lakes, which were projected to develop non-survival conditions in summer for the cold-water fish assemblages under the 2xCO2 scenario. These differences are due to: (1) the outputs of the CCC and GISS General Circulation Models, and (2) the inclusion of sediment-water heat exchange in the year-round water quality model which makes simulated water temperatures lower, and therefore reduces the number of lakes with non-survival conditions in summer under the 2xCO2 CCC scenario. Suitable cold-water fish habitat is projected in only two small and medium large eutrophic, shallow lake types under a 2xCO2 CCC climate scenario; winterkill is simulated in these same lakes under past climate conditions.

c) Validation of simulated fish habitat with fish observations. Fish habitats estimated from simulated "normal" daily water temperature and DO profiles for the open-water season were previously validated against fish observations. This validation was extended to year-round estimates of fish habitats. Model predictions of suitable fish habitats were tested against fish observations for 2231 lakes in northern Minnesota and for 771 lakes in southern Minnesota. Fish observation data were available in the Minnesota Lakes Fisheries Database (ERLD/MFLDB, 1990). The parameter used for model validation is the non-survival length (NSL). NSL is defined as the total number of days when either temperature or dissolved oxygen does not meet the fish presence criteria at all depths. NSL gives the length of both summer and winter periods of non-survival conditions at all depths. When NSL for a fish guild in a lake is estimated to be more than 7 days, it is concluded that the fish guild is unable to exist in that type of lake.

If at least one fish species representative of a particular fish assemblage is observed in a particular type of lake for which the model predicts suitability (of fishes of that fish assemblage in that lake type), agreement between model and observation can be claimed. Similarly, if one fish species representative of a guild is observed when the model simulates non-survival conditions at all depths over more than seven days, disagreement between model and observation is concluded. If fishes of a certain fish assemblage were not observed in a lake, no conclusions can be drawn by comparison with model simulations. Model predictions of fish habitat were found to agree with field observations for all fish guilds in all medium-depth (Hmax = 13 m) and deep (Hmax = 24 m) lakes in both northern and southern Minnesota. This is a better agreement with observations than in the previous validation. For shallow lakes, there are only two types, which are predicted to have suitable fish habitat, that agrees with field observations. These are small and medium oligotrophic lakes in northern Minnesota and small mesotrophic and oligotrophic lakes in southern Minnesota. With the DO survival limits set at 2.5 and 3.0 mg/l, winterkill was projected in all eutrophic and some mesotrophic shallow lakes), which is in disagreement with observations.

The reasons for the partial disagreement between model and observation for shallow lakes can be divided into two broad categories: (1) critical life processes which affect fish presence, but are not included in the simulation model, and (2) uncertainties in the fish database and the model predictions. The main reason for the disagreement found is the lumping of many fish species into a guild. The least sensitive fish species deviates from the guild average.

d) Sensitivity of winter fish habitat to DO survival limits. The generally accepted cause for winterkill of fish is oxygen depletion (suffocation) due to prolonged periods of snow/ice cover. Under low-temperature conditions, thresholds of many species of fresh-water fishes are believed to lie between 1.0 and 2.0 mg/l. Some of the less tolerant species may require up to 3.0 mg/l. However, some researchers found that tolerance g/l for largemouth bass and bluegills, 0.3 - 0.4 mg/l for yellow perch, mud pickerel, pumpkinseeds, pike, and chubsuckers, and 0.2-0.3 mg/l for bullheads and golden shiners.

Data from previous studies are still inadequate to establish specific tolerance limits of dissolved oxygen for fish winterkill. Therefore the sensitivity of estimated fish habitats to three DO survival limits during the ice cover period was investigated. For the winter ice-cover period, DO survival criteria were set at 0.0, 0.5, 1.0 mg/l for the three sensitivity tests. These winter limits were set for each of the three fish guilds without change of the 2.5 and 3.0 mg/l DO survival

limits during the open water season. Results of the sensitivity analysis show an appreciable change in fish habitat availability in lakes of northern and southern Minnesota.

e) Projected good-growth conditions for fish in Minnesota lakes. Fish-habitat parameters extracted from the water temperature and DO simulations include the GSL (days) defined as the total number of days when simultaneously any value of temperature is between the lower and upper good-growth bound and dissolved oxygen exceeds the survival limit. GSL quantifies the length of the good-growth period in part or all of a lake.

GSL can be used together with a lake's bathmetry (depth-area-volume) to integrate good-growth habitat areas or volumes over time since temperature and DO vary with depth in stratified lakes. Good-growth habitat area (GGHA) is defined as integrated lake bottom area (m2) over time (days) where and when (a) water temperatures are within the upper and lower good-growth temperature limits and (b) dissolved oxygen is greater than the survival limit. Integration begins at the first exceedance of the lower growth limit and the dissolved oxygen survival limit. The final value is divided by surface area AS, to give a normalized GGHA in units of days. Similarly good-growth habitat lake volume (GGHV) is defined as integrated lake volume (m3) over time (days) where and when (a) water temperatures are within the upper and lower good-growth temperature limits and (b) dissolved oxygen is greater than the survival limit. The final value of GGHV is divided by surface area AS, to give a normalized GGHV in units of meter-days. These two parameters are the summation of possible good growth bottom areas and good growth volumes. They are calculated for each fish guild over a growing season. Good-growth habitat areas and volumes measure thermal habitat space and were originally developed as predictor variables to estimate the total sustained yields of four commercially important fish species (lake trout, lake whitefish, walleye, and northern pike).

The dependence of good-growth parameters for three fish guilds on lake geometry (surface area and maximum depth), trophic state and meteorological conditions was established. When non-survival conditions at all depths exist in a lake type either during summer or winter, good-growth parameters are meaningless. Simulated good-growth lengths (GSL) for cold-water fish guilds are about five months (150 days) in stratified medium-depth and deep lakes in northern Minnesota (Duluth) under past climate conditions. Climate warming is projected to increase GSL by up to 40 days in deep (Hmax = 24 m) lakes due to an increase of the stratification period, but to decrease GSL by up to 32 days in medium-depth (Hmax = 13 m) lakes due to the increase of water temperature. Good-growth lengths for cool- and warmwater fish species were predicted to be about 3 months and less than 2 months in medium-depth and deep lakes under past (1961-1979) climate conditions, respectively. Climate warming is projected to increase GSL by up to 50 days for both cool- and warmwater fish species.

Simulated good growth habitat areas (GGHA) and volumes (GGHV) are more strongly dependent on lake geometry than on trophic state. Climate change was projected to increase the good-growth period, as well as good growth areas and volumes for cool-water and warmwater fish guilds in all lake types in Minnesota. Projected increases for GGHA in seasonally stratified lakes are on the average 50 percent and 115 percent for cool-water and warmwater fishes, respectively. For cold-water fishes, good growth spaces were projected to have a small increase (gain) in deep, strongly stratified lakes, due to an increase of the stratification period, and to have a small decrease (loss) in weakly stratified lakes due to projected higher water temperatures and lower DO near the lake bottom under the 2xCO2 climate scenario.

f) Fish survival in lakes of the contiguous U.S. Previous fish habitat estimates in small lakes (up to 10 km2) for three fish guilds were extended from Minnesota to 209 locations in the contiguous U.S. The locations in the contiguous U.S. where lakes experience winterkill or summerkill were identified for both past and projected 2xCO2 climate scenarios. Maps showing the impact of climate warming on suitable habitat for three fish guilds in the entire contiguous U.S. were produced. In this study the used for warmwater fish was 34.5°C. This lethal temperature limit is much higher than the LT (31.4°C) used in the earlier analysis and derived from data within the U.S. only.

The meteorological database used as input to the long-term lake simulations consisted of 19 years (1961-1979) of data from 209 data locations in the contiguous U.S. Input data included measured daily air temperature, dew point temperature, wind speed, solar radiation, total cloud cover, and precipitation (both rainfall and snowfall). The recorded daily weather data were obtained from the Solar and Meteorological Surface Observation Network (SAMSON). The period from 1961 to 1979 was chosen because it is long enough to give a representative average and variance of past conditions. Weather data from the 1980's and 1990's were available but not used to characterize the present climate because several years in that period were exceptionally warm in some regions.

Projected changes in climate conditions were obtained from the output of the Canadian Climate Center (Second Generation) General Circulation Model (CCC GCM) for a doubling of atmospheric CO2 The second generation CCC GCM includes higher spatial resolution (3.75° x 3.75°) than previous models (e.g. GISS, Goddard Institute of Space Studies at Columbia University) and full diurnal annual cycles. Mean monthly increments (for air temperature) or ratios (for solar radiation, wind speed, relative humidity, and precipitation) were calculated from the output of the CCC GCM under 1xCO2 and 2xCO2 climate scenarios. These monthly climate parameter increments or ratios were then applied to measured daily climate conditions (1962-1979) month by month to generate the projected daily 2xCO2 climate scenario. This protocol eliminates the effect of any GCM biases in the modern (1xCO2) climate, and was proposed by the USEPA. Monthly increments (or ratios) from the grid center point closest to a weather station were used for that station. Therefore the same monthly adjustments could be used for two or more weather stations if they are all close to a particular grid center point. The average annual increase in air temperature at these 209 locations ranged from 2.5°C to 6.5°C, and the average seasonal increase for the period January to March (winter) ranged from 2°C to 10°C.

Under past (1962-1979) climate conditions, medium deep (HMAX= 13 m) lakes near Duluth, Minnesota, and Austin, Texas, support cool-water fish habitat year-round, even though low DO concentrations near the lake bottom (< 3 mg/l) cause portions of the lake to be uninhabitable during the summer. Thermal stratification starts earlier under a 2xCO2 CCC scenario, and sedimentary and water column oxygen demands increase with water temperature; hence DO concentrations decrease earlier and more rapidly over time. Lake volumes where DO is less than the survival limit for cool-water fish are therefore greater than under a 2xCO2 scenario. This and the increases of surface water temperatures under a projected 2xCO2 scenario destroy cool-water fish habitat in medium-depth lakes near Austin, Texas, during the summer, but not near Duluth, Minnesota. The shift from year round suitable habitat under past (1962-1979) climate conditions to summerkill under a 2xCO2 CCC climate scenario is a significant negative impact of climate warming on all three fish guilds.

One may wonder how the geographic distribution of suitable fish habitat in lakes of the contiguous U.S. may change as the climate changes from past (1962-1979) conditions to a warmer 2xCO2 climate scenario. Under past climate conditions there are only 18 locations (out of 209 investigated) and mostly at high elevation, where cold-water fish habitat is found in all lake types. Deep or medium-depth lake types can support cold-water fish at northern latitudes of the contiguous U.S. while most of the shallow lakes experience either winterkill or summerkill.

Under a projected 2xCO2 climate scenario, fewer lake types are projected to support cold-water fish in the northern contiguous U.S.. Regions where no lakes can support cold-water fish extend significantly further north under a 2xCO2 climate scenario. This shows the severe impact of climate warming on cold-water fish in small lakes (up to 10 km2 surface area) in the contiguous U.S.

Under past climate conditions more than 20 lake types support year-round cool-water fish habitat at the northernmost latitudes of the contiguous U.S. About 3 to 6 eutrophic and/or mesotrophic, shallow lake types have winterkill. Shallow lakes in Texas, Louisiana, southern Arkansas, Mississippi, and Alabama, do not support cool-water fish under past climate conditions. All deep lakes in the contiguous U.S. can support cool-water fish under past climate conditions.

Under a projected 2xCO2 climate scenario, all lakes at northern latitudes can support cool-water fish because winterkill is projected to disappear from shallow lakes. Only few lake types are projected to support cool-water fish in most of the south-central and southeastern U.S. after climate warming.

Under past climate conditions, shallow eutrophic and/or mesotrophic lakes are projected to experience winterkill of warmwater fish at up to 25 locations in northern latitudes. With elimination of winterkill due to climate warming, all shallow lake types are projected to support warmwater fish under a projected 2xCO2 climate scenario. All lakes at southern latitudes of the contiguous U.S. can support warmwater fish habitat under both past and projected 2xCO2 climate scenarios.

The geographic distribution of the total number of lake types with suitable or unsuitable (summerkill/winterkill) fish habitat shows how well or how poorly different regions of the U.S. are suited for cold-, cool-and warmwater lake fisheries. It is also of interest to know how fish habitat in different lake types is affected by climate and potential climate change.

Changes from winterkill due to lack of oxygen to summer kill due to elevated temperature, or from winterkill due to lack of oxygen to suitable habitat, or from suitable habitat to summerkill due to elevated temperatures have been identified for specific lake types. There is no change to winterkill because the simulations indicate that shorter ice cover periods will eliminate winterkill under warmer climate. There is also no change from summerkill due to elevated water temperatures to suitable habitat because water temperatures in shallow lakes under a warmer climate are projected to get warmer, i.e. even less tolerable.

The positive impact of global climate warming on fish habitat is to eliminate winterkill in all shallow lakes for all three fish guilds, but a shift from winterkill to summerkill due to climate warming is projected to occur for cold-water fish. For cool-water and warmwater fish, all locations with winterkill are projected to support year-round fish habitat after climate change. Loss of cold-water and cool-water fish habitat in many locations is indicated by the numbers and the geographic distribution. Maps that present this information more clearly have been prepared and published.

g) Projected good-growth conditions for lakes in the U.S. U.S. maps for the good-growth parameter GSL clearly indicate regions where good fish habitat for specific fish guilds exists.

Cold-water fish in seasonally stratified lakes find at present habitat as far south and north of NV, UT, CO, KS, IO, OH, PA. Climate warming will change that, leaving only a small fringe along the Canadian border and the West Coast, where the good-growth period for cold-water fishes will shorten roughly from 150 to 100 days. Cool-water fish presently find habitat in seasonally stratified lakes in most parts of the country. Climate warming will introduce summerkill of cool-water species in the southeastern states (FL, TX, LA, MS, AL, GA, NC, and SC). The good growth period will be lengthened by up to 50 days in the north and at mid-latitudes of the contiguous U.S. Warmwater fish are projected to exist at all locations in seasonally stratified lakes under both climate scenarios.

Good-growth lengths (GSL) for cold-water fish decrease from north to south in the contiguous U.S., while GSL for cool-water and warmwater fish guilds increase from north to south. GSL is higher for cool-water fish than for warmwater fish.

Statistics (maximum and minimum values, mean and standard deviation) of good-growth parameters were developed for all 27 lake types and for the three fish guilds investigated. Good-growth lengths, although variable with lake types, last on average 122-243 days, 142-170 days and 131-139 days in all 27 lake types investigated for cold-water, cool-water and warmwater fish, respectively. Variations (standard deviation, STD) of GSL from the mean are on the order of 33 to 54 days under past climate conditions. Climate warming due to doubling of atmospheric CO2 is projected to increase GSL for cool-water and warmwater fish on average by 11 to 37 and 21 to 40 days, respectively. Good-growth lengths for cold-water fish species are projected to change by mean lengths ranging from ?25 days (decrease) to +38 days (increase) under the projected 2xCO2 climate scenario. Under the projected 2xCO2 climate scenario, the good-growth period for warmwater fishes extends from about 72 days at northern latitudes to an entire year (365 days) at southern latitudes. Climate warming is projected to increase the good-growth period of warmwater fishes on average by 30 to 40 days in all 27 lakes investigated. There are, however, large variations by lake type and from location to location as indicated by STD values of 14 to 24 days.

h) Summary of fish habitat projections. The following conclusions stand out among the many fish habitat study results for the contiguous U.S.

(1) Climate warming is projected to eliminate winterkill (due to low DO) in most shallow, eutrophic/mesotrophic lakes in the contiguous U.S. This would be a beneficial effect of climate warming.

(2) Summerkill due to elevated temperature is likely to eliminate suitable cold-water fish habitat in most shallow lakes of the contiguous U.S. Cold-water fish will continue to find some habitat in almost deep lakes at northern latitude.

(3) Summerkill under the projected 2xCO2 climate scenario is a projected significant negative impact, especially on cold-water fish in northern lakes and on cool-water fish in southern lakes of the contiguous U.S., where suitable habitat exists under present conditions. The strongest negative impact of climate warming on cold-water fish is in medium deep (max. depth = 13 m) lakes, and on cool-water fish in shallow (max. depth = 4 m) lakes.

(4) Climate warming is projected to reduce substantially the geographic area (number of locations), where lakes have suitable cold-water fish habitat. Cold-water fish have the best chance to survive in deep, stratified lakes near the northern border of the contiguous U.S.

(5) Climate warming is projected to reduce the geographic areas (number of locations), where lakes have suitable cool-water fish habitat, by up to one third.

(6) Simulated good-growth periods decrease for cold-water fish, but increase for cool-water and warmwater fish from northern to southern latitudes in the contiguous U.S. Under current climate conditions, good-growth periods are highly variable with lake types. Under the projected 2xCO2 climate scenario, good-growth periods for cool-water fish species are projected to increase on average by up to 37 days (average of 209 locations) with a maximum standard deviation of up to 30 days, dependent on lake type and location.

(7) Cool-water fish can exist in many lakes of the contiguous U.S. Exceptions are shallow, eutrophic lakes in the north-central U.S. (winterkill) and in south-central U.S. e.g. Louisiana, part of Texas, south Arkansas and Alabama, parts of Georgia and Florid (summerkill). Summerkill of cool-water fish in shallow lakes will expand significantly in south-central and southeastern states under a 2xCO2 climate scenario.

(8) No summerkill of warmwater fish, due to elevated temperature and/or dissolved oxygen deficiency, is projected to occur in any lake and at any location of the contiguous U.S. investigated under both climate scenarios. Under the projected 2xCO2 climate scenario, the good-growth period for warmwater fishes extends from about 72 days at northern latitudes to an entire year (365 days) at southern latitudes. Climate warming is projected to increase the good-growth period of warmwater fishes on average by up to 40 days in all 27 lake types investigated, with a maximum standard deviation of up to 34 days dependent on lake types.

Barbara M. Levinson and Brian Sidlauskas were the USEPA/ORD project officers. This work used a methodology developed in cooperation with John G. Eaton, J. Howard McCormick, Kenneth E. Hokanson, Brian Goodno, and other members of the Mid-Continent Ecology Division, USEPA, Duluth, MN. The Minnesota Supercomputer Institute, University of Minnesota, provided a resource grant.


Journal Articles on this Report : 20 Displayed | Download in RIS Format

Other project views: All 43 publications 34 publications in selected types All 20 journal articles
Type Citation Project Document Sources
Journal Article Ellis CR, Champlin J, Stefan HG. Density current intrusions in an ice-covered urban lake. Journal of the American Water Resources Association 1997;33(6):1363-1374. R824801 (Final)
not available
Journal Article Fang X, Stefan HG. Simulated climate change effects on dissolved oxygen characteristics in ice-covered lakes. Ecological Modelling 1997;103(2-3):209-229. R824801 (Final)
  • Full-text: Science Direct Full Text
    Exit
  • Abstract: Science Direct Abstract
    Exit
  • Journal Article Fang X, Stefan HG. Potential climate warming effects on ice covers of small lakes in the contiguous U.S. Cold Regions Science and Technology 1998;27(2):119-140. R824801 (Final)
  • Abstract: sciencedirect
    Exit
  • Journal Article Fang X, Stefan HG. Temperature variability in lake sediments. Water Resources Research 1998;34(4):717-729. R824801 (Final)
    not available
    Journal Article Fang X, Stefan HG, Alam SR. Simulation and validation of fish thermal DO habitat in north-central US lakes under different climate scenarios. Ecological Modelling 1999;118(2-3):167-191. R824801 (Final)
  • Full-text: sciencedirect
    Exit
  • Abstract: sciencedirect.com
    Exit
  • Journal Article Fang X, Stefan HG. Projections of climate change effects on water temperature characteristics of small lakes in the contiguous US. Climatic Change 1999;42(2):377-412. R824801 (Final)
  • Abstract: springerlink
    Exit
  • Journal Article Fang X, Stefan HG. Projected climate change effects on winterkill in shallow lakes in the northern contiguous United States. Environmental Management 2000;25(3):291-304. R824801 (Final)
  • Abstract from PubMed
  • Journal Article Fang X, Stefan HG, Eaton JG, McCormick JH, Alam SR. Simulation of thermal/dissolved oxygen habitat for fishes in lakes under different climate scenarios-Part 1. Cool-water fish in the contiguous US. Ecological Modelling 2004;172(1):13-37. R824801 (Final)
  • Full-text: sciencedirect
    Exit
  • Abstract: sciencedirect.com
    Exit
  • Journal Article Fang X, Stefan HG, Eaton JG, McCormick JH, Alam SR. Simulation of thermal/dissolved oxygen habitat for fishes in lakes under different climate scenarios-Part 2. Cold-water fish in the contiguous US. Ecological Modelling 2004;172(1):39-54. R824801 (Final)
  • Full-text: sciencedirect
    Exit
  • Abstract: sciencedirect
    Exit
  • Journal Article Fang X, Stefan HG, Eaton JG, McCormick JH, Alam SR. Simulation of thermal/dissolved oxygen habitat for fishes in lakes under different climate scenarios-Part 3. Warm-water fish in the contiguous US. Ecological Modelling 2004;172(1):55-68. R824801 (Final)
  • Full-text: sciencedirect
    Exit
  • Abstract: sciencedirect
    Exit
  • Journal Article Gao SB, Stefan HG. Multiple linear regression for lake ice and lake temperature characteristics. Journal of Cold Regions Engineering ASCE, June 1999;13(2):59-77. R824801 (Final)
  • Abstract: ascelibrary
    Exit
  • Journal Article Gao SB, Stefan HG. Potential climate change effects on ice covers of five freshwater lakes. Journal of Hydrologic Engineering 2004;9(3):226-234. R824801 (Final)
  • Abstract: ascelibrary
    Exit
  • Journal Article Henneman HE, Stefan HG. Snow and ice albedo measured with two types of pyranometers. Journal of the American Water Resources Association 1998;34(6):1487-1494. R824801 (Final)
  • Abstract: interscience
    Exit
  • Journal Article Henneman HE, Stefan HG. Albedo models for snow and ice on a freshwater lake. Cold Regions Science and Technology 1999;29(1):31-48. R824801 (Final)
  • Abstract: sciencedirect
    Exit
  • Journal Article Stefan HG, Fang X. Simulated climate change effects on ice and snow covers on lakes in a temperate region. Cold Regions Science and Technology 1997;25(2):137-152. R824801 (Final)
  • Abstract: sciencedirect
    Exit
  • Journal Article Stefan HG, Fang X, Hondzo M. Simulated climate change effects on year-round water temperatures in temperate zone lakes. Climatic Change 1998;40(3-4):547-576. R824801 (Final)
  • Abstract: springerlink
    Exit
  • Journal Article Stefan HG, Fang X, Eaton JG. Simulated fish habitat changes in North American lakes in response to projected climate warming. Transaction of the American Fisheries Society 2001;130(3):459-477. R824801 (Final)
    not available
    Journal Article Stefanovic DL, Stefan HG. An accurate and efficient algorithm for numerical simulation of conduction-type problems. Mathematics and Computers in Simulation 1998;47(1):37-46. R824801 (Final)
  • Abstract: sciencedirect
    Exit
  • Journal Article Stefanovic DL, Stefan HG. Simulation of transient cavity flows driven by buoyancy and shear. Journal of Hydraulic Research 2000;38(3):181-195. R824801 (Final)
  • Abstract: iahr
    Exit
  • Journal Article Stefanovic DL, Stefan HG. Accurate two-dimensional simulation of advective-diffusive-reactive transport. Journal of Hydraulic Engineering-ASCE 2001;127(9):728-737. R824801 (Final)
  • Abstract: ascelibrary.aip.org
    Exit
  • Supplemental Keywords:

    water, watershed, global climate, vulnerability, sensitve population, aquatic, habitat, hydrology, limnology., RFA, Scientific Discipline, Air, Water, Ecosystem Protection/Environmental Exposure & Risk, Hydrology, Water & Watershed, exploratory research environmental biology, Ecosystem/Assessment/Indicators, Ecosystem Protection, climate change, Ecological Effects - Environmental Exposure & Risk, Wet Weather Flows, Environmental Monitoring, Atmospheric Sciences, Ecological Risk Assessment, Watersheds, Ecological Indicators, sensitive populations, water resources, limnology, aquatic ecosystem, collaborative hydrologic modeling, ecosystem modeling, hydrologic dynamics, ecological exposure, fish habitats, climate change impact, fish habitat, watershed, precipitation monitoring, water availability, ecological modeling, hydrologic models, precipitation patterns, ecosystem effects, ecological impacts, climate models, cold regions, alternative climate conditions, rainfall patterns, aquatic ecosystems, water quality, land and water resources, fish , temperature variables, wet weather modeling, climate variability, Global Climate Change

    Progress and Final Reports:

    Original Abstract
  • 1996
  • 1997