Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Potential Effects of Climate Change on Inland Glacial Lakes and Implications for Lake Dependent Biota in Wisconsin
Final ReportApril 2013
PREPARED BY:
MICHAEL W. MEYER WISCONSIN DEPARTMENT OF NATURAL RESOURCES, SCIENCE SERVICES, RHINELANDER, WI
JOHN F. WALKERU.S. GEOLOGICAL SURVEY, WISCONSIN WATER SCIENCE CENTER, MADISON, WI
KEVIN P. KENOWU.S. GEOLOGICAL SURVEY, UPPER MIDWEST ENVIRONMENTAL SCIENCES CENTER, LA CROSSE, WI
PAUL W. RASMUSSENWISCONSIN DEPARTMENT OF NATURAL RESOURCES, SCIENCE SERVICES, MADISON, WI
PAUL J. GARRISONWISCONSIN DEPARTMENT OF NATURAL RESOURCES, SCIENCE SERVICES, MADISON, WI
PAUL C. HANSONUNIVERSITY OF WISCONSIN, CENTER FOR LIMNOLOGY, MADISON, WI
RANDALL J. HUNT U.S. GEOLOGICAL SURVEY, WISCONSIN WATER SCIENCE CENTER, MADISON, WI
Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.
ENVIRONMENTAL AND ECONOMIC RESEARCH AND DEVELOPMENT PROGRAM
This report was funded through the Environmental and Economic Research and Development Program of Wisconsin’s Focus on Energy.
ii
Contents
Executive Summary ii
Introduction 1
Hydrologic and Lake Quality Modeling 21
Historic Changes in Lake Condition 61
Historic Distribution of Common Loons in Wisconsin 89
Predicting Future Loon Occurrence in Wisconsin 109
Appendix – Figures Depicting Future Loon Occurrence in Wisconsin 151
iii
Executive Summary
Overview
The economic vitality and quality of life of many northern Wisconsin communities is closely
associated with the ecological condition of the abundant water resources in the region. Climate change
models predict warmer temperatures, changes to precipitation patterns, and increased evapotranspiration in
the Great Lakes region. Recently (1950-2006), many regions of Wisconsin have experienced warming, and
precipitation has generally increased except in far northern Wisconsin. Modeling conducted by the
University of Wisconsin Nelson Environmental Institute Center for Climate Research predicts an increase
in annual temperature by the middle of the 21st century of approximately 6oF statewide, and an increase in
precipitation of 1”–2”. However, summer precipitation in the northern part of the state is expected to be
less and winter precipitation will be greater. By the end of the 21st century, the magnitude of changes in
temperature and precipitation are expected to intensify.
Such climatic changes have altered, and would further alter hydrological, chemical, and physical
properties of inland lakes. Lake-dependent wildlife sensitive to changes in water quality, are particularly
susceptible to lake quality-associated habitat changes and are likely to suffer restrictions to current breeding
distributions under some climate change scenarios. We have selected the common loon (Gavia immer) to
serve as a sentinel lake-dependent piscivorous species to be used in the development of a template for
linking primary lake-dependent biota endpoints (e.g., decline in productivity and/or breeding range
contraction) to important lake quality indicators. In the current project, we evaluate how changes in
freshwater habitat quality (specifically lake clarity) may impact common loon lake occupancy in Wisconsin
under detailed climate-change scenarios. In addition, we employ simple land-use/land cover and habitat
scenarios to illustrate the potential interaction of climate and land-use/land cover effects. The methods
employed here provide a template for studies where integration of physical and biotic models is used to
project future conditions under various climate and land use change scenarios. Findings presented here
project the future conditions of lakes and loons within an important watershed in northern Wisconsin – of
importance to water resource managers and state citizens alike.
iv
Projecting Future Lake Conditions in the Trout Lake basin of Northern Wisconsin
We used a coupled ground-water/surface-water model with the capability of incorporating
predictions of temperature and precipitation from physical climate models to predict changes in the
hydrologic cycling within the Trout Lake basin in Vilas County. The Trout Lake basin has been the focus
of numerous modeling studies that represent stages in the development and refinement of a coupled
groundwater/surface-water model using the U.S. Geological Survey’s GSFLOW code. The basin also lies
within the heart of the current common loon breeding range in Wisconsin. The GSFLOW model, which is
driven by precipitation and temperature inputs, was calibrated using the Parameter ESTimation (PEST)
suite of calibration software.
A lake water quality model, adapted from a simple lake carbon cycling equilibrium model, was
used to predict the consequences of changing climate to in-lake dissolved carbon concentrations and
resulting changes to water clarity. The model is the steady state solution to a relatively simple differential
equation. The load to a lake is the sum of areal input from the lake perimeter, groundwater (GW) inflow,
precipitation and surface water (SW) inflow. The mass load of carbon from GW and SW is the product of
their respective concentrations and their inflow volumes described below. The areal load is assumed to be
the sum of shoreline with canopy and shoreline without canopy.
Given the uncertainty in climate modeling, it is desirable to use more than one global climate
model (GCM) in order to obtain a range of potential future climatic conditions. Daily precipitation and
temperature output from six GCMs were considered for one current and three future emissions scenarios.
The GCM output from the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Special
Report on Emission Scenarios was downscaled for the Trout Lake area by the Wisconsin Initiative on
Climate Change Impacts. The downscaled climate data for the period 1961-2100 was used as input to the
GSFLOW model, and the corresponding lake budgets from the GSFLOW runs were used as input for the
Lake Model.
Changes in Lake Levels
According to the model results, lowland drainage lakes in the watershed show relatively minor
changes due to the future climate, in part because drainage lakes are relatively low in the watershed, and
v
have large contributing areas relative to the lake area. Headwater drainage lakes, however, may be more
impacted (see Historic Changes in Lake Condition). Seepage lakes show a gradual decline in net
precipitation (precipitation minus evaporation) with time, a slight increase in net groundwater, and variable
net surface-water inputs, resulting in a relatively steady to moderate decline in water levels with time. In
almost all cases the variability of the various components increases in the later part of the 21st century.
Beginning around 2060, there appears to be an increase in variability and relatively large changes in the net
groundwater, which are reflected in the increase in variability in the lake stages. Because the predicted
increase in net groundwater is less than the decrease in net precipitation, the result is a declining water
level.
Changes in Water Clarity
The general trend in lakes was toward slightly higher water clarity, as indicated in model results
by increased Secchi depth. In all lakes, except Allequash and Edith, dissolved organic carbon (DOC)
concentrations decreased. Decreases in DOC had commensurate increases in clarity for these lakes, even
under conditions of elevated total phosphorus. The competing effect on clarity, from DOC and
chlorophyll, is the result of light extinction due to DOC which is 3 times greater than the light extinction
due to chlorophyll. Given the log relationship between total phosphorus and chlorophyll, the scenario with
a 25% increase in total phosphorus translates to approximately a 40% increase in chlorophyll for these
lakes. The ensuing decrease in water clarity was offset by a concurrent decrease in DOC of 15% or more.
Historic Changes in Lake Condition
To further illustrate the potential impact of a warmer and drier climate, a sediment core was
extracted from Max Lake, Vilas County, WI to assess the ecological changes experienced during the mid-
Holocene dry period. During the mid-Holocene (5,000-7,000 years ago) the climate in the Great Lakes
region was warmer and effectively drier than at the present. During this time the climate was similar to that
predicted by climate change models. The sediment core encompassed the entire 13,000 year history of the
lake since the retreat of the glaciers. The sediment was examined for charcoal to assess the general climate
of the region. An increase in the frequency of wildfires indicates a drier and perhaps warmer climate. The
vi
diatom community was examined to determine changes in pH and phosphorus levels. These variables
provide an estimate of the hydrologic regime of the lake.
Our results indicated that during the first few centuries following the establishment of Max Lake,
it functioned as a drainage lake that was likely part of a much larger lake. The pH levels were consistent
with those found at the present time in headwater drainage lakes. Charcoal levels were relatively low
indicating a cool, moist climate. Around 7,000 years ago, the frequency of wildfires greatly increased,
signally the beginning of the mid-Holocene warming period. At that time, effective reduction in
precipitation caused the lake to change from a drainage lake to a seepage lake. The change in the
hydrologic regime resulted in a decline in the pH values of about 0.5 units. Around 6,000 years ago, low
charcoal levels indicate a return to moist conditions and the lake briefly returned to a drainage lake. This
was short lived and for the last 6,000 years the lake has been a seepage lake with lower pH values and
lower phosphorus concentrations. Even though the charcoal indicates a drier climate during the last four
millennia, Max Lake has remained a seepage lake. This work demonstrates that if the future climate
becomes warmer and effectively drier as predicted, there may be substantial changes in the ecology of
headwater drainage lakes. These lakes may become seepage lakes resulting in potentially lower pH and
nutrient levels which may result in a simplification of the fish community and a reduction in the lake’s
overall productivity.
Historic Range Contraction of the Common Loon in Wisconsin
The breeding range of the common loon has shifted northward in the Great Lakes states from
former southern range limits that are known to once extend throughout southern Minnesota to northern
Iowa, throughout southern Wisconsin to northern Illinois, and throughout southern Michigan to northern
Indiana and Ohio. A review was conducted of historical and recent water quality data for Wisconsin lakes
that formerly supported but no longer support breeding loons. Historical phosphorus concentrations
inferred from paleolimnological data, and land use change that occurred within the drainage basins of
former loon nesting lakes were considered in the analysis. The review indicated that the landscape of
southern Wisconsin has changed dramatically since Common Loons last nested in this part of Wisconsin.
A number of factors have likely contributed to the decreased appeal of southern Wisconsin lakes to
vii
breeding Common Loons, including changes to water quality, altered trophic status resulting from nutrient
enrichment, and reductions in suitable nesting habitat stemming from shoreline development and altered
water levels. Increased nutrient and sediment inputs from agricultural and developed areas likely
contributed to a reduction in habitat quality. We contemplated whether similar shifts in trophic state of
lakes across Wisconsin might occur under a warming climate pattern accompanied with increased
evapotranspiration rates and have negative consequences on loon occupancy.
Projecting Future Loon Occurrence in the Trout Lake Basin of Northern Wisconsin
Wisconsin Loon Habitat Model
Research across North America has shown that common loons select breeding territories as a
function of lake physical and chemical characteristics. We developed a predictive regional common loon
breeding habitat suitability model based on loon populations in northern and central Wisconsin, and
assessed the potential effects of future climate change on loon habitat quality and breeding pair occupancy
of lakes within the Trout Lake basin in northern Wisconsin.
We used presence or absence of loon territorial pairs as the response variable in logistic regression
analyses with multiple habitat predictor variables. Water quality data including Secchi depth, color, total
phosphorus, temperature, pH, conductivity, and dissolved oxygen, nest habitat quality, and land cover
within 150 m and 500 m of each lake or reservoir were considered in the analysis. The best fitting logistic
regression model included as predictor variables nest habitat, number of waterbodies within 10 km, Secchi
depth, log of lake area, and the interaction between Secchi depth and log of lake area. The two next best
models included these same variables plus proportion of forest within either the 500 m or 150 m buffer.
Future Loon Occurrence in the Trout Lake Basin
We used the best-fitting loon habitat model to predict loon occurrence for 27 of the lakes in the
Trout Lake basin under projected future climatic conditions. Linked climate and lake hydrology model
(described above) were used to simulate lake conditions for all years from 1962 to 2100 under three
emissions scenarios and six general circulation models. The results of hydrology and lake modeling of the
Trout Lake basin in northern Wisconsin indicates that these lakes have the potential to become clearer
viii
under future climate conditions. This is due to the fact that DOC, not total phosphorus , is the parameter
most closely associated with lake clarity in the watershed that is likely to change under future climate
scenarios, and DOC levels are predicted to be lower. Secchi depth estimates from the linked climate and
hydrology models for the period between 2010 and 2090 are predicted to increase slightly in the 27 lakes in
the Trout Lake basin. Considering water clarity as a primary factor in loon habitat suitability, estimated
probabilities of loon occurrence are expected to stay the same or increase very slightly between 2010 and
2090 for all 27 lakes as Secchi depth slowly increases over time. Changes in nest habitat can have large
effects on the probability of loon occurrence in lakes. Lake stage is predicted to be lower at Trout basin
seepage lakes under future climate scenarios, potentially altering nest habitat quality. Step changes in nest
habitat generally overwhelm the effect of estimated changes in Secchi depth or possible changes in total
phosphorus. These results point to the critical need to conserve and enhance common loon nest habitat
within the Trout Lake basin, and throughout the current breeding range of loons in Wisconsin. Adaptation
strategies to reduce potential negative consequences of a changing climate should include preserving
existing critical nest habitat by managing shoreline development and habitat loss.
Acronyms used in the report
AIC Akaike’s Information Criterion
AUC Area under the ROC curve
Chl Chlorophyl-a concentration
COOP National Weather Service Cooperative
DEM Digital elevation model
DOC Dissolved organic carbon
EPA Legacy STORET U.S. Environmental Protection Agency Legacy
STOrage and RETrieval Data Warehouse
GCM Global Climate Model
GHB MODFLOW General Head Boundary condition
GIS Geographic Information System
GPS Geographical Positioning System
ix
Acronyms used in the report (continued)
GRTS Generalized Random-Tesselation Stratified
GSFLOW USGS coupled groundwater/surface-water model
HRU Hydrologic response units
IPCC Intergovernmental Panel on Climate Change
LAK MODFLOW Lake Package
LEC Light extinction coefficient of water
LU/LC Land use and land cover
MODFLOW USGS Modular Groundwater Flow Model
NLDC National Land Cover Database
PEST Software used for parameter estimation
PRMS USGS precipitation-runoff modeling system
ROC Receiver operating characteristic
Secchi Secchi depth in the lake
SFR2 MODFLOW Streamflow Routing Package
SRES Special Report on Emission Scenarios
SWIMS Wisconsin Surface Water Integrated Monitoring
System
TIGER/Line files Topologically Integrated Geographic Encoding and
Referencing Database Line Shapefiles
TP Total phosphorus concentration
UMESC U.S. Geological Survey Upper Midwest
Environmental Sciences Center
USEPA U.S. Environmental Protection Agency
USEPA STAR U.S. Environmental Protection Agency Science to
Achieve Results Research Grants program
USGS U.S. Geological Survey
x
Acronyms used in the report (continued)
UZF MODFLOW Unsaturated Zone Flow Package
WDNR Wisconsin Department of Natural Resources
Wisconsin DNR Wisconsin Department of Natural Resources
1
Chapter 1–Introduction
MICHAEL W. MEYER1, KEVIN P. KENOW2, and JOHN F. WALKER3 1Wisconsin Department of Natural Resources, Science Services, Rhinelander, Wisconsin 2U.S. Geological Survey, Upper Midwest Environmental Sciences Center, La Crosse, Wisconsin 3U.S. Geological Survey, Wisconsin Water Science Center, Middleton, Wisconsin
Contents
Abstract ........................................................................................................................................................................... 3
Introduction ...................................................................................................................................................................... 4
Project Goals ................................................................................................................................................................... 5
Background ..................................................................................................................................................................... 7
Potential Climate Change Impacts on Hydrology in Wisconsin ................................................................................... 7
Potential Climate Change Impacts on Wisconsin Lakes .............................................................................................. 8
Nutrient Enrichment ............................................................................................................................................... 10
Changes in Species Composition .......................................................................................................................... 11
Aquatic Invasive Species ....................................................................................................................................... 11
Changes in Ice Cover ............................................................................................................................................. 14
Potential Climate Change Impacts on Loons Breeding in Wisconsin ........................................................................ 15
Direct Impacts. ....................................................................................................................................................... 16
Indirect Impacts. ..................................................................................................................................................... 17
Program Area Addressed .............................................................................................................................................. 18
References .................................................................................................................................................................... 19
2
Figures
Figure 1–1. Measured change in the annual average temperature and precipitation in Wisconsin from 1950 to
2006under the A2 emission scenario; reprinted from WICCI (2011b). ............................................................................ 9
Figure 1–2. Projected change in the annual average temperature and precipitation from 1980 to 2090 under the A2
emission scenario in Wisconsin; reprinted from WICCI (2011b). ................................................................................... 9
Figure 1–3. Projected change in the annual average temperature and precipitation in Wisconsin from 1980 to 2090
under the A2 emission scenario; reprinted from WICCI (2011b). .................................................................................. 10
Figure 1–4. Measured changes stage of Anvil Lake, Vilas County, 1934-2010; reprinted from WICCI (2011c). ...... 13
Figure 1–5. Changes in lake ice-cover around Wisconsin; reprinted from WICCI (2011c). ..................................... 14
Figure 1–6. Projected changes in common loon distribution under various GCM/emission scenarios; from USDA
Bird Atlas http://nrs.fs.fed.us/atlas/bird/index.html. ....................................................................................................... 15
Figure 1–7. Relationship of territorial loon occupancy and lake water clarity (Secchi depth) in the Wisconsin
Northern Highlands Ecological Landscape (Meyer 2006). ............................................................................................ 18
3
Abstract
Climate change models predict warmer temperatures, changes to precipitation patterns, and
increased evapotranspiration in the Great Lakes region. Recently (1950-2006), many regions of
Wisconsin have experienced warming, and precipitation has generally increased except in far northern
Wisconsin. By the middle of the 21st century, current modeling conducted by the University of
Wisconsin Nelson Environmental Institute Center for Climate Research predicts an increase in annual
temperature of approximately 6o F statewide, and an increase in precipitation of 1”–2”. However,
summer precipitation in the northern part of the state is expected to be less and winter precipitation will
be greater. By the end of the 21st century, the magnitude of changes in temperature and precipitation
will intensify.
Most of Wisconsin is predicted to experience more frequent and severe precipitation events. As
a consequence, the hydrologic cycle may become more dynamic and there may also be thermal impacts
to water resources if atmospheric and water temperatures rise. Some of the most significant impacts on
lakes resulting from climate change may be from increased sediment and nutrient loads due to increased
precipitation, problems related to proliferation of aquatic invasive species, changes in species
composition, decreases in ice cover throughout the state, physical impacts on lakes, and changes in lake
levels.
Lake-dependent wildlife sensitive to changes in water quality, are particularly susceptible to lake
quality-associated habitat changes and are likely to suffer restrictions to current breeding distributions
under some climate change scenarios. We have selected the common loon (Gavia immer) to serve as a
sentinel lake-dependent piscivorous species to be used in the development of a template for linking
primary lake-dependent biota endpoints (e.g., decline in productivity and/or breeding range contraction)
4
to important lake quality indicators. In the current project, we evaluate how changes in freshwater
habitat quality (specifically lake clarity) may impact common loon breeding distribution in Wisconsin.
Introduction
Climate change models predict warmer temperatures, changes to precipitation patterns, and
increased evapotranspiration in the Great Lakes region (Kling et al. 2003). Such climatic changes have
altered, and will further alter hydrological, chemical, and physical properties of inland lakes (Schindler
1997, Magnuson et al. 1997, Kling et al. 2003). Lake-dependent wildlife sensitive to changes in water
quality, are particularly susceptible to lake quality-associated habitat changes and are likely to suffer
restrictions to current breeding distributions under some climate change scenarios.
A recent report by the Wisconsin Initiative on Climate Change Impacts (WICCI 2011a,
http://www.wicci.wisc.edu/publications.php) concluded that future climatic conditions will change the
nature of Wisconsin lakes, with impacts on lake-dependent biota. The common loon (Gavia immer)
exemplifies a lake-dependent wildlife species that is likely sensitive to climate change. The breeding
range of the common loon has shifted northward in the Great Lakes states from former southern range
limits that once (as recent as 100-150 years ago) extended throughout southern Minnesota to northern
Iowa, throughout southern Wisconsin to northern Illinois, and throughout southern Michigan to northern
Indiana and Ohio (McIntyre 1988). It is probable that changes in habitat quality on southern lakes
within the species’ former range are associated with the range constriction and likely that a warming
climate could induce similar habitat quality changes to northern lakes, threatening the remnant breeding
loon population in the Great Lakes Region.
We have selected the common loon to serve as a sentinel lake-dependent piscivorous species to
be used in the development of a template for linking primary lake-dependent biota endpoints (e.g.,
5
decline in productivity and/or breeding range contraction) to important lake quality indicators.
Ultimately, the information will be available to predict future breeding range limits of the common loon
in the Great Lakes region (with applicability elsewhere) and our approach will serve as a model for
developing monitoring plans for other primary biota endpoints.
Project Goals
The over-arching goal of this collaborative research endeavor is to project the impact of climate
change on common loon habitat quality, thus future loon occurence within the Trout Lake basin inVilas
County, Wisconsin. The project includes projections of future lake hydrology, chemistry, and clarity
under varying climatic conditions and assumed changes to lake total phosphorus concentrations within
the watershed; analysis of sediment core data from a seepage lake in the Trout Lake basin to map the
impact of past climate on lake conditions since post-glacial formation; an analysis of historic records on
the breeding distribution of the common loon to determine the association of breeding range contraction
to change in habitat characteristics sensitive to climate change; and development of a loon habitat
suitability model identifying lake factors associated with loon habitat use in central and northern
Wisconsin, plus projections of climate change impacts on lake factors associated with loon habitat in the
Trout Lake basin..
Specifically, in Chapter 2 a coupled ground-water/surface-water model with the capability of
incorporating predictions of temperature and precipitation from physical climate models is used to
predict changes in the hydrologic cycling within the Trout Lake basin. These hydrologic alterations
include changes to streamflow, lake and ground-water levels, and complete characterization of the
hydrologic budgets for lakes within the model domain. This approach, which links physical climate
models to basin-wide hydrologic models, is further developed to predict the consequences of changing
climate to in-lake dissolved carbon concentrations and resulting changes to water clarity. In Chapter 3,
6
results of analysis of a sediment core from Max Lake, Vilas County, provide a preview of what to
expect in softwater seepage lakes if the climate becomes warmer and effectively drier. The sediment
core covers the entire lifespan of the lake, circa 14,000 years. The diatom community is used to
reconstruct phosphorus and pH to estimate the impact of the warmer and effectively drier period during
the mid-Holocene. Charcoal fragments preserved in the sediments were used to assess the relative
frequency of fires which provide a surrogate for drought conditions. These conditions may be similar to
that predicted by GCM climate models for the next century. Chapter 4 examines whether factors
associated with the northward loon range restriction in the Upper Midwest may be accelerated under
conditions of a warming climate. The breeding range of the common loon has shifted northward in the
Great Lakes states from former southern range limits that once (as recent as 100-150 years ago)
extended throughout southern Minnesota to northern Iowa, throughout southern Wisconsin to northern
Illinois, and throughout southern Michigan to northern Indiana and Ohio. It is likely that changes in
habitat characteristics of southern lakes within the species’ former range are associated with the range
constriction. We investigate changes in water quality and land-use change associated with lakes that are
south of the current breeding range of common loons in Wisconsin but that historically supported
breeding loons. This chapter offers insights into how a changing climate might impact common loon
habitat quality and the future breeding distribution of loons in Wisconsin. In Chapter 5, a predictive
regional breeding habitat suitability model for common loon populations is developed using data
collected from lakes in both north and central Wisconsin. The central WI cohort represents the most
southerly extent of the current breeding distribution of common loons in the Upper Midwest U.S.
Metrics of habitat quality as well as anthropogenic change and influence are quantied as explanatory
factors. Selection of breeding territories by loons has been shown to be related to water quality and
nesting habitat quality in recent studies in Wisconsin and New Hampshire. The second objective in
7
Chapter 5 is to assess the potential effects of future climate change on habitat quality and loon
distribution – specifically, to incorporate the findings of hydrology and lake modeling (Chapter 2) with
that of loon habitat modeling, to estimate the change in probability of loon lake occupancy within the
Trout Lake basin, which represents the core of Wisconsin breeding loon distribution. We also assess the
potential impact of climate change on habitat quality throughout the loon breeding range in Wisconsin.
The inevitability of continued climate warming makes it imperative that natural resource
managers are provided the information and tools needed to develop adaptive management strategies for
expected changes in climate.
Background
In this section, we describe the potential impacts of climate change on hydrology, lakes, and
common loons in Wisconsin, primarily drawing upon information from the WICCI Assessment Reports
for Water Resources (WICCI 2011c; http://www.wicci.wisc.edu/report/Water-Resources.pdf) and Wildlife
(WICCI 2011d; http://www.wicci.wisc.edu/report/Wildlife.pdf).
Potential Climate Change Impacts on Hydrology in Wisconsin
Changes in annual temperature and precipitation regimes will undoubtably impact hydrological
regimes in Wisconsin. From the WICCI Water Resources Assessment Report (WICCI 2011c) “An
understanding of how hydrologic processes are affected by the driving forces of a changing climate is
necessary to assess future potential impacts on water resources. Spatially, the Wisconsin’s hydrologic
processes will not be affected uniformly. The differences reflect variations in land use, soil type and
surface deposits, groundwater characteristics, and runoff and seepage responses to precipitation”
“There will likely be system-wide changes in hydrologic processes but these may not be completely
predictable. There may even be times when abrupt and long-term changes take place. For example,
8
groundwater tables may rise as much as 12 feet in one season leaving formerly dry ground inundated
for the foreseeable future, streams may dry up due to lack of recharge, and change may result in lake
vegetation species composition”
Most of the state is predicted to experience more frequent and severe precipitation events (Figure
12 of the WICCI Climate Working Group Report). As a consequence, the following hydrological
impacts are predicted byWICCI (2011c).
• Where heavy rainfall becomes more intense, the hydrologic cycle may become more dynamic, with
increased flushing of lakes and a decreased mean residence time.
• Increased infiltration may lead to increased groundwater recharge to the point of groundwater flooding.
• Increased spring precipitation may increase spring runoff. With increased winter temperatures, it is not likely that snowfall will increase in the winter, but precipitation may occur instead as rain or freezing rain.
• Increased intensity and increased magnitude of precipitation may affect the depth of plant propagation,
water availability in the root zone, and deep drainage systems. The duration between rain events may strongly affect the interception of water. When intercepted water is returned to the atmosphere, it does not lead to vegetation productivity.
• There are also thermal impacts to water resources if atmospheric and water temperatures rise. Aquatic systems can suffer if streams or other resources are taxed with a heavy influx of warm water, either from overland flow, such as water picking up heat along a parking lot, or stormwater discharged from a stormwater management facility.
The length of time that the ground is frozen is an important variable in alteration of the
hydrologic cycle. Groundwater recharge depends on the timing of precipitation and warming
temperatures (WICCI 2011b).
Potential Climate Change Impacts on Wisconsin Lakes
The two climate change drivers that may impact Wisconsin lakes are increasing temperatures
and shifting precipitation patterns. Recently (1950-2006), many regions of Wisconsin have experienced
warming, and precipitation has generally increased except in far northern Wisconsin (Figure 1-1). By
9
the middle of the 21st century, current modeling conducted by the University of Wisconsin Nelson
Environmental Institute Center for Climate Research (based on downscaled, debiased output from IPCC
Global Circulation Models) predicts an increase in annual temperature of approximately 6o F statewide,
and an increase in precipitation of 1”–2” (Figure 1-2). However, summer precipitation in the northern
part of the state is expected to be less and winter precipitation will be greater (WICCI 2011b).
Figure 1–1. Measured change in the annual average temperature and precipitation in Wisconsin from 1950 to
2006under the A2 emission scenario; reprinted from WICCI (2011b).
Figure 1–2. Projected change in the annual average temperature and precipitation from 1980 to 2090 under the A2
emission scenario in Wisconsin; reprinted from WICCI (2011b).
10
By the end of the 21st century, the magnitude of changes in temperature and precipitation will
intensify. The magnitude of the change, as modeled, is also a function of emission scenario. The
projections displayed in Figure 1-3 represent the IPCC GCM A2 model, describing a future with
greatest greenhouse gas emissions (e.g. increasing rate of CO2 emissions).
Figure 1–3. Projected change in the annual average temperature and precipitation in Wisconsin from 1980 to 2090
under the A2 emission scenario; reprinted from WICCI (2011b).
Some of the most significant impacts on lakes resulting from climate change may be from
increased sediment and nutrient loads due to increased precipitation, problems related to proliferation of
aquatic invasive species, changes in species composition, decreases in ice cover throughout the state,
physical impacts on lakes, and changes in lake levels. The following includes a description of potential
changes to lakes as described in the WICCI Water Resources Assessment Report (WICCI 2011c).
Nutrient Enrichment
An increase in large intensity rainfall events may result in increased sediment and nutrient loads
to lakes from increased runoff from surrounding land and/or streams. Regional differences across the
11
state will depend on land use, soils, and geology. Drainage lakes and impoundments may experience
more impacts than seepage lakes which are governed by groundwater. The timing of increased
precipitation is critical to the potential impacts on lakes. If the ground is frozen but precipitation is in
the form of rain, it will run off into surface waters, potentially carrying nutrients associated with soil.
In contrast, when the ground is not frozen, precipitation is more likely to infiltrate into the ground and
recharge groundwater. However, as rainfall intensity increases, more of the precipitation will run off
as the ground becomes saturated. An increase in nutrient and sediment load to a lake will likely
contribute to an increase in trophic status and over time, may reduce lake water quality. Variables like
decreased water clarity can have negative effects on many species such as loons and walleyes (Sander
vitreus) that rely on locating prey by sight to feed. Turbid, shallow lakes are often dominated by fish
that are not sight feeders such as common carp (Cyprinus carpio) and bullheads (Ictalurus spp.). These
fish will further stir up the bottom sediments leading to even greater turbidity.
Changes in Species Composition
Increased temperatures may contribute to a change in the biological composition of a lake.
Species native to warmer areas may survive in a future warmer Wisconsin. Species composition may
shift from a predominance of green algae towards more blue-green algae. Coldwater fish species may
shift north and be locally extirpated due to warmer water. With increased temperatures, moderately
shallow lakes may no longer stratify, but mix continually. Internal phosphorus loading would then play
a dominating force in a lake’s dynamics and affect its trophic status .
Aquatic Invasive Species
Climate change may increase the spread of aquatic invasive species within and among lakes.
During flooding events, waterbodies may become interconnected, allowing invasive species to spread.
12
In contrast, drought may bring about new habitat suited for an invasive species that may be introduced.
Increased temperatures may also lead to new introductions and survival of aquatic invasive species not
previously recorded in Wisconsin. Species native to warmer areas may be more likely to survive when
temperatures rise because many species will be able to over-winter ( e.g., Hydrilla [Hydrilla sp.],
water hyacinth [Eichhornia sp.], red swamp crayfish [Procambarus clarkii]). These species are native
or well-established in the southern U.S., but are thought to be limited by cold temperatures and ice
cover. .
Lake Levels
Climate forecasts show that precipitation changes in the state will not be uniform, and changes
in lake levels are difficult to predict because of the balance between evaporation versus increased
precipitation. If evapotranspiration is greater than precipitation and/or recharge, lake levels will drop.
If precipitation and recharge are greater, lake levels will rise. Or, they may balance each other out.
Predicted seasonal and decade-scale variation in precipitation also affect lake levels. Changes in lake
levels, either increase or decrease, will have impacts on aquatic habitats, including plants, substrate
and coarse woody vegetation. Shallow lake systems would be most affected by lowered water levels, as
would be the littoral zones of deep lakes. Low lake levels leave important fish habitat out of the water,
such as emergent vegetation and downed trees. Human disturbance and removal of this habitat during
times of low water could lead to permanent changes in ecosystem functioning. In contrast, high water
conditions could result in redistribution of substrate and structural features to deeper water, and also
uproot vegetation.
Seepage lakes are the most sensitive to changes in precipitation and groundwater elevations. A
lake is either surface- or groundwater- dominated, and its water chemistry is influenced by the relative
contribution of each. If the dominant water source shifts from precipitation and overland flow to
13
groundwater, the lake will shift from being a soft water to hard water lake. If on the other hand,
groundwater sources are reduced due to long-term declines in the water table, lake chemistry could
become more dependent on precipitation. Changes in water chemistry could have major implications
for food webs, phytoplankton and zooplankton communities, algal and rooted aquatic plant growth, as
well as nutrient cycling and methyl mercury production.
The 74-year water level record for Anvil Lake (Figure 1-4), a northern Wisconsin seepage lake,
demonstrates pronounced, recurring highs and lows. The record appears to indicate that lake levels are
getting progressively lower during each succeeding dry period and especially during the present period.
The low levels reached between 2004 and 2010 are the lowest observed to date and are associated with
the low precipitation in recent years. (Source: U.S. Geological Survey; data prepared by Dale
Robertson; WICCI 2011c).
Figure 1–4. Measured changes stage of Anvil Lake, Vilas County, 1934-2010; reprinted from WICCI (2011c).
14
Changes in Ice Cover
Climate models predict that ice cover will decrease throughout the state and there may even
come a time when Wisconsin’s lakes are ice-free all winter. The environmental consequences of this are
great. Decreased ice cover may lead to an extended growing season in lakes, not just for aquatic
plants, but for green and blue-green algae. If there is more plant biomass, there may also be an
increase in aerobic decomposition during the summer, a shorter period of stratification, and an overall
increase in water temperatures. Aquatic plant growth is likely to accelerate with increased
temperatures. Less snow and ice cover will allow plants to get a head start in the spring and become a
nuisance during the summer. The following illustration (Figure 1-5) shows that over the past 150 years
ice cover on four southern and two northern lakes occurs later and breaks up earlier. Lake Geneva did
not freeze during two winters around 2000. (Source: University of Wisconsin - Madison, Center for
Limnology. Prepared by Christina Wolbers; WICCI 2011c)
Figure 1–5. Changes in lake ice-cover around Wisconsin; reprinted from WICCI (2011c).
15
Potential Climate Change Impacts on Loons Breeding in Wisconsin
The USDA Forest Service Climate Change Bird Atlas predicts substantial change in common
loon distribution in the Upper Midwest under various carbon emission and climate change scenarios
(Figure 1-6; http://www.nrs.fs.fed.us/atlas/bird/RFbirdmod_0070.html). This atlas establishes the current
relationship between tree species and forest birds, then simulates future distribution of tree species and
associated birds. In the case of an aquatic waterbird such as the common loon, we feel the approach is
inappropriate and modeling the relationship between aquatic habitat and loons better informs predictions
of future distribution.
Figure 1–6. Projected changes in common loon distribution under various GCM/emission scenarios; from USDA
Bird Atlas http://nrs.fs.fed.us/atlas/bird/index.html.
16
For common loons, the impacts of climate change may be direct or indirect, or both. In most
cases, these impacts will be mediated through the marine and freshwater aquatic environment in which
they spend their entire life cycle. In the current project, we evaluate how changes in freshwater habitat
quality (specifically lake clarity) may impact common loon breeding distribution in Wisconsin. In the
following paragraphs we describe additional potential impacts of climate change on common loons.
Direct Impacts.
For wildlife species with a direct life history linkage to temperature, precipitation, and other
ambient conditions, direct impacts of climate change are of concern. With observed changes in climate
patterns, some wildlife populations are experiencing weather-climate conditions for which they are no
longer suited. There is a common set of direct climate impacts that will alter the behavior, distribution,
development, reproduction, and/or survival of many animal populations – and there is a suggestion that
some of these predicted changes may directly impact common loons:
• Advance of spring conditions – affecting migration, breeding, and life-cycle timing (phenology). During
recent years when lakes became ice-free early (March) in northern Wisconsin, common loons were not observed on the breeding grounds until early April, leaving territories undefended for several days. Typically, individual resident loons occupy specific territories for several years. Prior to ice-out, loons stage on nearby rivers and open water, flying over their breeding territories daily.
• Spatial shift in suitable climate conditions – affecting the distribution of a species on the landscape. Common loons are the most southerly nesting of five Gavia species breeding in the northern hemisphere – ranging from the arctic (yellow-billed loon to 78N latitude) south to 42N latitude (southern current extent of breeding common loons) (http://bna.birds.cornell.edu/bna). It is unknown whether current distribution is limited by critical temperature during breeding season or other factors.
• High temperature events – causing physiological stress or death. Common loons have highly dense feather plumage, likely insulating loons from the cold waters upon arrival on the breeding grounds. Due to lack of a brood patch, common loons incubate their eggs against their legs and upon their feet, and are often visibly stressed by heat while incubating - vacating their eggs often to enter the water when temperatures are in excess of 80o F (M. Meyer, pers. obs.).
• Altered snow cover – increasing exposure to cold. Snow cover will likely have no direct impact on breeding common loons in Wisconsin, however the length of ice-cover on lakes may.
• Drought – causing physiological stress or death. The current moderate-severe drought in northern Wisconsin is altering nest success in the population. Because adults are not suited for movement on land, they nest within 3-6 feet (0.9-1.8m) of the water’s edge. In some northern Wisconsin lakes, the water line has receded >20 feet (6m) over the past 5 years (M. Meyer, pers. obs.) and, consequently, suitable nesting sites are inaccessible. The drought in northern Wisconsin is the product of declining precipitation
17
(15-20%) over the past 50 years (Kucharik et al. 2010) and climate projections indicate more frequent and serve droughts (IPCC 2007). Both the increase in heavy precipitation events (3-5 days/decade, Kucharik 2010b) in the loon's range and the increasing frequency of drought may likely reduce loon productivity in Wisconsin (WICCI 2011d).
• Heavy precipitation/ flooding events – destroying habitat or injuring and killing wildlife. Heavy precipitation is a primary factor limiting reproductive success in loons. Resulting flooding and elevated water levels have the potential to destroy nests. Loons may renest, but when heavy rains occur in mid to late June, loons have no time for a second nesting attempt (McIntyre and Barr 1997). In Minnesota's Voyageurs National Park, on Rainy, Namakan, and Kabetogoma Lakes, an average of 60-70% of loon nests failed due to water level fluctuations (Reiser 1988) and in one year, flooding caused 53% of nest failures in New Hampshire (Taylor and Vogel, unpublished data). (WICCI 2011d)
Indirect Impacts.
The indirect impacts of climate change on wildlife are equally important to consider:
• Changes in habitat.The distribution and abundance of terrestrial wildlife species are largely defined by the type, amount, and quality of suitable vegetation. The response of vegetation to climate change may be rapid and how this will affect animal populations is a major concern (WICCI 2011d). Aquatic-dependent wildlife species are defined by their relationship with the type, chemistry, and size of aquatic habitat. During the breeding season, loons strongly prefer freshwater lakes >10ha, with moderate to deep lake depth, complex shorelines, high water clarity and pH, and low alkalinity and total phosphorus (Meyer 2006, Found et al. 2008, Kuhn et al. 2011). Climate change, with increased temperatures and altered precipitation patterns, has the potential to alter lake morphometry and chemistry, thus loon habitat quality. Loons are more frequently found on lakes with high water clarity (Figure 1-7)
• Species interactions. Climate change will likely alter how species interact with each other. This may break, intensify, or establish novel relationships among species with consequences to ecosystems and society. During nesting, loons are vulnerable to parasitism by black flies (Simulium euryadminiculum). It is common to observe loons covered with feeding black flies resulting in subsequent nest abandonment from the disturbance (McIntyre 1988). The documented impacts of black flies on other bird species include decreased productivity (Bukacinski and Bukacinska 2000) and the transmission of disease (Hunter et al. 1997). Simulium annulus (Lundström) (junior synonym S. euryadminiculum Davies), is host-specific, and feeds preferentially on the common loon (Adler et al. 2004, Weinandt et al. In Press). The presence of swarming, biting black flies directly affects loon productivity and fitness. Weinandt (2007) found increased disease prevalence in loons with high blood mercury levels and indicated black flies could be the vector. During warm springs, black fly swarms peak in mid-late May in northern Wisconsin, coinciding with the peak of loon nest initiation. In such spring conditions, nest abandonment from black fly predation is most severe (M. Meyer, pers. obs.). Because of the close linkage between insect populations and temperature and precipitation patterns (Bale et al. 2002), changes in the timing and severity of black fly outbreaks are anticipated (WICCI 2011d).
18
Figure 1–7. Relationship of territorial loon occupancy and lake water clarity (Secchi depth) in the Wisconsin
Northern Highlands Ecological Landscape (Meyer 2006).
Program Area Addressed
The results of this research project are relevant to Focus on Energy Environmental and
Economic Research and Development Program RFP research interest area C2, Impacts of climate
change on watersheds in Wisconsin, including changes in characteristics of lakes, rivers and wetlands,
and levels and supplies of ground water. This project develops a model that predicts how climate
change can produce changes in Northern Wisconsin lake hydrology that could ultimately impact lake-
dependent wildlife. This research will provide information to increase the preparedness of resource
managers for expected changes to inland glacial lake ecosystems.
19
The project builds upon a coupled ground-water/surface-water model (GSFLOW) developed by
the U.S. Geological Survey (USGS) for the Trout Lake, WI watershed and a Wisconsin Loon Habitat
Suitability model recently developed by the Wisconsin Department of Natural Resources (WDNR) and
USGS Upper Midwest Environmental Science Center using field data collected within the highest
density of breeding loons currently in Wisconsin (Northern Highlands Ecological Landscape) and the
most southerly distribution of breeding loons in Wisconsin (Central Wisconsin). This is a joint USGS
Upper Midwest Environmental Science Center, USGS Water Science Center, and WDNR Science
Service project.
References
Adler, P., D. Currie, and D. Wood. 2004. The Black Flies (Simuliidae) of North America. Comstock Pub. Associates, Ithaca, NY.
Bale, J. S., G. J. Masters, I. D. Hodkinson, C. Awmack, T. M. Bezemer, V. K. Brown, J. Butterfield, A. Buse, J.
C. Coulson, J. Farrar, J. E. G. Good, R. Harrington, S. Hartley, T. H. Jones, R. L. Lindroth, M. C. Press, I. Symrnioudis, A. D. Watt, and J. B. Whittaker. 2002. Herbivory in global climate change research: direct effects of rising temperature on insect herbivores. Global Change Biology 8:1-16.
Bukacinski, D, and M. Bukacinska. 2000. The impact of mass outbreaks of black flies (Simuliidae) on the
parental behaviour and breeding output of colonial common gulls(Larus canus). Annales Zoologici Fennici 37:43-49.
Found, C., S. Webb, and M. Boyce. 2008. Selection of lake habitats by waterbirds in the boreal transition zone of
northeastern Alberta. Canadian Journal of Zoology 86:277-285. Hunter, D., C. Rohner, and D. Currie. 1997. Mortality in fledgling great horned owls from black fly hematophaga
and leucocytozoonosis. Journal of Wildlife Diseases 33:486-491. IPCC, 2007. Climate Change 2007: Synthesis Report. Contribution of Working Groups I, II and III to the Fourth
Assessment Report of the Intergovernmental Panel on Climate Change. Pachauri, R. K. and Reisinger, A. (Eds). Cambridge University Press, Cambridge, UK
Kling, G.W., K. Hayhoe, L.B. Johnson, J.J. Magnuson, S. Polasky, S.K. Robinson, B.J. Shuter, M.M Wander,
D.J. Wuebbles, D.R. Zak, F.L. Lindroth, S.C. Moser, and M.L. Wilson. 2003. Confronting Climate Change in the Great Lakes Region : Impacts on our Communities and Ecosystems. Union of Concerned Scientists, Cambridge, Massachusetts, and Ecological Society of America, Washington, D.C.
Kucharik, C. J., D. J. Vimont, et al. 2010b. Wisconsin Initiative on Climate Change Impacts Climate Working
Group Report: Climate Change in Wisconsin. http://www.wicci.wisc.edu/report/Climate.pdf
20
Kuhn, A., J. Copeland, J. Cooley, H. Vogel, K. Taylor, D. Nacci, and P. August. 2011. Modeling habitat
associations for the Common Loon (Gavia Immer) at multiple scales in northeastern North America. Avian Conservation and Ecology 6(1): 4. [online] URL: http://www.ace-eco.org/vol6/iss/art4/
Magnuson, J.J., K.E. Webster, R.A. Assel, C.J. Bouser, P.J. Dillon, J.G. Eaton, H.E. Evans, E.J. Fee, R.I. Hall,
L.R. Mortsch, D.W. Schlinder, and F.H. Quinn. 1997. Potential effects of climate changes on aquatic systems : Laurentian Great Lakes and Precambrian Shield Region. Hydrological Processes 11:825-872.
Matthews, S.N., L. R. Iverson, A.M. Prasad, A. M., and M.P. Peters. 2007-ongoing. A Climate Change Atlas for
147 Bird Species of the Eastern United States [database]. http://www.nrs.fs.fed.us/atlas/bird, Northern Research Station, USDA Forest Service, Delaware, Ohio.
McIntyre, J.W. 1988. The Common Loon. Spirit of Northern Lakes. University of Minnesota Press, Minneapolis,
MN (USA). Meyer, M.W. 2006. Evaluating the Impact of Multiple Stressors on Common Loon Population Demographics -
An Integrated Laboratory and Field Approach. Final Report to U.S. EPA under EPA STAR Co-operative Agreement Number: R82-9085.
Reiser, M. H. 1988. Effects of regulated lake levels on the reproductive success, distribution and abundance of
the aquatic bird community in Voyageurs National Park, Minnesota. National Park Service Research/Resources Management Report MWR-13. Omaha, NB.
Schindler, D.W. 1997. Widespread effects of climatic warming on freshwater ecosystems in North America.
Hydrological Processes 11:1043-1067. WICCI 2011a Wisconsin's Changing Climate: Impacts and Adaptation. 2011. Wisconsin Initiative on Climate
Change Impacts. Nelson Institute for Environmental Studies, University of Wisconsin-Madison and the Wisconsin Department of Natural Resources, Madison, Wisconsin.
WICCI 2011b Wisconsin's Changing Climate: Impacts and Adaptation. 2011. Wisconsin Initiative on Climate
Change Impacts- Climate Working Group Report. http://www.wicci.wisc.edu/report/Climate.pdf Nelson Institute for Environmental Studies, University of Wisconsin-Madison and the Wisconsin Department of Natural Resources, Madison, Wisconsin.
WICCI 2011c Wisconsin's Changing Climate: Impacts and Adaptation. 2011. Wisconsin Initiative on Climate
Change Impacts- Water Resources Working Group Report. http://www.wicci.wisc.edu/report/Water-Resources.pdf Nelson Institute for Environmental Studies, University of Wisconsin-Madison and the Wisconsin Department of Natural Resources, Madison, Wisconsin.
WICCI 2011d Wisconsin's Changing Climate: Impacts and Adaptation. 2011. Wisconsin Initiative on Climate
Change Impacts- Wildlife Working Group Report. http://www.wicci.wisc.edu/report/Climate.pdf Nelson Institute for Environmental Studies, University of Wisconsin-Madison and the Wisconsin Department of Natural Resources, Madison, Wisconsin.
Weinandt, M.L., M. Meyer, M. Strand, and A.R. Lindsay. in press. Cues used by the black fly, Simulium
annulus, for attraction to the common loon (Gavia immer). J. Vector Ecology.
21
Chapter 2–Hydrologic and Lake Quality Modeling
JOHN F. WALKER1 , PAUL C. HANSON2 and RANDALL J. HUNT1
1U.S. Geological Survey, Wisconsin Water Science Center, Middleton, Wisconsin 2University of Wisconsin Center for Limnology, Madison, Wisconsin
Contents
Abstract ........................................................................................................................................................ 24
Introduction ................................................................................................................................................... 26
Methods ........................................................................................................................................................ 28
The Hydrologic Model ............................................................................................................................... 28
The Lake Quality Model ............................................................................................................................ 30
Climate Change Scenarios ....................................................................................................................... 34
Results .......................................................................................................................................................... 35
Hydrologic Model Calibration .................................................................................................................... 35
Hydrologic simulations .............................................................................................................................. 39
Lake Quality Simulations .......................................................................................................................... 48
References ................................................................................................................................................... 55
22
Tables
Table 2–1. Selected General Circulation Models (GCMs) used to simulate future climate conditions ..... 35
Table 2–2. Selected emission scenarios used to simulate future climate conditions ............................... 35
Figures
Figure 2–1. Calibration results for bi-weekly lake levels for the LTER lakes in the model ....................... 36
Figure 2–2. Calibration results for average lake levels for the LTER lakes (panel A) and other lakes
(panels B and C) in the model ...................................................................................................................... 37
Figure 2–3. Resulting climate forcings from the WICCI downscaled data for the three emissions
scenarios and the 6 GCMs ........................................................................................................................... 38
Figure 2–4. Final climate-change simulations for Allequash Lake showing precipitation minus
evaporation (panel A), net groundwater inflow (panel B), net surface-water inflow (panel C) and resulting
lake level (panel D) ....................................................................................................................................... 40
Figure 2–5. Final climate-change simulations for Big Musky Lake showing precipitation minus
evaporation (panel A), net groundwater inflow (panel B), net surface-water inflow (panel C) and resulting
lake level (panel D) ....................................................................................................................................... 42
Figure 2–6. Final climate-change simulations for Crystal Lake showing precipitation minus evaporation
(panel A), net groundwater inflow (panel B), net surface-water inflow (panel C) and resulting lake
level (panel D) ........................................................................................................................................... 43
Figure 2–7. Final climate-change simulations for Sparkling Lake showing precipitation minus
evaporation (panel A), net groundwater inflow (panel B), net surface-water inflow (panel C) and resulting
lake level (panel D) ....................................................................................................................................... 44
23
Figure 2–8. Final climate-change simulations for Edith Lake showing precipitation minus evaporation
(panel A), net groundwater inflow (panel B), net surface-water inflow (panel C) and resulting lake
level (panel D) ............................................................................................................................................. 46
Figure 2–9. Final climate-change simulations for Nichols Lake showing precipitation minus evaporation
(panel A), net groundwater inflow (panel B), net surface-water inflow (panel C) and resulting lake
level (panel D) ........................................................................................................................................... 47
Figure 2–10. Final climate-change simulations for Allequash Lake showing in-lake DOC concentrations
(panel A) and resulting secchi depths for 0, 10 and 25 percent increases in average in-lake Total
Phosphorus concentrations (panels B–D, respectively) ............................................................................... 49
Figure 2–11. Final climate-change simulations for Big Musky Lake showing in-lake DOC concentrations
(panel A) and resulting secchi depths for 0, 10 and 25 percent increases in average in-lake Total
Phosphorus concentrations (panels B–D, respectively) ............................................................................... 50
Figure 2–12. Final climate-change simulations for Crystal Lake showing in-lake DOC concentrations
(panel A) and resulting secchi depths for 0, 10 and 25 percent increases in average in-lake Total
Phosphorus concentrations (panels B–D, respectively) ............................................................................... 51
Figure 2–13. Final climate-change simulations for Sparkling Lake showing in-lake DOC concentrations
(panel A) and resulting secchi depths for 0, 10 and 25 percent increases in average in-lake Total
Phosphorus concentrations (panels B–D, respectively) ............................................................................... 52
Figure 2–14. Final climate-change simulations for Edith Lake showing in-lake DOC concentrations (panel
A) and resulting secchi depths for 0, 10 and 25 percent increases in average in-lake Total Phosphorus
concentrations (panels B–D, respectively) ................................................................................................... 53
Figure 2–15. Final climate-change simulations for Nichols Lake showing in-lake DOC concentrations
(panel A) and resulting secchi depths for 0, 10 and 25 percent increases in average in-lake Total
Phosphorus concentrations (panels B–D, respectively) ............................................................................... 54
24
Abstract
Climate change models predict warmer temperatures, changes to precipitation patterns,
and increased evapotranspiration in the Great Lakes region. Such climatic changes have altered,
and would further alter hydrological, chemical, and physical properties of inland lakes. The
Trout Lake basin has been the focus of numerous modeling studies that represent stages in the
development and refinement of a coupled groundwater/surface-water model using the U.S.
Geological Survey’s GSFLOW code. The GSFLOW model, which is driven by precipitation and
temperature inputs, was calibrated using the PEST suite of calibration software.
The lake water quality model was adapted from a simple lake carbon cycling equilibrium
model based on a manuscript in review. The model is the steady state solution to a relatively
simple differential equation. The load to a lake is the sum of areal input from the lake perimeter,
groundwater inflow, precipitation and surface water inflow. The mass load of carbon from GW
and SW is the product of their respective concentrations and their inflow volumes described
below. The areal load is assumed to be the sum of shoreline with canopy and shoreline without
canopy.
Given the uncertainty in climate modeling, it is desirable to use more than one GCM in
order to obtain a range of potential future climatic conditions. Daily precipitation and
temperature output from six GCMs were considered for one current and three future emissions
scenarios. The GCM output from the Intergovernmental Panel on Climate Change (IPCC) Fourth
Assessment Special Report on Emission Scenarios was downscaled for the Trout Lake area by
the Wisconsin Initiative on Climate Change. The downscaled climate data for the period 1961-
25
2100 was used as input to the GSFLOW model, and the corresponding lake budgets from the
GSFLOW runs were used as input for the Lake Model.
Drainage lakes in the watershed show relatively minor changes due to the future climate,
in part because drainage lakes are relatively low in the watershed, and have large contributing
areas relative to the lake area. Seepage lakes show a gradual decline in net percipitation
(precipitation minus evaporation) with time, a slight increase in net groundwater, and variable
net surface-water inputs, resulting in a relatively steady to moderate decline in water levels with
time. In almost all cases the variability of the various components increases in the later part of
the 21st century. Beginning around 2060 there appears to be an increase in variability and
relatively large changes in the net groundwater, which are reflected in the increase in variability
in the lake stages. Because the increase in net groundwater is less than the decrease in net
precipitation, the result is a declining water level.
The general trend in lakes was toward slightly higher water clarity, as indicated by
increased Secchi depth. In all lakes, except Allequash and Edith, DOC concentrations decreased.
Decreases in DOC had commensurate increases in clarity for these lakes, even under conditions
of elevated total phosphorus. The competing effects on clarity from DOC and chlorophyll is the
result of light extinction due to DOC which is 3 times greater than the light extinction due to
chlorophyll. Given the log relationship between total phosphorus and chlorophyll, the scenario
with a 25% increase in total phosphorus translates to approximately a 40% increase in
chlorophyll for these lakes. The ensuing decrease in water clarity was offset by a concurrent
decrease in DOC of 15% or more.
26
Introduction
Climate change models predict warmer temperatures, changes to precipitation patterns,
and increased evapotranspiration in the Great Lakes region. Such climatic changes have altered,
and would further alter hydrological, chemical, and physical properties of inland lakes. Previous
work has shown that climate change has the potential to alter the hydrologic cycle in Northern
Wisconsin and could have profound consequences for lakes (Walker et al., 2009). Earlier studies
have demonstrated that groundwater inflow is important in regulating the acid/base chemistry of
softwater seepage lakes (Anderson and Bowser 1986, Eilers et al. 1993), because the alkalinity
concentration of groundwater is much higher than that of precipitation. In Michigan’s Nevins
Lake, a drought-induced reduction in the input of solute-rich groundwater resulted in a
substantial decrease in cation concentration, leading to lower alkalinity and pH values (Webster
et al. 1990, Krabbenhoft and Webster 1995).
The U.S. Geological Survey GSFLOW model is an integration of the USGS
Precipitation-Runoff Modeling System (PRMS; Leavesley and others, 1983; Leavesley and
others, 2005) with the 2005 version of the USGS Modular Groundwater Flow Model
(MODFLOW-2005; Harbaugh, 2005). In GSFLOW, separate equations are coupled to simulate
horizontal and vertical flow through the soil zone, gravity-driven vertical flow through the
unsaturated zone, and three-dimensional groundwater flow through the saturated zone. GSFLOW
was designed to simulate the most important processes using a numerically efficient algorithm,
thus allowing coupled simultaneous simulation of flow in and across one or more watersheds.
GSFLOW incorporates physically based methods for simulating runoff and infiltration from
snow and rain precipitation, as well as groundwater/ surface water interaction. It is intended to
be used on watershed-scale problems that can range from a few square kilometers to several
27
thousand square kilometers, and for time periods that range from months to several decades
(Markstrom and others 2008).
The lake water quality model was adapted from a simple equilibrium model described in
Hanson et al (in review). The premise is that water clarity in this region is primarily a function of
colored dissolved organic carbon (DOC) and phytoplankton biomass (Carpenter et al. 1998,
Hanson et al. 2011), the proxy of which is chlorophyll concentration. Colored DOC is assumed
to be allochthonous in origin and chlorophyll is assumed to be driven by phosphorus
concentration. The challenge we face is that neither carbon nor phosphorus fluxes are explicitly
included in the future climate scenarios. Changes in the export of nutrients from terrestrial
ecosystems to lakes would likely derive from changes in land use and land cover (LU/LC), in
addition to meteorological forcings.
Our approach to the absence of terrestrial nutrient export in future scenarios is simple.
DOC cycling in our model includes three major fluxes: loads, which are a function of hydrology
and surrounding vegetation; internal processing, which is controlled, in part, by water
temperature; and export, which is a function of hydrology. The model is calibrated to current
conditions, then run under future climate scnearios, which have some influence on all three
fluxes because of the importance of hydrology and temperature. LU/LC remain static in our
simulations. Our approach to modeling chlorophyll is even simpler. We assume three different
phosphorus concentrations and calculate chlorophyll accordingly. This parsimonious approach
captures the major changes to water clarity in the lakes, and provides for a tractable analytical
solution that integrates easily into the results from the hydrologic modeling.
28
Methods
The Hydrologic Model
The Trout Lake basin has been the focus of numerous modeling studies (Cheng 1994,
Hunt and others 1998, Champion and Anderson 2000, Pint 2002, Pint and others 2003, John
2005; Muffels 2008; Hunt and others, 2008) that represent stages in the development and
refinement of a regional groundwater model, which will be used in future studies to address a
variety of research problems including the effects of climate change. We modified the model of
Muffels (2008) and Hunt and others (2008) for the work described here.
The general model construction is described briefly here and in detail in Walker et al. (in
review). An inset model was extracted from an analytic element model constructed for Trout
Lake area (Hunt and others, 1998). The regional-scale output was used to assign constant flux
boundary conditions along the perimeter of the inset model; the inset grid is 230 rows, 240
columns, where each grid cell is 75 m on a side. The model has 6 layers; 2 for each of the glacial
sediment units of Attig (1985). In the inset model of the basin, all streams inside the area of
interest (=nearfield) used the Streamflow Routing Package (SFR2 - Niswonger and Prudic,
2005), to allow accounting of streamflow, and limits to the amount of water a stream can lose to
the aquifer to the amount of water captured upstream. The MODFLOW Lake (LAK) Package
was used to simulate nearfield seepage and drainage lakes. Outside the area of interest
(=farfield), streams and lakes were simulated used the General Head Boundary (GHB) head-
dependent flux boundary condition. Simulation of recharge to the groundwater system is
automatically derived from the surface-water-model soil zone to the Unsaturated Zone Flow
(UZF) Package.
29
The surface-water modules included in the GSFLOW framework require that the model
domain be split into discrete subareas, known as hydrologic response units (HRUs). Each HRU
ideally has similar hydrologic characteristics, such as slope, vegetation, land use, or soil type.
The HRU configuration was generated by use of the GIS Weasel (Viger and Leavesley, 2007).
The 30m digital elevation map (DEM) of Wisconsin (Gesch and others, 2002) was clipped,
rotated, and resampled to coincide with the bounds of the groundwater model. Because the
hydrologic system is dominated by groundwater flow, a combination of surface elevation
differences and variations in the water table were used to construct the drainage network which
was further subdivided into the final HRUs.
The PRMS model contains nearly one hundred user-specified parameter values that can
be used to tailor the model to the specific area of study. Some of these parameters are more
important than others. Initial parameter values were generated by means of the GIS Weasel
(Viger and Leavesley, 2007). The GIS Weasel has been developed to assist in generating model
parameter values for PRMS and GSFLOW models; a project-specific digital elevation model is
used to generate HRUs and other physical model parameters, while general, nationwide soils and
landuse GIS datasets are used to generate other parameter values.
Because of the relative homogeneous nature of the soils and geology in the Trout Lake
watershed, it was felt that detailed spatial variability of the soil-zone and parameters controlling
runoff was not warranted. Instead, the HRUs were aggregated into 8 subbasins and a far-field
area. The initial parameter estimates derrived by the GIS Weasel were averaged over the
subbasins, and each HRU within a subbasin was assigned the average initial value.
The PRMS model uses external climate data as input to the hydrologic system. Daily
values of precipitation and temperature (maximum and minimum) are required inputs. For this
30
study, we chose to calculate solar radiation using the cloud-cover solar-radiation algorithm and
potential evapotranspiration using the Jenson-Haise formulation.
The closest National Weather Service Cooperative (COOP) weather station is located at
Rest Lake (National Weather Service station ID 477092), approximately 12 miles northwest of
Trout Lake. To alleviate the difficulties of missing data and spatial variability of precipitation, a
total of six COOP weather stations were chosen to serve as input to the model. The temperature
and precipitation from the six weather stations was distributed to the HRUs using an algorithm
based on the inverse of distance from the HRU centroid and each particular weather station (see
Walker et. al (in review) for details).
The growing season determines the period when evapotranspiration from the vegetation
portion of the HRUs can occur. It is specified in the model by two parameters: spring_frost and
fall_frost. The spring_frost parameter specifies the date for each HRU at the beginning of the
growing season. Likewise, the fall_frost parameter specifies the date for each HRU at the end of
the growing season. The two frost parameters were pre-processed using an algorithm described
in Christiansen and others (2011) and documented in Markstrom and others (2011). During
calibration the frost parameters were determined as average values for the calibration period.
During climate-change scenarios the frost parameters were determined for each simulation
period using input for each specific GCM and emissions scenario.
The Lake Quality Model
The model was adapted from another project on lake carbon cycling and is fully
described in a manuscript in preparation by Hanson et al (in review). An abbreviated description
follows.
31
The model is the steady state solution to a relatively simple differential equation. The
load to a lake is assumed to be a distribution with a mean value equal to the sum of areal input
from the lake perimeter (Plake), groundwater inflow (GWin) and precipitation (Precip), and
surface water inflow (SWin). Groundwater inflow and precipitation are combined because their
concentrations of DOC are relatively low and nearly equal at approximately 2 mg C L-1(Hanson
et al. 2004). The mass load of carbon from GW and SW is the product of their respective
concentrations and their inflow volumes described below. The areal load is assumed to be the
sum of two components, discriminated by the proportion of shoreline that is canopy. The first is
shoreline with canopy, which has an annual input of 1.5 g m-1(shoreline) y-1 (parameter AOC),
and the second is shoreline without canopy, which has an input of 0.2 g m-1(shoreline) y-1
(Preston et al. 2008). Because shoreline canopy data were not available for this project, we
assumed lakes had 90% canopy coverage around their perimeters.
The DOC load from groundwater and precipitation is given as
LoadGW&ppt = DOCGW[ ]• Precip+GWin( )
where
LoadGW&ppt is the load from groundwater inflow and precipitation in g/m2/y,
[DOCGW] is the concentration of DOC in groundwater in g/m3,
Precip is the total annual precipitation in m/y, and
GWin is the total annual groundwater inflow in m/y.
The DOC load from surface water is given as
LoadSW = DOCSW[ ]• LAKin + SWin( )
where
32
LoadSW is the load from surface-water inflow in g/m2/y,
[DOCSW] is the concentration of DOC in surface water in g/m3,
LAKin is the total annual direct runoff into the lake in m/y, and
SWin is the total annual streamflow in m/y.
The shoreline load from the tree canopy is given as
Loadshore =Plake • AOC[ ]•PC
Alake
where
Loadshore is the load from the shoreline canopy in g/m2/y,
Plake is the perimeter of the lake in m,
AOC is the shoreline load factor in g/m/y,
PC is the proportion of shoreline as canopy and wetlands, and
Alake is the surface area of the lake in m2.
The outflow factor is calculated as
Outflowfactor =GWout + SWout
Zmean
where
Outflowfactor is the inverse of the outflow residence time in 1/y,
SWout is the total annual streamflow from the lake in m/y, and
Zmean is the mean depth of the lake in m.
33
The temperature-adjusted retention factor is calculated as
Retentionfactor = RDOC *1.07
T−20( )
where
Retentionfactor is the retention of DOC in the lake in 1/y,
RDOC is the retention coefficient in 1/y, and
T is the average annual temperature in °C.
The DOC concentration is thus calculated as
DOC[ ] =LoadGW&ppt + LoadSW + Loadshore
Zmean • Outflowfactor + Retentionfactor( )
where
[DOC] is the concentration of DOC in the lake in g/m3,
LoadGW&ppt is the load from groundwater inflow and precipitation in g/m2/y,
LoadSW is the load from surface-water inflow in g/m2/y,
Loadshore is the load from the shoreline canopy in g/m2/y,
Outflowfactor is the inverse of the outflow residence time in 1/y,
Retentionfactor is the retention of DOC in the lake in 1/y, and
Zmean is the mean depth of the lake in m.
T is the average annual temperature in °C.
The Chlorophyl-a concentration is calculated from total phosphorus concentration as
Chl[ ] =10 1.583log10 TP[ ]−1.134( )
where
34
[Chl] is the chlorophyl concentration in the lake in mg/m3, and
[TP] is the total phosphorus concentration in the lake in mg/m3.
Total secchi depth is thus calculated as
Secchi = 1.45LECW + LECDOC • DOC[ ]+ LECChl • Chl[ ]
where
Secchi is the secchi depth in the lake in m,
LECW is the light extinction coefficient of water in 1/m,
LECDOC is the light extinction coefficient of DOC in 1/m/gC/m3,
[DOC] is the concentration of DOC in the lake in g/m3,
LECChl is the light extinction coefficient of chlorophyl in 1/m/mgChl/m3, and
[Chl] is the concentration of chlorophyl in the lake in mgChl/m3,
Climate Change Scenarios
Given the uncertainty in climate modeling, it is desirable to use more than one GCM in
order to obtain a range of potential future climatic conditions. Monthly precipitation and
temperature output from six GCMs were considered (Table 2–1). For each GCM, one current
and three future scenarios were used and are described in Table 2–2. The GCM output from the
Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Special Report on
Emission Scenarios (SRES) (IPCC, 2007) was downscaled for the Trout Lake area by the
Wisconsin Initiative on Climate Change Impacts (Notaro and others, 2011; WICCI, 2011). This
dataset was reformatted for input to the GSFLOW model.
35
Table 2–1. Selected General Circulation Models (GCMs) used to simulate future climate conditions
Model abbreviation Model ID Organization cccma_cgcm3_1 CGCM3.1(T47), 2005 Canadian Centre for Climate Modeling and Analysis, Canada
cnrm_cm3 CNRM-CM3, 2004 Centre National de Recherches Meteorologiques, France csiro_mk3_0 CSIRO-MK3.0, 2001 Commonwealth Scientific and Industrial Research Organization
(CSIRO) Atmospheric Research, Australia gfdl_cm2_0 GFDL-CM2.0, 2005 National Oceanic and Atmospheric Administration (NOAA)
Geophysical Fluid Dynamics Laboratory (GFDL) miroc3_2_medres 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_2a MRI-CGCM2.3.3, 2003 Meteorological Research Institute, Japan
Table 2–2. Selected emission scenarios used to simulate future climate conditions
Scenario Description
A1B Rapid economic growth, global population peaking in mid-century and declining thereafter, and introduction of new and efficient technologies with a balance across all sources
A2 Very heterogeneous world with self-reliance and preservation of local identies with gradual population growth, and slow regional economic growth and technological change
B1 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
The GSFLOW model was run for 18 combinations of the 6 GCMs and the 3 emissions
scenarios. Based on results from the model for several lakes, one of the GCM/scenario
combinations (cccma_cgcm3_1 GCM and the A2 emissions scenario) was deemed to produce
input that was too extreme for the GSFLOW model, resulting in nonsense predictions. This
GCM/scenario combination was eliminated from further consideration.
Results
Hydrologic Model Calibration
The final GSFLOW model was calibrated using the universal parameter estimation
computer code PEST (Doherty, 2010). The PEST optimization algorithm automatically adjusted
36
input parameters (for example, hydraulic conductivity, riverbed conductance) in a series of
model runs. After each model run, simulated model outputs (groundwater levels, vertical
gradients, streamflows) were automatically compared to equivalents measured in the field. Runs
continue until a best fit between simulated and measured targets was attained. Selected
calibration results are shown in Figures 2–1 and 2–2.
Figure 2–1. Calibration results for bi-weekly lake levels for the LTER lakes in the model
0 24 48 73 97 121
Observation number
493.0
493.5
494.0
494.5
495.0
Wat
er le
vel a
bove
MS
L, in
m
A. Allequash LakeN-S coefficient: 0.9497
0 21 42 64 85 106
Observation number
499.0
499.5
500.0
500.5
501.0
Wat
er le
vel a
bove
MS
L, in
m
B. Big Musky LakeN-S coefficient: 0.9838
0 22 43 65 86 108
Observation number
500.0
500.5
501.0
501.5
502.0
Wat
er le
vel a
bove
MS
L, in
m
C. Crystal LakeN-S coefficient: 0.9884
0 23 46 68 91 114
Observation number
494.0
494.5
495.0
495.5
496.0
Wat
er le
vel a
bove
MS
L, in
m
D. Sparkling LakeN-S coefficient: 0.0000
37
Figure 2–2. Calibration results for average lake levels for the LTER lakes (panel A) and other lakes (panels
B and C) in the model
The resulting climatic forcings for the three emissions scenarios across the GCMs are
depicted in Figure 2–3. In general, all of the climate forcings show a gradual increase of
maximum and minimum temperatures (panels A and B), with variability increasing in the latter
ms_
al
ms_
bm
ms_
cr
ms_
sp
ms_
tr490
493
496
499
502
505
Mea
n la
ke le
vel a
bove
MS
L, in
mA. LTER lakes
ms_
day
ms_
diam
ms_
edith
ms_
esc
ms_
fall
ms_
ffly
ms_
fran
k
ms_
jag
ms_
ljohn
ms_
lrock
ms_
lcan
oe
ms_
man
n
490
493
496
499
502
505
Mea
n la
ke le
vel a
bove
MS
L, in
m
B. Non-LTER lakes
ms_
nebi
sh
ms_
nich
ol
ms_
palle
t
ms_
rudy
ms_
star
et
ms_
stre
et
ms_
unam
ed
ms_
vcoo
k
ms_
wbi
rch
ms_
wild
w
490
495
500
505
510
515
Mea
n la
ke le
vel a
bove
MS
L, in
m
C. Non-LTER lakes
38
half of the 21st century. Note that maximum temperature increases faster than minimun
temperature. The temperature changes result in average increases in the growing season ranging
from 20-45 days. The results for precipitation (panel C) are less clear; there appears to be a
possible slight upward trend in precipitation, although the inter-annual variability far
overshadows the slight trend. As with temperature, the variability across the GCMs and
scenarios increases in the latter half of the 21st century, although not as much as with
temperature.
Figure 2–3. Resulting climate forcings from the WICCI downscaled data for the three emissions scenarios
and the 6 GCMs. Values shown are 10-year moving annual average values.
39
Hydrologic simulations
The resulting climate simulations for a drainage lake (Allequash) and selected seepage
lakes are depicted in Figures 2–4 through 2–9. In general, the drainage lakes show relatively
minor changes due to the future climate, in part because drainage lakes are relatively low in the
watershed, and have large contributing areas relative to the lake area. Secondly, the lakes in the
Trout Lake model have natural outlets to the stream, thus the outlet tends to hold the lake level
within a relatively narrow range; changes in the inputs to the lake are reflected in changes to the
flow from the outlet. Third, the drainage lakes generally have a net inflow of groundwater, thus
are less sensitive to changes in the groundwater system.
40
Figure 2–4. Final climate-change simulations for Allequash Lake showing precipitation minus evaporation
(panel A), net groundwater inflow (panel B), net surface-water inflow (panel C) and resulting lake level
(panel D)
41
The results for the seepage lakes tend to be similar, with subtle differences from lake to
lake. In general, the seepage lakes show a gradual decline in net percipitation (precipitation
minus evaporation) with time, a slight increase in net groundwater, and variable net surface-
water inputs, resulting in a relatively steady to moderate decline in water levels with time. In
almost all cases the variability of the various components increases in the later part of the 21st
century. Beginning around 2060 there appears to be an increase in variability and relatively large
changes in the net groundwater, which are reflected in the increase in variability in the lake
stages. Because the increase in net groundwater is less than the decrease in net precipitation, the
result is a declining water level.
A comparison of Big Musky and Crystal Lakes, two seepage lakes relatively high in the
flow system, indicates that the lake with a surface-water source (Big Musky) has a slightly
higher variability in lake stage in the latter portion of the 21st century, but the decline in the lake
level for the lake without surface-water input (Crystal) is a bit larger (Figure 2–5 and Figure 2–6,
respectively). Likewise, a seepage lake with a larger contribution from surface water (Sparkling
Lake, Figure 2–7) shows a more mitigated decline in lake level, in spite of a larger net
groundwater loss of water leaving the Trout Lake watershed.
42
Figure 2–5. Final climate-change simulations for Big Musky Lake showing precipitation minus evaporation
(panel A), net groundwater inflow (panel B), net surface-water inflow (panel C) and resulting lake level
(panel D)
43
Figure 2–6. Final climate-change simulations for Crystal Lake showing precipitation minus evaporation
(panel A), net groundwater inflow (panel B), net surface-water inflow (panel C) and resulting lake level
(panel D)
44
Figure 2–7. Final climate-change simulations for Sparkling Lake showing precipitation minus evaporation
(panel A), net groundwater inflow (panel B), net surface-water inflow (panel C) and resulting lake level
(panel D)
45
The results are somewhat different for Edith and Nichols Lakes, two seepage lakes high
in the flow system. The lake with a substantial contribution from surface water (Edith) shows a
higher variability of lake-stage change (Figure 2-8) compared to the lake without surface-water
input (Nichols; Figure 2-9). This is likely due to variability in the surface-water input to Edith
Lake, which reflects variability in the precipitation input. Note that in this comparison, Edith
Lake has a higher loss of water due to net groundwater, hence the response in lake level decline
is similar to Nichols Lake, although the overall variability in the response in lake level is lower
for the lake without surface-water input (Nichols). The differences in net groundwater output are
likely due to differences in the groundwater flow system in the vicinity of the two lakes.
46
Figure 2–8. Final climate-change simulations for Edith Lake showing precipitation minus evaporation (panel
A), net groundwater inflow (panel B), net surface-water inflow (panel C) and resulting lake level (panel
D)
47
Figure 2–9. Final climate-change simulations for Nichols Lake showing precipitation minus evaporation
(panel A), net groundwater inflow (panel B), net surface-water inflow (panel C) and resulting lake level
(panel D)
48
Lake Quality Simulations
Results of the lake-quality simulations for the climate-change scenarios for a drainage
lake (Allequash) and selected seepage lakes are depicted in Figures 2-10 through 2-15. In each
figure, resulting projections are shown for DOC (panel A), no increase in average in-lake total
Phosphorus (panel B) a 10 percent increase in total Phosphorus (panel C) and a 25 percent
increase in total Phosphorus (panel D). A 25-percent change in in-lake total Phosphorus was
considered an upper limit given the current level of development adjacent to the lakes. However,
if the development density were to increase considerably and land management were to result in
substantial increases of nutrients reaching the lakes, the increase in average in-lake Phosphorus
could be greater than 25 percent. The general trend in lakes was toward slightly higher water
clarity, as indicated by increased Secchi depth. In all lakes, except Allequash and Edith, DOC
concentrations decreased. Big Muskellunge and Nichols lakes showed marked decreases of
approximately 25%. Decrease in DOC had commensurate increase in clarity for these lakes, even
under conditions of elevated TP. Allequash and Edith lakes had little change in DOC and no
noticeable change in water clarity.
49
Figure 2–10. Final climate-change simulations for Allequash Lake showing in-lake DOC concentrations
(panel A) and resulting secchi depths for 0, 10 and 25 percent increases in average in-lake Total
Phosphorus concentrations (panels B–D, respectively)
50
Figure 2–11. Final climate-change simulations for Big Musky Lake showing in-lake DOC concentrations
(panel A) and resulting secchi depths for 0, 10 and 25 percent increases in average in-lake Total
Phosphorus concentrations (panels B–D, respectively)
51
Figure 2–12. Final climate-change simulations for Crystal Lake showing in-lake DOC concentrations (panel
A) and resulting secchi depths for 0, 10 and 25 percent increases in average in-lake Total Phosphorus
concentrations (panels B–D, respectively)
52
Figure 2–13. Final climate-change simulations for Sparkling Lake showing in-lake DOC concentrations (panel
A) and resulting secchi depths for 0, 10 and 25 percent increases in average in-lake Total Phosphorus
concentrations (panels B–D, respectively)
53
Figure 2–14. Final climate-change simulations for Edith Lake showing in-lake DOC concentrations (panel A)
and resulting secchi depths for 0, 10 and 25 percent increases in average in-lake Total Phosphorus
concentrations (panels B–D, respectively)
54
Figure 2–15. Final climate-change simulations for Nichols Lake showing in-lake DOC concentrations (panel
A) and resulting secchi depths for 0, 10 and 25 percent increases in average in-lake Total Phosphorus
concentrations (panels B–D, respectively)
An important driver of the decreasing DOC is the increasing water temperature (data not
shown). DOC mineralization increases exponentially with temperature, according to our assumed
Arrhenius relationship. What remains uncertain is the parameterization of that relationship. We
have assumed a theta of 1.07, which approximates a Q10 of 2. Recent work in non aquatic
55
systems suggest a globally consistent Q10 closer to 1.4 (Mahecha et al., 2010), which is a theta
of approximately 1.04. Had we used a lower theta, we would have seen less of a decrease in
DOC and less of an increase in water clarity.
The competing effects on clarity from DOC and chlorophyll can be understood by
comparing two parameters in the model. The light extinction due to DOC (LECDOC = 0.051 m-1
(gC m-3)-1) has nearly 3x the effect on clarity per unit change when compared to that of
chlorophyll (LECChl = 0.018 m-1 (mgChl m-3)-1). Given the log relationship between TP and Chl,
the scenario with a 25% increase in TP translates to approximately a 40% increase in chlorophyll
for these lakes. The ensuing decrease in water clarity would have to be offset by an approximate
15% decrease in DOC. Most simulations had a decrease in DOC of this magnitude or greater.
References
Attig, J.W. 1985. Pleistocene geology of Vilas County, Wisconsin. Wisconsin Geological and
Natural History Survey Information Circular 50.
Carpenter S, Cole J, Kitchell J. Impact of dissolved organic carbon, phosphorus, and grazing on
phytoplankton biomass and production in experimental lakes. Limnology and [Internet]. 1998
[cited 2012 Apr 16];43(I). Available from: http://www.jstor.org/stable/10.2307/2838941
Champion, G., and M.P. Anderson. 2000. Assessment of impacts on groundwater/lake systems
using MODFLOW with a lake package: Application to the Trout Lake Basin, northern
Wisconsin. Water Resources Institute, University of Wisconsin-Madison, Groundwater
Research Report WRI GRR 00-05.
Cheng, X. 1994. Numerical analysis of groundwater and lake systems with application to the
Trout River Basin, Vilas County, Wisconsin. Ph.D. dissertation, Department of Geology and
Geophysics, University of Wisconsin-Madison.Gelman, A., J.B. Carlin, H.S. Stern, and D.B.
56
rubin. 2004. Bayesian data analysis. Second edition. Chapman and Hall, Boca Raton, Florida,
USA.
Gesch, D., Oimoen, M., Greenlee, S., Nelson, C., Steuck, M. and Tyler, D., 2002, The national
elevation dataset: Photogrammetric Engineering and Remote Sensing, v. 68, p. 5–12.
Hanson, P. C., A. Pollard, D. L. Bade, K. Predick, S. R. Carpenter, and J. Foley. 2004. A model
of carbon evasion and sedimentation in temperate lakes. Global Change Biology. 10: 1285-
1298.
Hanson, P.C., S.R. Carpenter, J.A. Cardille, M.T. Coe, and L.A. Winslow. 2007. Small lakes
dominate a random sample of regional lake characteristics. Freshwater Biology. 52: 814-822.
Hanson, P.C., D.P. Hamilton, E.H. Stanley, N. Preston, O.C. Langman, E.L. Kara. 2011. Fate of
Allochthonous Dissolved Organic Carbon in Lakes: A Quantitative Approach. Plos One.
6:E21884. http://dx.plos.org/10.1371/journal.pone.0021884.
Hanson, P.C., I. Buffam, J. Rusak, E.H. Stanley, C. Watras. In review. A simple dissolved
organic carbon equilibrium model for lakes.
Harbaugh, A.W. 2005. MODFLOW-2005, the U.S. Geological Survey modular ground-water
model - the Ground-Water Flow Process: U.S. Geological Survey Techniques and Methods 6-
A16.
Hunt, R.J., Anderson, M.P., and Kelson, V.A., 1998, Improving a complex finite-difference
ground water flow model through the use of an analytic element screening model: Ground
Water, v. 36, no. 6, p. 1011-1017.
Hunt, R.J., Prudic, D., Walker, J.F., and Anderson, M.P., 2008, Importance of unsaturated zone
flow for simulating recharge in a humid climate: Ground Water, v. 46, no. 4, p. 551-560.
57
John, R. 2005. Simulating response rates and historical transience of surface water and
groundwater: Trout Lake Basin, northern Wisconsin. University of Wisconsin-Madison
Department of Geology and Geophysics Masters of Science Thesis, 153 p.
Leavesley, G.H., Lichty, R.W., Troutman, B.M., and Saindon, L.G., 1983, Precipitation-runoff
modeling system—User’s manual: U.S. Geological Survey Water-Resources Investigation
Report 83–4238, 207 p.
Leavesley, G.H., Markstrom, S.L., Viger, R.J., and Hay, L.E., 2005, USGS Modular Modeling
System (MMS)— Precipitation-Runoff Modeling System (PRMS) MMS-PRMS, in Singh, V.,
and Frevert, D., eds., Watershed Models: Boca Raton, Fla., CRC Press, p. 159-177.
Lunn, D.J., A. Thomas, N. Best, and D. Spiegelhalter. 2000. WinBUGS: a Bayesian modelling
framework: concepts, structure, and extensibility. Statistics and Computing 10:235-337.
Mahecha, M. D., Reichstein, M., Carvalhais, N., Lasslop, G., Lange, H., Seneviratne, S. I.,
Vargas, R., et al. (2010). Global convergence in the temperature sensitivity of respiration at
ecosystem level. Science (New York, N.Y.), 329(5993), 838-40. doi:10.1126/science.1189587
Markstrom, S.L., Niswonger, R.G., Regan, R.S., Prudic, D.E., and 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 6–D1, 240 p.
Muffels, C.T., 2008, Application of the LSQR Algorithm to the Calibration of a Regional
Groundwater Flow Model: Trout Lake Basin, Vilas County, WI: Madison, WI, University of
Wisconsin, M.S. Thesis, 106 p.
58
Niswonger, R.G., and D.E. Prudic. 2005. Documentation of the Streamflow-Routing (SFR2)
Package to include unsaturated flow beneath streams – a modification to SFR1: U.S.
Geological Techniques and Methods 6-A13.
Notaro, M., D. Lorenz, D. Vimont, S. Vavrus, C. Kucharik, and K. Franz, 2011, 21st century
Wisconsin snow projections based on an operational snow model driven by
statistically downscaled climate data, International Journal of Climatology, 31, 1615-1633.
Patrick, R. and C.W. Reimer. 1966. The Diatoms of the United States. Volume 1. Monograph 13,
Academy of Natural Sciences of Philadelphia, Philadelphia, Pennsylvania, USA. 688 pp.
Pint, C.D., 2002, A Groundwater Flow Model of the Trout Lake Basin, Wisconsin: Calibration
and Lake Capture Zone Analysis: Madison, WI, University of Wisconsin, MS, 123 p.
Pint, C.N., Hunt, R.J., and Anderson, M.P., 2003, Flow Path Delineation and Ground Water Age,
Allequash Basin, Wisconsin: Groundwater, v. 41, no. 7, p. 895-902.
Preston ND, Carpenter SR, Cole JJ, Pace ML. 2008. Airborne carbon deposition on a remote
forested lake. Aquatic Sciences 70:213-224.Anderson, M. P., and C. J. Bowser. 1986. The role
of groundwater in delaying lake acidification. Water Resour. Res. 22:1101-1108.
Viger, R.J. and Leavesley, G.H., 2007, The GIS Weasel User’s Manual: U.S. Geological Survey
Techniques and Methods 6-B4, http://pubs.usgs.gov/tm/2007/06B04/.
Walker, J.F., Hunt, R.J., Markstrom, S.L., Hay, L.E., and Doherty, J., 2009, Using a Coupled
Ground-Water/Surface-Water Model to Predict Climate-Change Impacts to Lakes in the Trout
Lake Watershed, Northern Wisconsin, in 3rd Interagency Conference on Research in the
Watersheds, Estes Park, CO, U.S. Geological Survey, p. 155-161.
WICCI, 2011, Wisconsin’s Changing Climate: Impacts and Adaptation, Wisconsin Initiative on
Climate Change Impacts, Nelson Institute for Environmental Studies, University of
59
Wisconsin-Madison and the Wisconsin Department of Natural Resources, Madison,
Wisconsin. 217 pp.
60
Page intentionally left blank
61
Chapter 3– Historic Changes in Lake Condition
PAUL J. GARRISON1, SAMANTHA KAPLAN2, AND GINA LALIBERTE1
1Wisconsin Department of Natural Resources, Madison, WI 2University of Wisconsin-Stevens Point. Stevens Point, WI
Contents
Abstract ....................................................................................................................................................................... 64
Introduction .................................................................................................................................................................. 65
Methods ....................................................................................................................................................................... 67
Results ......................................................................................................................................................................... 69
Lithology and chronology ......................................................................................................................................... 69
Loss on ignition ........................................................................................................................................................ 70
Diatom community ................................................................................................................................................... 72
Charcoal ................................................................................................................................................................... 78
Discussion ................................................................................................................................................................ 82
References .................................................................................................................................................................. 85
Figures
Figure 3–1. Photographs of the 5 drives from the Max Lake sediment core. The first 2 drives were largely
uniform in color and mucky in texture. The third drive displayed more banding. The fourth drive was composed
62
largely of clay material and was reddish brown in color. The fifth drive was composed largely of sand with some silt
near the top and likely consists of material deposited as the glacier receded. ............................................................. 70
Figure 3–2. Loss on ignition values are typical from many lakes in northern Wisconsin. Silicates/ash are high in
the Late Glacial and Early Holocene due to detrital siliclastics material in glacial meltwater and outwash.
Ameliorating temperatures circa 11,000 yr BP resulted in increased in lake productivity and an elevated proportion
of organic matter in the lake sediment throughout the Holocene. ................................................................................ 71
Figure 3–3. Loss on ignition values for the last two centuries, like the CHAR records for this period, also reflect
an anthropogenic pattern of activity. An increase in silicates/ash are the likely result of land-disturbing activities
that contribute slope-wash and soil to the lake basin. The decrease in organic matter is likely just relative to the
increase in siliciclastic material and not reflective of reduced primary productivity in the lake. .................................... 72
Figure 3–4. Cluster analysis of the diatom samples in the sediment core. The analysis used the Ward’s
clustering method with squared chord distance. The scale is the square root of the squared chord distance. The
table on the right indicates calibrated 14C years BP. .................................................................................................... 73
Figure 3–5. Principal Components Analysis ordination plot of downcore samples. Circled samples correspond
with the 4 groups identified with the cluster analysis. Groups 1 and 2 have higher pH and phosphorus values. The
ordination plot shown in the inset is a Redundancy Analysis (RDA) illustrating the relationship between two
environmental variables, pH and phosphorus. ............................................................................................................. 75
Figure 3–6. Relative abundance of common diatom taxa. The four color groups are the 4 groups identified in the
cluster analysis. ........................................................................................................................................................... 76
Figure 3–7. Diatom inferred pH and phosphorus for the portion of the core that was dated. Both pH and
phosphorus levels declined after the mid-Holocene warming period ........................................................................... 77
Figure 3–8. Charcoal accumulation rates (CHAR) for Max Lake from the Late Glacial to the present. CHAR
values are lowest in the late-Glacial and early Holocene and highest in the last 2000 years. ..................................... 79
63
Figure 3–9. Max Lake CHAR record excluding the last 2000 years to better illustrate charcoal variability in the
early and mid-Holocene. The largest charcoal accumulation rates in the Holocene are centered on 10,000 years
ago, 7000 years ago and 4000 years ago. .................................................................................................................. 79
Figure 3–10. Max Lake CHAR values for the last 2000 years. During this interval the most wildfire activity in the
Max Lake basin occurred between 1100 and 550 yrs BP, concurrent with the Medieval Warm Period and likely
reflecting regional drought. .......................................................................................................................................... 80
Figure 3–11. Max Lake CHAR values for the same period of record (previous figure) illustrating the relative
contribution of charcoal pieces >250 µm and those between 125 µm and 250 µm in diameter. An elevation of the
finer charcoal during the Medieval Warm Period suggests that fires were regional in nature and consistent with
drought, versus local, isolated phenomena.................................................................................................................. 80
Figure 3–12. Max Lake CHAR values for the same period of record as (previous two figures) showing the
contribution charcoal derived from arboreal/canopy component of vegetation and charcoal from grasses and
understory vegetation. During the Medieval Warm Period (1100 and 550 yrs BP) charcoal in Max Lake suggests
fires were larger arboreal/canopy events. .................................................................................................................... 81
Figure 3–13. In the last 200 years CHAR in Max Lake reflects human activities in the area. In the early and
middle parts of the 19th Century very little charcoal accumulated in the lake, but logging at the turn of the 20th
Century as well as settlement in the area has resulted in increased fire frequencies and elevated CHAR. A rise in
CHAR values in the last two decades may reflect a period of sustained drought in the region. ................................... 82
Figure 3–14. Annual PDSI Index for the instrumental record for northern Wisconsin. Negative PDSI values
represent drought conditions. Years with low regional precipitation and low PDSI values occur in the early 1930's
and mid-1970's, both corresponding to episodes of increased CHAR (Figure 3-13). .................................................. 82
Figure 3–15. Hypothesized extent of lake basin that Max Lake was a part of prior to the mid-Holocene warming
period. At the present time most of this historical lake basin is a wetland.................................................................... 84
64
Abstract
Climate change models predict warmer temperatures and increased evapotranspiration in the
Great Lakes region. During the mid-Holocene (5-7000 years ago) the climate in the Great Lakes region
was warmer and effectively drier than at the present. During this time the climate was similar to that
predicted by climate change models. A sediment core was extracted from Max Lake, Vilas County, WI
to use the ecological changes experienced during the mid-Holocene dry period to assess the impact of
future climate change.
The sediment core encompassed the entire 13,000 year history of the lake since the retreat of the
glaciers. The sediment was examined for charcoal to assess the general climate of the region. An
increase in the frequency of wildfires indicates a drier and perhaps warmer climate. The diatom
community was examined to determine changes in pH and phosphorus levels. These variables provide
an estimate of the hydrologic regime of the lake.
During the first few centuries following the establishment of Max Lake, it was a drainage lake
that was likely part of a much larger lake. The pH levels were consistent with those found at the present
time in headwater drainage lakes. Charcoal levels were relatively low indicating a cool, moist climate.
Around 7000 years ago the frequency of wildfires greatly increased signally the beginning of the mid-
Holocene warming period. At this time effective reduction in precipitation caused the lake to change
from a drainage to a seepage lake. The change in the hydrologic regime resulted in a decline in the pH
values of about 0.5 units. Around 6000 years ago, low charcoal levels indicate a return to moist
conditions and the lake briefly returned to a drainage lake. This was short lived and for the last 6000
years the lake has been a seepage lake with lower pH values and lower phosphorus concentrations. Even
65
though the charcoal indicates a drier climate during the last four millennia, Max Lake has remained a
seepage lake.
This study demonstrates that if the future climate becomes warmer and effectively drier as
predicted, there may be substantial changes in the ecology of headwater drainage lakes. These lakes
may become seepage lakes resulting in potentially lower pH and nutrient levels which may result in a
simplification of the fish community and a reduction in the lake’s overall productivity. The ecology of
some of these lakes may be altered sufficiently to impact common loon (Gavia immer) habitation as
their food resources are negatively impacted.
Introduction
Global climate models for the Upper Midwest indicate a warmer and drier climate with more
intensive precipitation events will likely occur in the future in Wisconsin (Magnuson et al. 1997, WICCI
2011). Such climatic changes have altered, and are expected to further alter hydrological, chemical, and
physical properties of inland lakes. Earlier studies have demonstrated that groundwater inflow is
important in regulating the acid/base chemistry of softwater seepage lakes (Anderson and Bowser 1986,
Eilers et al. 1993), because the alkalinity concentration of groundwater is much higher than that of
precipitation. In Michigan’s Nevins Lake, a drought-induced reduction in the input of solute-rich
groundwater resulted in a substantial decrease in cation concentration, leading to lower alkalinity and
pH values (Webster et al. 1990, Krabbenhoft and Webster 1995).
Max lake, Vilas County, is a seepage lake in the Northern Highlands ecoregion. It is not used by
territorial common loons though individual loons are observed foraging on the lake – likely a function
of its small size and lack of suitable nesting habitat. If it indeed was a drainage lake in the past, as
implied by the core evaluation, it very well may once have been used by territorial loons, but now is not
due to small size and nest habitat limitations – however this must remain speculation.
66
The objective of this study was to examine changes in the historical water chemistry of Max
Lake, Vilas County, Wisconsin, which is a softwater seepage lake. At the present time, this lake
receives low amounts of groundwater inflow and thus its acid-base chemistry is susceptible to alteration
by changing climate conditions. Under wet conditions more groundwater may enter the lake and result
in increased pH values, but with a reduction in groundwater input pH should decline. During the mid-
Holocene the climate in the Upper Midwest was warmer and effectively drier than at the present time
(Valero-Garcés et al. 1997). By analyzing a sediment core from Max Lake for a time period covering
the Holocene (14,000 years BP to the present) we intended to document changes in the acid-base
chemistry of the lake and thus better predict how a future warmer climate with increased
evapotranspiration will affect softwater seepage lakes.
This study used the diatom community to reconstruct the pH and nutrient (phosphorus) history
of Max Lake since its creation after the recession of the midcontinent glacier over 14,000 years ago.
Charcoal analysis was used to estimate the local climate, as a drier climate results in the increased
incidence of forest fires.
Fire frequency in north-temperate, forested ecosystems is dependent on fuel loads, ignition
source, and antecedent moisture in soils and fuels. Increased wildfire activity is positively correlated to
periods of moisture stress, while the size of any one fire will depend on moisture stress and available
fuels. Sedimentary charcoal was used to identify episodes of past fire in the Max Lake watershed.
Numerous studies demonstrate that charcoal >125 μm is abundant in lakes found in a 10 km radius of a
wildfire and all but absent at greater distances (Whitlock and Millspaugh 1996, Whitlock and Larsen
2001). Charcoal >250 μm is generally considered too heavy to be wind transported and represents
fluvial or slopewash transport of material from fires in the immediate drainage basin (Carcaillet et al.
2001).
67
Methods
A long core was collected 26 and 27 March 2009 from the deep area of Max Lake. The total
length of the core was 5.515 meters. The core penetrated into the glacial outwash of clay and sand,
resulting in recovery of a record spanning circa 13,000 years. Split core sections are archived in the
National Lacustrine Core Repository, University of Minnesota. An additional shorter core (117 cm)
through the sediment water interface was collected nearby and extruded into 1 and 2 cm increments.
The chronology of the long core was based on nine accelerator mass spectrometry (AMS) radiocarbon
dates determined at the Center for Applied Isotope Studies, University of Georgia. AMS dates were
converted to calendar years before present (BP). A polynomial function derived from the converted 14
C
dates was used to establish the sediment chronology. The short core was used to establish the
chronology for the last 180 years using 210
Pb analysis aand the constant rate of supply (CRS) model
(Appleby and Oldfield 1978).
Loss on ignition (LOI) was performed by extracting 1 cc of lake core sediment at increments of
1cm. The wet samples were dried in a 50°C oven overnight to establish water content. Dried samples
were burned at 550°C for 4 hours to establish organic matter content, and then burned at 950°C for 2
hours to derive the proportion of carbonates. Organic matter generally corresponds to levels of lake
productivity and algal growth, while carbonates can arise from various factors affecting base saturation
of lake water and calcite abundance. Seepage lakes in northern Wisconsin typically have low calcite
saturations due to the absence of carbonate bedrock. Residual minerals, mostly silicates, typically
reflect the proportion of terrigenous materials which have washed into the lake.
Diatoms are a type of algae which possess siliceous cell walls and are usually abundant, diverse,
and well preserved in sediments. They are especially useful for reconstructing past lake conditions as
they are ecologically diverse and their ecological optima and tolerances can be quantified. Samples for
68
diatom analysis were cleaned with hydrogen peroxide and potassium dichromate (van der Werff 1956).
Cover slips on which a portion of the diatom suspension was dried were mounted on microscope slides
with Naphrax®. Specimens were identified and counted under oil immersion objective (1400X) until at
least 600 valves had been encountered. Diatoms were identified to species level whenever possible
using references which included Patrick and Reimer (1966, 1975), Krammer and Lange-Bertalot (1986,
1988, 1991a,b), Camburn and Charles (2000), Krammer (2000), Lange-Bertalot (2001), and Siver et al.
(2005) as well as primary species literature.
The diatom data were analyzed using a variety of multivariate statistical techniques.
Ecologically relevant statistical methods have been developed to infer environmental conditions from
diatom assemblages. These methods are based on multivariate ordination and weighted averaging
regression and calibration (Birks et al. 1990). Ecological preferences of diatom species are determined
by relating modern limnological variables to surface sediment diatom assemblages. The species-
environment relationships are then used to infer environmental conditions from fossil diatom
assemblages found in the sediment core. A modified form of weighted averaging was used to
reconstruct the pH and phosphorus history during the Holocene period in Max Lake. Phosphorus and
pH preferences for individual taxa, cited in Siver et al. (2005), were applied to the diatom community
found in the core to estimate historical trajectories of pH and phosphorus.
A cluster analysis was performed to explore for unique stratigraphic areas based upon the diatom
community composition. Ward’s clustering method was used with squared chord distance in the
program R (R Development Core Team 2006). An exploratory detrended correspondence analysis
showed that the gradients of species responses were relatively short so a linear model was used for
unconstrained (principal components analysis; PCA) and constrained (redundancy analysis; RDA)
69
ordinations (CANOCO 4.5 software, ter Braak and Smilauer 2002). The environmental variables used
for the RDA analysis were diatom-inferred phosphorus and pH.
When identifying potential periods of past drought from charcoal in lake sediment it is important
to identify not only the abundance of charcoal in any given level, but the frequency of influx peaks (of
any size) over a period of time. For this reason, fine sampling resolution of sediments is required in
order to identify episode of past drought. Charcoal isolation and counting followed the method of
Huber and Markgraf (2003) as adapted from Whitlock and Millspaugh (1996). 1 cc samples were taken
contiguously every 0.5 cm along the entire Max Lake core. Samples were bleached with 6% hydrogen
peroxide, washed with distilled water through 250μm and 125μm sieves, and allowed to dry in a 90°C
oven. Upon completion of drying, careful inspection of charcoal pieces was performed in which total
number in each size class was counted and individual morphology was noted. Morphological
classifications follow those established by Umbanhowar and McGrath (1998) and Jensen et al. (2005)
and allow charcoal to be categorized as having been derived from arboreal/canopy or grass/understory
vegetation. Charcoal concentrations (number of particles/cm3 sediment) were converted to charcoal
accumulation rates (CHAR) by dividing the concentration by the number of years per centimeter as
derived through radiocarbon and lead-210 dating.
Results
Lithology and chronology
The total length of the core was 5.5 m which was collected in 5 drives. The deepest drive (4.6-
5.5 m) consisted mostly of coarse and fine-grained sand (Figure 3-1) and likely represented glacial
outwash. The next shallower drive (4.2-4.6 m) consisted mostly of fine material largely composed of
clay and silt. The third drive (2.7-4.2 m) consisted of gyttja (muck) with intermittent different colored
70
banding which likely represents high energy hydrologic events. The top two drives (0-2.7 m) were
mucky and did not possess colored banding.
The deepest sample that was radiocarbon dated was at 4.6 m and revealed a calibrated age of
12,827 BP. Max Lake was apparently established about 13,000 years ago as the midcontinent glacier
receded.
Figure 3–1. Photographs of the 5 drives from the Max Lake sediment core. The first 2 drives were largely uniform in
color and mucky in texture. The third drive displayed more banding. The fourth drive was composed largely of
clay material and was reddish brown in color. The fifth drive was composed largely of sand with some silt near
the top and likely consists of material deposited as the glacier receded.
Loss on ignition
Loss on ignition measures organic matter in the sample. Loss on ignition follows a pattern
typical of lakes in this region with very low carbonate values throughout (Figure 3-2). Ash/silicates are
71
very high in the late-Glacial and early Holocene due to detrital siliciclastic material in the glacial
meltwater and outwash. When temperatures increased in the Holocene, lake productivity rapidly
increased and has remained high with further increases in the last 4000 years. While carbonate values
are negligible, they do parallel the pattern of organic matter, except with more variability in the last
4000 years, and are probably tied to a combination of base cation saturation of groundwater and algal
productivity. An episode of lowered carbonate percentages corresponds to an increase in ash/silicates
circa 2500 years BP and may indicate a drought event in which there were increased contributions of
terrigenous materials to the lake.
Figure 3–2. Loss on ignition values are typical from many lakes in northern Wisconsin. Silicates/ash are high in the
Late Glacial and Early Holocene due to detrital siliclastics material in glacial meltwater and outwash.
Ameliorating temperatures circa 11,000 yr BP resulted in increased in lake productivity and an elevated
proportion of organic matter in the lake sediment throughout the Holocene.
72
In the last 200 years loss on ignition values reflect human activity in the lake basin (Figure 3-3).
Increases in ash/silicates likely correspond to human activities and mobilization of detrital material into
the lake through logging and land clearing. The decrease in organic matter is probably a result of the
relative increase in ash/silicates and not a reduction in lake productivity. Carbonates are absent from the
sediment in the last 200 years.
Figure 3–3. Loss on ignition values for the last two centuries, like the CHAR records for this period, also reflect an
anthropogenic pattern of activity. An increase in silicates/ash are the likely result of land-disturbing activities
that contribute slope-wash and soil to the lake basin. The decrease in organic matter is likely just relative to
the increase in siliciclastic material and not reflective of reduced primary productivity in the lake.
Diatom community
Diatoms were only preserved in the upper 4.3 m (11,775 yrs BP) of the core and were generally
well preserved and abundant. Below this depth, diatoms either were not present or were too poorly
preserved to be identified.
73
The core samples were classified into 4 groups in cluster analysis, based upon the similarities of
the diatom samples (Figure 3-4). Groups 1 and 2 were interspersed at the bottom of the core and were
deposited between 7200 and 12,000 years ago. The third group represented more recent deposition
between 700 and 3700 years ago. The fourth group was deposited between 3700 and 7200 years ago but
also included the very upper part of the core.
Figure 3–4. Cluster analysis of the diatom samples in the sediment core. The analysis used the Ward’s clustering
method with squared chord distance. The scale is the square root of the squared chord distance. The table on
the right indicates calibrated 14C years BP.
74
Ordination of the fossil diatom assemblages using PCA revealed temporal trends that supported
those found in the cluster analysis. The sample scores were divided into four groups. The most widely
separated groups were from the early- to mid-Holocene and later Holocene (Figure 3-5). The other two
groups were clustered near each other but did not overlap. These groups represented much of the
Holocene except for the last 3.5 centuries. The first PCA axis explained 36.1% of variation in species
while the second axis explained an additional 21.1% of the variation. Ordination using RDA suggests
that the influence of phosphorus and pH were similarly important in ordering the diatom community but
their influence was different enough to allow reconstruction of both these variables. Groups 1 and 2
which encompassed the first half of the Holocene had higher phosphorus and pH values. Groups 3 and 4
from the diatom community deposited after 7200 years BP had lower phosphorus and pH values.
75
Figure 3–5. Principal Components Analysis ordination plot of downcore samples. Circled samples correspond with
the 4 groups identified with the cluster analysis. Groups 1 and 2 have higher pH and phosphorus values. The
ordination plot shown in the inset is a Redundancy Analysis (RDA) illustrating the relationship between two
environmental variables, pH and phosphorus.
The diatom flora was diverse throughout the core. Navicula heimansoides was the only taxon
found throughout the entire core except for the two earliest samples (Figure 3-6). The diatom diversity
indicates that Max Lake has experienced significant environmental changes throughout its 13,000 year
history. The diatom community in the first half of the Holocene consisted largely of taxa which prefer
circumneutral pH and moderate nutrient concentrations. Pseudostaurosira trainorii was common and
Siver et al. (2005) reported it from slightly acidic to slightly alkaline and mostly mesotrophic ponds on
Cape Cod. The flora found in the sections from 3700 to 600 years BP was representative of acidic, low
nutrient environments. The diatom Aulacoseira pseudoamericana was primarily found in this part of the
76
core (Figure 3-6). The diatom community in the latter half of the Holocene outside the years 6200 to
5900 years BP was more diverse than 3700 to 700 years BP, and no single taxon was dominant in these
sections of the core.
Figure 3–6. Relative abundance of common diatom taxa. The four color groups are the 4 groups identified in the
cluster analysis.
The majority of the pH and phosphorus reconstructions were based on the use of over 80 percent
of the diatom community. At a few depths where the community was more diverse a smaller percentage
of the taxa were used, but it was always greater than 60 percent.
In the first half of the Holocene the pH was greater than 6.5 but it declined to around 6.1 by 7400
years ago, and concurrently there was a large decline in the pH of 0.5 units (Figure 3-7). Higher pH
values returned for a short time around 6000 years ago, then pH values declined by about 1 pH unit
77
lower in the second half of the Holocene. The pH further declined to 5.3 2000 years ago but then
increased to 5.4-5.5 near the top of the core.
Figure 3–7. Diatom inferred pH and phosphorus for the portion of the core that was dated. Both pH and phosphorus
levels declined after the mid-Holocene warming period
The nutrient status of the lake was more variable than the pH with low LOI-estimated
productivity in the late-glacial period even with phosphorus levels increasing 9000 years BP to 15
µg L-1
(Figure 3-7). Between 9000 and 2000 years BP the lake’s productivity declined to 9 µg L-1
. For
the last two centuries phosphorus concentrations have increased, however they remain below
concentrations present in the early Holocene.
78
Charcoal
Charcoal concentrations in Max Lake are low compared to lakes in the southern and western part
of Wisconsin and likely reflect the higher snowfall levels, colder temperatures, and generally more
mesic conditions in the northernmost part of the state. However, nearby Sparkling and Allequash Lakes
do have elevated charcoal concentrations in the last 300 years (Jennifer Schmidt personal
communication). Charcoal in Max Lake was isolated and classified by size and as grass/understory or
arboreal/canopy based on morphology. Charcoal that derives from grass, forbs and understory
vegetation has a porous, fibrous or cellular appearance, while charcoal originating from wood, bark and
tree leaves will be dark and glossy and/or have a lattice, spongy or branched structure.
Overall, charcoal accumulation rates (CHAR) are lowest in the late-Glacial and early Holocene
and highest in the last 2000 years (Figure 3-8). With the exception of the last 2000 years, the largest
charcoal peaks in the Holocene are centered on 10,000 years ago, 7000 years ago and 4000 years ago
(Figure 3-9). In the last 2000 years the most wildfire activity in the Max Lake basin occurred between
1100 and 550 years BP, overlapping the Medieval Warm Period (MWP) (Figure 3-10). The MWP was
a documented drought event occurring between 700-1450 AD which extended throughout the western
states and into the western Great Lakes region (Cook et al., 2004; Booth et al, 2006). While the number
of large charcoal particles (>250μm) are at a minimum, suggesting fires within the immediate Max-
Lake basin were limited during this interval, the sustained increase in the number of small charcoal
particles (>125µm) indicates a prolonged period of regional wildfire (Figure 3-11). Morphologically,
the majority of charcoal from this period is arboreal, meaning larger crown fires predominated (Figure
3-12).
79
Figure 3–8. Charcoal accumulation rates (CHAR) for Max Lake from the Late Glacial to the present. CHAR values
are lowest in the late-Glacial and early Holocene and highest in the last 2000 years.
Figure 3–9. Max Lake CHAR record excluding the last 2000 years to better illustrate charcoal variability in the early
and mid-Holocene. The largest charcoal accumulation rates in the Holocene are centered on 10,000 years
ago, 7000 years ago and 4000 years ago.
80
Figure 3–10. Max Lake CHAR values for the last 2000 years. During this interval the most wildfire activity in the Max
Lake basin occurred between 1100 and 550 yrs BP, concurrent with the Medieval Warm Period and likely
reflecting regional drought.
Figure 3–11. Max Lake CHAR values for the same period of record (previous figure) illustrating the relative
contribution of charcoal pieces >250 µm and those between 125 µm and 250 µm in diameter. An elevation of
the finer charcoal during the Medieval Warm Period suggests that fires were regional in nature and consistent
with drought, versus local, isolated phenomena.
81
Figure 3–12. Max Lake CHAR values for the same period of record as (previous two figures) showing the
contribution charcoal derived from arboreal/canopy component of vegetation and charcoal from grasses and
understory vegetation. During the Medieval Warm Period (1100 and 550 yrs BP) charcoal in Max Lake
suggests fires were larger arboreal/canopy events.
The Lead-210 age model spans the period from 1810 to the present and captures the time
immediately prior to European settlement in the area. Not surprisingly, fire activity is much lower
during the first half of this interval, but increases dramatically in the late 1800's through the present
when the human presence increased (Figure 3-13). Charcoal from logging in the late 1800's followed
by additional land clearing and settlement in the 20th century likely swamps any climatic signal in the
fire record. However, years with low regional precipitation and low Palmer Drought Severity Index
(PDSI) values occur in the early 1930's and mid-1970's, both corresponding to episodes of increased
CHAR (Figure 3-14). A rise in charcoal in the latter part of the 1990's and early 2000's corresponds to a
period of reduced precipitation and lowered lake levels in the region.
82
Figure 3–13. In the last 200 years CHAR in Max Lake reflects human activities in the area. In the early and middle
parts of the 19th Century very little charcoal accumulated in the lake, but logging at the turn of the 20th Century
as well as settlement in the area has resulted in increased fire frequencies and elevated CHAR. A rise in
CHAR values in the last two decades may reflect a period of sustained drought in the region.
Figure 3–14. Annual PDSI Index for the instrumental record for northern Wisconsin. Negative PDSI values represent
drought conditions. Years with low regional precipitation and low PDSI values occur in the early 1930's and
mid-1970's, both corresponding to episodes of increased CHAR (Figure 3-13).
83
Discussion
The diatom community indicates significant changes have occurred in Max Lake during the last
13,000 years. Changes during the late-Glacial to early Holocene are expected as the climate transitioned
from a cool wet environment to a warmer and drier one. At this time Max Lake had higher pH and
phosphorus levels. Studies have shown that in seepage lakes, the amount of input of solute-rich
groundwater has a significant effect on the lake’s pH (Anderson and Bowser 1986, Eilers et al. 1993).
While Max Lake at the present time is a seepage lake with little groundwater input, the historically
higher pH values indicate that either during the early Holocene groundwater input was much greater, or
the lake was a drainage lake with inflow from surface water sources. The stratigraphy evident in drive 3
of the core (Figure 3-1) indicates that Max Lake was a drainage lake and perhaps part of a much larger
lake. Frequent banding that is light gray in color indicates that high energy hydrologic events occurred
which are typical of flooding. During the mid-Holocene the climate in the Upper Midwest was warmer
and effectively drier than previous centuries and 7100 years ago was the most arid period since the
Holocene began (Valero-Garcés et al. 1997). The charcoal analysis in the Max Lake core shows that
7000 years ago wildfire frequency in the Max Lake area was at the highest levels experienced since the
recession of the glaciers (Figure 3-9). This indicates a prolonged drought and likely drop in lake level.
At this time the pH of the lake declined 0.5 pH units, indicating isolation of Max Lake from the
surrounding hydrologic regime, either from substantial reduction of groundwater input or transition to a
seepage lake. There was a brief return to higher pH levels 6000 years BP, when a flood event likely
occurred, as indicated by a light colored band in the core with many degraded diatoms. The diatoms
may have become degraded as a result of high energy environment that would have been present in
flowing water. After this episode the region apparently became more hydrologically stable and there is
no additional evidence of floods. The pH of the lake has remained low during the last 5500 years
84
indicating that after the lake was hydrologically isolated it remained a seepage lake, even though the
charcoal data indicate a wetter environment.
Although Max Lake receives little ground water or stream inflow today, prior to 7200 years BP
it may have been part of a large shallow lake. At the present time there is a large wetland near Max Lake
that may have been part of the lake before the mid-Holocene. Figure 3-15 shows the current topography
of the area near Max Lake and the hypothesized larger lake. The wetland currently has an outflowing
stream that drains to the Trout River and it is possible that in the early Holocene this was the outlet for
the historical large lake.
Figure 3–15. Hypothesized extent of lake basin that Max Lake was a part of prior to the mid-Holocene warming
period. At the present time most of this historical lake basin is a wetland.
The results from the Max Lake core indicate that a changing future climate may have surprising
consequences for headwater drainage lakes. If the future climate is warmer and effectively drier, these
water bodies may become hydrologically isolated. This could lead to a reduction in pH and a decline in
nutrient levels, and therefore change lake productivity. The loss of connectivity with streams will isolate
85
the fish community and lead to a simplification of the fishery and a greater chance of extirpation (Rahel
1984, Magnuson et al. 1998).
References
Anderson, M. P., and C. J. Bowser. 1986. The role of groundwater in delaying lake acidification. Water
Resour. Res. 22:1101-1108.
Appleby, P.G., and F. Oldfield. 1978. The calculation of lead-210 dates assuming a constant rate of
supply of unsupported 210
Pb to the sediment. Catena. 5:1-8.
Birks, H.J.B., J.M. Line, S. Juggins, A.C. Stevenson, and C.J.F. ter Braak. 1990. Diatoms and pH
reconstruction. Phil. Trans. R. Soc., Lond., series B. 327:263-278.
Booth, R.K., and S.T. Jackson. 2003: A high-resolution record of late-Holocene moisture variability
from a Michigan raised bog, USA. The Holocene. 13;865-78.
Camburn, K.E. and D.F. Charles. 2000. Diatoms of low-alkalinity lakes in the northeastern United
States. Special Publication 18. The Academy of Natural Sciences of Philadelphia. Scientific
Publications. Philadelphia. 152 pp.
Carcaillet, C., M. Bouvier, B. Frechette, A.C. Larouche, and P.J.H. Richard. 2001. Comparison of
pollen slide and sieving methods in lacustrine charcoal analyses for local and regional fire history.
The Holocene. 11:467-476.
Cook, E.R., C.A. Woodhouse, C.M. Eakin, D.M. Meko, and D.W. Stahle. 2004. Long-term aridity
changes in the western United States. Science. 306: 1015-18.
Eilers, J.M., G. E. Glass, K. E. Webster, and J. A. Rogalla. 1983. Hydrologic control of lake
susceptibility to acidification. Can. J. Fish. Aquat. Sci. 40:1896-1904
Huber, U.M. and V. Markgraf. 2003. European impact on fire regimes and vegetation dynamics at the
steppe-forest ecotone of southern Patagonia. Holocene. 13:567–579.
86
Jensen, K., E. Lynch, R. Calcote, and S. Hotchkiss. 2005. Morphology of charcoal from lake sediments:
a clue to prehistoric fire regimes. Ecological Society of American Annual Meeting.
Krabbenhoft, D. P., and K. E. Webster. 1995. Transient hydrogeological controls on the chemistry of a
seepage lake. Water Resour. Res. 31: 2295-2305.
Krammer, K. 2000. The genus Pinnularia. In: Diatoms of Europe. Diatoms of the European inland
waters and comparable habitats. (H. Lange-Bertalot Ed.) A.R.G. Ganter Verlag K.G. 703 pp.
Krammer, K. and H. Lange-Bertalot. 1986. Bacillariophyceae. 1. Teil: Naviculaceae. In: Ettl, H., J.
Gerloff, H. Heynig, and D. Mollenhauer (Eds.) Süßwasserflora von Mitteleuropa. Gustav Fisher
Verlag, Stuttgart. Band 2/1: 876 pp.
Krammer, K. and H. Lange-Bertalot. 1988. Bacillariophyceae. 2. Teil: Bacillariaceae, Epithemiaceae,
Surirellaceae. In: Ettl, H., J. Gerloff, H. Heynig, and D. Mollenhauer (Eds.) Süßwasserflora von
Mitteleuropa. Gustav Fisher Verlag, Stuttgart. Band 2/2: 876 pp.
Krammer, K. and H. Lange-Bertalot. 1991a. Bacillariophyceae. 3. Teil: Centrales, Fragilariaceae,
Eunotiaceae. In: Ettl, H., J. Gerloff, H. Heynig, and D. Mollenhauer (Eds.) Süßwasserflora von
Mitteleuropa. Gustav Fisher Verlag, Stuttgart. Band 2/3: 576 pp.
Krammer K. and Lange-Bertalot H. 1991b. Bacillariophyceae. 4. Teil: Achnanthaceae. In: Ettl, H., G.
Gärtner, J. Gerloff, H. Heynig, and D. Mollenhauer (Eds.) Süßwasserflora von Mitteleuropa. Gustav
Fisher Verlag, Stuttgart.
Lange-Bertalot, H. 2001. Diatoms of Europe, Volume 2, Navicula sensu stricto 10 Genera Separated
from Navicula sensu lato Frustulia. A.R.G. Ganter Verlag K.G. 526 pp.
Lynch, E.A., F. Calcote, and S. Hotchkiss. 2006. Late-Holocene vegetation and fire history from Ferry
Lake, northwestern Wisconsin, USA. The Holocene. 16:495-504.
87
Magnuson, J.J., W.M. Tonn, A. Banerjee, J. Toivonen, O. Sanchez, and M. Rask. 1998. Isolation
versus extinction in the assembly of fishes in small northern lakes. Ecology. 79:2941-2956.
Magnuson, J.J., K.E. Webster, R.A. Assel, C.J. Bowser, P.J. Dillon, J.G. Eaton, H.E. Evans, E.J. Fee,
R.I. Hall, L.R. Mortsch, D.W. Schindler, and F.H. Quinn. 1997. Potential effects of climate changes
on aquatic systems: Laurentian Great Lakes and Precambrian Shield region. Hydrological Processes.
11:825-871.
Patrick, R. and C.W. Reimer. 1966. The Diatoms of the United States. Volume 1. Monograph 13,
Academy of Natural Sciences of Philadelphia, Philadelphia, Pennsylvania, USA. 688 pp.
Patrick, R. and C.W. Reimer. 1975. The Diatoms of the United States. Volume 2, Part 1. Monograph 13,
Academy of Natural Sciences of Philadelphia, Philadelphia, Pennsylvania, USA. 213 pp.
R Developmentment Core Team. 2006. R: a language and environment for statistical computing,
version 2.4.1. The R foundation for statistical computing. ISBN 3-900051-07-0. http://www/r-project.
org.
Rahel, F.J. 1984. Factors structuring fish assemblages along a bog lake successional gradient. Ecology
65:1276-1289.
Siver, P.A., P.B. Hamilton, K. Stachura-Suchoples, and J.P. Kociolek. 2005. Diatoms of North America
The Freshwater Flora of Cape Cod. Iconographia Diatomologica. (H. Lange-Bertalot Ed.) A.R.G.
Ganter Verlag K.G. 463 pp.
ter Braak, C.J.F. and P. Smilauer. 2002. CANOCO Reference manual and CanoDraw for Windows
User’s Guide: Software for Canonical Community Ordination (version 4.5). Microcomputer Power
(Ithaca, NY, USA). 500 pp.
Umbanhowar Jr., C. E., and M.J. McGrath. 1998. Experimental production and analysis of microscopic
charcoal from wood, leaves, and grasses. The Holocene. 8:341-346.
88
Valero-Garcés, B.L., K. R. Laird, S.C. Fritz, K. Kelts, E. Ito, and E.C. Grimm. 1997. Holocene climate
in the Northern Great Plains inferred from sediment stratigraphy, stable isotopes, carbonate
geochemistry, diatoms,a and pollen at Moon Lake, North Dakota. Quaternary Research. 48:359-369.
van der Werff, A. 1956. A new method of concentrating and cleaning diatoms and other organisms.
Internationale Vereinigung fuer Theoretische und Angewandte Limnologie. 12:276-277.
Webster, K. E., A. D. Newell, L. A. Baker, and P. L. Brezonik. 1990. Climatically induced rapid
acidification of a softwater seepage lake. Nature. 347: 374-376.
Whitlock, C. and C.P.S. Larsen. 2001. Charcoal as a Fire Proxy, J.P. Smol, H.J.B. Birks, W.M. Last,
Editors, Tracking Environmental Change Using Lake Sediments: Volume 3 Terrestrial, Algal, and
Siliceous Indicators, Kluwer Academic Publishers, Dordrecht. pp. 75–97.
Whitlock, C., and S.H. Millspaugh. 1996. Testing assumptions of fire history studies: an examination
of modern charcoal accumulation in Yellowstone National Park. The Holocene. 6:7-15.
WICCI. 2011. Wisconsin’s Changing Climate: Impacts and Adaptation. Wisconsin Initiative on Climate
Change Impacts. Nelson Institute for Environmental Studies, University of Wisconsin-Madison and
the Wisconsin Department of Natural Resources. Madison, Wisconsin. 217 pp.
89
Chapter 4–Historic Distribution of Common Loons in
Wisconsin in Relation to Changes in Lake Characteristics
and Surrounding Land Use
KEVIN P. KENOW1, PAUL J. GARRISON2, TIMOTHY J. FOX1, and MICHAEL W. MEYER3
1U.S. Geological Survey, Upper Midwest Environmental Sciences Center, La Crosse, Wisconsin 2Wisconsin Department of Natural Resources,Science Services, Madison, Wisconsin 3Wisconsin Department of Natural Resources, Science Services, Rhinelander, Wisconsin
Contents
Abstract ........................................................................................................................................................................91
Introduction ...................................................................................................................................................................91
Methods ........................................................................................................................................................................95
Results ..........................................................................................................................................................................96
Delavan Lake (Walworth County) .............................................................................................................................97
Island Lake (Dane County) .......................................................................................................................................99
Lake Five (Waukesha and Washington counties) ...................................................................................................100
Lake Koshkonong (Jefferson, Dane, and Rock counties) .......................................................................................100
Lakes Mendota and Monona (Dane) .......................................................................................................................101
Lake Winnebago (Clumet, Fond du Lac, and Winnebago counties) .......................................................................102
Lake Pewaukee (Waukesha County) ......................................................................................................................102
Discussion...............................................................................................................................................................103
Acknowledgements.....................................................................................................................................................105
90
References .................................................................................................................................................................105
Tables
Table 4–1. Wisconsin lakes historically supporting breeding common loons that occur south of their current
breeding distribution. ................................................................................................................................................... 96
Table 4–2. Historic (late-1800s) and recent (2007-2011) total phosphorus concentrations, Secchi readings, and
trophic states of southern Wisconsin lakes that historically supported breeding common loons based on
paleolimnological assessment estimates and recently measured values. .................................................................. 97
Table 4–3. Change in land use from the time of the original Wisconsin land use survey (ca 1830) to current (2006)
land cover within drainage areas of lakes that historically supported breeding common loons. .................................. 98
Table 4–4. Lakeshore house density (2006) within 100 m buffer of southern Wisconsin lakes that historically
supported breeding common loons. ............................................................................................................................. 99
Figures
Figure 4–1. Distribution of the common loon (from Evers 2007). ............................................................................. 92
Figure 4–2. Breeding extent records for the common loon in Wisconsin prior to 1976. ........................................... 93
Figure 4–3. Breeding extent records for the common loon in Wisconsin, 1976-2000. ............................................. 94
91
Abstract
A study was conducted to evaluate changes in water quality and land-use change associated with
lakes that are south of the current breeding range of Common Loons in Wisconsin but that historically
supported breeding loons. Museum collection records and published accounts were examined to
identify lakes in southern Wisconsin with a former history of loon nesting activity. Historical and recent
water quality data were obtained from state and USEPA databases for the former loon nesting lakes that
were identified and paleolimnological data were acquired for these lakes from sediment cores used to
infer historical total phosphorus concentrations from diatom assemblages. U.S. General Land Office
notes and maps from the original land survey conducted in Wisconsin during 1832-1866 and the
National Land Cover Database 2006 were utilized to assess land use changes that occurred within the
drainage basins of former loon nesting lakes. Our results indicate that the landscape of southern
Wisconsin has changed dramatically since Common Loons last nested in the region. A number of
factors have likely contributed to the decreased appeal of southern Wisconsin lakes to breeding
Common Loons, including changes to water quality, altered trophic status resulting from nutrient
enrichment, and reductions in suitable nesting habitat stemming from shoreline development and altered
water levels. Increased nutrient and sediment inputs from agricultural and developed areas likely
contributed to a reduction in habitat quality.
Introduction
The common loon (Gavia immer) is a lake-dependent wildlife species that is sensitive to changes
in habitat quality, including water quality and availability of suitable nesting substrate. The breeding
range of the common loon has shifted northward in the Great Lakes states from former southern range
92
limits that once (as recent as 100-150 years ago) extended throughout southern Minnesota to northern
Iowa, throughout southern Wisconsin to northern Illinois, and throughout southern Michigan to northern
Indiana and Ohio (Bent 1919; Figure 4-1, Evers 2007). The common loon was described as a common
breeder on small lakes from the southern tier of Wisconsin counties northward at the turn of the
twentieth century (Kumlien and Hollister 1951).
Figure 4–1. Distribution of the common loon (from Evers 2007).
In Wisconsin, the breeding extent of common loons has more recently been limited to the
northern two-thirds of the state. Figures 4-2 and 4-3 depict the southern breeding extent of the common
loon in Wisconsin established for the period 1966-1975 by Robbins (1977), resident loon locations
reported by Zimmer (1979), and the primary range for nesting loons as incorporated into the Project
93
LoonWatch Common Loon Survey in Wisconsin (Gostomski and Rasmussen 2000).
Figure 4–2. Breeding extent records for the common loon in Wisconsin prior to 1976.
Resident loons were documented within central Wisconsin (St. Croix, Brown, Green Lake, and
Sheboygan counties) as recently as 1954-1975 in Wisconsin Society for Ornithology field notes
(Wisdom et al. 1975); and in Waupaca, Waushara, and Fond du Lac counties during 1946-1953
(Zimmer 1979). Resident common loons have also been present in the Glacial Lake Wisconsin Sand
94
Plain ecoregion since at least the mid-1960s (K.P. Kenow, unpublished data).
Figure 4–3. Breeding extent records for the common loon in Wisconsin, 1976-2000.
It is likely that changes in habitat characteristics of southern lakes within the species’ former
range are associated with the range constriction. We investigated changes in water quality and land-use
change associated with lakes that are south of the current breeding range of common loons in Wisconsin
but that historically supported breeding loons. This paper offers insights into how a changing climate
might impact common loon habitat quality and the future breeding distribution of loons in Wisconsin.
95
Methods
Museum collection records (UW-Madison Zoological Museum, Louisana State University
Museum of Natural Science, Chicago Field Museum, National Museum of Natural History, Smithsonian
Institution, Milwaukee Public Museum, Bell Museum of Natural History, Webster House Museum, US-
Stevens Point Museum of Natural History and Biology Research Collection, UW-Whitewater Science
Department, Walworth County Historical Society, Hoard Museum-Ft. Atkinson), published accounts
(Robbins 1977, Kumlien and Hollister 1951, Zimmer 1979, Robbins 1991, Schorger 1929, Wisdom et
al. 1975, Keeler 1890, Carr 1890), and Wisconsin Breeding Bird Atlas (Cutright et al. 2006) common
loon nesting data were examined to identify lakes in southern Wisconsin with a former history of loon
nesting activity.
Historical and recent water quality data were obtained from state and USEPA databases (e.g.,
Wisconsin Surface Water Integrated Monitoring System [SWIMS], EPA Legacy STORET) for the
former loon nesting lakes that were identified. Paleolimnological data were acquired for these lakes
from sediment cores used to infer historical total phosphorus concentrations from diatom assemblages
(Garrison et al. 2000, Garrison et al. 2008). Trophic state index values (Carlson 1977) were calculated
based on historic and recent total phosphorus concentrations.
The original vegetation cover of Wisconsin, compiled from U.S. General Land Office notes and
maps from the original land survey conducted in Wisconsin during 1832-1866 (Finley 1976), and the
National Land Cover Database 2006 (NLDC2006; Fry et al. 2011) were utilized to assess land use
changes that occurred within the drainage basins of former loon nesting lakes. Lakeshore development
was based on house density determined within a 100-meter buffer surrounding each lake and derived
from USDA National Agriculture Imagery Program 2005 high-resolution imagery.
96
Results
We obtained historical (late 1800s-early 1900s) records that documented common loon breeding
activity on eight southern Wisconsin lakes that are south of the current breeding range (Table 4-1,
Figure 4-2). Evidence was based on museum egg collections at three lakes (Island, Winnebago,
Pewaukee lakes), an adult collected during the breeding season at Lake Five in Waukesha and
Washington counties, and published accounts concerning lakes Koshkonong, Delavan, Mendota, and
Monona.
Table 4–1. Wisconsin lakes historically supporting breeding common loons that occur south of their current breeding
distribution.
Lake Name County
Location (UTM
Easting,
Northing)
Evidence of historic loon nesting
(year; reference) Record type
Last recorded
occurrence
Koshkonon
g
Jefferson,
Dane, Rock
339823 E,
4748895 N
ca
1921
Kumlein and Holister 1951 Account ca 1921
Delavan Walworth 367905 E,
4717975 N
1888 Kumlein and Holister 1951 Account ca 1890
Island Dane 312650 E,
4755130 N
1888 C.K. 1890, UW-Madison
Zoological Museum records
Eggs not present in
1889-1890
Winnebago Calumet,
Fond du Lac,
Winnebago
386110 E,
4875990 N
1884 UW-Madison Zoological
Museum records
Eggs
Pewaukee Waukesha 393050 E,
4769305 N
1869
,
1874
Smithsomian records Eggs no summer
sightings in
Waukesha County
since 1958
Five Waukesha,
Washington
396560 E,
4783260 N
1925 Milwaukee Public Museum
records
adult
present on
1 May
Mendota Dane 302520 E,
4775380 N
ca
1890
Carr 1890 Account no records by
1929
Monona Dane 308190 E,
4771400 N
ca
1890
Carr 1890 Account no records by
1929
Historic phosphorus levels were inferred for six of the eight lakes from sediment-core analyses
either from the particular lake or from lakes within the area with similar morphometry, landscape
position, surface/groundwater interfaces, and soil conditions. All of the lakes of interest fall within the
97
Southeastern Wisconsin Till Plains Level III ecoregion. Average summer total phosphorus readings for
seven of the eight lakes were extracted from the SWIMS database.
Table 4–2. Historic (late-1800s) and recent (2007-2011) total phosphorus concentrations, Secchi readings, and
trophic states of southern Wisconsin lakes that historically supported breeding common loons based on
paleolimnological assessment estimates and recently measured values.
Lake
Core
collected
Total phosphorus (µg
L-1
)
Secchi
reading
(m)
Trophic State
Index (TSITP)c
Trophic class shift
Historic
level Recent levelb
Recent
levelb
Historic Recent Historic → Recent
Delavan no 15-20a
25-30 1.9-2.9 43-47 51-53 mesotrophic → eutrophic
Island yes na na na - -
Five no na 14-31 2.5-4-0 - 42-54 -
Koshkonong no na 104-492 0.3-0.5 - 71-94 -
Mendota yes 15-20a 22-75 1.2-2.8
d 43-47 49-66 mesotrophic → eutrophic
Monona no 15-20a 30-94 1.2-1.8 43-47 53-70 mesotrophic → eutrophic
Winnebago yes 40-50a
93-213 0.5-1.2 57-61 70-81 eutrophic→ hypereutropphic
Pewaukee yes 20 14-22 0.5-1.1 47 42-49 mesotrophic→ mesotrophic aGarrison et al. 2008.
bSource: Wisconsin DNR Surface Water Integrated Monitoring System average annual summertime records for
2007-2011. cCalson Trophic State Index (Carlson 1977) based on total phosphorus concentration.
dSource: North Temperate Lakes Long Term Ecological Research program (http://lter.limnology.wisc.edu), NSF,
Center for Limnology, University of Wisconsin-Madison.
Delavan Lake (Walworth County)
Kumlien and Hollister (1951) documented a “few” nesting loons on Delavan Lake through 1888
with “possibly a straggling pair now and then for a few years later.” A sediment core was not collected
from this lake, but similar lakes have been sampled in the vicinity. Historically, lakes of this type had
phosphorus levels around 15-20 µg L-1
(Garrison et al. 2008). Average summer total phosphorus levels
in Delavan Lake over the past 5 years have been 25-30 µg L-1
, and recent average summer Secchi
readings have ranged between 1.9-2.9 m (Table 4-2). Recently, the lake is more eutrophic than it was
historically. Land use change analysis indicates cultivation and development have resulted in a drastic
98
Table 4–3. Change in land use from the time of the original Wisconsin land use survey (ca 1830) to current (2006) land cover within drainage areas of
lakes that historically supported breeding common loons.
Lake
Drainage
area (km2)
Open water Wetland Forest Herbaceous Cultivated Developed
1830 2006 Δ 1830 2006 Δ 1830 2006 Δ 1830 2006 Δ 1830 2006 Δ 1830 2006 Δ
Delavan 81,080 9.2 8.9 -0.3 2.0 1.4 -0.6 46.6 8.0 -38.6 42.2 1.1 -41.1 0 62.0 +62.0 0 18.6 +18.6
Island 75,558 0.2 1.4 +1.2 7.9 3.9 -4.0 73.9 8.1 -65.8 17.9 0.6 -17.3 0 70.4 +70.4 0 14.8 +14.8
Five 108,179 3.6 4.4 +0.8 10.3 10.3 0 42.5 15.9 -26.6 43.6 0.4 -43.2 0 33.2 +33.2 0 33.8 +33.8
Koshkonong 143,923 28.0 29.3 +1.3 7.7 9.7 +2.0 14.6 10.5 -4.1 49.7 0.6 -49.1 0 43.6 +43.6 0 5.3 +5.3
Mendota 82,276 41.6 44.4 +2.8 4.4 2.1 -2.3 51.8 6.7 -45.1 2.1 <0.1 -2.1 0 7.4 +7.4 0 39.2 +39.2
Monona 110,332 13.2 13.6 +0.4 19.2 6.9 -12.3 62.6 7.1 -55.5 5.0 <0.1 -5.0 0 14.0 +14.0 0 58.3 +58.3
Winnebago 478,384 97.8 99.8 +2.0 0.4 <0.1 -0.4 1.5 <0.1 -1.5 0.3 <0.1 -0.3 0 <0.1 0 0 <0.1 0
Pewaukee 89,242 9.8 10.1 +0.3 16.7 3.7 -13.0 18.6 13.8 -4.8 54.8 0.3 -54.5 0 27.6 +27.6 0 44.3 +44.3
99
reduction in forest and herbaceous cover within the drainage area of Delavan Lake (Table 4-3). The
lakeshore house density in 2006 was 18.8 houses per km of shoreline (Table 4-4).
Table 4–4. Lakeshore house density (2006) within 100 m buffer of southern Wisconsin lakes that historically
supported breeding common loons.
Lake Number of houses Perimeter (km) House density/ km shoreline
Delavan 492 26.1 18.8
Island 4 4.2 1.0
Five 45 3.3 13.5
Koshkonong 410 48.4 8.5
Mendota 699 38.4 18.2
Monona 813 27.1 30.0
Winnebago 3,592 181.2 19.8
Island Lake (Dane County)
Eggs collected from a loon nest at Island Lake on 20 May 1888 are housed at the UW-Madison
Zoology Museum collection. An account of the collection indicated that loons were not present on
Island Lake during the two years subsequent to the collection (Keeler 1890). A sediment core was
collected from this lake. However, this lake is a deep water marsh and the sediment at the bottom of the
core was representative of a wetland rather than a lake. Consequently, the diatom communities in the
top (recent) and bottom (historical) slices of the core were very different. This difference was largely
the result of different habitats and not necessarily related to nutrients. Island Lake appears to have
alternated between a deep water marsh with abundant emergent vegetation and an open water system
dependent upon abundant rainfall (P. Garrison, unpubl. data). Therefore, it was not possible to
accurately reconstruct the historical phosphorus concentrations. Recent phosphorus data are not
available. While the lakeshore housing density at Island Lake is relatively low (1.1 houses per km
shoreline; Table 4), 70% of the lake drainage area is now cultivated and 15% developed (Table 4-3). It
100
is likely that agricultural practices during the past 60 years have resulted in higher sediment and
phosphorus delivery to the lake.
Lake Five (Waukesha and Washington counties)
An adult loon was collected on Lake Five on 1 May 1925 according to Milwaukee Public
Museum records. This collection likely occurred during the breeding season; accordingly, we have
included Lake Five as a lake that potentially supported breeding loons at one time. A sediment-core has
not been collected at Lake Five, but similar deep seepage lakes in the area have a pre-settlement
phosphorus concentration of about 13-15 µg L-1
. Recent summer phosphorus concentrations averaged
14-31 µg L-1
with Secchi readings averaging 2.5-4.0 m (Table 4-2). Primary changes in land cover
since historic times are reductions in forest (-27%) and herbaceous (-43%) land cover, and increases in
cultivated (+33%) and developed (+34%) land cover (Table 4-3). Shoreline house density was 13.5
houses per km shoreline in 2006 (Table 4-4).
Lake Koshkonong (Jefferson, Dane, and Rock counties)
Kumlien and Hollister (1951) indicated that common loons bred on Lake Koshkonong in about
1921. Subsequent to this time, loons occurred on Lake Koshkonong and other small lakes in southern
Wisconsin during migration. A core was not collected from this lake because of significant habitat
changes. Lake Koshkonong has undergone a major change in water quality and habitat alteration since
1900. Prior to the early twentieth century, Lake Koshkonong was a deep water marsh with dense
emergent vegetation and was renowned for its abundant waterfowl (Kumlien 1877, Main 1945). Market
hunting for waterfowl was very popular on this lake prior to about 1917. At that time, the water level
was raised via a dam and much of the emergent vegetation was inundated (Hylan 1923, Sinclair 1924).
The water quality became significantly degraded beginning in the 1940-50s as changing agricultural
101
practices resulted in increased delivery of sediments and nutrients to the lake by the Rock River. This
agricultural-related degradation has been documented with sediment cores in lakes Nagawicka and
Ripley, which are located in southeastern Wisconsin (Garrison 2004, Garrison and Pillsbury 2009).
Recent concentrations of total phosphorus are relatively high in this hypereutrophic lake and summer
water clarity is poor (Secchi readings average 0.3-0.5 m; Table 4-2). Land use change analysis
indicated that most of the herbaceous land cover historically present within the drainage area of the
Lake Koshkonong has been placed under cultivation (Table 4-3). The shoreline housing density in 2006
was 8.5 houses/km (Table 4-4).
Lakes Mendota and Monona (Dane)
Carr (1890) reported that common loons were nesting on the Madison lakes in 1890. By 1929,
loons in adult plumage were frequently observed in Dane County during the summer, but there were no
recent breeding records (Schroger 1929). A sediment-core was analyzed from Lake Mendota, but the
diatom community had very few taxa and an accurate determination could not be made of historical
phosphorus concentrations. Historical phosphorus concentrations in southern Wisconsin lakes similar
to these stratified drainage lakes were likely 15-20 µg L-1
(Garrison et al. 2008). Recent average
summer phosphorus concentrations are much higher (22-75 µg L-1
in Lake Mendota; 30-94 µg L-1
in
Lake Monona) and the trophic state of these lakes has shifted from a historical mesotrophic state to
eutrophic (Table 4-2). Recent summer Secchi readings averaged 1.2-1.8 m on Lake Monona and 1.2-
2.8 m on Lake Mendota. The historic land cover within the drainage area of these lakes was
predominately oak forest, but is now largely developed (Table 4-3). Shoreline house densities in 2006
were 18.2 houses/km on Lake Mendota and 30.0 houses/km surrounding Lake Monona (Table 4-4).
102
Lake Winnebago (Clumet, Fond du Lac, and Winnebago counties)
Four common loon eggs were collected from Lake Winnebago in 1884 and are housed in the
UW-Madison Zoology Museum collection. It is uncertain how recently breeding loons have used Lake
Winnebago. Resident loons were noted in Fond du Lac County as recently as about 1950 (Zimmer
1979). Sediment-core analysis indicated the phosphorus concentration at the top (recent) was 145 µg L-
1 and 85 µg L
-1 at the bottom (historic). Radiochemical analysis indicated the bottom sample was
deposited during the middle of the twentieth century, so phosphorus concentrations in the early part of
the century were likely lower (they probably were around 40-50 µg L-1
based upon the results of a core
taken in 2007; P. Garrison, unpubl. data). Recent average summer phosphorus concentrations ranged
from 93 to 213 µg L-1
and the average Secchi depth was 0.5-1.2 m (Table 4-2). Lake Winnebago is now
classified as hypereutrophic based on total phosphorus concentrations. Shoreline housing density was
19.8 houses/km in 2006 (Table 4-4).
Lake Pewaukee (Waukesha County)
Eggs were collected from common loon nests on Pewaukee Lake in 1869 and in 1874 (both
collections by B. Goss). These eggs reside in the National Museum of Natural History Vertebrate
Zoology Birds Collection, Smithsonian Institution. Robbins (1991) indicated that no June or July
sightings of common loons had been reported in Waukesha County since 1958. Analysis of a sediment-
core slice representing the period in the late-1800s indicates that phosphorus concentrations at that time
about 20 µg L-1
, while recent summer phosphorus concentrations have averaged between 14 and 22 µg
L-1
and Secchi readings averaged between 0.5 and 1.1 m (Table 4-2). The predominant pre-settlement
vegetation within the Lake Pewaukee drainage area was characterized by oak openings. Most of this
103
land cover has since been replaced, cultivated, or developed (Table 4-3). The recent shoreline housing
density was 34.5 houses/km – highest among the lakes we evaluated (Table 4-4).
Discussion
A number of factors have likely contributed to the decreased appeal of southern Wisconsin lakes
to breeding common loons, including changes to water quality, altered trophic status resulting from
nutrient enrichment, and reductions in suitable nesting habitat stemming from shoreline development
and altered water levels. The landscape of southern Wisconsin has changed dramatically since common
loons last nested in the region. Watershed land use of lakes assessed in this study have undergone
substantial alteration, with a median of 30% (maximum = 70%) of watershed areas now under
cultivation and a median of 26% (maximum = 58%) having been developed. While it is not clear if
landscape changes alone deter loons from using these lakes for breeding, increased nutrient and
sediment inputs from agricultural and developed areas likely contributed to a reduction in habitat
quality.
Since the 1800s, most of the lakes we evaluated in this study have experienced increased
concentrations of phosphorus, a nutrient that typically limits biotic growth in midwestern lakes
(Robertson et al. 2002). Phosphorus enrichment promotes phytoplankton growth, leading to
eutrophication and a corresponding reduction in water clarity. Common loons are generally considered
to be visual predators. Consequently, lake suitability for loons is influenced by water clarity. Meyer et
al. (Chapter 5) considered several characteristics of Wisconsin lakes and reservoirs in development of a
regional common loon habitat model. Presence of nesting habitat, number of waterbodies within 10 km,
Secchi depth, log of lake area, and the interaction between Secchi depth and log of lake area were all
related to the presence of territorial loons on lakes. The importance of water clarity, as indexed by
104
Secchi disk depth, was related to lake size; where the probability of territorial loon presence was
positively related to increasing water clarity on large lakes.
Apparently, factors associated with shoreline development also negatively affect the value of
lakes to breeding common loons. Meyer (2006) presented evidence of a threshold of shoreline building
density of 25 buildings per km that appears to preclude territorial pair occupancy on highly developed
lakes. Two of the lakes we evaluated (Monona and Pewaukee) had housing densities in excess of this
threshold and three lakes exceeded 18 houses per km. It is not likely that the buildings themselves are
the ultimate factor causing lack of lake use by loons, but building densities are likely indices of the
cumulative impact of multiple factors such as nest habitat alteration, disturbance, or increased predator
densities. Southern Wisconsin lakes are in close proximity to large cities and towns and serve
recreational boating of surrounding population centers, contributing to disturbance.
How might a changing climate impact common loon habitat quality and the future breeding
distribution of loons in Wisconsin? Climate models project a warming climate pattern and increased
evapotranspiration rates in the Great Lakes region (Kling et al. 2003, Magnuson et al. 1997). These
changes to the region’s climate are expected to result in changes to land cover within boreal and mixed
forest ecosystems, impacting hydrological, chemical, and physical properties which, in turn, affect
inland lake productivity (Schindler 1997, Magnuson et al. 1997, Kling et al. 2003). Depending on the
lake type, trophic states could be changed in northern temperate lakes of the Great Lakes region, where
most lakes are currently oligotrophic or mesotrophic, potentially resulting in cascading impacts
throughout the foodweb (Kling et al. 2003). Lake habitat characteristics have been altered in lakes
within the former range of the common loon. However, it is not possible to tie this range shift directly
to the effects of a warming climate, with so many other anthropomorphic changes affecting the
landscape.
105
Acknowledgements
We acknowledge L. Halverson (UW-Madison Zoological Museum), B. Herderson (Milwaukee
Public Museum), C. Yahnke (UW-Stevens Point Museum of Natural History and Biology Research), E.
Davis (UW-Whitewater) for searching for and/or providing museum collection records. J. Filbert
(Wisconsin DNR) provided relevant water quality data from the Wisconsin Surface Water Integrated
Monitoring System. S. Matteson reviewed an earlier version of this portion of the report and provided
useful suggestions.
References
Carr, CF. 1890. A list of birds known to nest within the boundaries of Wisconsin, with a few notes
thereon. Wisconsin Naturalist 1:28-29.
Cutright, N.J, B.R.Harriman, and R.W. Howe. 2006. Atlas of the Breeding Birds of Wisconsin. The
Wisconsin Society for Ornithology, Inc. 602pp.
Evers, D. C. 2007. Status assessment and conservation plan for the Common Loon (Gavia immer) in
North America. BRI Report 2007-20. U.S. Fish and Wildlife Service, Hadley, MA.
Fry, J., Xian, G., Jin, S., Dewitz, J., Homer, C., Yang, L., Barnes, C., Herold, N., and Wickham, J.,
2011. Completion of the 2006 National Land Cover Database for the Conterminous United States,
PE&RS, Vol. 77(9):858-864.
Garrison, P.J. 2000. Paleoecological study of Geneva Lake, Walworth County: Wisconsin Department
of Natural Resources, Bureau of Integrated Science Services, PUB-SS-952, 25 pp.
Garrison, P.J. 2004. Paleoecological Study of Nagawicka Lake, Waukesha County. Wisconsin
Department of Natural Resources, Bureau of Integrated Science Services, PUB-SS-993 2004, 20 pp.
106
Garrison, P.J. and R. Pillsbury. 2009. Paleoecological Study of Lake Ripley, Jefferson County.
Wisconsin Department of Natural Resources, Bureau of Science Services, PUB-SS-1062 2009, 18 pp.
Garrison, P., M. Jennings, A.Mikulyuk, J.Lyons, P.Rasmussen, J.Hauxwell, D. Wong, J.Brandt, and
G.Hatzenbeler. 2008. Implementation and Interpretation of Lakes Assessment Data for the Upper
Midwest. Final Report to the U.S. EPA Grant No. X7-83124601. PUB-SS-1044 2008, 72 pp.
Hylan, O.R. 1923. A field report on Lake Koshkonong. Wisconsin Conservation Department Files.
Keeler, CA. 1890. The nesting of the loon. Wisconsin Naturalist 4:60-62.
Kling, G.W., K. Hayhoe, L.B. Johnson, J.J. Magnuson, S. Polasky, S.K. Robinson, B.J. Shuter, M.M
Wander, D.J. Wuebbles, D.R. Zak, F.L. Lindroth, S.C. Moser, and M.L. Wilson. 2003. Confronting
Climate Change in the Great Lakes Region : Impacts on our Communities and Ecosystems. Union of
Concerned Scientists, Cambridge, Massachusetts, and Ecological Society of America, Washington,
D.C.
Kumlien, T. 1877. Lake Koshkonong. By and old settler. Pp 628-631 in: Madison, Dane County and
surrounding towns: being a history and guide to places of scenic beauty and historical note found in
the town of Dane County and surroundings, including the organization of the towns, and early
intercourse of the settlers with the Indians, their camps, trails, mounds, etc., with a complete list of
county supervisors and officers, and legislative members, Madison village and city council. Wm. J.
Park & Co., Madison.
Main, A.K. 1945. Studies in ornithology at Lake Koshkonong and vicinity by Thure Kumlien from 1843
to July 1850. Wisconsin Academy of Sciences, Arts, and Letters 37: 91-109.
Magnuson, J.J., K.E. Webster, R.A. Assel, C.J. Bouser, P.J. Dillon, J.G. Eaton, H.E. Evans, E.J. Fee,
R.I. Hall, L.R. Mortsch, D.W. Schlinder, and F.H. Quinn. 1997. Potential effects of climate changes
107
on aquatic systems : Laurentian Great Lakes and Precambrian Shield Region. Hydrological
Processes 11:825-872.
Meyer, M.W. 2006. Evaluating the Impact of Multiple Stressors on Common Loon Population
Demographics - An Integrated Laboratory and Field Approach. Final Report to U.S. EPA under EPA
STAR Co-operative Agreement Number: R82-9085.
Robbins, SD. 1991. Wisconsin Birdlife: Population and distribution past and present. The University
of Wisconsin Press, Madison.
Robertson, D.M., G.L. Goddard, E.A. Mergerer, W.J. Rose, and P.J. Garrison. 2002. Hydrology and
water quality of Geneva Lake, Walworth County, Wisconsin: U.S. Geological Survey Water
Resources Investigations Report 02-4039, 73 p.
Schindler, D.W. 1997. Widespread effects of climatic warming on freshwater ecosystems in North
America. Hydrological Processes 11:1043-1067.
Sinclair, F. 1924. Lake Koshkonong. Outdoor America 1924.
108
Page intentionally left blank.
109
Chapter 5 – Development of a Regional Habitat Suitability
Model for Common Loons and Predicting Consequences
of Expected Future Climate Conditions on Loon
Occurrence in Wisconsin
MICHAEL W. MEYER1, PAUL S. RASSMUSSEN2, and KEVIN P. KENOW1 1Wisconsin Department of Natural Resources, Science Services, Rhinelander, Wisconsin 2Wisconsin Department of Natural Resources,Science Services, Madison, Wisconsin 3U.S. Geological Survey, Upper Midwest Environmental Sciences Center, La Crosse, Wisconsin
Contents
Abstract ..................................................................................................................................................................... 112
Introduction ................................................................................................................................................................ 113
Methods ..................................................................................................................................................................... 115
Loon habitat suitability model development ........................................................................................................... 115
Predicted probability of occurrence following climate change ................................................................................ 127
Results ....................................................................................................................................................................... 128
Loon habitat suitability model ................................................................................................................................. 128
Predicted probability of occurrence following climate change ................................................................................ 139
Discussion ................................................................................................................................................................. 140
Acknowledgements .................................................................................................................................................... 145
110
References ................................................................................................................................................................ 146
Tables
Table 5–1. Habitat variables used in modeling, with abbreviations used in other tables and summary
statistics ................................................................................................................................................................. 117
Table 5–2. Common loon territorial pairs and nest habitat by ecoregion ........................................................... 130
Table 5–3. AICc for logistic regression models fit to data from all three ecoregions (276 lakes with no missing
values for these predictors). See abbreviations in Table 5-1. Log-transformed variables are log of total phosphorus
(Ltp), log of lake area (Lh), log of color (Lcol), log of shoreline development index (Lsdi), and log of lake perimeter
(Lper). Interactions are identified by ‘*’ between the variables included in the interaction. K is the number of
parameters in the model. ....................................................................................................................................... 131
Table 5–4. Coefficient estimates and standard errors for the best fitting models from Table 5-3. These models
were fit to the largest data set without any missing values for these predictor variabiles, which consisted of 285
observations. ......................................................................................................................................................... 134
Table 5–5. Estimates of prediction accuracy for selected logistic regression models ........................................ 136
Table 5–6. The change in the percentage of lakes occupied by loon territorial pairs as predictor variables change
from current values. These were estimated for all 285 lakes without any missing values for these predictor
variables. Currently, 102 of 285 lakes in the sample, or 35.8%, were occupied by territorial loon pairs. The %
changes in the table represent additive changes from the current state – thus, a change of -10% would result in
35.8% - 10% = 25.8% of lakes occupied. Note that the % changes in predictor variables were applied to each lake
separately – thus, a 10% change in Secchi depth for a very clear lake would be much larger than a 10% change for
a murky lake. ......................................................................................................................................................... 137
Table 5–7. Predicted probability of loon occupancy with changes in nest habitat for lakes in the Manitowish River
watershed. A decrease in habitat quality is a decrease of one category (e.g., from excellent to good, or good to
111
fair); an increase in habitat quality is an increase of one category. Current habitat quality is designated as E
(excellent), G (good), F (fair), and P (poor). Bold indicates observed or predicted presence of territorial pairs. ... 139
Figures
Figure 5–1. Box plots for lake area, maximum depth, lake density and secchi depth within each ecoregion. ... 120
Figure 5–2. Box plots for lake pH, color, total phosphorus and dissolved oxygen within each ecoregion. ......... 121
Figure 5–3. Box plots for selected land-use classifications within each ecoregion. ........................................... 122
Figure 5–4. Variable importance as determined by the random forests procedure for the data set including all
three ecoregions. Variable abbreviations are listed on the y-axis; the x-axis is the mean % decrease in prediction
accuracy for loon presence if the variable is not included in the model. Predictor variables are listed in order from
most important at the top. ...................................................................................................................................... 133
Figure 5–5. Predicted probability of territorial loon presence based on the best fitting model for data from all
three ecoregions. Predictions are for lakes with a fixed number of lakes within 10 km (equal to the observed mean
in the data set). Each line is for one nest habitat category (excellent – solid line; good – dashed line; fair – dotted
line; poor – dotted/dashed line). Numbers in the strips above each panel are the log of lake area that the
predictions in the panel assume. ........................................................................................................................... 135
112
Abstract
Research across North America has shown that common loons (Gavia immer) select breeding
territories as a function of lake physical and chemical characteristics. Climate change models predict
warmer temperatures, changes to precipitation patterns, and increased evapotranspiration in the Great
Lakes region. Such climatic changes have altered, and would further alter hydrological, chemical, and
physical properties of inland lakes. In this project we developed a predictive regional common loon
breeding habitat suitability model based on loon populations in northern and central Wisconsin, and
assessed the potential effects of future climate change on loon habitat quality and breeding pair
occupancy of lakes within the Trout Lake basin in northern Wisconsin.
We used presence or absence of loon territorial pairs as the response variable in logistic
regression analyses with multiple habitat predictor variables. Water quality data including Secchi depth,
color, total phosphorus, temperature, pH, conductivity, and dissolved oxygen, nest habitat quality, and
land cover within 150 m and 500 m of each lake or reservoir were considered in the analysis. The best
fitting logistic regression model included as predictor variables nest habitat, number of waterbodies
within 10 km, Secchi depth, log of lake area, and the interaction between Secchi depth and log of lake
area. The two next best models included these same variables plus proportion of forest within either the
500 m or 150 m buffer.
We used the best-fitting loon habitat model to predict loon occurrence for 27 of the lakes in the
Trout Lake basin under projected future climatic conditions. Linked climate and lake hydrology models
(see Chapter 2) were used to simulate lake conditions for all years from 1962 to 2100 under 3 emissions
scenarios and 6 general circulation models. The results of hydrology and lake modeling in the Trout
Lake basin in northern Wisconsin indicates that these lakes have the potential to become clearer under
future climate conditions. This is due to the fact that dissolved organic carbon (DOC), not total
113
phosphorus (TP), is the parameter most closely associated with lake clarity in the watershed that is
likely to change under future climate scenarios, as DOC levels are predicted to be lower. Secchi depth
estimates from the linked climate and hydrology models for the period between 2010 and 2090 are
predicted to increase slightly in the 27 lakes in the Trout Lake basin. Considering water clarity as a
primary factor in loon habitat suitability, estimated probabilities of loon occurrence are expected to stay
the same or increase very slightly between 2010 and 2090 for all 27 lakes as Secchi depth slowly
increases over time. Changes in nest habitat can have large effects on the probability of loon
occurrence in lakes. Lake stage is predicted to be lower at Trout basin seepage lakes under future
climate scenarios, potentially altering nest habitat quality. Step changes in nest habitat generally
overwhelm the effect of estimated changes in Secchi depth or possible changes in total phosphorus.
These results point to the critical need to conserve and enhance common loon nest habitat within the
Trout Lake basin, and throughout the current breeding range of loons in Wisconsin. Adaptation
strategies to reduce potential negative consequences of a changing climate should include preserving
existing critical nest habitat by managing shoreline development and habitat loss.
Introduction
Research across North America has shown that common loons select breeding territories as a
function of lake physical and chemical characteristics (Meyer 2006, Kuhn et al. 2011, Found et al.
2008). With grant support from USEPA (STAR grant R82-9085), a loon habitat model was developed
for northern Wisconsin, designed to identify factors associated with territorial loon presence or absence
on area lakes (Meyer 2006). The Wisconsin Loon Habitat Suitability model was developed using 2002-
2004 field data collected on 377 randomly selected lakes. The goal of the modeling effort was to assess
the relative importance of natural and anthropogenic features in defining and predicting loon
114
distribution, abundance, and productivity. This data analysis simultaneously evaluated the relationship
between loon lake use (territorial pair presence/absence) and reproductive success (fledged chick
presence/absence) with lake morphometry, water chemistry, shoreline habitat characteristics, as well as
factors associated with human activity including shoreline building number and density as well as
boating activity. Logistic regression analysis was used to identify important natural features associated
with loon lake use and reproductive success while also exploring whether human settlement density or
boating disturbance has a measurable impact.
It was found that the presence of territorial adults was more likely on lakes within the study area
with increasing convolutedness of shoreline (shoreline development index), lake pH, and Secchi disk
depth, and where good nest habitat was present. The lakes most likely to have territorial loons are
those that are convoluted, clear, with neutral pH, along with the presence of nest islands or other
physical features associated with loon nest locations (sphagnum mats, marshy hummocks, narrow
peninsulas, shorelines with extensive wetland area).
Our first objective in the current project was to expand the scope of the 2002-2004 habitat
modeling effort to develop a predictive regional breeding habitat suitability model for common loon
populations in both north and central Wisconsin. The central Wisconsin cohort represents the most
southerly extent of the current breeding distribution of common loons in the Upper Midwest U.S. We
include metrics of habitat quality as well as anthropogenic change and influence as explanatory factors.
We hypothesized that the loon is influenced by both habitat and anthropogenic cues when selecting
breeding habitat in northern and central Wisconsin. Our second objective was to assess the potential
effects of future climate change on habitat quality and loon distribution – specifically, to incorporate the
findings of hydrology and lake modeling (see Chapter 2) with that of the regional loon breeding habitat
115
suitability model, to estimate the change in probability of loon lake occupancy within the core of
Wisconsin’s breeding loon distribution.
Methods
Loon habitat suitability model development
We surveyed lakes across the range of environmental conditions in which common loons
currently breed within Wisconsin with the objective of developing a loon habitat suitability model to
assess the consequences of climate change on common loon distribution in the state. Lakes were
surveyed for loon occupancy (territorial pairs), nesting, and nest success in the center of their
distribution in northern Wisconsin (Northern Highlands ecoregion), as well as in areas at the edge of
their distribution in the state (Central Sand Ridges and Glacial Lake Wisconsin Sand Plain ecoregions;
ecoregions are described in Omernik et al. 2000). A variety of sampling strategies have been used for
habitat suitability modeling, including random and systematic samples of all sampling units, or separate
samples of used and available units (Hirzel and Guisan 2002; Johnson et al. 2006). Johnson et al.
(2006) found that habitat suitability models can be developed using data from both types of samples.
Because of advantages in coordinating loon sampling with ongoing, pre-existing studies in the
Northern Highlands ecoregion, we surveyed all lakes greater than 10 ha in size within or adjacent to the
Lac du Flambeau Indian Reservation in Vilas County (91 lakes), Wisconsin and all lakes within the
Trout Lake basin, also in Vilas County (31 lakes). We supplemented this sample with 24 additional
lakes in the Northern Highlands ecoregion that were similar in morphometry to the other lakes, but with
low water clarity, so that water clarity conditions found in other regions of Wisconsin were represented
in our sample. We used Generalized Random-Tesselation Stratified (GRTS) sampling to select lakes in
the Central Sands and Glacial Lake Wisconsin ecoregions (Stevens and Olsen, 2004) from lakes and
116
reservoirs >10 ha in size based on Wisconsin DNR 1:24,000-Scale Surface Water Hydrography
(http://dnr.wi.gov/maps/gis/datahydro.html). We divided these lakes into two strata: one for which we
had knowledge of past breeding use by loons, the other all remaining lakes. We stratified in this way
because the density of loons in this area of Wisconsin is low and we wanted to ensure that we surveyed
some lakes with loons present. We surveyed 46 lakes in the Central Sands ecoregion and 100 lakes in
the Glacial Lake Wisconsin ecoregion.
All lakes used in the model were surveyed for loon occupancy once per week during 01 May
through 15 July. At each visit, the observer used motorboat, canoe, or kayak to travel the perimeter of
the lake at low speed; or shoreline observations were conducted when an entire lake or reservoir was
viewable from shore. Binoculars and/or spotting scope were used to scan the lake for the presence of
adult loons. Adult loons are highly conspicuous and all staff had training or experience with the survey
protocol prior to conducting surveys. Once located, adult loons were observed for a minimum of 10
minutes to determine whether they were territorial adults or unpaired. Territorial adult loon pairs are
typically found in close proximity to one another during the early breeding season and exhibit non-
aggressive behavior including foraging, locomoting, resting, courtship, and nest building. Unpaired
loons either occurred singly or in small groups in undisputed lakes or areas of lakes, or participated in
aggressive encounters with resident loons. At the end of each survey (typically 1-2 hours per lake), the
observer recorded the total number of territorial and/or unpaired loons observed and general behavior.
Observers also maintained a record of nesting attempts, nest locations, outcome of each attempt, number
of eggs laid, and number of chicks hatched. Loon chicks were captured and banded at 5 to 7 weeks
post-hatch. Chicks alive at this age were considered "fledged". Observers also collected water quality
data including Secchi depth, color, total phosphorus, temperature, pH, conductivity, and dissolved
oxygen (Table 5-1).
117
Table 5–1. Habitat variables used in modeling, with abbreviations used in other tables and summary statistics
Abbrevia-
Variable tion N Mean St Dev Median Min Max
Morphometry
Lake area, ha Lh 292 96.69 228.47 29.22 4.11 2028.53
Lake perimeter, m Lp 292 6296.30 8919.66 3251.70 837.99 62804.88
Shoreline development index sdi 292 2.00 1.16 1.66 1.04 8.87
Maximum depth, m Md 289 7.97 8.52 6.10 0.30 108.20
Number of waterbodies Nw 292 36.15 21.10 35.00 2.00 70.00
> 10 ha within 10 k
Water quality
Secchi depth, m S 285 2.41 1.81 2.13 0.15 9.14
pH ph 286 7.35 1.13 7.41 4.19 9.48
Color Col 286 88.22 115.65 30.00 0.00 500.00
Conductivity, us/cm Con 286 0.100 0.108 0.068 0.006 0.800
Dissolved oxygen, % Dop 286 99.61 29.63 102.85 4.60 168.90
Dissolved oxygen, mg/L Dom 286 8.54 2.64 8.84 0.37 15.25
Total phosphorus, mg/L tp 282 0.023 0.030 0.014 0.000 0.237
Development
Houses/km in 100 m buffer Hk 286 6.42 9.36 1.28 0.00 54.80
Total access points Tap 292 0.57 0.95 0.00 0.00 7.00
Number improved accesses Imp 292 0.47 0.90 0.00 0.00 7.00
Number unimproved accesses Uni 292 0.10 0.32 0.00 0.00 2.00
Percentages of land cover in
150 m buffer of lakeshore
Roads Pr150 292 0.03 0.02 0.03 0.00 0.12
Open water Po150 292 10.89 8.60 8.77 0.53 51.85
Developed Pd150 292 10.90 8.84 8.69 0.00 58.58
Forest Pf150 292 39.41 24.12 39.17 0.00 90.50
Shrub/herbaceous Ph150 292 2.24 4.62 0.00 0.00 32.08
Agriculture Pa150 292 7.39 13.24 0.00 0.00 71.15
Wetland Pw150 292 29.08 20.44 23.25 0.00 97.11
Barren Pb150 292 0.10 0.90 0.00 0.00 10.76
Percentages of land cover in
500 m buffer of lakeshore
Roads Pr 292 0.03 0.02 0.02 0.00 0.36
Open water Po 292 9.79 8.53 7.61 0.30 41.17
Developed Pd 292 8.84 6.95 7.29 0.00 61.86
Forest Pf 292 43.77 22.81 45.55 0.00 94.71
Shrub/herbaceous Ph 292 2.56 4.79 0.44 0.00 38.68
Agriculture Pa 292 9.89 14.97 0.10 0.00 71.26
Wetland Pw 292 25.05 18.62 19.31 0.59 90.77
Barren Pb 292 0.10 0.93 0.00 0.00 15.06
Observers classified the shoreline of lakes surveyed into categories of nest habitat quality by
traveling the lake perimeter by boat and segmenting the shoreline on 24K USGS topography maps by
habitat type. The quality of nest habitat was evaluated on a 4-level scale for each lake or reservoir
surveyed. Nest habitat quality was categorized for each lake/reservoir with one level assigned to each
segment. All islands were classified into habitat quality category as well. Habitat quality was assessed
according to the following criteria:
118
EXCELLENT - Lake/reservoir contains islands in the I-1 and I-2 categories, vegetation
hummocks >3m from shore, artificial nest platforms, floating vegetation masses (i.e. root masses of
lilypads, sphagnum mats) >3m from shore.
GOOD - Lake/reservoir contains islands in the I-3 category. Extensive shoreline wetland areas
(floating bog mats, marshes, swamp, etc) that are >35m in width and >10m in depth along the lake
shoreline. This also includes vegetation hummocks or floating vegetation masses that are <3m from
shore. This extent of wetland reduces the likelihood of mammalian predation but also provides sites for
easy nest construction.
FAIR - Lake/reservoir contains Islands in the I-4 category. Limited shoreline wetland areas
(<35m in width or <10m in depth or both) present.
POOR - Lake/reservoir contains only upland and/or developed shoreline habitat without wetland
area fringe.
Island habitat quality was assessed according to the following criteria:
I-1 - Small islands (< 5m diameter) with shore substrate amenable (not continuously rocky or
physical barrier to loon access such as rock ledge) to nest construction and > 10m from shore with few
or no perch trees (for eagles).
I-2 - Small islands (<5m diameter with perch trees present or < 10m from shore with or without
perch trees) and medium islands (5-50m diameter, with or without perch trees) > 10m from mainland
shoreline with shore substrate amenable (not continuously rocky or physical barrier to loon access such
as rock ledge) to nest construction.
I-3 - Medium islands < 10m from shore or large islands (>50m diameter) any distance from
shore.
119
I-4 - All islands with continuous or periodic human activity at >50% of available nest sites (i.e.
camping, cabin, etc) or with shore substrate NOT amenable (continuously rocky or physical barrier to
loon access such as rock ledge) to nest construction.
The density and total surface area of open water bodies >10 ha in size that occurred within 10
km of each lake/reservoir’s centroid was determined from the Wisconsin DNR 1:24,000-Scale Surface
Water Hydrography datalayer. Road density within 150 m and 500 m of each lake/reservoir was
determined using the U.S. Census Bureaus's year 2000 TIGER/Line files
(ftp://dnrftp01.wi.gov/geodata/US_Census_2000_Roads/). The Wisconsin DNR boat access data layer
(http://dnr.wi.gov/org/land/facilities/boataccess/) combined with observer experience was used to
determine the number of improved and unimproved boat access points on each lake/reservoir. House
density (houses/km shoreline) within the northern Wisconsin study area was based on the occurrence of
houses within a 100 m buffer of lakes according to county housing layers. Within the central Wisconsin
study area, we tallied houses within a 100 m buffer of the shoreline of each lake/reservoir that appeared
in high-resolution aerial imagery (http://goto.arcgisonline.com/maps/I3_Imagery_Prime_World_2D).
120
Figure 5–1. Box plots for lake area, maximum depth, lake density and secchi depth within each ecoregion.
Water quality parameters were assessed at each site one time per month. Collection time, GPS
coordinates, Secchi depth, and total depth were recorded first at each site. A YSI 6600 sonde was used
to measure temperature (oC), conductivity (mS/cm
3), dissolved oxygen percent saturation, dissolved
oxygen (mg/L), and pH. Each measurement was taken at every half meter depth interval for lakes less
than five meters in depth, and every meter interval for lakes greater than or equal to five meters in
depth. Apparent water color was determined using a Hach model CO-1 color test kit. A water sample
Central Sand Glacial Lake Northern Highland
510
20
50
100
200
500
2000
Ecoregion
Lake s
ize,
ha
Central Sand Glacial Lake Northern Highland
05
10
15
20
25
30
35
Ecoregion
Max d
epth
, m
Central Sand Glacial Lake Northern Highland
010
20
30
40
50
60
70
Ecoregion
Num
ber
of
lakes w
ithin
10 k
m
Central Sand Glacial Lake Northern Highland
02
46
8
Ecoregion
Secchi depth
, m
121
for total phosphorous analysis was taken once per month at each site by inverting a 100 mL Nalgene
sample bottle and submerging it to approximately 0.5 m, then reverting and filling to about the neck of
the bottle. Samples were preserved with sulfuric acid and then maintained on ice until they were
submitted for analysis to the USGS water quality lab at the Upper Midwest Environmental Sciences
Center, La Crosse, WI.
Figure 5–2. Box plots for lake pH, color, total phosphorus and dissolved oxygen within each ecoregion.
Central Sand Glacial Lake Northern Highland
45
67
89
Ecoregion
Lake p
H
Central Sand Glacial Lake Northern Highland
12
510
20
50
100
200
500
Ecoregion
Colo
r
Central Sand Glacial Lake Northern Highland
0.0
00.0
50.1
00.1
50.2
0
Ecoregion
Tota
l phosphoru
s,
mg/L
Central Sand Glacial Lake Northern Highland
050
100
150
Ecoregion
Dis
solv
ed o
xygen,
%
122
Land cover within 150 m and 500 m of each lake or reservoir was estimated using the 2006
National Land Cover Dataset 30-m resolution land cover classification scheme (Fry et al. 2009). We
considered seven land cover categories based on the Anderson Level I classification scale (Anderson et
al. 1976): open water, developed, forest, shrub/herbaceous, agricultural, wetland, and barren.
Figure 5–3. Box plots for selected land-use classifications within each ecoregion.
Central Sand Glacial Lake Northern Highland
010
20
30
40
50
60
70
Ecoregion
% a
griculture
, 500 m
buff
er
Central Sand Glacial Lake Northern Highland
020
40
60
80
Ecoregion
% f
ore
st,
500 m
buff
er
Central Sand Glacial Lake Northern Highland
020
40
60
80
Ecoregion
% w
etland,
500 m
buff
er
Central Sand Glacial Lake Northern Highland
010
20
30
40
50
60
Ecoregion
% d
evelo
ped,
500 m
buff
er
123
Statistical analysis
We used presence or absence of loon territorial pairs as the response variable in logistic
regression analyses with multiple habitat predictor variables. Because of the conspicuousness of loons
and the intensity of sampling, detection rates were considered extremely high and we did not estimate
them. Most of the lakes occupied by loons had only one loon pair, although a few of the large lakes in
the Northern Highlands ecoregion did have multiple pairs. Before fitting any models we examined the
distribution of all predictor variables individually for highly skewed distributions and outlying
observations. We log-transformed predictor variables that were positive and with long upper tails; this
is generally appropriate for predictor variables in both linear and generalized linear models (Fox 2008).
We computed pairwise correlations and examined pairwise scatterplots among predictor variables to
determine if some pairs of variables were highly collinear; while collinearity among predictor variables
is not necessarily a problem for predictive models, it can lead to difficulties in interpreting model
coefficients and decreased precision in their estimates (Fox 2008).
Logistic regression models with multiple habitat predictors have been used extensively for
habitat selection and species distribution models (e.g., Johnson et al. 2006, Manly et al. 2002, Guisan et
al. 2002) and they have been used specifically to develop such models for Common Loons and for
related waterbirds (Kuhn et al. 2011, Found et al. 2008). A number of recently developed
computationally intensive methods have also been used for similar modeling purposes (Cutler et al.
2007, Prasad et al. 2006). We used the method of random forests primarily as a way of evaluating
predictor variable importance. This method involves fitting classification or regression trees to multiple
bootstrap samples of observations, with a random subset of predictor variables included for each tree.
This is repeated a large number of times (e.g., 500) and the prediction for each observation is
determined as the majority class (if the response is binary or categorical) or mean response (if the
124
response is quantitative) over the total set of trees (Berk 2010). Predictor variable importance can be
evaluated as the decrease in prediction accuracy when a predictor is not allowed to contribute to the
predictions from the model. This provides an alternative method to significance testing or Akaike’s
Information Criterion (AIC) for evaluating the importance of predictors (Cutler et al. 2007).
The method of random forests does not result in a single model, or even a small number of
possible models, but instead provides predictions or estimates based on summaries over a large number
of models. It can be very effective for prediction when all predictor variables that were involved in the
initial modeling are available for future predictions. This was not the case for our situation where we
could not obtain values of all predictor variables under different future climate scenarios. Random
forests can also be used to provide a visual description of the relationship between each predictor and
the response. This can reveal nonlinear relationships or thresholds that would be difficult to fit and
evaluate with a method such as logistic regression (Berk 2010). We used random forests as an
exploratory tool to assess predictor variable importance and to describe the nature of relationships
between the response (loon presence) and predictors.
For logistic regression modeling, we followed the overall modeling strategy suggested by
Burnham and Anderson (2002) that involves developing an a priori set of candidate models, fitting them
by maximum likelihood, and ranking the models by AIC. We developed a set of candidate models
based on previous research on Common Loons carried out in New Hampshire and Wisconsin (Kuhn et
al. 2011, Meyer 2006), research on yellow-billed (Earnst et al. 2006) and Pacific loons (Heglund et al.
1994), and studies of other northern waterbirds (Found et al. 2008). These studies suggested that
models for loon occupancy should include some measure related to water clarity (Secchi depth, color,
total P), some measure of lake size (lake area, lake perimeter, maximum depth), one or more variables
describing nest habitat, and a measure of spatial distance among loons. Because we were unable to
125
determine the distance to the nearest lake with loons present, as was done in New Hampshire (Kuhn et
al. 2011), we used the number of lakes greater than 10 ha in size within 10 km of each lake’s centroid.
We also examined the effect of human development and nearby land cover, because of concern over
impending changes in both in northern Wisconsin. We used the random forests procedure to determine
variable importance both as a check on our a priori assessment of variables to consider in models and as
an aid in selecting land cover and human development variables to include in larger models. We realize
this is not strictly in keeping with the a priori model development philosophy advocated by Burnham
and Anderson (2002), but it is consistent with the goal of assessing the potential impact of human
development and land cover metrics on loon habitat use.
We fit models to data from all three ecoregions (292 lakes) and to the Northern Highlands
ecoregion alone (146 lakes), where loon density and occupancy is greatest. Our emphasis is on models
for data from all three ecoregions because this data set encompasses the environmental conditions that
may occur in northern Wisconsin if current climate trends continue. We used AIC corrected for small
sample size (AICc) to compare models (Burnham and Anderson 2002), although our sample size was
fairly large. For purposes of generating predictions under various climate change scenarios, we selected
one best model based on AICc, the scientific interpretation of the models, and predictive accuracy. We
also examined residuals from the best fitting models to ensure that assumptions of the models were
satisfied, and we computed the variance inflation factor to examine possible effects of multicollinearity
on estimates (Fox 2008).
Predictions from logistic regression models are probabilities (p) or logits (log(p/(1-p)); in our
models these are probabilities that one or more territorial loon pairs are present on a lake. For assessing
classification, these probabilities can be converted to presence if the predicted probability is greater than
a threshold, usually 0.5, although the probability threshold may be adjusted depending on the prevalence
126
of the species (Manel et al. 2001, Boyce et al. 2002). We used the 0.5 probability threshold for
assessing predictive accuracy of models by methods that required a threshold (note that prevalence of
loon territorial pairs was 35% for our entire data set, and 51% for the Northern Highlands data).
We used four methods to assess the predictive accuracy of models. Three of the methods
required specifying a probability threshold: these are the proportion correctly classified, a cross-
validation estimate of the proportion correctly classified, and Cohen’s kappa statistic. The cross-
validation estimate of the proportion correctly classified was computed by dividing the data set
randomly into 10 groups, fitting the specified model using data from 9 of the groups, classifying lakes in
the 10th group that was not used in fitting the model, and repeating this for all 10 groups (Boyce et al.
2002, Maindonald and Braun 2007). Because this proportion varies depending on the random allocation
of observations to groups, we repeated it 10 times and computed the prediction accuracy as the mean of
the 10 estimates. Cohen’s kappa is a measure of agreement of the predicted classes with the observed
classes that is adjusted for chance agreement, because some agreement will occur by chance even for
two completely independent measures of classification (Agresti 2007). Values of Cohen’s kappa range
from 0 (no better than chance agreement) to 1 (perfect agreement). Values greater than 0.75 are
considered excellent, and values between 0.4 and 0.75 are considered good (Fielding and Bell 1997)
A fourth method of assessing classification accuracy that does not require specifying a
probability threshold is the receiver operating characteristic (ROC) curve. The area under the ROC
curve (AUC) provides an index of prediction accuracy that summarizes performance independent of any
threshold. An AUC of 1.0 indicates perfect classification, while an AUC of 0.5 is the lowest possible.
AUC values of 0.7 to 0.9 are considered useful, while values greater than 0.9 are excellent (Manel et al.
2001).
127
We evaluated effects of changes in predictor variables on the presence of loon territorial pairs by
estimating loon presence for our sample of lakes under various percentage changes in predictor
variables from current conditions. The changes in predictor variables were applied to each lake
individually, so that, for instance, a 10% change in Secchi depth for a clear lake would be much larger
than a 10% change for a murky lake. For lakes in the Trout Lake basin where individual lake conditions
were estimated under several climate change scenarios, we also obtained predictions for all lakes in the
watershed under each scenario.
Predicted probability of occurrence following climate change
We used the best-fitting loon habitat model to predict loon occurrence for 27 of the lakes in the
Trout Lake basin under projected future climatic conditions. Linked climate and lake hydrology models
(see Chapter 2) were used to simulate lake conditions for all years from 1962 to 2100 under 3 emissions
scenarios and 6 general circulation models (GCMs). One of the 18 combinations gave unrealistic results
and was dropped, leaving 17 model outputs for each year. The loon habitat model included 4 predictor
variables plus an interaction: nest habitat quality, number of waterbodies within 10 km, secchi depth,
and log of lake area, as well as the interaction between secchi depth and log of lake area. The only one
of these predictor variables directly simulated by the linked climate and lake hydrology model was
secchi depth. Measured values of secchi depth used in the development of the loon habitat model
differed somewhat from the simulated values for 2010, but our goal here is to assess the effects of
climate change on loon occurrence without any confounding due to differences in current evaluations of
secchi depth. Changes in precipitation and lake stage could potentially affect lake area and loon nesting
habitat, but the linkages, especially with nesting habitat, are not well enough understood to be
simulated.
128
To evaluate the potential impact of enhanced total phosphorus input to lakes due to changes in
future landcover or landuse, two total phosphorus change scenarios were also incorporated (10% and
25% increases), resulting in 51 model outputs for each year between 1962 and 2100. We estimated the
probability of loon occurrence, based on our loon habitat model with simulated values of secchi depth,
for each of the 27 lakes in the Trout Lake basin in 2010, 2050, and 2090 under all 51 climate model
outputs. We evaluated possible effects of changes in nest habitat on loon occurrence as step changes by
one level among the four discrete levels of nest habitat. Nest habitat could be affected by changes in
precipitation and lake stage as well as by human development along shorelines. Note that nest habitat in
the best category cannot improve beyond that category, and nest habitat in the worst category cannot
further decline. We summarized the predicted probabilities as the minimum, median, and maximum
probability over the 17 predictions for each of the 3 levels of change in total phosphorus and for an
increase or decrease in nest habitat quality
Results
Loon habitat suitability model
Of 33 habitat variables measured, 24 were available for all 292 lakes sampled (variables listed in
Table 5-1 plus nest habitat). Because of correlations greater than 0.9 between some pairs of variables,
we used in our analyses % dissolved oxygen but not dissolved oxygen (mg/L), number of improved boat
ramps but not total number of access sites, and number of waterbodies greater than 10 ha within 10 km
but not the surface area of those waterbodies. The % land cover variables in the 150m buffer were
strongly correlated with the same variables in the 500m buffer, but we used results of random forests
(below) to make decisions about which of these variables to include in other analyses. We log
transformed (base 10) conductivity, color (after adding 1), total phosphorus (mg/L), lake area (ha), lake
129
perimeter (m), and number of houses (after adding 1) because these variables all exhibited highly
skewed distributions with long upper tails.
Territorial loon pairs were present in 103 of 292 lakes sampled (35%; Table 5-2). The
percentage of lakes occupied by territorial pairs was 51% in the Northern Highland ecoregion, 27% in
the Glacial Lake Wisconsin ecoregion, and 4% in the Central Sand ecoregion (Table 5-2). Note that
percentages of occupied lakes are biased upward for the Glacial Lake and Central Sand ecoregions
because for these ecoregions we included one stratum of lakes with historical information indicating
past use by loons. This does not bias the habitat selection model, however. Lakes in all four nest
habitat categories were sampled, with the fewest lakes (39) in the poor category (Table 5-3). The
percentage of lakes occupied by territorial loons was strongly related to nest habitat category, with 67%
of lakes in the excellent category occupied, but none of the lakes in the poor category occupied (Table
5-3). Most lakes in the Central Sand ecoregion were in the poor or fair nest habitat categories, while
most in the Glacial Lake and Northern Highland ecoregions were in the good or excellent nest habitat
categories (Table 5-4). There were large differences among ecoregions in many of the other habitat
variables as well (Figures 5-1 through 5-3). Lakes in the Northern Highland ecoregion have many more
neighboring lakes within 10 km, a lower proportion of surrounding land cover in agriculture, and
greater proportion in forest than lakes in the other two ecoregions. Lakes in the Glacial Lake ecoregion
are largely reservoirs associated with cranberry growing operations and have shallower Secchi depths,
shallower maximum depths, more color, and more total phosphorus than lakes in the other ecoregions.
130
Table 5–2. Common loon territorial pairs and nest habitat by ecoregion
Territorial pairs
Ecoregion absent present % present All
Central Sand 44 2 4 46
Glacial Lake 73 27 27 100
Northern Highland 72 74 51 146
All 189 103 35 292
Table 3. Common loon territorial pairs by nest habitat
Territorial pairs
Nest habitat absent present % present Al
Excellent 28 57 67 85
Good 62 35 36 97
Fair 60 11 15 71
Poor 39 0 0 39
All 189 103 35 292
Table 4. Nest habitat quality among ecoregions
Nest habitat
Ecoregion excellent good fair poor All
Central Sand 5 4 18 19 46
Glacial Lake 45 25 22 8 100
Northern Highland 35 68 31 12 146
All 85 97 71 39 292
Random forests analyses indicated that the most important predictor variables for loon presence
in the full data set were: nest habitat, number of waterbodies within 10 km, log of total phosphorus, log
of lake perimeter, Secchi depth, log of conductivity, proportion of surrounding land cover in agriculture
(% agriculture) within either 150m or 500m of the shoreline, and log of lake area (Figure 5-4). Of these
variables, nest habitat and number of waterbodies within 10 km were the most important. Based on
results of these analyses we included % forest and % agriculture in both 150 m and 500 m buffers in
logistic regression models, but not any other land cover or development variables.
131
Table 5–3. AICc for logistic regression models fit to data from all three ecoregions (276 lakes with no missing values
for these predictors). See abbreviations in Table 5-1. Log-transformed variables are log of total phosphorus
(Ltp), log of lake area (Lh), log of color (Lcol), log of shoreline development index (Lsdi), and log of lake
perimeter (Lper). Interactions are identified by ‘*’ between the variables included in the interaction. K is the
number of parameters in the model.
AICc Cumul. Log
Model K AICc Δ_AICc Weight Weight likelihood
H+S+Lh+Nw+S*Lh 8 204.49 0.00 0.49 0.49 -93.98
H+S+Lh+Nw+Pf+S*Lh 9 205.83 1.33 0.25 0.74 -93.58
H+S+Lh+Nw+Pf150+S*Lh 9 205.98 1.48 0.23 0.97 -93.65
H+S+Lh+Nw+Pf 8 213.39 8.89 0.01 0.98 -98.42
H+S+Lh+Nw 7 214.18 9.69 0.00 0.98 -99.88
H+S+Lh+Nw+S*Nw 8 215.18 10.68 0.00 0.98 -99.32
H+S+Lh+Nw+Lh*Nw 8 215.24 10.75 0.00 0.99 -99.35
H+S+Lh+Nw+Lcol+Pf 9 215.36 10.87 0.00 0.99 -98.34
H+S+Lh+Nw+Ltp+Pf 9 215.39 10.90 0.00 0.99 -98.36
H+S+Lh+Nw+Ltp 8 215.68 11.19 0.00 0.99 -99.57
H+S+Lh+Nw+Lcol 8 216.02 11.52 0.00 0.99 -99.74
H+S+Lh+Nw+Pa 8 216.15 11.65 0.00 1.00 -99.80
H+S+Lh+Nw+Pa 150 8 216.30 11.80 0.00 1.00 -99.88
H+S+Lper+Nw 7 216.72 12.22 0.00 1.00 -101.15
H+Ltp+Lh+Nw+Ltp*Lh 8 216.99 12.49 0.00 1.00 -100.22
H+S+Lh+Nw+Ltp+Nw*Ltp 9 217.76 13.27 0.00 1.00 -99.54
H+S+Lh+Nw+Lcol+Lcol*Nw 9 218.10 13.60 0.00 1.00 -99.71
H+S+Nw+Pf 7 222.20 17.71 0.00 1.00 -103.89
H+S+Nw 6 222.63 18.14 0.00 1.00 -105.16
H+S+Lsdi+Nw 7 223.22 18.73 0.00 1.00 -104.40
H+Ltp+Lh+Nw 7 228.35 23.85 0.00 1.00 -106.96
H+Lcol+Lh+Nw 7 230.02 25.52 0.00 1.00 -107.80
H+Ltp+Lh+Nw+Nw*Ltp 8 230.40 25.91 0.00 1.00 -106.93
Because of missing data for some predictor variables, we used data from 276 lakes for model
comparisons, instead of the full data set of 292 lakes from the three ecoregions. Model comparisons
using AIC require that all models must be fit to the same data set, so the same number of observations
must be available for all predictor variables. Three of the logistic regression models fit substantially
better than any of the other models, as judged by AICc (Table 5-3). Of these, the best fitting model
included as predictor variables nest habitat, number of waterbodies within 10 km, Secchi depth, log of
lake area, and the interaction between Secchi depth and log of lake area. The two next best models
132
included these same variables plus % forest within either the 500 m or 150 m buffer. In all three
models, the predicted probability of loon presence increased as the number of nearby waterbodies
increased (Table 5-6). The effect of the interaction between Secchi depth and log of lake area in all the
models was to reduce the effect of Secchi depth on the probability of loon presence in small lakes, while
increasing its effect in larger lakes (Figure 5-5). Small lakes with good or excellent nest habitat were as
likely to be occupied by loons when they were shallow and stained as when deep and clear. Conversely,
in the largest lakes, the probability of loon presence increased very rapidly with increasing water clarity,
unless there was only poor nest habitat available.
133
Figure 5–4. Variable importance as determined by the random forests procedure for the data set including all three
ecoregions. Variable abbreviations are listed on the y-axis; the x-axis is the mean % decrease in prediction
accuracy for loon presence if the variable is not included in the model. Predictor variables are listed in order
from most important at the top.
Mean % decrease in prediction accuracy
pwet150
pwet500
ncanoe
pshrub150
roads150
nramp
pdevel150
pdevel500
loghouse
popen500
dopct
roads500
pshrub500
pforest500
popen150
logdepth
pforest150
ph
logcolor
logsdi
logha
pag150
pag500
logconduct
secchi
logperim
logtp
nwater10k
flkhab
0 1 2 3 4
134
Table 5–4. Coefficient estimates and standard errors for the best fitting models from Table 5-3. These models were
fit to the largest data set without any missing values for these predictor variabiles, which consisted of 285
observations.
Model Model
without % forest with % forest
Predictor estimate SE estimate SE
Intercept -0.51 1.052 -0.65 1.065
Nest habitat – good -2.24 0.498 -2.17 0.499
Nest habitat – fair -3.98 0.650 -3.91 0.653
Nest habitat – poor -24.99 1261.54 -24.80 1262.64
Secchi depth, meters -0.86 0.473 -0.71 0.496
Log of lake area, ha -0.55 0.636 -0.41 0.654
# lakes within 10 km 0.033 0.012 0.040 0.014
Secchi - log lake area 1.05 0.334 0.98 0.341
interaction
% forest, 500 m buffer -0.011 0.012
The coefficients for % forest in the 2nd
and 3rd
best fitting models were negative (Table 5-4
includes estimated coefficients for the two models with minimum AICc). This is an unexpected result
and difficult to explain. Exploratory analyses using random forests and classification trees suggested a
complicated nonlinear relationship between % forest and loon presence that may not be possible to
represent easily in a logistic regression model. Coefficients for other predictors in the two models with
minimum AICc were similar, so inclusion of % forest as a predictor did not change other relationships in
the model. We used the simpler model with minimum AICc and without % forest for predictions.
135
Figure 5–5. Predicted probability of territorial loon presence based on the best fitting model for data from all three
ecoregions. Predictions are for lakes with a fixed number of lakes within 10 km (equal to the observed mean in
the data set). Each line is for one nest habitat category (excellent – solid line; good – dashed line; fair – dotted
line; poor – dotted/dashed line). Numbers in the strips above each panel are the log of lake area that the
predictions in the panel assume.
Secchi depth, meters
Pre
dic
ted
pro
ba
bility o
f lo
on
pre
se
nce
0.0
0.2
0.4
0.6
0.8
1.0
2 4 6 8
0.75 1.25
2 4 6 8
1.75
2.25
2 4 6 8
2.75
0.0
0.2
0.4
0.6
0.8
1.0
3.25
136
Because nest habitat is a categorical variable, the coefficients for levels of this variable change
the intercept of the logistic regression model. Here the ‘excellent’ category is considered the baseline
level, and coefficient estimates for other levels determine deviations from that baseline (as illustrated in
Figure 5-5). The estimated coefficient for the poor nest habitat category has an extremely large standard
error (Table 5-4) because there were no loons present in lakes with poor nest habitat. A logistic
regression coefficient of negative infinity would be required to produce a predicted probability of zero
for this nest habitat category and in this situation the maximum likelihood estimate is not uniquely
defined (Fox 2008). We compared this model with a simpler one in which the fair and poor nest habitat
categories were combined into one so that every nest habitat category had loons present and absent.
While the coefficients for this new nest category differed from any in the full model, all other
coefficients were almost identical, suggesting that estimated coefficients from the full model were
appropriate. Residual plots indicated that assumptions of the analysis were approximately satisfied, and
variance inflation factors suggested no problems with multicollinearity.
Table 5–5. Estimates of prediction accuracy for selected logistic regression models
Cross-validation
estimate of
Proportion proportion Cohen’s
Ecoregions Model correct correct kappa AUC
All H+S+Lh+Nw+S*Lh 0.832 0.813 0.625 0.917
All H+S+Lh+Nw+Pf+S*Lh 0.818 0.808 0.598 0.919
We fit additional models that included maximum depth or the log of maximum depth, and
interactions of these two predictor variables with Secchi depth. These models were suggested because
the interaction term might help account for loon occupancy of shallow, stained reservoirs such as those
in the Glacial Lake ecoregion where loons sometimes occur, as opposed to large, deep lakes with low
Secchi values that rarely have loon territorial pairs (but also see comments on the interaction of Secchi
137
depth and log of lake area above). These models are not included in Table 5-3 because maximum depth
was missing for some lakes included in data used to construct that table. Comparisons of the models
including maximum depth were carried out using the reduced data set of 274 observations. Models
including maximum depth as a predictor had much larger AICc values than the simpler model
corresponding to the best model in Table 5-3.
The two best fitting models for data from all three ecoregions appeared to predict loon
occupancy quite well (Table 5-5). Cross-validation estimates of the proportion correctly classified were
above 0.8 and AUC values were above 0.9. The simpler model without % forest as a predictor
performed slightly better.
Table 5–6. The change in the percentage of lakes occupied by loon territorial pairs as predictor variables change
from current values. These were estimated for all 285 lakes without any missing values for these predictor
variables. Currently, 102 of 285 lakes in the sample, or 35.8%, were occupied by territorial loon pairs. The %
changes in the table represent additive changes from the current state – thus, a change of -10% would result in
35.8% - 10% = 25.8% of lakes occupied. Note that the % changes in predictor variables were applied to each
lake separately – thus, a 10% change in Secchi depth for a very clear lake would be much larger than a 10%
change for a murky lake.
% change in the percentage of lakes occupied
% change in Predictors
predictor # waterbodies log lake area Secchi depth
-50% -9.5% -16.5% -12.6%
-40% -8.1% -14.4% -8.4%
-30% -6.3% -11.9% -6.3%
-20% -4.6% -7.0% -3.9%
-10% -1.4% -2.8% -2.1%
None - - -
+10% +1.4% +2.1% +1.1%
+20% +3.5% +5.6% +3.9%
+30% +6.0% +7.4% +5.3%
+40% +9.8% +10.2% +6.7%
+50% +12.3% +13.0% +9.5%
138
We assessed the relative importance of predictor variables in the best model by comparing the
effect of measured increases among the three predictor variables, Secchi depth, log of lake area, and
number of waterbodies within 10 km, on predicted loon occupancy (Table 5-6). The resulting change in
loon occupancy was similar among the three variables. The effect of a step change in nest habitat
quality is more dramatic, however (Table 5-7). If nest habitat were to decrease by one category on all
lakes (excellent habitat becomes good, good becomes fair, fair becomes poor, and poor remains poor),
then the percentage of lakes occupied by loons is estimated by the model to decrease from the current
value of 35.8% to 12.3% for the set of 292 lakes surveyed. If nest habitat increased by one category on
all lakes, then the percentage of lakes occupied is estimated to increase from the current value of 35.8%
to 54.7%.
139
Table 5–7. Predicted probability of loon occupancy with changes in nest habitat for lakes in the Trout Lake basin. A
decrease in habitat quality is a decrease of one category (e.g., from excellent to good, or good to fair); an
increase in habitat quality is an increase of one category. Current habitat quality is designated as E (excellent),
G (good), F (fair), and P (poor). Bold indicates observed or predicted presence of territorial pairs.
Predicted probability
Current of loon presence
habitat loon Change in habitat quality
Lake quality presence decrease increase
Allequash Lake E 1 0.936 0.993
Big Muskellunge Lake F 1 0.000 1.000
Crystal Lake P 0 0.000 0.799
Day Lake F 1 0.000 0.996
Diamond Lake P 0 0.000 0.826
Edith Lake F 0 0.000 0.625
Escanaba Lake G 1 0.878 0.997
Fallison Lake G 1 0.380 0.970
Firefly Lake P 0 0.000 0.223
Frank Lake G 1 0.557 0.985
Jag Lake F 1 0.000 0.668
Little John Lake G 0 0.407 0.973
Little Rock Lake F 1 0.000 0.753
Lost Canoe Lake P 0 0.000 0.981
Mann Lake G 0 0.230 0.941
Nebish Lake G 1 0.841 0.996
Nichols Lake F 0 0.000 0.274
Pallette Lake P 0 0.000 0.825
Rudolph Lake G 0 0.091 0.842
Sparkling Lake P 0 0.000 0.978
Starrett Lake P 0 0.000 0.146
Street Lake F 0 0.000 0.392
Trout Lake G 1 1.000 1.000
Unnamed (Max) P 0 0.000 0.073
Vandercook Lake G 0 0.431 0.976
White Birch Lake G 1 0.738 0.993
Wildwood Lake G 0 0.032 0.640
Predicted probability of occurrence following climate change
Secchi depth estimates from the linked climate and hydrology models for the period between
2010 and 2090 increase slightly in the 27 lakes in the Trout Lake basin (Chapter 2). To focus on effects
of changing secchi depth, we examined the probability of loon occurrence under climate change
scenarios for the years 2050 and 2090 (Appendix Figures A-1 through A-5). Estimated probabilities of
loon occurrence stay the same or increase very slightly between 2010 and 2090 for all 27 lakes as secchi
140
depth slowly increases over time. Higher levels of variability in the probability of loon occurrence for
some lakes (e.g., Jag Lake) are due to greater variability among projected secchi depth estimates from
the linked climate and lake hydrology models for those lakes.
Changes in nest habitat can have large effects on the probability of loon occurrence in lakes
(Appendix Figures A-6 through A-12). Step changes in nest habitat generally overwhelm the effect of
estimated changes in secchi depth or possible changes in total phosphorus. The exception to this is in
lakes with very high or very low current probabilities of loon occurrence, where single step changes in
nest habitat have little effect. These are either large, clear lakes with excellent nest habitat or small
lakes with low water clarity and poor habitat; very large changes in habitat or water clarity would be
required to significantly alter probabilities of loon occurrence in these lakes. Increasing lake total
phosphorus concentrations up to 25% reduces the probability of loon occurrence on all 27 lakes by
decreasing water clarity, however the change in secchi depth is generally less than 10% in the lakes of
the Trout Lake basin.
Discussion
Our regional common loon habitat model was based on data from lakes and reservoirs across
diverse landscapes, from the heart of the loon’s breeding range in northcentral Wisconsin (Northern
Highlands ecoregion) to the southern edge of the current breeding distribution in central Wisconsin
(Glacial Lake and Central Sand ecoregions). Consequently, lakes and reservoirs were represented
across a wide continuum of physical and chemical characteristics, degree of isolation, surrounding land
cover types and composition, and level of human development. The model performed very well in
predicting common loon occupancy based on availability of nest habitat, water clarity, lake size, density
of neighboring lakes, and Secchi-lake size interaction, with a prediction accuracy of better than 80%.
141
Kuhn et al. (2011) modeled common loon habitat suitability in New Hamphsire, and found similar
factors associated with loon occupancy. Overall optimal breeding habitat for loons across New
Hampshire was identified as having the following characteristics: clear, higher elevation lakes with
islands; further away from human population centers; with lower surrounding road densities; and with
nearby lakes occupied by other loons (this parameter was not measured in this study as all adjacent
lakes were not surveyed). Water clarity has long been associated with loon habitat preference, and New
Hamphsire results are consistent with previous studies (Vermeer 1973, McIntyre 1983, 1988, Blair
1992, Meyer 2006). The significance of the minimum-distance-to-nearest-lake-with-loon-presence
metric supports anecdotal observations and published research describing loon natal dispersal. This
biotic interaction metric describes the philopatric and conspecific attraction loons display and reveals
the importance of the spatial configuration of lakes in the selection of loon breeding habitat (Strong et
al. 1987, Evers 2001, Piper et al. 2006, 2008). The unique habitat feature that distinguishes the best
model among the New Hampshire statewide models was the presence of islands, which contribute to
high-quality breeding habitat for loons (Vermeer 1973, Jung 1991, McIntyre and Barr 1997, Piper et al.
2006, Evers 2007). Based on odds ratio estimates, islands and lake clarity were the most important
factors distinguishing loon breeding habitat in New Hampshire. The presence of islands was also
reflected in loon nest habitat quality in our Wisconsin model, the factor most closely associated with
loon occupancy.
A positive relationship was found between lake size and loon occupancy in both Wisconsin (this
study) and New Hampshire (Kuhn et al. 2011). Piper et al. (2012) found juvenile loon survival (hatch
through age 3-years) was lower, and 2-chick broods showed greater disparity in body mass and lower
survival through week 5, in chicks reared on small, acidic lakes in Wisconsin. Loon pairs are often food-
limited in small lakes (Alvo et al. 1988: Barr 1996), and Piper et al. attributed their findings to lower
142
prey availability (adult loons rearing chicks on small lakes also occasionally led chicks overland to
larger lakes). Winter die-offs of fishes (common on small, shallow lakes in the region; Magnuson et al.
1985) prevents fledging of chicks from lakes that supported chick production in previous years or
permit only one of two chicks to survive. Availability of forage fish is undoubtedly an important
component influencing the suitability of lakes and reservoirs for species that are primarily piscivores,
such as the common loon. Unfortunately, we did not obtain a metric that reflects fish forage base of the
lakes and reservoirs included in our study, nor is there an obvious assessment method that could be
applied across water bodies that vary in size and morphometry.
Future climate conditions will likely change the distribution and structure of fish communities in
Wisconsin, but not necessarily fish abundance. Coldwater fishes, such as cisco (Corgeonus artedii),
occur within the Trout Lake watershed (Big Muskellunge, Pallette, and Trout Lakes), require cold water
temperatures, high dissolved oxygen concentrations, and oligotrophic conditions. Under climate change
scenarios, as air temperatures increase, epilimnion and hypolimnion water temperatures are expected to
increase. As water temperatures increase, the duration of the lake stratification period is expected to
increase, isolating the deep waters from exchanges with the atmosphere, making it more likely that
metabolic activity will reduce dissolved oxygen concentrations in the hypolimnion to stressful or lethal
levels (De Stasio et al. 1996). The combination of warmer water temperatures and lower dissolved
oxygen concentrations under climate change scenarios in larger, deeper lakes typically suitable for
coldwater fishes could result in their extirpation in some Wisconsin lakes (Sharma et al. 2011);
however, this outcome is not anticipated in the Trout Lake watershed (J.Lyons, WDNR Science
Services, pers. comm.). As for cool and warm water species, it is expected that percids and esocids will
decline and centrarchids will increase as a general pattern in Wisconsin. Populations of small fishes will
also change, with most minnow species persisting, but some darters (Etheostoma sp.) and sculpins
143
(Cottus sp.) declining (J. Lyons. pers. comm.). Changes will likely be most pronounced in smaller
shallow lakes, where winter fish kills may also occur.
We also acknowledge that common loons may have thermal constraints under which they are
unable to incubate or maintain core body temperatures normally under hot weather/warm water
conditions. However, an evaluation of loon thermal dynamics was beyond the scope of this study.
Of the factors associated with loon habitat selection in Wisconsin, we hypothesized that lake
clarity and nest habitat quality are most likely to be influenced by a warming climate. An increase in
large intensity rainfall events (predicted for Wisconsin by the WICCI Climate Change Working Group,
see Chapter 1) may result in increased sediment and nutrient loads to lakes from increased runoff from
surrounding land and/or streams. An increase in nutrient and sediment load to a lake will likely
contribute to an increase in trophic status and, over time, may reduce water clarity of the lake (e.g.,
Robertson et al. 2002). Regional differences across the state will depend on land use, soils, and
geology. Drainage lakes and impoundments may experience more impacts than seepage lakes which
are governed by groundwater. Nest habitat quality is most likely to be compromised by fluctuating
water levels, specifically flooding events and/or drought leading to low lake levels levels (Evers et al.
2010).
The results of hydrology and lake modeling in the Trout Lake basin in northern Wisconsin
indicates that these lakes have the potential to become clearer under future climate conditions (Chapter
2). This is due to the fact that dissolved organic carbon (DOC), not total phosphorus (TP), is the
parameter most closely associated with lake clarity in the watershed that is likely to change under future
climate scenarios, as DOC levels are predicted to be lower. There is a low probability of increased TP
loading in this watershed because the soils have low nutrient content, the area is mostly forested, and the
majority is in public ownership (Northern Highlands State Forest), thus unlikely to convert to
144
agriculture or be developed for housing. Such a relation involving reduction in DOC and concomitant
increase in clarity is consistent with the pattern observed in a long-term data set from oligotrphic lakes
monitored in Killarney Park, Ontario (Gunn et al. 2001). Output from the hydrology modeling indicates
that lake levels (e.g. stage) are predicted to be lower under future climate conditions that are 1) warmer
and drier and 2) much warmer and drier scenarios; however, lake stage will remain mostly unchanged
under warmer and wetter conditions. Seepage lakes will be most impacted, with some predicted to have
lake stage drop over one meter during the modeled period. Loons often nest on associated wetlands on
seepage lakes, and it is very likely this nest habitat will be lost under warmer and drier conditions,
diminishing nest habitat quality.
In 2009, 11 Trout Lake basin lakes (out of 27) were occupied by territorial common loons.
Counter to our original hypothesis, under 17 climate change scenarios and 3 total phosphorus starting
conditions for lakes in the Trout Lake basin, we project none of these lakes are likely to change their
occupancy status due to changes in water clarity between the present and 2090. For these same lakes
we also predicted loon occupancy following a step change up or down in habitat quality. The effects of
such habitat changes are much greater, with only 6 of the lakes predicted to be occupied following a
decrease in habitat quality, but 22 of the lakes are predicted to be occupied following an increase in
habitat quality.
These results point to the critical need to conserve and enhance common loon nest habitat within
the Trout Lake basin, and throughout the current breeding range of loons in Wisconsin. Adaptation
strategies to reduce potential negative consequences of a changing climate should include preserving
existing critical nest habitat by managing shoreline development and habitat loss. Artificial nest
platforms have been shown to be readily accepted by breeding loons and have been shown to markedly
increase loon nest success in areas prone to fluctuating water levels (Piper et al. 2002). It is likely that a
145
combination of nest site conservation, augmented with appropriate placement of artificial nest
platforms, will be a powerful tool with which to conserve breeding common loons in Wisconsin.
Shoreland management should also focus on reducing nutrient inputs to lakes to maintain high
water clarity. Though changes in total P concentrations and future climate conditions will not likely
reduce water clarity thus loon habitat suitability within the Trout Lake basin, this may not be the case in
other Wisconsin watersheds. Where lakes are already more productive, and secchi depth is influenced
by total P (via algae growth) not dissolved organic carbon, future climate conditions have the potential
to have a much greater impact on future water clarity. However even in these watersheds, protection
and enhancement of nest habitat will have a larger influence on the probability of future loon occupancy
of lakes.
Acknowledgements
We appreciate the dedicated efforts of several field staff who collected the data necessary to
construct the Wisconsin loon habitat model, including L. Clausen, D. Killian, A. East, B. Fevold, J.
Borgen, D. Stockwell (all with WDNR Science Services), Josh Smith (Northland College), and Pete
Boma, Steve Houdek, Patrick Kelly, and Luke Fara (all with USGS UMESC La Crosse). Funding was
through Pitman Robertson Project W-160-P (WDNR), a grant from the Wisconsin Focus on Energy
Environmental Research Program, and USGS base funds.
146
References
Agresti, A. 2007. An introduction to categorical data analysis. 2nd
ed. Wiley.
Alvo, R., Hussell, D. J. T. and Berrill, M. 1988. The breeding success of common loons (Gavia immer)
in relation to alkalinity and other lake characteristics in Ontario. Can. J. Zool. 66:746–752.Anderson,
J.R., E.E. Hardy, J.T. Roach, and R.E. Witmer. 1976. A land use and land cover classification
system for use with remote sensor data. Geological Survey Professional Paper 964. United States
Government Printing Office, Washington.
Barr, J. F. 1996. Aspects of common loon (Gavia immer) feeding biology on its breeding ground.
Hydrobiologia 321: 119–144.Berk, R.A. 2010. Statistical learning from a regression perspective.
Springer.
Boyce, M.S., P.R. Vernier, S.E. Nielsen, and F.K.A. Schmiegelow. 2002. Evaluating resource
selection functions. Ecological Modelling 157:281-300
Burnham, K.P., and D.R. Anderson. 2002. Model selection and multimodel inference: a practical
information-theoretic approach. 2nd
ed. Springer.
Cutler, D.R., T.C. Edwards, Jr., K.H. Beard, A. Cutler, K.T. Hess, J. Gibson, and J.J. Lawler. 2007.
Random forests for classification in ecology. Ecology 88:2783-2792.
De Stasio, B.T., Hill, D.K., Kleinhans, J.M., Nibbelink, N.P., Magnuson, J.J. 1996. Potential effects of
global climate change on small north-temperate lakes: Physics, fish, and plankton. Limnol. Oceanogr.,
41: 1136-1149
Earnst, S.L., R. Platte, and L. Bond. 2006. A landscape-scale model of yellow-billed loon (Gavia
adamsii) habitat preferences in northern Alaska. Hydrobiologia 567:227-236.
147
Evers, D.C., J.D. Paruk, J.W. Mcintyre, and J.F. Barr. 2010. Common Loon (Gavia immer), The Birds
of North America Online (A. Poole, Ed.). Ithaca: Cornell Lab of Ornithology; Retrieved from the
Birds of North America Online: http://bna.birds.cornell.edu/bna/species/313 doi:10.2173/bna.313
Fielding, A.H., and J. F. Bell. 1997. A review of methods for the assessment of prediction errors in
conservation presence/absence models. Environmental Conservation 24:38-49.
Found, C., S.M. Webb, and M.S. Boyce. 2008. Selection of lake habitats by waterbirds in the boreal
transition zone of northeastern Alberta. Canadian Journal of Zoology. 86:277-285.
Fox, J. 2008. Applied regression analysis and generalized linear models. 2nd
ed. Sage Publications.
Fry, J., Xian, G., Jin, S., Dewitz, J., Homer, C., Yang, L., Barnes, C., Herold, N., and Wickham, J.,
2011. Completion of the 2006 National Land Cover Database for the Conterminous United States,
PE&RS, Vol. 77(9):858-864.
Guisan, A., T.C. Edwards, Jr., and T. Hastie. 2002. Generalized linear and generalized additive models
in studies of species distributions: setting the scene. Ecological Modelling 157:89-100.
Gunn, J.M, E. Snucins, N.D. Yan, M.T. Arts. 2001. Use of water clarity to monitor the effects of
climate change and other stressors on oligotrophic lakes. Environmental Monitoring and Assessment
67:69-88.
Heglund, P.J., J.R. Jones, L.H. Fredrickson and M.S. Kaiser. 1994. Use of boreal forested wetlands by
Pacific loons (Gavia pacific Lawrence) and horned grebes (Podiceps auritus L.): relations with
limnological characteristics. Hydrobiologia 279/280:171-183.
Hirzel, A. and A. Guisan. 2006. Which is the optimal sampling strategy for habitat suitability
modeling. Ecological Modelling. 157:331-341.
148
Johnson, C.J., S.E. Nielsen, E.H. Merrill, T.L. McDonald, and M.S. Boyce. 2006. Resource selection
functions based on use-availability data: Theoretical motivation and evaluation methods. Journal of
Wildlife Management. 70:347-357.
Kuhn, A., J. Copeland, J. Cooley, H Vogel, K. Taylor, D. Nacci, and P. August. 2011. Modeling
habitat associations for the Common Loon (Gavia Immer) at multiple scales in northeastern North
America. Avian Conservation and Ecology 6(1): 4. [online] URL: http://www.ace-
eco.org/vol6/iss/art4/
Maindonald, J. and W.J. Braun. 2007. Data analysis and graphics using R: An example-based
approach. Cambridge University Press.
Manel, S., H.C. Williams, and S.J. Ormerod. 2001. Evaluating presence-absence models in ecology:
the need to account for prevalence. 2001. Journal of Applied Ecology 38:921-931.
Manly, B.F.J., L.L. McDonald, D.L. Thomas, T.L. McDonald, and W.P. Erickson. 2002. Resource
selection by animals: statistical design and analysis for field studies. Kluwer Academic Publishers.
Meyer, M.W. 2006 Evaluating the impact of multiple stressors on Common Loon population
demographics – an integrated laboratory and field approach. EPA STAR co-operative Agreement
Number: R82-9085. Environmental Protection Agency, Washington, D.C., USA.
Omernik, J.M., S.S. Chapman, R.A. Lillie, and R.T. Dumke. 2000. Ecoregions of Wisconsin.
Transactions of the Wisconsin Academy of Sciences, Arts, and Letters 88:77-103.
Piper, W.H., J.S. Grear, and M.W. Meyer, 2012. Juvenile survival in common loons Gavia immer:
effects of natal lake size and pH. Journal of Avian Biology 43:1–9.
Piper, W.H., M.W. Meyer, M. Klich, K.B. Tischler and A. Dolsen. 2002. Floating platforms increase
reproductive success of common loons. Biological Conservation 104:199-203.
149
Prasad, A.M., L.R. Iverson, and A. Liaw. 2006. Newer classification and regression tree techniques:
Bagging and random forests for ecological prediction. Ecosystems 9:181-199.
Robertson, D. M., G. L. Goddard, E. A. Mergerer, W. J. Rose, and P. J. Garrison. 2002. Hydrology and
water quality of Geneva Lake, Walworth County, Wisconsin. U.S. Geological Survey Water
Resources Investigations Report 02-4039, 73 pp.
Ryder, R.A. 1982. The morphoedaphic index--use, abuse, and fundamental concepts. Transactions of
the American Fisheries Society 111:154-164.
Sharma S., M.J. Vander Zanden, J.J. Magnuson, J. Lyons. 2011. Comparing Climate Change and
Species Invasions as Drivers of Coldwater Fish Population Extirpations. PLoS ONE 6(8):
e22906. doi:10.1371/journal.pone.0022906.Stevens, D.L. and A.R. Olsen. 2004. Spatially
balanced sampling of natural resources. Journal of the American Statistical Association.
99:262-278.
150
Page intentionally left blank.
151
Appendix – Figures Depicting Predicted Consequences of
Expected Future Climate Conditions on Loon Occurrence
in Wisconsin
Figures
Figure A–1. Predicted probability of loon occurrence for Frank, Starrett, Vandercook, Big Musky, Crystal and
Day Lakes under 17 projected climate scenarios. For each lake listed along the x-axis, the figure gives predictions
for 2010, 2050, and 2090. For each year, the open circle is the median of the 17 projections, and the vertical line
extends from the minimum to the maximum of the 17 projections. ........................................................................ 155
Figure A–2. Predicted probability of loon occurrence for Diamond, Edith, Fallison, Firefly, Jag and Little Rock
Lakes under 17 projected climate scenarios. For each lake listed along the x-axis, the figure gives predictions for
2010, 2050, and 2090. For each year, the open circle is the median of the 17 projections, and the vertical line
extends from the minimum to the maximum of the 17 projections. ........................................................................ 156
Figure A–3. Predicted probability of loon occurrence for Nebish, Nichols, Pallette, Rudolph, Sparkling and Street
Lakes under 17 projected climate scenarios. For each lake listed along the x-axis, the figure gives predictions for
2010, 2050, and 2090. For each year, the open circle is the median of the 17 projections, and the vertical line
extends from the minimum to the maximum of the 17 projections. ........................................................................ 157
Figure A–4. Predicted probability of loon occurrence for Unnamed, Wildwood, Trout, Mann, Little John and
Allequash Lakes under 17 projected climate scenarios. For each lake listed along the x-axis, the figure gives
152
predictions for 2010, 2050, and 2090. For each year, the open circle is the median of the 17 projections, and the
vertical line extends from the minimum to the maximum of the 17 projections. ..................................................... 158
Figure A–5. Predicted probability of loon occurrence for Lost Canoe, Escanaba and White Birch Lakes under 17
projected climate scenarios. For each lake listed along the x-axis, the figure gives predictions for 2010, 2050, and
2090. For each year, the open circle is the median of the 17 projections, and the vertical line extends from the
minimum to the maximum of the 17 projections. .................................................................................................... 159
Figure A–6. ..... Predicted probability of loon occurrence for Frank, Starret, Vandercook and Big Musky Lakes in the
Trout Lake basin in 2050 and 2090 under 17 projected climate scenarios. Plotting symbols are median predictions
for 0% change in total P (open circle), 10% increase (closed circle), and 25% increase (x). Vertical bars extend
from the minimum to the maximum of the projections for the 17 climate scenarios. For each year (2050 and 2090),
predictions are shown for a 1 step decrease in nesting habitat (leftmost), unchanged habitat (central), and a 1 step
increase in nesting habitat (rightmost). The horizontal dotted line is the median of predictions for each lake in 2010.
............................................................................................................................................................................... 160
Figure A–7. ... Predicted probability of loon occurrence for Crystal, Day, Diamond and Edith Lakes in the Trout Lake
basin in 2050 and 2090 under 17 projected climate scenarios. Plotting symbols are median predictions for 0%
change in total P (open circle), 10% increase (closed circle), and 25% increase (x). Vertical bars extend from the
minimum to the maximum of the projections for the 17 climate scenarios. For each year (2050 and 2090),
predictions are shown for a 1 step decrease in nesting habitat (leftmost), unchanged habitat (central), and a 1 step
increase in nesting habitat (rightmost). The horizontal dotted line is the median of predictions for each lake in 2010.
............................................................................................................................................................................... 161
Figure A–8. ....... Predicted probability of loon occurrence for Fallison, Firefly, Jag and Little Rock Lakes in the Trout
Lake basin in 2050 and 2090 under 17 projected climate scenarios. Plotting symbols are median predictions for
0% change in total P (open circle), 10% increase (closed circle), and 25% increase (x). Vertical bars extend from
the minimum to the maximum of the projections for the 17 climate scenarios. For each year (2050 and 2090),
153
predictions are shown for a 1 step decrease in nesting habitat (leftmost), unchanged habitat (central), and a 1 step
increase in nesting habitat (rightmost). The horizontal dotted line is the median of predictions for each lake in 2010.
............................................................................................................................................................................... 162
Figure A–9. ..... Predicted probability of loon occurrence for Nebish, Nichols, Pallett and Rudolph Lakes in the Trout
Lake basin in 2050 and 2090 under 17 projected climate scenarios. Plotting symbols are median predictions for
0% change in total P (open circle), 10% increase (closed circle), and 25% increase (x). Vertical bars extend from
the minimum to the maximum of the projections for the 17 climate scenarios. For each year (2050 and 2090),
predictions are shown for a 1 step decrease in nesting habitat (leftmost), unchanged habitat (central), and a 1 step
increase in nesting habitat (rightmost). The horizontal dotted line is the median of predictions for each lake in 2010.
............................................................................................................................................................................... 163
Figure A–10. ... Predicted probability of loon occurrence for Sparkling, Street, Unnamed and Wildwood Lakes in the
Trout Lake basin in 2050 and 2090 under 17 projected climate scenarios. Plotting symbols are median predictions
for 0% change in total P (open circle), 10% increase (closed circle), and 25% increase (x). Vertical bars extend
from the minimum to the maximum of the projections for the 17 climate scenarios. For each year (2050 and 2090),
predictions are shown for a 1 step decrease in nesting habitat (leftmost), unchanged habitat (central), and a 1 step
increase in nesting habitat (rightmost). The horizontal dotted line is the median of predictions for each lake in 2010.
............................................................................................................................................................................... 164
Figure A–11. Predicted probability of loon occurrence for Trout, Mann, Little John and Allequash Lakes in the Trout
Lake basin in 2050 and 2090 under 17 projected climate scenarios. Plotting symbols are median predictions for
0% change in total P (open circle), 10% increase (closed circle), and 25% increase (x). Vertical bars extend from
the minimum to the maximum of the projections for the 17 climate scenarios. For each year (2050 and 2090),
predictions are shown for a 1 step decrease in nesting habitat (leftmost), unchanged habitat (central), and a 1 step
increase in nesting habitat (rightmost). The horizontal dotted line is the median of predictions for each lake in 2010.
............................................................................................................................................................................... 165
154
Figure A–12.Predicted probability of loon occurrence for Lost Canoe, Escanaba and White Birch Lakes in the Trout
Lake basin in 2050 and 2090 under 17 projected climate scenarios. Plotting symbols are median predictions for
0% change in total P (open circle), 10% increase (closed circle), and 25% increase (x). Vertical bars extend from
the minimum to the maximum of the projections for the 17 climate scenarios. For each year (2050 and 2090),
predictions are shown for a 1 step decrease in nesting habitat (leftmost), unchanged habitat (central), and a 1 step
increase in nesting habitat (rightmost). The horizontal dotted line is the median of predictions for each lake in 2010.
............................................................................................................................................................................... 166
155
Figure A–1. Predicted probability of loon occurrence for Frank, Starrett, Vandercook, Big Musky, Crystal and Day
Lakes under 17 projected climate scenarios. For each lake listed along the x-axis, the figure gives predictions
for 2010, 2050, and 2090. For each year, the open circle is the median of the 17 projections, and the vertical
line extends from the minimum to the maximum of the 17 projections.
Lake and year
Pre
dic
ted
pro
ba
bility o
f o
ccu
rre
nce
2010 2090 2010 2090 2010 2090 2010 2090 2010 2090 2010 2090
FrankLk StarretLk VandercookL BigMuskyLk CrystalLk DayLk
0.2
0.4
0.6
0.8
156
Figure A–2. Predicted probability of loon occurrence for Diamond, Edith, Fallison, Firefly, Jag and Little Rock Lakes
under 17 projected climate scenarios. For each lake listed along the x-axis, the figure gives predictions for
2010, 2050, and 2090. For each year, the open circle is the median of the 17 projections, and the vertical line
extends from the minimum to the maximum of the 17 projections.
Lake and year
Pre
dic
ted
pro
ba
bility o
f o
ccu
rre
nce
2010 2090 2010 2090 2010 2090 2010 2090 2010 2090 2010 2090
DiamondLk EdithLk FallisonLk FireflyLk JagLk LittleRockL
0.2
0.4
0.6
0.8
157
Figure A–3. Predicted probability of loon occurrence for Nebish, Nichols, Pallette, Rudolph, Sparkling and Street
Lakes under 17 projected climate scenarios. For each lake listed along the x-axis, the figure gives predictions
for 2010, 2050, and 2090. For each year, the open circle is the median of the 17 projections, and the vertical
line extends from the minimum to the maximum of the 17 projections.
Lake and year
Pre
dic
ted
pro
ba
bility o
f o
ccu
rre
nce
2010 2090 2010 2090 2010 2090 2010 2090 2010 2090 2010 2090
NebishLk NicholsLk PalletteLk RudolphLk SparklingLk StreetLk
0.2
0.4
0.6
0.8
158
Figure A–4. Predicted probability of loon occurrence for Unnamed, Wildwood, Trout, Mann, Little John and
Allequash Lakes under 17 projected climate scenarios. For each lake listed along the x-axis, the figure gives
predictions for 2010, 2050, and 2090. For each year, the open circle is the median of the 17 projections, and
the vertical line extends from the minimum to the maximum of the 17 projections.
Lake and year
Pre
dic
ted
pro
ba
bility o
f o
ccu
rre
nce
2010 2090 2010 2090 2010 2090 2010 2090 2010 2090 2010 2090
UnnamedLk WildwoodLk TroutLk MannLk LittleJohnL AllequashLk
0.2
0.4
0.6
0.8
159
Figure A–5. Predicted probability of loon occurrence for Lost Canoe, Escanaba and White Birch Lakes under 17
projected climate scenarios. For each lake listed along the x-axis, the figure gives predictions for 2010, 2050,
and 2090. For each year, the open circle is the median of the 17 projections, and the vertical line extends from
the minimum to the maximum of the 17 projections.
Lake and year
Pre
dic
ted
pro
ba
bility o
f o
ccu
rre
nce
2010 2050 2090 2010 2050 2090 2010 2050 2090
LostCanoeLk EscanabaLk WhiteBirchL
0.2
0.4
0.6
0.8
160
Figure A–6. Predicted probability of loon occurrence for Frank, Starret, Vandercook and Big Musky Lakes in the
Trout Lake basin in 2050 and 2090 under 17 projected climate scenarios. Plotting symbols are median
predictions for 0% change in total P (open circle), 10% increase (closed circle), and 25% increase (x). Vertical
bars extend from the minimum to the maximum of the projections for the 17 climate scenarios. For each year
(2050 and 2090), predictions are shown for a 1 step decrease in nesting habitat (leftmost), unchanged habitat
(central), and a 1 step increase in nesting habitat (rightmost). The horizontal dotted line is the median of
predictions for each lake in 2010.
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
FrankLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
StarretLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
VandercookL
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
BigMuskyLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
161
Figure A–7. Predicted probability of loon occurrence for Crystal, Day, Diamond and Edith Lakes in the Trout Lake
basin in 2050 and 2090 under 17 projected climate scenarios. Plotting symbols are median predictions for 0%
change in total P (open circle), 10% increase (closed circle), and 25% increase (x). Vertical bars extend from
the minimum to the maximum of the projections for the 17 climate scenarios. For each year (2050 and 2090),
predictions are shown for a 1 step decrease in nesting habitat (leftmost), unchanged habitat (central), and a 1
step increase in nesting habitat (rightmost). The horizontal dotted line is the median of predictions for each
lake in 2010.
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
CrystalLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
DayLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
DiamondLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
EdithLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
162
Figure A–8. Predicted probability of loon occurrence for Fallison, Firefly, Jag and Little Rock Lakes in the Trout Lake
basin in 2050 and 2090 under 17 projected climate scenarios. Plotting symbols are median predictions for 0%
change in total P (open circle), 10% increase (closed circle), and 25% increase (x). Vertical bars extend from
the minimum to the maximum of the projections for the 17 climate scenarios. For each year (2050 and 2090),
predictions are shown for a 1 step decrease in nesting habitat (leftmost), unchanged habitat (central), and a 1
step increase in nesting habitat (rightmost). The horizontal dotted line is the median of predictions for each
lake in 2010.
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
FallisonLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
FireflyLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
JagLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
LittleRockL
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
163
Figure A–9. Predicted probability of loon occurrence for Nebish, Nichols, Pallett and Rudolph Lakes in the Trout
Lake basin in 2050 and 2090 under 17 projected climate scenarios. Plotting symbols are median predictions
for 0% change in total P (open circle), 10% increase (closed circle), and 25% increase (x). Vertical bars extend
from the minimum to the maximum of the projections for the 17 climate scenarios. For each year (2050 and
2090), predictions are shown for a 1 step decrease in nesting habitat (leftmost), unchanged habitat (central),
and a 1 step increase in nesting habitat (rightmost). The horizontal dotted line is the median of predictions for
each lake in 2010.
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
NebishLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
NicholsLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
PalletteLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
RudolphLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
164
Figure A–10. Predicted probability of loon occurrence for Sparkling, Street, Unnamed and Wildwood Lakes in the
Trout Lake basin in 2050 and 2090 under 17 projected climate scenarios. Plotting symbols are median
predictions for 0% change in total P (open circle), 10% increase (closed circle), and 25% increase (x). Vertical
bars extend from the minimum to the maximum of the projections for the 17 climate scenarios. For each year
(2050 and 2090), predictions are shown for a 1 step decrease in nesting habitat (leftmost), unchanged habitat
(central), and a 1 step increase in nesting habitat (rightmost). The horizontal dotted line is the median of
predictions for each lake in 2010.
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
SparklingLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
StreetLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
UnnamedLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
WildwoodLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
165
Figure A–11. Predicted probability of loon occurrence for Trout, Mann, Little John and Allequash Lakes in the Trout
Lake basin in 2050 and 2090 under 17 projected climate scenarios. Plotting symbols are median predictions
for 0% change in total P (open circle), 10% increase (closed circle), and 25% increase (x). Vertical bars extend
from the minimum to the maximum of the projections for the 17 climate scenarios. For each year (2050 and
2090), predictions are shown for a 1 step decrease in nesting habitat (leftmost), unchanged habitat (central),
and a 1 step increase in nesting habitat (rightmost). The horizontal dotted line is the median of predictions for
each lake in 2010.
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
TroutLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
MannLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
LittleJohnL
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
AllequashLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
166
Figure A–12. Predicted probability of loon occurrence for Lost Canoe, Escanaba and White Birch Lakes in the Trout
Lake basin in 2050 and 2090 under 17 projected climate scenarios. Plotting symbols are median predictions
for 0% change in total P (open circle), 10% increase (closed circle), and 25% increase (x). Vertical bars extend
from the minimum to the maximum of the projections for the 17 climate scenarios. For each year (2050 and
2090), predictions are shown for a 1 step decrease in nesting habitat (leftmost), unchanged habitat (central),
and a 1 step increase in nesting habitat (rightmost). The horizontal dotted line is the median of predictions for
each lake in 2010.
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
LostCanoeLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
EscanabaLk
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8
Year - habitat change
Pre
dic
ted p
robability o
f occurr
ence
WhiteBirchL
worse same better worse same better
2050 2090
0.2
0.4
0.6
0.8