GREAT LAKES FISHERY COMMISSION

2011 Project Completion Report1

A Decision Analysis for Multispecies Harvest Management of Lake Huron Commercial

Fisheries

by:

Michael L. Jones

2 and Brian J. Langseth

2

2 Quantitative Fisheries Center Department of Fisheries & Wildlife Michigan State University East Lansing, MI, 48824-1222

January 2012

1 Project completion reports of Commission-sponsored research are made available to the Commission’s

Cooperators in the interest of rapid dissemination of information that may be useful in Great Lakes fishery

management, research, or administration. The reader should be aware that project completion reports have not

been through a peer-review process and that sponsorship of the project by the Commission does not

necessarily imply that the findings or conclusions are endorsed by the Commission. Do not cite findings

without permission of the author.

2

ABSTRACT:

Tradeoffs between achieving desired yield objectives for lake whitefish (Coregonus clupeaformis) and restoration

objectives for lake trout (Salvelinus namaycush) were assessed for harvest policies affecting coldwater

commercial fisheries in Lake Huron. Lake whitefish are targeted in the majority of commercial fisheries operating

in Lake Huron, but these fisheries also capture lake trout as bycatch. Lake trout were nearly extirpated from Lake

Huron by the 1950s, and substantial stocking efforts have been underway for decades to aid in recovery. Ongoing

or expanded harvest of lake whitefish may negatively affect rehabilitation efforts for lake trout. Additionally,

Lake Huron has undergone substantial changes to its food web in the last two decades. Dreissenid mussels and

round gobies (Neogobius melanostomus) have invaded and thrived in Lake Huron. Abundance of several prey

fish species has declined as has abundance of Diporeia, a primary food source for lake whitefish. These

ecosystem changes may also affect tradeoffs for coldwater commercial fishing policy in Lake Huron. To assess

these tradeoffs, we developed a food-web model (Ecopath with Ecosim: EwE) for the coldwater community in the

main basin of Lake Huron and used this model to compare harvest policies and evaluate the importance of key

system uncertainties to policy rankings. We engaged Lake Huron fishery stakeholders in two workshops to help

guide model development. Obtaining a balanced EwE model, and appropriately including invasive species in the

dynamic simulations both proved difficult, and prompted additional simulation studies. We found that in general,

dynamic simulations in Ecosim are not highly sensitive to ad hoc balancing adjustments, but that sensitivity

increases as the strength of trophic interaction among groups increases. We compared four methods for

incorporating invasive species into the EwE model and concluded that initializing invasive species biomasses

before actual invasion at very low biomasses, and maintaining them at low levels by imposing an ad hoc mortality

until the time of invasion was reasonably good at reproducing observed time series of all groups. The completed

EwE model was used to simulate changes to fishing mortality targets, to the season in which fishing occurred, and

to the type of gear used. Conversions of gill nets to trap nets resulted in the maintenance of lake whitefish harvest

and 15% increases in lake trout biomass over the status quo policy. Changes in fishing seasons varied among

policies, but resulted in at most a 14% increase in lake trout biomass, and a 39% increase in lake whitefish

harvest. Changing fishing mortality targets revealed the expected tradeoffs between lake whitefish harvest and

lake trout biomass. In general, changes in harvest were greater than changes in biomass as fishing mortality

targets changed, suggesting increases in harvest could be achieved without large decreases in biomass, but raising

questions about the model’s representation of compensatory processes for both species. Our assessment of the

significance of uncertainties about future environmental productivity, diet, and strength of trophic interactions

revealed that the first of these had the greatest effect on model outcomes, but did not alter the relative

performance of the policies. The other two uncertainties had lesser effects than changes in productivity, and

influenced lake whitefish harvest and biomass much more than that of lake trout. We found little evidence for

substantial indirect interactions between lake trout and lake whitefish, leading us to conclude that the commercial

fishery is the primary factor that links these two groups. Future work on balancing tradeoffs in the commercial

fisheries should therefore focus on direct interactions with the fishery (i.e. bycatch reduction), rather than on

indirect interactions through the food web.

3

INTRODUCTION:

This report describes work completed on a project to evaluate harvest options for multispecies commercial

fisheries in Lake Huron, using a food-web modeling approach. Our work focused on interactions between lake

whitefish (Coregonus clupeaformis) and lake trout (Salvelinus namaycush) exploitation. Lake whitefish are the

primary exploited coldwater species in Lake Huron, and lake trout are commonly harvested as bycatch in the lake

whitefish fishery. Lake trout are also the object of a native-species restoration program in Lake Huron, so impacts

of the lake whitefish fishery on lake trout recovery, either directly through bycatch, or indirectly through food-

web interactions, are a significant management concern. Our research sought to identify whether certain harvest

policies would be preferable to others in simultaneously meeting the objectives of commercial fishers and of other

stakeholders who view lake trout restoration as a top priority.

Consideration of direct and indirect interactions among exploited species is a hallmark of multispecies and

ecosystem-based management (EPAP 1999; Pikitch et al. 2004). Ecosystem-based approaches have received

increased attention within the past decades and often build upon the techniques of single-species management. A

variety of quantitative approaches and modeling tools have been developed (and continue to be developed) for

considerations of such multispecies and ecosystem-based management (Jackson et al. 2000; Smith et al. 2007).

Within the Great Lakes region an ecosystem approach to management is central to the guiding principles of the

FCOs (DesJardine et al. 1995), the fundamental concept of the Strategic Vision of the Great Lake Fishery

Commission for the First Decade of the New Millenium (GLFC 2001).

A number of studies have looked at interactions among multiple species in Great Lakes fish communities

(Fontaine and Stewart 1992; Jones et al. 1993; Jones et al. 1995; Hebert and Sprules 2002), including implications

for management. Fewer studies have attempted to understand Great Lakes ecosystems as a whole (but see Flint

1986; Halfon et al. 1996; Kitchell et al. 2000; Jaeger 2006), and these studies often place less emphasis on

management options (with the exception of Kitchell et al. 2000). Elsewhere, harvest strategies incorporating

interactions among multiple species and the fisheries that target them have also been discussed, including issues

of bycatch (De Oliviera et al. 1998; Pascoe 2000), competing management objectives (Cochrane et al. 1998;

Matsuda and Abrams 2006) and trophic interactions (Arreguin-Sanchez et al. 2004; Zetina-Rejon et al. 2004; Eby

et al. 2006).

Viewing fisheries management at an ecosystem level may be particularly important given the substantial changes

that have occurred recently in Lake Huron (Ebener 2005). One of the most noticeable recent changes to Lake

Huron’s fish community has been the collapse of the alewife (Alosa psuedoharengus) population (Riley et al.

2008). The potential for native species to fill the void left by alewives has yet to be realized in Lake Huron,

however recent pulses of lake herring (Coregonus artedi; Schaeffer and Warner 2008) and bloater (Coregonus

hoyi; Roseman and Riley 2009) are promising. In the meantime, the alewife collapse has raised serious concerns

about the ability of the forage base in Lake Huron to support predators, possibly including lake trout, at desired

levels. Consequently, efforts to rehabilitate lake trout may necessitate reducing other forms of mortality on the

prey fish community, including fishing. In addition, increased densities of Dreissenid mussels (Nalepa et al. 2007)

raise concerns that Lake Huron’s energy pathways have changed, leading to uncertain consequences for growth

and production of harvested fish populations, especially lake whitefish (Nalepa et al. 2009). Dreissenids have

been implicated in the declines of the macroinvertebrate, Diporeia (Nalepa et al. 2007). Diporeia are a primary

food source for lake whitefish, and thus declines in Diporeia density could result in changes to diet content and

production of lake whitefish (Pothoven and Nalepa 2006). This context provides a strong argument in favor of

using a food-web approach to consider the performance of harvest strategies applied to the multiple fisheries

operating in Lake Huron.

OBJECTIVES:

Our project was guided by five objectives:

4

1. Determine an appropriate set of management objectives and harvest strategy options for Lake Huron’s

coldwater commercial fishery and identify the key uncertainties about this system that limit our ability to

make confident predictions. These are essential components of any decision analysis;

2. Develop a food-web model for the Lake Huron coldwater fish community that reflects our understanding of

trophic interactions within this community as well as key environmental drivers of these interactions;

3. Incorporate a multispecies commercial fishery into the food-web model to simulate the effects of alternative

harvest strategies that result from completion of objective 1;

4. Critically assess the consequences of uncertainties identified in objective 1 on expected harvest policy

performance by using alternative parameterizations of the food-web model;

5. Provide stakeholders with an evaluation of the relative performance of a range of harvest strategy options

measured in terms of their success at achieving the management objectives identified through completion of

project objective 1.

As detailed below, we have completed four of these five objectives. Completion of objective 2 (food-web model)

proved very challenging, and led to us addressing two other unanticipated sub-objectives:

2a. Evaluate the sensitivity of Ecosim projections to the assumptions made in developing a balanced Ecopath

model; and

2b. Assess alternative strategies for incorporating invasive species into an Ecopath with Ecosim model when

these species are not present in the ecosystem at the initial time step of the model.

The credibility of our food-web model for Lake Huron depended on us addressing these two sub-objectives

before moving on to objectives 3 and 4.

The fifth objective will be completed in part by distribution of this completion report to the stakeholders we

engaged earlier in the project. However, we also intend to convene a final workshop during spring 2012, where

we will share our findings with stakeholders, and discuss implications for management.

METHODS:

Objective 1. We convened two workshops with Lake Huron fishery stakeholders to discuss management

objectives, management options, and food-web model development issues. The first workshop was held on April

20-21, 2009 and was attended by 15 stakeholder participants. The second workshop was held on April 22, 2010,

and was attended by 13 stakeholder participants. The workshops followed a Structured Decision Making (SDM)

format (Irwin et al. 2011). At the first workshop we focused on introducing the participants to the food-web

modeling approach we proposed to undertake, and developed an initial list of management objectives and options.

At the second workshop we presented preliminary model results, and refined our list of management objectives

and options. We also discussed key areas of uncertainty to consider in our decision analysis. Lists of participants

at both meetings are provided in Appendix 1.

Objective 2. We developed a food-web model for the main basin of Lake Huron using the Ecopath with Ecosim

(EwE) modeling platform (Christensen and Walters 2004). Graduate student Brian Langseth visited the

University of British Columbia for several weeks in 2009 and 2010 to collaborate with EwE experts (Carl Walters

and Villy Christensen) on model development. We used an extensive literature search and consultation with Great

Lakes food-web experts on all trophic levels to obtain estimates of biomass, production, and diet for all trophic

groups included in the model, as well as time series estimates of biomass for model fitting. The model

development methods are described in detail in Appendices 2 and 3 which also describe the methods for sub-

objectives 2a and 2b, respectively.

Objective 3. We included five commercial fisheries in the food web model: US treaty water lake whitefish (trap

and gill nets); Canadian (non-treaty) lake whitefish gill net fishery; Canadian (non-treaty) lake whitefish trap net

fishery; US treaty water Chinook salmon (Oncorhynchus tshawytscha) fishery; and Canadian bloater fishery. We

focused our policy analysis on lake whitefish fisheries and lake trout bycatch in these fisheries. We evaluated

three types of policies: varying levels of a fixed fishing mortality control rule; conversion of gill net fisheries to

5

trap nets; and restrictions on seasons during which fishing occurred. The gear conversion and seasonal fishing

policies included extreme scenarios that were unrealistic but allowed examination of whether these policies have

much impact on the achievement of lake trout restoration objectives. The implementation of these policies is

described in detail in Appendix 4.

Objective 4. We focused on three key areas of uncertainty: future ecosystem productivity; strengths of trophic

interactions; and diet contributions for relatively rare prey. The EwE modeling strategy did not allow a statistical

assessment of these uncertainties, so we used an ad hoc approach to evaluate how sensitive our policy results

were to each of these uncertainties. We used retrospective model fitting results, which included estimates of

“production anomalies”, to bound possible future ecosystem productivities. Specifically, we projected into the

future with three different productivity levels, representing the first quartile, the median, and the third quartile of

the range of production anomalies estimated for the period 1981-2008. We represented uncertainty concerning the

strength of trophic interactions by deliberately modifying values of vulnerability (parameters that describe the

strength of trophic interaction). We changed vulnerability values for the oldest age groups of lake whitefish, lake

trout, and Chinook salmon to values that differed as much as possible from those estimated in the retrospective

model fitting process without leading to substantially worse model fits (i.e., vulnerability values that were

plausible, given the data). Finally, we evaluated sensitivity to diet by including young lake whitefish as a small

(2%) component of older lake trout diets, and by including age 1+ alewife and age 1+ rainbow smelt as small (1%

each) components of older lake whitefish diets. Our choice of which modifications to make to our “best guess”

diets was based on discussions with stakeholders at our second workshop. When we modified the vulnerabilities

or the diets to explore sensitivity, we re-fit the remaining model parameters to the historical data before projecting

forward. Further details are provided in Appendix 4.

Objective 5. As noted earlier, this objective will be partially addressed by distribution of this report, in addition to

a third stakeholder workshop planned for spring 2012.

RESULTS:

Objective 1. Management objectives and harvest strategy options were identified during the first stakeholder

workshop and refined during the second workshop (Table 1). Details of these objectives and options are in

Appendix 1.

Table 1: Refined management objectives and harvest strategy options from the two stakeholder workshops. Order

in table does not suggest order of preference.

Management objectives Harvest strategy options

Minimize operating costs of fishing Constant catch harvest control rule

High catch rates for fish for sports fishermen Constant fishing mortality (F) harvest control rule

Diversity of the fishery Seasonal adjustments incorporating both control

rules

Minimum bycatch induced restrictions on harvest of

target species

Gear conversions from gill net to trap incorporating

both control rules; constant catch and constant F

Maintain lake whitefish yield objective from FCOs (3.8

million kg yield of lake whitefish, bloater, and herring)

Market development to improve demand and

therefore price

Maximize landed value of catch incorporating species

mix

Policies based on desired species ratios in catches,

biomass, trends, or reference points

Stability of yield, but not at the expense of foregone

harvestable surplus

Maximize harvest while maintaining lake trout

biomass above a threshold.

Lake trout annual mortality < 40%

Key uncertainties in the system were also discussed and refined during the two stakeholder workshops (Table 2).

A subset of the options and uncertainties were then implemented in the modeling work and are discussed further

under objective 4.

6

Table 2: Key uncertainties in the understanding of food-web dynamics in Lake Huron.

Key uncertainties Extent that fouling by filamentous algae limits lake whitefish harvest

Diet proportions of rare items in lake whitefish and lake trout

Diets of difficult to monitor groups, e.g. larval fish

Lamprey predation trends based on biomass estimates or marking rates

Indirect effects of alewife on predators, predatory zooplankton on zooplankton,

and Dreissenids on predator-prey interactions

Predatory impact of walleye

Lake whitefish dynamics in US, non-treaty waters

Production of age 0 lake whitefish

Stochasticity in recruitment

Future levels of environmental productivity

Effects of balancing on model results

Effects of accounting for invasive species on model results

Effects of vulnerabilities on model results

Reasons for declines in key trophic groups (alewife, Diporeia, other prey fish)

Extent of predation mortality by round gobies

Response of recreational fishery to future biomass levels

Are FCO targets achievable given new food-web?

Objective 2a (see Appendix 2 for supporting details). Developing a balanced food-web model proved more

challenging than expected given the wide availability of data for Great Lakes aquatic organisms. The initial model

was unbalanced, and in general consumption by predators exceeded the available production of their prey. A

balanced model was required before policy simulations could be run and balancing required ad hoc changes to

model inputs (i.e., assumptions about initial biomass and consumption). It was uncertain how great an effect these

changes would have on future biomass dynamics. To assess the effect of balancing on future biomass dynamics,

the model was balanced in two different ways, simple policy simulations were run for both cases, and the

resulting biomass dynamics were compared. The first balancing approach preferentially adjusted consumption of

predators downward, whereas the second preferentially adjusted production of prey upward. Alternative values of

vulnerabilities were also used in each balanced model to explore the effect of vulnerabilities on future biomass

dynamics. We found that changes in model input and changes in vulnerabilities both affected future biomass

dynamics. The proportional difference in biomass dynamics between the two balanced models was 4% for default

values of vulnerabilities, but changed approximately four-fold when vulnerabilities were adjusted. The

proportional difference was 15% when vulnerabilities were increased and 1% when vulnerabilities were

decreased. To conclude, differences in model inputs affected biomass dynamics most when vulnerabilities of prey

to their predators were large, however the overall affect of these changes appeared small.

Objective 2b (see Appendix 3 for supporting details). An additional challenge encountered while developing the

food-web model was how to treat recent invasive species. Dreissena sp., predatory zooplankton (Bythotrephes

longimanus), and round goby (Neogobius melanostomus) invaded Lake Huron after 1981, the year in which the

food-web model was initialized, and would therefore have zero biomass at the time the model was initialized.

Ecopath with Ecosim models require modeled groups to have positive biomasses, which presented a challenge for

these groups. These three invasive species have considerably altered food-web dynamics in Lake Huron since

their invasion, so it was essential to incorporate their dynamics in the model. With the help of additional

collaborators, we developed four methods of incorporating invasive species into our EwE model. These methods

included 1) forcing invasive species biomass to observed levels, 2) starting invasive species biomass at very low

levels and allowing them to increase, 3) starting invasive species biomass at recent (high) levels and artificially

removing them until the time at which they invade, and 4) adjusting vulnerabilities of invasive species over time

to match biomass dynamics. The ability of each method to reproduce the observed dynamics of each invasive

species, as well as of non-invasive groups was assessed. All methods could reproduce “invasion”, i.e. an increase

in biomass, while maintaining reasonable fits to other groups. Method 2, however, did this somewhat better than

the other methods and with greater simplicity. We recommend using this method for including invasive species in

EwE models.

7

Objective 3 (see Appendix 4 for supporting details). As fishing mortality targets were incrementally increased

from the lowest rate (-75% of status quo) to the highest (+100% of status quo), biomass of lake whitefish and lake

trout declined by 54% and 28%, respectively, while harvest increased by 271% and 266%, respectively. Increases

in harvests were greater for both species than declines in biomasses, suggesting compensatory responses to

fishing. Biomass and harvest of lake whitefish remained unchanged in the gear conversion policy, while lake

trout biomass increased 15% above status quo levels for a total conversion of gill nets to trap nets. Partial

conversions of gill nets to trap nets resulted in smaller increases in biomass of lake trout. Fishing only in winter

resulted in the greatest changes to biomass and harvest of lake whitefish and lake trout. Biomass of lake trout

increased by at most 14% over the status quo policy when lake whitefish targets were maintained. Harvest of lake

whitefish increased by at most 39% over the status quo policy when lake whitefish targets were increased while

lake trout harvest was maintained at current levels. Harvest policy simulations therefore suggested that doubling

fishing mortality targets for lake whitefish best achieved lake whitefish yield objectives, whereas total conversion

of gill nets to trap nets best achieved lake trout biomass objectives. Multiple objectives exist for the policies

however, and although lake whitefish harvest was increased under changes to fishing mortality targets, lake trout

biomass declined. In contrast, gear conversion and seasonal adjustment policies balanced higher lake trout

biomass with unchanged or greater lake whitefish harvests. Consequently, although extreme, gear conversion and

seasonal adjustment policies show promise in meeting the competing objectives of high lake whitefish harvest

and high lake trout biomass.

Objective 4 (see Appendix 4 for supporting details). Uncertainties in future levels of environmental productivity

caused the largest changes in expected levels of harvest and biomass of all the uncertainties considered. Increased

productivity positively affected lake whitefish biomass to a greater extent than lake trout biomass. This is likely

due to lake whitefish feeding on a greater proportion of lower trophic organisms than do lake trout. Although the

magnitude of policy outcomes changed with environmental productivities, the relative performance of each policy

did not change. In addition, alewife biomass was predicted to recover when productivity was increased, whereas

under median or low productivity, alewife biomass remained very low. Changes in diet and trophic interaction

strengths resulted in less change to policy outcomes than did changes in productivity, and similarly affected lake

whitefish more than lake trout. Changes to diet and trophic interaction strengths resulted in greater sensitivity of

lake whitefish harvest and less sensitivity of lake whitefish biomass to changes in fishing mortality. Uncertainties

in diet also altered the extent of direct and indirect interactions between lake whitefish and lake trout. In general,

increases in the biomass of one species resulted in small decreases in biomass to the other. Changes in biomass of

lake trout minimally impacted biomass of lake whitefish; when diet of lake trout was adjusted (see Methods,

Objective 4), the interaction become slightly stronger. Changes in biomass of lake whitefish affected biomass of

lake trout more than vice versa, likely due to the greater absolute biomass of lake whitefish in the system.

Adjusting lake whitefish diets mitigated some of the effect on lake trout biomass, but overall the indirect

interactions between these two species were found to be weak.

DISCUSSION:

This project sought to examine harvest policy options for Lake Huron commercial fisheries by developing a

food-web model that would allow consideration of both direct and indirect effects on targeted and incidentally

harvested species. We focused our analysis on lake whitefish and lake trout, because the former is the primary

target for coldwater commercial fisheries in Lake Huron and the latter is incidentally harvested in lake whitefish

fisheries and is the object of a native species restoration program. We developed the food web model using

Ecopath with Ecosim (EwE), which uses a mass-balance approach to describe the food web and its constituent

trophic interactions, and a foraging-arena predation model to enable dynamic simulations that project food-web

changes over time. Despite the large amounts of data that exist for the Lake Huron food web, development of a

model proved very challenging, and considerable effort in this project was devoted to addressing two key model

development issues: obtaining a balanced Ecopath model, and incorporating invasive species into the dynamic

Ecosim model. Nevertheless, we were successful in developing a model that allowed evaluation of a variety of

harvest policies for lake whitefish commercial fisheries and consideration of three key areas of uncertainty

regarding food-web dynamics.

8

Our EwE model included 20 groups of species, several of which were divided into age-based stanzas, resulting in

a total of 36 modeled groups. Initial attempts to balance the Ecopath model using available data were

unsuccessful, primarily due to estimated predator consumption demands greatly exceeding estimates of available

prey production. This made it necessary for us to make ad hoc adjustments to the Ecopath model inputs, raising

concerns about the sensitivity of our results to these ad hoc choices. Our comparison of two contrasting strategies

for ad hoc model balancing indicated that in general dynamic simulation results are not highly sensitive to the

choice of strategy, although the sensitivity is substantially greater if the strength of trophic interactions among

species is high (i.e., high vulnerabilities, top-down control). We concluded from this analysis that correctly

determining the degree of top-down control in a food web was more important than the choice of strategy for

making ad hoc adjustments to model inputs (Appendix 2).

Our EwE model was initialized in 1981. We chose this early period to allow use of as much historical biomass

time series data as possible to fit the Ecosim model. Between 1981 and the present, three new taxa have invaded

Lake Huron and had substantial food web effects: Dreissenid mussels, Bythotrephes, and round gobies. EwE is

not well-suited to additions of new species groups to the model after the initial period, so we needed to develop a

method for including these species in the model. Working in collaboration with other Great Lakes EwE modelers,

we developed and evaluated four alternative methods for incorporating invasive species (Appendix 3). We

concluded that a method which initializes the biomass of each invasive species at a very low level in 1981

resulted in the best fit of estimated Ecosim biomass dynamics to observed data. This method required

adjustments of predator diets so that the invasive species’ contribution to consumption is negligible until the

species abundance reaches levels observed during the actual invasion/establishment period. The other methods

also performed reasonably well, but implementation of this method was comparatively simple and allowed the

model fitting process to estimate the strength of trophic interactions between the invasive species and their prey.

After addressing these two key model development issues, we fitted an EwE model for the main basin of Lake

Huron to time series of biomass estimates for 18 species groups. These time series included any data that were

available during the period 1981-2008. The best-fit model suggested that control in Lake Huron is mostly

bottom-up. Estimated vulnerabilities for most major predator groups were relatively low implying that there is

little evidence for strong trophic interactions among these groups. Instead, we found that the best model fits

resulted when we allowed for substantial changes to environmental productivity during the 1981-2008 period.

When we considered a range of possible future productivity levels during our harvest policy analysis, we found

that the outcomes were strongly influenced by this uncertainty. If productivity remains low in the future, our

simulations suggest that alewife biomass is very unlikely to recover, and that future biomass and yield of

harvested species will be much lower than if productivity returns to higher levels observed prior to the early

2000s.

Our analysis of harvest policy alternatives revealed the expected trade-off between harvest and biomass.

Increased exploitation rates for the targeted lake whitefish resulted in higher harvests of both lake whitefish and

lake trout, together with reduced biomasses of both groups. Harvests tended to increase more with increased

fishing mortality than biomass was reduced, suggesting some form of compensation in both species. For lake

trout this is at least partially due to the maintenance of recruitment through stocking, regardless of the removal of

spawning stock biomass through harvest. Our model was not able to consider the potential linkage between the

biomass of hatchery-derived lake trout that reach maturity and subsequent recruitment of wild lake trout – rather

hatchery and wild lake trout were treated as distinct groups in the model. The model results also did not reveal a

decline in harvest of either species at the highest fishing mortality level considered (double the current levels),

suggesting that there is little risk of overfishing at current harvest rates. However, we are concerned that the EwE

model may be overestimating the capacity for compensation in these species, especially lake whitefish. This

result warrants further examination. For lake whitefish, the pattern of declining biomass and increasing harvest

with increasing fishing mortality was sensitive to our uncertainties about diet and vulnerabilities, with biomass of

lake whitefish declining less when additional fish prey were added to lake trout and lake whitefish diets, and

when vulnerabilities for lake trout and lake whitefish were reduced. This sensitivity was much less evident for

lake trout.

Our simulations of gear conversion and seasonal restrictions on fishing indicated that these policies can provide

9

benefits for lake trout restoration by either reducing lake trout bycatch while maintaining lake whitefish harvests

at current levels, or increasing lake whitefish harvests without increasing lake trout removals. However the

predicted effects on lake trout biomass of complete conversion to trap nets or complete seasonal closures were

not very large (less than 20% increases in biomass). These estimated ecological benefits would need to be

weighed against the economic consequences of these policies in terms of gear conversion costs and market

consequences of shifting the seasonal patterns of lake whitefish harvests.

We also used the EwE model to assess the importance of indirect effects of harvest policies on non-target species.

In general the indirect interactions between lake trout and lake whitefish were quite weak. Substantially

increasing harvest of lake whitefish (and thus reducing lake whitefish biomass) without changing lake trout

harvest resulted in only small (< 10%) increases in lake trout biomass. Changes to lake trout harvest had even

less of an indirect effect on lake whitefish biomass. The magnitude of these indirect effects was sensitive to our

assumptions about diets – in particular whether lake trout consume lake whitefish – but remained small

regardless of the diet assumptions. From this we conclude that harvest policies for lake whitefish and lake trout in

Lake Huron should be more concerned with direct effects – through bycatch – than with indirect effects.

Our analysis represents the first attempt to explicitly link an evaluation of alternative harvest strategies with a

food-web model for a Great Lakes commercial fishery. The large changes that have occurred to the Lake Huron

food web in recent years point to considerable uncertainty about what commercial fishery yields and coldwater

fish biomasses can be expected in the future. If productivity remains low, expectations for future yields should be

adjusted downward to reflect this. Clearly this type of change in expectations has already occurred for the Lake

Huron recreational Chinook salmon fishery. On the other hand, our finding suggest that comparisons of

alternative harvest policies for balancing lake whitefish yield objectives with lake trout recovery objectives

should focus on direct effects – through management of incidental harvest of lake trout – rather than on indirect

effects manifested through food-web interactions.

REFERENCES:

Arreguin-Sanchez, F., Hernandez-Herrera, A., Ramirez-Rodriguez, M., Perez-Espana, H., 2004. Optimal management

scenarios for the artisanal fisheries in the ecosystem of La Paz Bay, Baja California Sur, Mexico. Ecol.

Model. 172, 373-382.

Christensen, V., Walters, C., 2004. Ecopath with Ecosim: methods, capabilities, and limitations. Ecol. Model. 172,

109-139.

Cochrane, K.L., Butterworth, D.S., De Oliveira, J.A.A., Roel, B.A., 1998. Management procedures in a fishery based

on highly variable stocks and with conflicting objectives: experiences in the South African pelagic fishery.

Rev. Fish. Biol. Fish. 8, 177-214.

De Oliveira, J.A.A., Butterworth, D.S., Roel, B.A., Cochrane, K.L., and Brown, J.P., 1998. The application of a

management procedure to regulate the directed bycatch fishery of South African sardine Sardinops sagax. S.

Afr. J. Mar. Sci. 19, 449-469.

DesJardine, R.L., Gorenflo, T.K., Payne, N.R., Schrouder, J.D., 1995. Fish-community objectives for Lake Huron.

Great Lakes Fish. Comm. Spec. Pub. 95-1. 38 p.

Ebener, M.P., 2005. The state of Lake Huron in 1999. Great Lakes Fish. Comm. Spec. Pub. 05-02, Ann Arbor. 140 p.

Eby, L.A., Roach, W.J., Crowder, L.B., Stanford, J.A., 2006. Effects of stocking-up freshwater food webs. Trends

Ecol. Evol. 21, 576-584.

EPAP (Ecosystem Principles Advisory Panel), 1999. Ecosystem-based fishery management. 54 p. (available online at

http://www.nmfs.noaa.gov/sfa/EPAPrpt.pdf).

Flint, R.W., 1986. Hypothesized carbon flow through the deepwater Lake Ontario food web. J. Great Lakes Res. 12,

344-354.

Fontaine, T.D., Stewart, D.J., 1992. Exploring the effects of multiple management objectives and exotic species on

Great Lakes food webs and contaminant dynamics. Environ. Manage. 16, 225-229.

GLFC, 2001. Strategic Vision of the Great Lakes Fishery Commission for the first decade of the new millennium.

Great Lakes Fish. Com., Ann Arbor. 40 p. (available online at:

http://www.glfc.org/pubs/SpecialPubs/StrategicVision2001.pdf).

10

Halfon E., Schito, N., Ulanowicz, R.E.. 1996. Energy flow through the Lake Ontario food web: Conceptual model

and an attempt to mass balance. Ecol. Model., 86, 1-36.

Hebert, C.E., Sprules, W.G., 2002. The relevance of seabird ecology to Great Lakes management. J. Great Lakes Res.

28, 91-103.

Irwin, B.J., M.J. Wilberg, M.L. Jones, and J.R. Bence. 2011. Applying Structured Decision Making to Recreational

Fisheries Management. Fisheries 36(3):7-16.

Jackson, L.J., Trebitz, A.S., Cottingham, K.L., 2000. An introduction to the practice of ecological modeling.

Bioscience 50, 694-706.

Jaeger, A.L., 2006. Invasive species impacts on ecosystem structure and function. Michigan State University. M.S.

Thesis. 199 p.

Jones, M.L., Koonce, J.F., O’Gorman, R., 1993. Sustainability of hatchery-dependent salmonine fisheries in Lake

Ontario – The conflict between predator demand and prey supply. Trans. Am. Fish. Soc. 122, 1002-1018.

Jones, M.L., Eck, G.W., Evans, D.O., Fabrizio, M.C., Hoff, M.H., Hudson, P.L., Janssen, J., Jude, D.,

O’Gorman, R., Savino, J.F., 1995. Limitations to lake trout (Salvelinus namaycush) rehabilitation in the Great

Lakes imposed by biotic interactions occurring at early life stages. J. Great Lakes Res. 21(S1), 505-517.

Kitchell, J.F., Cox S.P., Harvey C.J., Johnson T.B., Mason D.M., Schoen K.K., Aydin K., Bronte C., Ebener M. P.,

Hansen M., Hoff M., Schram S., Schreiner D., Walters C. J., 2000. Sustainability of the Lake Superior fish

community: Interactions in a food web context. Ecosystems 3, 545-560.

Matsuda, H., Abrams P.A., 2006. Maximal yields from multispecies fisheries systems: rules for

systems with multiple trophic levels. Ecol. Appl. 16, 225-237.

Nalepa, T.F., Fanslow D.L., Pothoven S.A., Foley III A.J., Lang G.A., 2007. Long-term trends in benthic

macroinvertebrate populations in Lake Huron over the past four decades. J. Great Lakes Res. 33, 421-436.

Nalepa, T.F., Pothoven, S.A., Fanslow, D.L., 2009. Recent changes in benthic macroinvertebrate populations in Lake

Huron and impact on the diet of lake whitefish (Coregonus clupeaformis). Aquat. Ecosyst. Health 12, 2-10.

Pascoe, S., 2000. Single species conservation in a multispecies fishery: the case of the Australian eastern gemfish.

Ecol. Econom. 32, 125-136.

Pikitch, E.K., Santora, C., Babcock, E.A., Bakun, A., Bonfil, R., Conover, D.O., Dayton, P., Doukakis, P., Fluharty,

D., Heneman, B., Houde, E.D., Link, J., Livingston, P.A., Mangel, M., McAllister, M.K., Pope, J., Sainsbury,

K.J., 2004. Ecosystem-based fishery management. Science 305, 346-347.

Pothoven, S.A., Nalepa T.F., 2006. Feeding ecology of lake whitefish in Lake Huron. J. Great Lakes Res. 32, 489-

501.

Riley, S.C., Roseman, E.F., Nichols, S.J., O’Brien, T.P., Kiley, C.S., Schaeffer, J.S., 2008. Deepwater demersal fish

community collapse in Lake Huron. Trans. Am. Fish. Soc. 137, 1879-1890.

Roseman, E.F., Riley, S.C., 2009. Biomass of deepwater demersal forage fishes in Lake Huron, 1994-4007:

Implications for offshore predators. Aquat. Ecosyst. Health 12, 29-36.

Schaeffer, J.S., Warner D.M., 2008. Status and trends of pelagic prey fish in Lake Huron, 2007. U.S. Geological

Survey. 11 p. (available online at http://www.glsc.usgs.gov/_files/reports/2007LakeHuronAcoustic.pdf).

Schaeffer, J.S., Warner D.M., O'Brien T.P., Roseman E.F., 2007. Status and trends of pelagic prey fish in Lake

Huron's main basin, 2006. U.S. Geological Survey. 9 p. (available online at

http://www.glsc.usgs.gov/_files/reports/2006LakeHuronAcoustic.pdf).

Smith, A.D.M., Fulton, E.J., Hobday, A.J., Smith, D.C., Shoulder, P., 2007. Scientific tools to support the practical

implementation of ecosystem-based fisheries management. ICES J. Mar. Sci. 64, 633-639.

Zetina-Rejon, M.J., Arreguin-Sanchez, F., Chavez, E.A., 2004. Exploration of harvesting strategies for the

management of a Mexican coastal lagoon fishery. Ecol. Model. 172, 361-372.

11

ACKNOWLEDGEMENTS:

This project depended on contributions from many individuals and organizations that we would like to

acknowledge. The participants at both of our stakeholder workshops, listed in Appendix 1, helped to guide model

development and the focus of our analysis. Valuable data and insights about these data were provided by Richard

Barbiero, Travis Brenden, Dave Caroffino, Adam Cottrill, Norine Dobiesz, Mark Ebener, Darryl Hondorp, Ji He,

Ora Johannsson, Jim Johnson, Dave Jude, Tracy Kolb, Ann Krause, Lloyd Mohr, Charles Madenjian, Andrea

Miehls, Steve Pothoven, Stephen Riley, Edward Roseman, Jeffrey Schaeffer, and Mike Siefkes. Brian Irwin

assisted in the early stages of the project and in writing the original grant. Matt Catalano assisted with analyses of

time series data. Carl Walters and Villy Christensen provided invaluable assistance to BJL in understanding the

EwE modeling software. David McLeish, Tammy Newcomb, Jim Bence, Yu-Chun Kao, Hongyan Zhang, Mark

Rogers, provided numerous helpful comments and insight during discussions of the model. Joe Buszowski and

Jeroen Steenbeek helped with refinements to EwE code. Our work was funded by a grant from the Great Lakes

Fishery Commission to MLJ, a fellowship (William E. Ricker Fellowship) from Michigan State University to BJL

to support his graduate research, and travel support from the Foreign Affairs and International Trade Canada to

BJL for travel to the University of British Columbia.

DELIVERABLES:

Progress reports in Dec 2009 and Dec 2010

Final report (this document)

Final EwE Lake Huron main basin model – available from Quantitative Fisheries Center

Participation in four EwE modeling workshops funded by the Great Lakes Fishery Commission Science

Transfer Program (Jun & Nov 2010, Feb & July 2011)

Oral presentations (IAGLR 2010 – Toronto; AFS 2010 – Pittsburgh; AFS 2011 – Seattle) – all by Brian

Langseth, PhD student and Ricker Fellow, Quantitative Fisheries Center

Three manuscripts (Appendices 2-4, this document): Appendix 2 is under revision for re-submission to

Ecological Modeling; Appendix 3 will be merged with results from a Lake Michigan EwE for submission

in 2012; Appendix 4 will be revised based on further analyses and submitted in 2012.

PhD dissertation – Brian Langseth, to be submitted in 2012.

PRESS RELEASE:

Researchers at Michigan State University have completed a three year study that uses a food-web model to

explore commercial fishing policies in Lake Huron. Mussels, gobies, and other invasive species have increased in

Lake Huron, while other prey of fished species have declined. These changes increase uncertainty about future

numbers of fish in the lake, as well as future fishery yields. The researchers used the food-web model to account

for these changes and forecast how different fishing policies might perform.

Lake whitefish (Coregonus clupeaformis) is the main species caught by the fisheries, but lake trout (Salvelinus

namaycush) is caught as well. Lake trout is being restored in the Great Lakes, adding objectives other than

harvest to the analysis. The researchers found that policies which change the type of fishing gear or change

fishing seasons maintained or increased lake whitefish harvest while at the same time maintaining or increasing

lake trout biomass, thereby balancing the tradeoff between objectives.

Future environmental productivity was found to be a very important source of uncertainty. Expected yields of

lake whitefish as well as expected biomass of lake trout and other species were much lower if productivity

remained low than if productivity increased to levels experienced in the 1990s. The researchers also found little

evidence that lake trout and lake whitefish interact indirectly through food-web dynamics. This led them to

conclude that the commercial fishery is the primary factor linking these two important species. Future work on

balancing tradeoffs in the commercial fisheries should therefore focus on direct interactions within the fishery –

that is, reducing bycatch – rather than on indirect interactions through the food web. Even so, the food-web model

can still be useful to managers for exploring other questions related to Lake Huron and its changing food web.

12

APPENDICES:

Appendix 1: Notes from first two stakeholder workshops

Appendix 2: Draft manuscript describing the effects of balancing on Ecosim output. To be resubmitted to Ecological

Modeling.

Appendix 3: Modeling species invasions in an Ecopath with Ecosim model of a Laurentian Great Lake, Lake Huron.

Appendix 4. Evaluation of harvest policies for Lake Huron coldwater commercial fisheries using an Ecopath with

Ecosim model.

1 - 1

Appendix 1: Notes from first two stakeholder workshops

Workshop 1

Context of this project: How to manage commercial fisheries that harvest multiple species to ensure long-term

sustainability of the fisheries?

Purpose of this project: Explore what harvest policies for Lake Huron commercial fisheries are optimal given trade-

offs.

Objective of this project: Determine possible management objectives and harvest options to consider. These do not

reflect objectives and options which should actually be done, but rather objectives and options which can be

considered in an exploratory modeling exercise. Determine key uncertainties in our knowledge of the system. Use a

spatially structured food-web model using Ecopath with Ecosim that makes forecasts into the future given a range of

possible management options. Evaluate the performance of the model forecasts in meeting the possible management

objectives.

Objective of this workshop: List possible management objectives and options. Discuss the spatial resolution and

species to include in the model.

Stakeholders and their affiliation:

Michael Jones Michigan State University (MSU)

Brian Irwin MSU

Brian Langseth MSU

Frank Krist Michigan Lake Huron Citizens Fishery Advisory Committee

Chris McLaughlin McMaster University

David Reid Ontario Ministry of Natural Resources (OMNR)

Adam Cottrill OMNR

Stephen Riley United States Geological Survey

David McLeish OMNR

Lloyd Mohr OMNR

David Carlson Commercial fisherman

Peter Meisenheimer Ontario Commercial Fisheries Association (OCFA)

Kevin Reid OCFA

Dennis Morrison Lake Huron Georgian Bay Fisheries Stewardship Council

Tim Purdy Commercial fisherman

Forrest Williams Michigan Fish Producers Association

Milford Purdy Commercial fisherman

George Purvis Commercial fisherman, President Algoma-Manitoulin Commercial

Fishing Association

Schedule of events:

Day one - April 20

1. Welcome and introductions

2. Background to project

3. Overview of project goals, objectives and approach

4. Discussion of management objectives

Break

5. Discussion of management options

6. Wrap-up, plan for day two

Day two – April 21

1. Foodweb/management models – Ecopath with Ecosim

1 - 2

2. Discussion of key model components and uncertainties

Break

3. Continued discussion

4. Wrap-up, review of next steps

Outcome of discussions pertinent to meeting the objectives of this workshop:

Management objectives – note that this list is a summary of ideas presented by workshop participants without further

interpretation or refinement of the MSU team. This list of objectives will be used by the MSU team to define

quantities that our forecasting model will need to include, so that we can assess the predicted performance of

management options.

Stability over time of TAC’s, but not at the expense of forgone harvestable

surplus

Lake trout total annual mortality < 40%

50% of lake trout are wild

Minimum bycatch induced restrictions on harvest of target species

Maximize landed value of catch incorporating species mix

Minimize operating costs of fishing

Diversity of the fishery (more harvestable species)

Maintain lake whitefish yield objective from Fish Community Objectives

Production of valued species not limited by native food base (but prefer low

abundance of alewives)

Higher yields of chubs (bloater), cisco, walleye, and yellow perch

Reduced uncertainty

Maintain high catch rates for fish for sports fishermen

Management options (things in the system that can be changed, particularly by fisheries management agencies) –

similarly, this list has not been interpreted or refined by the MSU team and will be used to inform the development of

our model..

Quotas

Seasonal closures/openings

Area closures/openings

Gear conversion when feasible (e.g. trap nets can’t keep up with lake whitefish

movements on Canadian side of southern main basin)

Gillnet length

Effort

Buy outs

Stocking - only stock natives, stock by ratios

Policies based on species ratios in catches, biomasses, reference points, trend

response

Quota zonation

Increased assessment

Active adaptive management – allow targeted effort for lake trout

Bycatch reduction

Market development

Enhanced sea lamprey control

Risk-based harvest control rules when species of importance influence each other

either positively or negatively (e.g. cisco harvest related to lake trout restoration)

Description of the fishery

Gillnet offshore in Ontario, trapnet nearshore in southern main basin (4-5) and off

Manitoulin (4-2)

No commercial operations in southern main basin on Michigan side up to tip of

1 - 3

thumb. Trapnets from tip to Hammond Bay, and mix of trapnet and gillnet in northern main basin.

Recreational fishing for lake trout, walleye, yellow perch, salmon, cisco in north

channel and northern main basin.

How to represent the spatial structure of Lake Huron in the model:

Can model all three basins, however focus should be on main basin.

Test the model using a single basin and then build other basins in if model is

useful

Nearshore and offshore zones are different in nutrients and species composition.

Should consider separate models for each

Saginaw bay species are not as important to the cold-water fishery, and thus likely

do not need to be included in the model. Exception would be if they play a significant role in the diet

of cold-water species

Species to be included in the model and relative uncertainty about them:

Species Uncertainty Species Uncertainty

lake whitefish low alewife Low

lake trout low/med smelt Low

bloater low/med slimy/deepwater sculpin Low

cisco med/high gobies low/med

walleye low/med emerald shiners High

yellow perch low/med spottail shiners Med

round whitefish med/high sticklebacks med/high

lamprey low trout perch med/high

cormorant low burbot med

salmon species med/high dreissenids med

pinks native molluscs high

chinook diporeia med

steelhead mysis high

coho zooplankton med/high

atlantic phytoplankton high

brown cladophora high

Potential drivers of recent changes in Lake Huron

Mussels – reduce diporeia and contribute to poor food quality, therefore reducing

prey availability

Gobies – predation pressure

Salmon and trout – predation pressure

Thiamine deficiency syndrome

--------------------------------------------------------------------------------------------------------------------------------------------

Workshop 2

Objectives of this workshop:

1) Review and comment on current EwE model

2) Detail the metrics we will use to report outcomes of the modeling work.

3) Discuss viable management options.

Attendees:

Michael Jones Michigan State University (MSU)

Brian Irwin MSU

Brian Langseth MSU

1 - 4

Frank and Teresa Krist Michigan Lake Huron Citizens Fishery Advisory Committee

David Reid Ontario Ministry of Natural Resources (OMNR)

Adam Cottrill OMNR

David McLeish OMNR

Lloyd Mohr OMNR

David and Paul Carlson Commercial fisherman

Peter Meisenheimer Ontario Commercial Fisheries Association (OCFA)

Kevin Reid OCFA

Dennis Morrison Lake Huron Georgian Bay Fisheries Stewardship Council

George Purvis Commercial fisherman, President Algoma-Manitoulin Commercial

Fishing Association

Jim Johnson Michigan Dept. of Natural Resources and the Environment

Agenda:

1) Update on food-web model (Langseth)

2) Discussion of management objective, what metrics should be reported (Irwin)

3) Discussion of management options (Jones)

Summary of agenda item 1:

-Main basin model built, parameterized in 1981.

-Reasonable fits of model dynamics to biomass time series can be generated by fitting vulnerabilities only, as

well as both vulnerabilities and primary production anomalies.

-Invasive species (round gobies, zebra and quagga mussels, and bythotrephes) are included in the model.

Mortalities of these groups are set very high in earlier years when invasives were not present, and reduced for

periods when they invade.

-Using values from the literature suggest an overconsumption by prey fish on lower trophic groups

(invertebrates and plankton). Consequently, to achieve mass balance, production of lower trophic groups was

increased and consumption of prey fish was decreased.

Summary of agenda item 2

Objectives (in no particular order) and initial identification of corresponding performance metrics:

1. Maintain lake whitefish yield objective from FCOs

Metrics: harvest of bloater, lake whitefish, and cisco.

2. High catch rates for fish for sports fishermen

Metrics: number per effort, with some measure of quality.

3. Minimum bycatch induced restrictions on harvest of target species

Bycatch defined as catch above that which is allowed, or catch of non-marketable fish.

Metrics: biomass of bycatch species.

4. Diversity of the fishery

Want to be able to pursue other species when beneficial to do so.

Metrics: age composition of harvest, and fishable biomass.

5. Lake trout annual mortality < 40%

Metrics: total mortality, spawning stock biomass, for fully selected ages

6. Maximize landed value of catch incorporating species mix

Value dependent on supply, also exchange rate.

Metrics: dockside values and harvest.

7. Minimize operating costs of fishing

1 - 5

Metrics: effort and price per effort for trap net and gill net.

8. Stability of yield, but not at the expense of foregone harvestable surplus

Metrics: total net income (harvest x price – costs).

Additional discussion item:

Need understanding that this model operates on a lake-wide scale but that an individual local effect could still

influence results.

Summary of agenda item 3:

Management Options (in no particular order) to focus on in model simulations.

a. Quotas

Include quotas in policy searches

b. Effort

Output controls (e.g. quotas) generally preferred over input controls (e.g. effort)

c. Gear adjustments or conversions

Where physically able to use gear

d. Seasonal closures/openings

Seasons where lake whitefish price is highest (winter) are preferable, but requires gill net

e. Market development

Market forces impact price, and thus increasing available harvest may not always be as profitable

Strategies on implementing these options were also discussed and include

f. Policies based on species ratios in catches, biomass, reference points, trends.

Explicit rule that says harvest related to some population measure

g. Risk-based harvest control rules when species of importance influence each other either + or - (e.g. cisco

harvest related to LT restoration)

(Maximize harvest without reducing lake trout biomass to some low level)

2 - 1

Appendix 2: Draft manuscript describing the effects of balancing on Ecosim output. To be resubmitted to Ecological

Modeling.

Title: The effect of different approaches to achieve mass balance on a food-web model of Lake Huron

Authors: Brian J. Langsetha*

(langset2@msu.edu), Michael L. Jonesa (jonesm30@anr.msu.edu), Stephen C. Riley

b

(sriley@usgs.gov)

Addresses: a Quantitative Fisheries Center, 153 Giltner, Michigan State University, East Lansing, MI, USA, 48824.

b Great Lakes Science Center, United States Geological Survey, 1451 Green Rd., Ann Arbor, MI, USA, 48105.

*Corresponding author: Tel: 1-517-355-0126, Fax: 1-517-355-0138

Highlights:

Ecopath models are rarely balanced. Imbalances are often in lower-trophic groups.

Biomass dynamics in Ecosim can also be used to assess the effect of balancing.

The effect of balancing is greatest when vulnerabilities are high.

The effect of balancing is greatest when changes in biomass through time are small.

Abstract:

The Ecopath with Ecosim (EwE) software is an increasingly popular modeling tool in fishery research and

management. Ecopath requires a mass-balanced snapshot of a food-web at a particular point in time, which Ecosim

then uses to simulate biomass through time. Initial inputs to Ecopath, including available estimates of biomass,

production, consumption, and diets, rarely produce mass balance, and thus changes to the inputs are required to

balance the model. There has been little discussion of whether these changes to achieve mass balance affect model

results. We constructed two EwE models for the offshore community of Lake Huron, balanced in two contrasting but

realistic ways. One placed more confidence in estimates of consumption; levels of production were increased to

achieve mass balance. The other placed more confidence in estimates of production; levels of predation were

decreased. To assess the effect of balancing approach on model results, we compared ecosystem metrics within

Ecopath (ascendency and system omnivory index (SOI)), as well as biomass dynamics within Ecosim. Within

Ecosim, we compared simulations given alternative assumptions about the type of control (top-down or bottom-up)

under two scenarios of (1) increased fishing mortality or (2) increased environmental production. Ascendency for the

first balancing approach was approximately four times that of the second, and was mostly due to greater assumed

production in planktonic groups. Values of SOI were nearly identical, suggesting very little difference between the

two approaches. Differences in overall biomass between the two balancing approaches were greatest under scenarios

assuming top-down control, and differed by at most a factor of 1.15. When expressed relative to the overall change in

biomass for each scenario, the differences between balancing approach represented at most 41%, but were much

lower for scenarios where biomass changed substantially. The importance of these differences appears to be small,

and comparisons with other models would be helpful to compare significance. Our findings suggest that approaches

to balancing Ecopath models have the greatest effect on model results when top-down control is prevalent in the

system and when simulated biomass dynamics are stable through time.

Keywords: Ecosystem models, Ecopath, Ecosim, mass balance, vulnerabilities, Great Lakes

2 - 2

1. Introduction:

Mathematical models are tools to explain complex processes in simple and understandable ways. Trophic

interactions within most aquatic ecosystems are complex, and attempts to understand these interactions for

applications to fisheries management can be aided by models. A variety of modeling frameworks have been used to

describe ecosystem processes, and even within an individual framework, the processes themselves can be described in

multiple ways (e.g. alternative assumptions about the strength of trophic interactions). When models are used to

inform decisions, and thereby influence management actions, it is important to determine whether alternative

descriptions of ecosystem processes within a modeling framework affect conclusions.

The use of models to inform fisheries management decisions is likely to increase in the future. This is due

partly to increased computing power, which allows for more explicit accounting of multiple species and the processes

that govern their interactions (Hilborn and Walters, 1992; Robinson and Frid, 2003), and also to greater awareness of

the need to base management on an understanding of the ecosystems within which fisheries operate (Bundy et al.,

2001; Walters et al., 2008). These have contributed towards the progression from single-species management to

ecosystem-based fisheries management (EBM), which incorporates objectives for many components within an

ecosystem, not just for harvested species (Marasco et al., 2007). Examples of ecosystem objectives include prey

availability for non-human components of the ecosystem, maintaining important habitats, and managing bycatch

(Ruckelshaus et al., 2008). Ecosystem models are a valuable tool for the assessment of EBM policies in meeting

desired objectives (Pikitch et al., 2004; Smith et al., 2007; Walters and Martell, 2004). Policies for Laurentian Great

Lakes fisheries have incorporated attributes of EBM, such as the Fish Community Objectives for Lake Huron

(DesJardines et al., 1995) and the Joint Strategic Plan (Gaden et al., 2008). As the use of ecosystem models increases,

the value of understanding their sensitivity to alternative process descriptions within the model will as well.

A variety of ecosystem models have been used for fisheries applications (Hollowed et al., 2000; Latour et al.,

2003; Robinson and Frid 2003; Whipple et al., 2000), all with strengths and weaknesses. The modeling framework

used here is the Ecopath with Ecosim (EwE) software. EwE consists of two modules, Ecopath and Ecosim. Ecopath

allows for a mass-balanced description of a food-web at a single point in time, which is parameterized with biomass,

growth, and consumption data for each modeled group (Christensen and Walters, 2004). The Ecopath model, along

with values reflecting the strength of interspecific interactions is then used as input to Ecosim, which simulates the

dynamics of the modeled groups through time (Christensen and Walters, 2004). Published Ecopath models in the

Great Lakes exist for Lake Superior (Cox and Kitchell, 2004; Kitchell et al., 2000) and Lake Ontario (Halfon et al.,

1996; Koops et al., 2006; Stewart and Sprules, 2011), while unpublished work exists for Lake Michigan (Ann Krause,

University of Toledo (UT), pers. comm.). Work is also underway on building models for other Great Lakes (David

Bunnell, United States Geological Survey (USGS), pers. comm.; Sara Adlerstein and Ed Rutherford, University of

Michigan, pers. comm.).

A challenge when constructing Ecopath models is obtaining representative average data inputs across a year

and system that meet the physical constraints of mass balance. Mass balance is a requirement in Ecopath models; the

amount of food consumed must be less than what is available, and the gain in weight from consumption must be less

than the weight of food consumed. Achieving mass balance in Ecopath models invariably requires modifying input

parameters (Christensen et al., 2005), but EwE users rarely discuss this process. Modifying input parameters can be

done in many ways, and thus multiple descriptions of the food web are possible as a result of the balancing process.

Studies that used Ecosim models of the Great Lakes focused on time dynamic simulations and only briefly discussed

balancing procedures (Cox and Kitchell, 2004; Kitchell et al., 2000). Studies without time dynamic simulations

focused on ecosystem descriptions, and discussed the balancing process only briefly (Halfon et al., 1996; Koops et al.,

2006), with the exception of Stewart and Sprules (2011).

There are two primary procedures to balance Ecopath models. Model inputs can be adjusted subjectively by

the modeler, based on their judgment of which inputs are least reliable or most likely to achieve mass balance

(Christensen et al., 2005). Alternatively, changes can be made objectively by the software, based on a quantification

of user-perceived reliability for input parameters and a formal objective function, such as minimizing changes to

initial inputs (Kavanagh et al., 2004). In one of the few publications in which the impact of balancing was discussed,

Essington (2007) found that randomly balancing a model produced a balanced model as close to the “true” model

(one in which all parameters were known) as when balancing was done using an objective function. Although the

overall variation in input parameters among balanced models was similar when using either balancing procedure, the

differences in model outputs such as ecosystem metrics or simulated biomasses in Ecosim were not assessed. When

using the subjective procedure, the modeler, rather than the software, makes changes to input parameters which can

2 - 3

provide a better understanding of the linkages within the system. Although more time consuming, consideration of

such linkages may reveal unexpectedly important groups or gaps in the understanding of the system.

Subjective balancing was used to balance the model described herein. Through iterative attempts to achieve

mass balance, a major source of imbalance was found to be due to a perceived overconsumption of intermediate

trophic-level groups. Conversations with other EwE modelers suggest this is a common problem, but the observation

has not been discussed in standard balancing documentation and has only recently been discussed in the literature

(Stewart and Sprules, 2011).

The objective of this research was to assess how alternative strategies for balancing an Ecopath model

affected EwE results. An Ecopath model of the offshore fish community of Lake Huron was constructed and

subjectively balanced using two approaches. The first approach assumed data for upper trophic-level groups were

more reliable than for lower trophic-level groups, whereas the second assumed the opposite. Differences in model

outputs of ecosystem metrics in Ecopath and time dynamic simulation results in Ecosim were used to compare the

two balancing approaches. The strength of trophic interactions, and the level of environmental productivity and

fishing mortality were adjusted within Ecosim to assess the impact of ecosystem conditions on model results. Possible

reasons for observed imbalances in the model are also described.

2. Methods:

2.1 Study site

Lake Huron is the second largest of the Laurentian Great Lakes with a surface area of 59,570 km2

(DesJardines et al., 1995), and is divided by the border between the United States and Canada (Fig. 1). Both

commercial and recreational fisheries exist in the lake; the majority of commercial effort occurs in Canadian waters,

whereas the majority of recreational activity occurs in US waters. Management responsibilities for the fisheries are

shared among Michigan, Ontario, and Native American agencies. These agencies collect information pertinent to

fisheries management, and are assisted in collection and coordination by the USGS, US Fish and Wildlife Service,

Environmental Protection Agency, Canadian Department of Fisheries and Oceans, and Great Lakes Fishery

Commission. Data to parameterize the model were provided by these agencies as well as from published sources, and

are described in detail in the online appendix (Table A.1).

2.2 The model

Ecoppath with Ecosim (EwE) version 6.1.1 was used to construct a food-web model of the offshore fish

community in the main basin of Lake Huron (Fig. 2). It is a freely downloadable software package that describes a

snapshot of an ecosystem at a particular point in time (Ecopath), and simulates the dynamics of modeled groups

through time (Ecosim). Details of the modeling software have been discussed previously (Christensen and Pauly,

1992; Christensen and Walters, 2004), but a summary of key equations and parameters is provided below. The mass

balance equation in Ecopath for each group i is

Pi = Yi + Ei + (BA)i + Bi(M2)i + Bi(M0)i , (1)

where P is production; Y is total harvest; E is net migration (emigration - immigration); BA is biomass accumulation;

B is biomass; M2 is predation mortality rate; and M0 is other mortality rate, which represents sources of mortality not

included in the model. For simplicity, both E and BA were assumed to be zero for this research. The Ecopath user

does not provide estimates for all parameters in equation (1) directly, but rather provides inputs from which

parameters in equation (1) are calculated. When expressed in terms of Ecopath user inputs, equation (1) becomes

ii

j

ijjji

iBBP

DCBBQY

EE)/(

/

, (2)

where EE is ecotrophic efficiency, which equals 1-M0i/(P/B)i and represents the proportion of total mortality

explained by sources in the model; Q/B is the consumption to biomass ratio; DCij is the percent contribution of prey i

to the diet of predator j; and P/B is the production to biomass ratio. Allen (1971) found that under standard

assumptions of exponential mortality and von Bertalanffy growth, P/B is equivalent to the total instantaneous

mortality rate, assuming BA is zero. It is assumed that all components of mortality are included in equation (1), and

thus conservation of matter requires that the sum of these terms equals the assumed level of production. In relation to

equation (2), this means total production (denominator) must be greater than the loses due to consumption and fishing

described in the model (numerator) so that EE<1. When this occurs, the Ecopath model is said to be balanced and can

be used as input into Ecosim.

2 - 4

Ecosim is governed by two primary equations. For groups with a single age stanza, the change in biomass of

each group is modeled as

iiiii

j

ij

j

jii

i BeFMIQQgdt

dB0 , (3)

where t is time in months; g is the gross food-conversion efficiency ratio (GCE), which is the ratio of P/B to Q/B; Qij

is consumption of prey i by predator j; I is immigration; e is the emigration rate; and F is the fishing mortality rate.

The solution to the differential equation is estimated using an Adams-Bashford method (Christensen et al., 2005).

More complicated versions of equation (3) are used for groups with multiple age stanzas (see Walters et al., 2008 for

details). The second Ecosim equation defines the value of consumption, which in its simplest form is modeled as

jijij

jiijij

ijBav

BBvaQ

2, (4)

where aij is the effective search rate of predator j for prey i, and vij is the vulnerability of prey i to predator j.

Vulnerabilities define the strength of trophic interaction between a predator and its prey and can be adjusted within

Ecosim. The user does not actually adjust vij, but rather a constant (Kij) from which vij is calculated using vij =

Kij(Qij/Bi) (Walters and Christensen, 2007).

2.2.1 Building an Ecopath model

The Ecopath model of Lake Huron was constructed in a series of three steps: choice of groups, choice of time

period, and choice of data sources. For choosing which groups to include in the model, Lake Huron biologists were

consulted, and 21 species or groups of species were identified based on their assumed importance to the offshore

community (Table 1). Individual species were further separated into age stanzas when appropriate data were

available. Each age stanza consisted of a set of age classes and reflected a difference in trophic ontogeny or mortality

schedule. The total number of modeled groups increased from 21 to 36 when age stanzas were included (Table 1).

Special considerations were required for modeling stocked and semelparous species. Ecopath requires the age

of the first stanza for each group to begin at 0 months. Species are stocked into Lake Huron at either 6 months

(fingerlings) or 12 months (yearlings) so a pre-stocking stanza with age either 0-6 or 0-12 months was included in the

model with a very low mortality rate and import-only diet. An import-only diet represents feeding outside the

modeled system and allows a stanza that is not actually in the system, but is required by the model, to be included

without affecting other species. Within Ecosim, the biomass of the first age stanza for stocked species (lake trout

Salvelinus namaycush, steelhead Oncorhynchus mykiss, and Chinook salmon Oncorhynchus tshawytscha) was set to

the initial Ecopath value for each simulation year. This was also done for sea lamprey Petromyzon marinus, whose

recruitment is heavily influenced by management (pest control). Ecopath also assumes that all species are iteroparous.

Therefore, the semelparous Chinook salmon was modeled with a terminal age stanza (age 6+) that had an import-only

diet. A terminal age stanza (age 6+) for steelhead was also included in the model.

The second step was to choose a representative time period, which was dictated by the availability of data.

Biomass estimates for nearly all modeled groups were available for 1999. The choice of time period should reflect a

period of stability in the food-web. Lake Huron has undergone substantial changes in recent years; zebra Dreissena

polymorpha and quagga mussels Dreissena bugensis have proliferated in the 1990s and early 2000s (Nalepa et al.,

2007) and alewife Alosa pseudoharengus abundance collapsed in 2003 (Riley et al., 2008). Despite the proximity of

these events, 1999 was chosen as the modeled year due to the greater availability of biomass estimates, particularly

for lower trophic-level groups.

Step three was to find estimates for Ecopath parameters centered around 1999 for the chosen groups. Ecopath

requires estimates of B, P/B, Q/B, and diet for all groups. For groups with age stanzas, B and Q/B are needed for only

one stanza. Ecopath then calculates B and Q/B for all other age stanzas based on an age-structured model that assumes

a stable age distribution and von Bertalanffy growth (Walters et al., 2008). Consequently, von Bertalanffy growth

coefficients (K) were also required for groups with multiple age stanzas. When estimates of parameters were not

available for Lake Huron in 1999, estimates were obtained from as similar a system and time period as possible.

Recent time periods in Lake Huron are generally divided by the invasion of dreissenid mussels in the early 1990s and

the collapse of alewife in 2003. Consequently, when data were not available explicitly for 1999, sources from 1990-

2003 were preferentially used. When estimates were not available for Lake Huron, estimates for other Great Lakes

were used. Given the geographical and biological similarities between Lake Huron and Lake Michigan, parameter

estimates were preferentially taken from Lake Michigan data, and as near to 1999 as possible. When Lake Michigan

2 - 5

estimates were not available, preference was given to studies on Lake Ontario, Lake Superior, and finally Lake Erie.

When estimates were not available from studies within the Great Lakes, estimates were taken from Ecopath models

built for Lake Michigan (Ann Krause, UT, pers. comm.), Lake Superior (Cox and Kitchell, 2004; Kitchell et al.,

2000), or from regression relationships (e.g. Pauly, 1980).

2.2.2 Balancing an Ecopath model

Once the Ecopath model was built, parameter estimates were adjusted so that the model was balanced. We are

not aware of any Ecopath model that met the requirement of mass balance without some adjustment to initial

parameter estimates. To achieve mass balance, initial parameter estimates were changed by an iterative process

following recommended practices (Christensen et al., 2005). Estimates of P/B, Q/B, and diet from other Great Lakes

and other time periods were consulted to supply a range of plausible parameter estimates as additional guidelines to

inform the balancing process. Lake Huron biologists were consulted throughout the balancing process to ensure that

only plausible estimates of parameters were considered.

The order in which parameters were changed to achieve mass balance reflected their assumed reliability. Diet

studies for Lake Huron species are not common, and studies within the Great Lakes themselves range over many

years, particularly for lower trophic-level groups. The inherent nature of diet studies also results in poor reliability.

Diet varies both temporally and spatially, and thus diet observations may differ substantially among samples, even

within the same system and year. Consequently, diet information was assumed to be least reliable and was changed

first during balancing. After reasonable adjustments to diets were made, Q/B and P/B ratios, and B were changed.

Changing Q/B ratios was particularly relevant when many groups consumed by the same predator were all

unbalanced.

The balancing process can produce many different, but balanced, models. Two contrasting but realistic

approaches to balancing were compared. The first approach was based on the assumption that data for top trophic-

level groups were most reliable. Therefore, to achieve mass balance, production (P/B or B) of lower trophic-level

groups was preferentially increased to meet the consumption demands of their predators. This approach will be

referred to as the “consumption-based” approach throughout this paper, reflecting the confidence placed on the

estimates of consumption by top trophic-level groups. The second approach was based on the assumption that data for

prey fish and other lower trophic-level groups were most reliable. Consumption (Q/B or B) by predators was therefore

lowered to meet the production of their prey. This approach will be referred to as the “production-based” approach,

reflecting the confidence placed on estimates of production by lower trophic-level groups.

2.3 Assess impacts of balancing approaches

Ecosystem metrics within Ecopath and biomass dynamics within Ecosim were used to compare the two

balancing approaches. Measures of ecosystem maturity (sensu Odum, 1969) have been used to compare ecosystem

models (Christensen, 1995). For this research, ascendency (as described in Christensen and Pauly, 1992) was used to

assess ecosystem maturity. Ascendency describes the size and organizational structure of an ecosystem, and is argued

to reflect ecosystem maturity where high values represent a mature ecosystem and low values represent an immature

ecosystem (Ulanowicz, 1986). The second ecosystem metric was system omnivory index (SOI) which is the average

of each consumer’s omnivory index weighted by the logarithm of their food intake (Christensen et al., 2005). This

index also describes the structure of the food web while accounting for the magnitude of consumption for each group,

and is therefore influenced by changes to parameters that define either food-web structure (diet) or food intake (B,

Q/B).

Ecosim dynamics were also used to compare the two balancing approaches. Ecosim was run for a duration of

40 years, and the proportional difference in biomass between the two balancing approaches was calculated as

index 1 =mtim

mtim

tiB

Bmedmed

,,

,,

min

max, (5)

where medi is the median value taken over the index (in this case i groups); maxm and minm are the maximum and

minimum values between the m balancing approaches, respectively; and Bi,t,m is the biomass of group i in month t for

balancing approach m. To assess the sensitivity of balancing, the median change in relative biomass across balancing

approach, time, and then groups was calculated as

index 2 = mti

mtimti relBmedmedmed ,,

,, , (6)

2 - 6

where relBi,t,m is the biomass of group i in month t for balancing approach m relative to the biomass in month 1, and

ζi,t,m is an indicator variable that was 1 if relBi,t,m > 1 and -1 if relBi,t,m < 1, and then the ratio of the two indices

(adjusted to reflect percentage changes) was calculated as

index 3 = (index 1 – 1)/(index 2 – 1). (7)

For all indices, sea lamprey and the first age stanza for stocked species were not included in the calculations because

the biomasses of these groups were forced at their initial Ecopath values. The last age stanzas for steelhead and

Chinook salmon were also excluded because they were modeling artifacts and were not of interest.

In the absence of any perturbation, the biomass of each group in Ecosim does not change from its initial

Ecopath value. Therefore, to compare the two balancing approaches, either the underlying productivity of the

environment (modeled as the P/B ratio of phytoplankton) or the total fishing mortality on all groups was doubled

during simulations. This caused the biomass for each group to vary across years in both balancing approaches. The

changes made to environmental productivity and fishing mortality were based on observed variability from past

estimates. Estimates based on chlorophyll a suggest that primary production during the late 1980s and early 1990s

was approximately twice that during 1999 (Barbiero et al., 2009). Similarly, estimates of fishing mortality for lake

whitefish Coregonus clupeaformis (Adam Cottrill, Onatrio Ministry of Natural Resources, unpublished data) and

other salmonines have varied more than two-fold through time, although doubling F produced greater values than

those observed in the past for lake trout (Ji He, Michigan Department of Natural Resources, unpublished data) and

Chinook salmon (Travis Brenden, Michigan State University, unpublished data).

Ecosim dynamics are heavily influenced by assumptions about the strength of species interaction

(vulnerabilities) (Christensen and Walters, 2004). Therefore, three different values for the vulnerabilities of prey to

their predators were also used: 1.01, 2, and 10. These values reflect low (1.01), intermediate (2), and high (10) impact

of predators on their prey. Low impact can be thought to represent bottom-up control, whereas high impact can be

thought to represent top-down control. Each vulnerability value was used on both types of forcing (environmental or

fishing) for a total of six different Ecosim scenarios.

3. Results:

As expected, the initial Ecopath model was unbalanced. Unbalanced groups included age 0 bloater

Corregonus hoyi (EE=2.1) and smelt Osmerus mordax (EE=3.2); less abundant prey fish including round goby

Neogobius melanostomus (EE=28.4), ninespine stickleback Pungitius pungitius (EE=7.4), and slimy Cottus cognatus

(EE=60) and deepwater sculpins Myoxocephalus thompsoni (EE=1.2); yearling lake trout (EE=2.1); age 1+ alewife

(EE=1.5); and phytoplankton (EE=3.2). Except for phytoplankton, these groups occupied intermediate trophic levels,

where demand on them was entirely predatory. Groups with demand from commercial or recreational harvest, on the

other hand, were all balanced.

To achieve mass balance, simple changes in diet contributions were made first. The final diet matrix for the

“consumption-based” approach describes the changes made during balancing (Table 2). The diet matrix for the

“production-based” approach was similar and is not shown. Predation on age 0 bloater was nearly entirely (99%) from

benthic prey-fish. Consumption of bloater eggs by slimy sculpin, deepwater sculpin, and stickleback was modeled as

consumption of age 0 bloater. Removing the contribution of age 0 bloater from the diets of these three species

resulted in mass balance for bloater (Table 2). Slimy sculpin had the largest pre-balance value of EE, and thus was the

most unbalanced group. Approximately 80% of its predation mortality was from age 1+ rainbow smelt. Lantry and

Stewart (1993) assumed that smelt consumed slimy sculpin, and although supported by Brandt and Madon (1986),

few other studies report predation on this species (Storch et al., 2007; Walsh et al., 2008). Consequently, slimy

sculpin was removed from the diet of age 1+ smelt (Table 2). This change alone, however, did not result in mass

balance.

Changes to more than just diet were required for balancing some species. Values of B, P/B, and Q/B for all

species are provided for models balanced with both the “consumption-based” (Table 3) and “production-based”

(Table 4) approaches. Slimy sculpin were particularly difficult to balance. Once predation by age 1+ smelt was

removed, the remaining mortality was from steelhead and burbot Lota lota. Moderate changes to slimy sculpin in the

diet of these species were made (Table 2), but slimy sculpin were still unbalanced. Balance was achieved by either

increasing slimy sculpin biomass (Table 3), or reducing consumption by its predators (Table 4). Stickleback and

round gobies also required more than just changes in diet to balance. Most consumption of stickleback (94%) and all

consumption of round goby was by adult lake whitefish. Although contributions of round goby and stickleback to the

diet of lake whitefish were small, the large biomass of lake whitefish resulted in high levels of consumption. Changes

to the diet of lake whitefish lowered the EE of these groups, but, as was the case for slimy sculpin, balance was

2 - 7

achieved only after either increasing biomass of round goby and stickleback (Table 3) or decreasing consumption by

lake whitefish (Table 4). Changes in diet were not used to balance phytoplankton. Zooplankton were responsible for

99% of mortality of phytoplankton, and because they fed nearly exclusively on phytoplankton, balance was achieved

by either increasing phytoplankton production (Table 3) or decreasing zooplankton consumption (Table 4).

Values for the ecosystem metrics showed some difference between the two balancing approaches.

Ascendency was 28884 for the “consumption-based” approach and 7840 for the “production-based” approach. The

nearly four-fold difference in ascendency between the two models was mostly due to the different levels of

production of phytoplankton and zooplankton. For the “consumption-based” and “production-based” balancing

approaches, phytoplankton and zooplankton contributed 56 and 47% of the total ascendency, respectively. Detritus

made up an additional 40% for each approach. Although the scale of the models as estimated by ascendency was

different, the structure was nearly identical; SOI was 0.085 for the “consumption-based” approach and 0.084 for the

“production-based” approach.

Differences in biomass dynamics between the two balancing approaches (index 1) also suggested that

balancing affected model results. The magnitude of the difference depended on the assumed strength of interaction

(vulnerabilities) between predators and prey. Ecosim dynamics in the two balancing approaches were least similar

under high vulnerabilities and most similar under low vulnerabilities (Fig. 3). Although this pattern was maintained

under both types of forcing, it was more pronounced when environmental productivity was doubled than when fishing

mortality was doubled. At most, the difference was 15% (Fig. 3).

Median changes in relative biomass (index 2) followed a similar trend as index 1. Doubling environmental

productivity produced the greatest change in relative biomass, especially when vulnerabilities were high (Fig. 4).

Although index 2 reflected both decreases and increases from initial biomass levels, biomass of nearly all groups

increased when productivity was doubled. Doubling fishing mortality resulted in much smaller changes, but these

were still greatest when vulnerabilities were high (Fig. 4). Index 2 was used to calculate index 3, which expressed the

difference in biomass between the balancing approaches relative to the change in biomass due to other sources of

perturbation. Values for index 3 were at most 41% and were again greatest under high vulnerabilities (Table 5), but in

contrast to other indices, were greatest when fishing mortality was doubled than when environmental productivity was

doubled (Table 5).

Indices were also calculated for each modeled group. The largest values for indices 1 and 2 were observed

when environmental productivity was doubled and vulnerabilities were high. Under these conditions, index 1 was

greatest for dreissenids (7.39) and stickleback (5.31), and followed by slimy sculpin (1.60). Index 2 was greatest for

age 3+ and age 0-3 burbot (21.2 and 21.0, respectively), and followed by dreissenids (12.3). In scenarios with low

vulnerabilities, index 1 was greatest for age 0 alewife when environmental productivity and fishing mortality were

doubled (1.19 and 1.27, respectively) and was followed by age 0.5-1 Chinook salmon (1.06 and 1.12, respectively). In

scenarios with intermediate vulnerabilities, index 1 was greatest for age 1-5 Chinook salmon (1.10) when fishing

mortality was doubled, and stickleback (1.19) when environmental productivity was doubled.

Values for index 3 were greatest when fishing mortality was doubled, but contrary to the pattern in overall

medians, the greatest individual group value occurred under intermediate vulnerabilities. Under intermediate and high

vulnerabilities, respectively, age 0 lake whitefish (203 and 161%) and Diporeia (200 and 170%) had the greatest

values when fishing mortality was doubled. Under low vulnerabilities, age 0 alewife (12.7 and 119%) and age 0.5-1

Chinook salmon (3.3 and 102%) had the greatest values when both environmental productivity and fishing mortality

were doubled, respectively. Stickleback had the greatest values (8.65 and 70%) under intermediate and high

vulnerabilities, respectively, when environmental productivity was doubled.

Biomass dynamics for age 4+ lake whitefish, Diporeia, age 1+ alewife, and sticklebacks illustrate the general

patterns observed among modeled groups (Fig. 5). Lake whitefish represent 80% of commercial harvest in Lake

Huron (Ebener et al., 2008), Diporeia has been their primary prey, and alewife are a major component of Chinook

salmon and lake trout diets. Stickleback was included because of its large values for index 1 and 3. Because indices

were greatest when vulnerabilities were high, biomass dynamics are shown only for these scenarios (Fig. 5).

4. Discussion:

4.1 Effect of balancing approach

The effect of balancing depended on the strength of trophic interactions and the magnitude of biomass change

in the system. Under both types of forcing, all indices were greatest when vulnerabilities were high. Although the

overall difference between balancing approaches (index 1) was smallest when fishing mortality was doubled (Fig. 3),

the effect of balancing when expressed relative to the overall change in biomass (index 3) was greatest (Table 5). This

2 - 8

occurred because changes in relative biomass were small when fishing mortality was doubled (Fig. 4), and therefore

differences between the two balancing approaches were also small in absolute terms (Fig. 3). When expressed relative

to the change in biomass, however, the effect of these differences appeared greater (Fig. 5). In other words, when

small changes to biomass occurred in the system, the effect of changes produced by differences in balancing becomes

more important. Based on these results, we conclude that the way in which our Ecopath model was balanced most

affected model conclusions when vulnerabilities were high and when biomass dynamics in the system were stable.

The significance of our results remains difficult to determine. Although the effect of balancing is different

among the six scenarios, and influenced by whether absolute or relative effects we considered, the question remains;

“Is a 41% effect important?”. Based on our results, the answer to this questions appears to be no. Under scenarios

with greater biomass change (i.e. environmental productivity was doubled) the absolute difference between balancing

approaches was at most 1.15 (Fig. 3). When put into the context of the overall change in biomass, the difference

between balancing approaches was 3.5% (Table 5). Although a 41% effect is greater than a 3.5% effect, the overall

change in biomass when fishing mortality was doubled was at most 1.08 and therefore small in absolute terms (Fig.

4). We understand that our findings reflect a single ecosystem, and that other ecosystems may have greater changes in

biomass, or require more contrast between balancing approaches when modeled. Additional comparisons would be

helpful to determine the significance of these results, and whether the general patterns we indentify hold.

Indices for some individual species (i.e. not averaging over i groups) were larger than those when averaging

over all groups. Age 0 lake whitefish, Diporeia, age 0 alewife, age 0.5-1 Chinook salmon, and dreissenids were often

the groups with the greatest values in various scenarios. Biomass dynamics of age 0 alewife and age 0.5-1 Chinook

salmon were highly oscillatory when vulnerabilities were low. A similar pattern existed for age 0 lake whitefish at all

vulnerabilities when environmental productivity was doubled, and can also be seen for age 4+ lake whitefish (Fig. 5).

A slight shift in phase for oscillating biomass trajectories inflated the difference between the two balancing

approaches, and thus these indices were large. Similarly, as biomass dynamics from either balancing approach neared

zero, the differences between the two approaches increased greatly. This occurred for dreissenids, and to a lesser

extent sticklebacks (Fig. 5), when vulnerabilities were high and environmental productivity was doubled. Although

biomass dynamics for sticklebacks did approach zero briefly, the large effect of balancing was likely because of high

predation mortality by lake whitefish. Under high vulnerabilities, oscillations in lake whitefish biomass likely

amplified oscillations in stickleback biomass.

Ecopath metrics were less informative in determining the effects of balancing on model results than indices

within Ecosim. Of the two Ecopath metrics that were used, ascendency changed the most. Values of ascendency for

both models were within the range (approximately 1000-100000) of those for 41 other Ecopath models (Christensen,

1994), and similar to Lake Superior (5221.3; Kitchell et al., 2000) and Lake Ontario (25630; Halfon and Schito,

1993). Christensen (1994) found that ascendency was highly correlated with total system throughput, which is a

measure of ecosystem size. Because production by lower trophic-level groups was increased in the “consumption-

based” approach, and consumption of higher trophic-level groups was decreased in the “production-based” approach,

the sizes of the models balanced by the two approaches were expected to be different. However, much of the

difference in size was due to differences in biomass of planktonic and detrital groups. Although the sizes of the Lake

Huron models were different, the values for SOI were very similar. This was likely because SOI is a metric for

consumers, and thus does not account for differences in biomass estimates of phytoplankton and detritus, which were

substantial. The values of SOI for Lake Huron were between those for Lake Ontario (0.0633; Halfon and Schito 1993)

and Lake Superior (0.108; Kitchell et al., 2000), and less than two models of the Bay of Quinte in Lake Ontario

(0.114 and 0.147; Koops et al., 2006).

4.2 Effect of vulnerabilities

Vulnerability parameters influenced the effect of balancing on Ecosim results. Values for each index were

greatest under scenarios with high vulnerabilities, and smallest under scenarios with low vulnerabilities. When

vulnerabilities are high, the effect that a predator can have on its prey is greater, and small changes to predator

biomass can cause large changes in the biomass of their prey. It makes sense then that differences in biomass

dynamics between the balanced models would be amplified when vulnerabilities were high. Vulnerabilities are known

to influence results when exploring management scenarios within Ecosim (Cox and Kitchell, 2004), but have not

previously been examined for their influence on balancing.

Vulnerabilities are commonly associated with the type of control in the food web, either top-down, bottom-

up, or a combination of the two. Because high vulnerabilities reflect top-down control, Ecopath models of systems

where predators control overall dynamics would be most affected by the way in which balancing is achieved. There is

2 - 9

no consensus on the type of control operating in Great Lakes food-webs (Bence et al., 2008). Lake trout were

historically the top-piscivore within the Great Lakes and were believed to exert strong top-down control on the food

web (Ryder et al., 1981). Through a combination of increased fishing mortality and predation mortality from the

invasive sea lamprey, lake trout were nearly extirpated from the Great Lakes (Eshenroder et al., 1995). Ongoing

stocking of introduced Pacific salmonines and lake trout has increased the number of species occupying top trophic

levels in several Great Lakes and has raised concerns about levels of consumption (Dobiesz et al., 2005; Jones et al.,

1993). This concern continues today, especially given the high level of wild Chinook salmon recruitment in Lake

Huron (Johnson et al., 2010) and because predation by salmonines was found to be an important factor in alewife

recruitment in Lake Michigan (Madenjian et al., 2005). Evidence of top-down control also exists at lower trophic

levels. Increased predation mortality by predatory zooplankton has been suggested by Bunnell et al. (2011) to have

caused declines in zooplankton abundance in Lake Huron (Barbiero et al., 2009).

Bottom-up control has also been hypothesized as a driver of recent changes in Lake Huron. Dreissenid

mussels, which are hypothesized to remove nutrients from the offshore community and bind them in nearshore areas

(Hecky, 2004) have been implicated in the decline of Diporeia (Nalepa et al., 2007) and consequently the reduction in

lake whitefish growth (Pothoven and Madenjian, 2008). The clearing of the water column by mussels has also been

suggested as a possible driver for large declines in cladoceran zooplankton biomass (Barbiero et al., 2009; 2011).

Significant declines in primary production have been observed in Lake Michigan (Fahnenstiel et al., 2010) and Lake

Erie (Depew et al., 2001), and both have been attributed to filtering of dreissenids. The declines in zooplankton are

implicated in the decline in the abundance of alewife as well as other deepwater prey-fish in Lake Huron (Riley et al.,

2008). Consequently, community dynamics within Lake Huron are likely a combination of top-down and bottom-up

control.

4.3 Model imbalance

A common observation in Ecopath models is that based on initial input values, consumption of intermediate

and lower-trophic groups often exceeds production by those groups, and that various adjustments must be made to

achieve mass balance. The model presented here supports these observations. Stewart and Sprules (2011) discuss

trophic imbalances in an Ecopath model of Lake Ontario and compare this to observations from stream ecology (see

Huryn, 1996). An informal poll of other EwE modelers revealed that they had imbalances in the lower-trophic groups

in their models. Apart from Stewart and Sprules (2011), little has been discussed in the EwE literature about whether

these common observations reveal a larger issue with the modeling framework or with sampling of aquatic

ecosystems in general.

If intermediate and lower trophic-level groups are commonly overconsumed in Ecopath models, it makes

sense to discuss possible reasons for why this should occur. An obvious reason is uncertainty in input parameters. A

critical assumption of sampling is that it is both spatially and temporally representative of the system being modeled.

Biomass estimates of prey fish were taken from area-swept methods based on bottom trawl surveys. These surveys

are best suited for benthopelagic species, and are likely to underestimate biomass of benthic species, such as

deepwater and slimy sculpins, and pelagic species such as rainbow smelt (Riley et al., 2008). Comparisons with

acoustic surveys however, suggest that trawl estimates are generally similar (Roseman and Riley, 2009; Warner et al.,

2009). Biomass estimates of phytoplankton were taken within the top 20 m of the water column and therefore ignored

the contribution in the deep chlorophyll layer, which has been estimated to provide a substantial contribution to

primary production (Barbiero and Tuchman, 2001). Biomass estimates of zooplankton were taken from surveys done

once in August, which likely captured a peak in consumptive demand on phytoplankton (Barbiero et al. 2009).

Increasing biomass of benthic and pelagic fish and phytoplankton (Table 3), or decreasing biomass of zooplankton

(Table 4) addressed some model imbalances as reflected by the two balancing approaches.

Although spatial and temporal constraints on sampling can be addressed by more representative techniques,

there still exist challenges that may contribute to imbalances when combining parameter estimates into a single

model. Often, the offshore zone is less productive than the nearshore, although some studies have found the opposite

(Depew et al., 2001). Estimates taken only from offshore sites ignore the greater contribution of food resources

provided in the nearshore zone. Consequently, resources are available to predators in the system that are not included

in an offshore model. In addition, predators rarely feed within an average area of a system, so predation is likely to be

concentrated in areas were food resources are at their highest. Thus, although a sampler sees an average value of food

availability, a predator may see an above-average value.

2 - 10

4.4 Subjective balancing

We elected to balance the Ecopath model using the subjective balancing procedure, despite the ease at which

balancing can be reproduced by the objective procedure (Kavanagh et al., 2004). Although two different objective

functions could have been used to perform the analyses for this research, we chose to balance the model in ways with

more biological appeal. Initial discussions with Lake Huron biologists about ways to address model imbalance often

focused on increasing production or decreasing consumption, and thus were used as the two approaches. While

subjectively balancing the model, problems in the modeling of diet contributions were found such as the consumption

of bloater eggs being attributed as age 0 bloater, and consumption of slimy sculpin by age 1+ smelt. These balancing

problems could have also been noticed when using the objective procedure, however, because changes can be made

automatically within the software it does not assist in their possible discovery.

4.5 Conclusions

Estimates used to parameterize Ecopath models commonly result in imbalance. The process of balancing an

Ecopath model is rarely described in the literature, especially with respect to its effect on simulations in Ecosim. This

research assessed the effect of contrasting but realistic approaches to balancing on biomass dynamics in Ecosim and

ecosystem metrics in Ecopath. Indices in Ecosim were more informative at assessing differences between balancing

approaches than were metrics in Ecopath. The difference in biomass dynamics between the two balancing approaches

was greatest when control of the ecosystem was top-down, providing further evidence that vulnerabilities are

important parameters in Ecosim. The effect of balancing was also greatest when changes in biomass caused by other

perturbations in the model were small. We encourage EwE modelers to consider the sensitivity of their models to

balancing when dynamics within their system are stable through time or when top-down control is expected, but

otherwise to focus their efforts on understanding the sensitivity of model results to vulnerability parameters rather

than alternative ways to achieve mass balance.

Acknowledgements:

This research was made possible by a grant from the Great Lakes Fishery Commission to MLJ, and with the

assistance of the Government of Canada for BJL to travel to the University of British Columbia (UBC). We’d also

like to thank members of the Fisheries Centre at the UBC for help in understanding the modeling software, members

of the Lake Huron Technical Committee for providing data and assistance during model construction, participants at

stakeholder meetings for assistance in model construction, and members of the Quantitative Fisheries Center (QFC) at

Michigan State University for assistance in model construction.

References:

Allen, K.R., 1971. Relation between production and biomass. Journal of the Fisheries Research Board of Canada 28,

1573-1581.

Barbiero, R.P., Balcer, M., Rockwell, D.C., Tuchman, M.L., 2009. Recent shifts in the crustacean zooplankton

community of Lake Huron. Canadian Journal of Fisheries and Aquatic Sciences 66, 816-828.

Barbiero, R.P., Lesht, B.M., Warren, G.J., 2011. Evidence for bottom–up control of recent shifts in the pelagic food

web of Lake Huron. Journal of Great Lakes Research 37, 78-85.

Barbiero, R.P., Tuchman, M.L., 2001. Results from the U.S. EPA's biological open water surveillance program of the

Laurentian Great Lakes: II. Deep chlorophyll maxima. Journal of Great Lakes Research 27, 155-166.

Bence, J.R., Dobiesz, N.E., Madenjian, C.P., Argyle, R.L., Barbiero, R.P., Bowlby, J.N., Claramunt, R.M.,

O'Gorman, R., Schaner, T., 2008. Top-down effects of open-water salmonine predators in the Great Lakes.

Technical Report, Quantitative Fisheries Center, East Lansing.

Brandt, S.B., Madon, S.P., 1986. Rainbow smelt (Osmerus mordax) predation on slimy sculpin (Cottus cognatus) in

Lake Ontario. Journal of Great Lakes Research 12, 322-325.

Bundy, A., 2001. Fishing on ecosystems: the interplay of fishing and predation in Newfoundland-Labrador. Canadian

Journal of Fisheries and Aquatic Sciences 58, 1153-1167.

Bunnell, D.B., Davis, B.M., Warner, D.M., Chriscinske, M.A., Roseman, E.F., 2011. Planktivory in the changing

Lake Huron zooplankton community: Bythotrephes consumption exceeds that of Mysis and fish. Freshwater

Biology 56, 1281-1296.

Christensen, V., 1994. On the behavior of some proposed goal functions for ecosystem development. Ecological

Modelling 75/76, 37-49.

Christensen, V., 1995. Ecosystem maturity - towards quantification. Ecological modelling 77, 3-32.

2 - 11

Christensen, V., Pauly, D., 1992. ECOPATH II - a software for balancing steady-state ecosystem models and

calculating network characteristics. Ecological Modelling 61, 169-185.

Christensen, V., Walters, C.J., 2004. Ecopath with Ecosim: methods, capabilities and limitations. Ecological

Modelling 172, 109-139.

Christensen, V., Walters, C.J., Pauly, D., 2005. Ecopath with Ecosim: A user's guide. Fisheries Centre, Vancouver.

www.ecopath.org

Cox, S.P., Kitchell, J.F., 2004. Lake Superior ecosystem, 1929-1998: Simulating alternative hypotheses for

recruitment failure of lake herring (Coregonus artedi). Bulletin of Marine Science 74, 671-683.

Depew, D.C., Guildford, S.J., Smith, R.E.H., 2001. Nearshore-offshore comparison of chlorophyll a and

phytoplankton production in the dreissenid-colonized basin of Lake Erie. Canadian Journal of Fisheries and

Aquatic Sciences 63, 1115-1129.

DesJardine, R.L., Gorenflo, T.K., Payne, R.N., Schrouder, J.D., 1995. Fish-community objectives for Lake Huron.

Special Publication 95-1, Great Lakes Fishery Commission, Ann Arbor.

Dobiesz, N.E., McLeish, D.A., Eshenroder, R.L., Bence, J.R., Mohr, L.C., Ebener, M.P., Nalepa, T.F., Woldt, A.P.,

Johnson, J.E., Argyle, R.L., Makarewicz, J.C., 2005. Ecology of the Lake Huron fish community, 1970-

1999. Canadian Journal of Fisheries and Aquatic Sciences 62, 1432-1451.

Ebener, M.P., Mohr, L.C., Riley, S., Roseman, E.F., Fielder, D.G., 2008. Whitefishes and ciscoes, in: Bence, J.R.,

Mohr, L.C. (eds.), The state of Lake Huron in 2004, Great Lakes Fishery Commission Special Publication 08-

01. pp. 37-46.

Eshenroder, R.L., Payne, N.R., Johnson, J.E., Bowen II, C., Ebener, M.P., 1995. Lake trout rehabilitation in Lake

Huron. Journal of Great Lakes Research 21, 108-127.

Essington, T.E., 2007. Evaluation the sensitivity of a trophic mass-balance model (Ecopath) to imprecise data inputs.

Canadian Journal of Fisheries and Aquatic Sciences 64, 628-637.

Fahnenstiel, G., Pothoven, S., Vanderploeg, H., Klarer, D., Nalepa, T., Scavia, D., 2010. Recent changes in primary

production and phytoplankton in the offshore region of southeastern Lake Michigan. Journal of Great Lakes

Research 36, 20-29.

Gaden, M., Krueger, C., Goddard, C., Barnhart, G., 2008. A joint strategic plan for management of great lakes

fisheries: a cooperative regime in a multi-jurisdictional setting. Aquatic Ecosystem Health & Management 11,

50-60.

Halfon, E., Schito, N., 1993. Lake Ontario food web, an energetic mass balance, in: Christensen, V., Pauly, D. (eds.),

Trophic models of aquatic ecosystems, vol. 26, 390 p., pp. 29-39.

Halfon, E., Schito, N., Ulanowicz, R.E., 1996. Energy flow through the Lake Ontario food web: conceptual model

and an attempt at mass balance. Ecological Modelling 86, 1-36.

Hecky, R.E., Smith, R.E.H., Barton, D.R., Guildford, S.J., Taylor, W.D., Charlton, M.N., Howell, T., 2004. The

nearshore phosphorus shunt: a consequence of ecosystem engineering by dreissenids in the Laurentian Great

Lakes. Canadian Journal of Fisheries and Aquatic Sciences 61, 1285-1293.

Hilborn, R., Walters, C.J., 1992. Quantitative fisheries stock assessment: Choice, dynamics, and uncertainty.

Chapman and Hall, New York.

Hollowed, A.B., Bax, N., Beamish, R., Collie, J., Fogarty, M., Livingston, P., Pope, J., Rice, J.C., 2000. Are

multispecies models an improvement on single-species models for measuring fishing impacts on marine

ecosystems? ICES Journal of Marine Science 57, 707-719.

Huryn, A.D., 1996. An appraisal of the Allen paradox in a New Zealand trout stream. Limnology and Oceanography

41, 243-252.

Johnson, J.E., DeWitt, S.P., Gonder, D.J., 2010. Mass-marking reveals emerging self-regulation of the chinook

salmon population in Lake Huron. North American Journal of Fisheries Management 30, 518-529.

Jones, M.L., Koonce, J.F., O’Gorman, R., 1993. Sustainability of hatchery-dependent salmonine fisheries in Lake

Ontario: the conflict between predator demand and prey supply. Transaction of the American Fisheries

Society 122, 1002-1018.

Kavanagh, P., Newlands, N., Christensen, V., Pauly, D., 2004. Automated parameter optimization for Ecopath

ecosystem models. Ecological Modelling 172, 141-149.

Kitchell, J.F., Cox, S.P., Harvey, C.J., Johnson, T.B., Mason, D.M., Schoen, K.K., Aydin, K., Bronte, C., Ebener,

M.P., Hansen, M., Hoff, M., Schram, S., Schreiner, D., Walters, C.J., 2000. Sustainability of the Lake

Superior fish community: Interactions in a food web context. Ecosystems 3, 545-560.

2 - 12

Koops, M.A., Irwin, B.J., MacNeil, J.E., Millard, E.S., Mills, E.L., 2006. Comparative modelling of the cosystem

impacts of exotic invertebrates and productivity changes on fisheries in the Bay of Quinte and Oneida Lake.

Project Completion Report, Great Lakes Fishery Commission, Ann Arbor.

Lantry, B.F., Stewart, D.J., 1993. Ecological energetics of rainbow smelt in the Laurentian great lakes: An interlake

comparison. Transactions of the American Fisheries Society 122, 951-976.

Latour, R.J., Brush, M.J., Bonzek, C.F., 2003. Toward ecosystem-based fisheries management: strategies for

multispecies modeling and associated data requirements. Fisheries 28, 10-22.

Madenjian, C.P., Höök, T.O., Rutherford, E.S., Mason, D.M., Croley II, T.E., Szalai, E.B., Bence, J.R., 2005.

Recruitment variability of alewives in Lake Michigan. Transactions of the American Fisheries Society 134,

218-230.

Marasco, R.J., Goodman, D., Grimes, C.B., Lawson, P.W., Punt, A.E., Quinn II, T.J., 2007. Ecosystem-based

fisheries management: some practical suggestions. Canadian Journal of Fisheries and Aquatic Sciences 64,

928-939.

Nalepa, T.F., Fanslow, D.L., Pothoven, S.A., Foley III, A.J., Lang, G.A., 2007. Long-term trends in benthic

macroinvertebrate populations in Lake Huron over the past four decades. Journal of Great Lakes Research 33,

421-436.

Odum, E.P., 1969. The strategy of ecosystem development. Science 164, 262-270.

Pauly, D., 1980. On the interrelationship between natural mortality, growth parameters, and mean environmental

temperature in 175 fish stocks. Journal du Conseil 39, 175-192.

Pikitch, E.K., Santora, C., Babcock, E.A., Bakun, A., Bonfil, R., Conover, D.O., Dayton, P., Doukakis, P., Fluharty,

D., Heneman, B., Houde, E.D., Link, J., Livingston, P.A., Mangel, M., McAllister, M.K., Pope, J., Sainsbury,

K.J., 2004. Ecosystem-based fishery management. Science 305, 346-347.

Pothoven, S.A., Madenjian, C.P., 2008. Changes in consumption by alewives and lake whitefish after dreissenid

mussel invasions in lakes Michigan and Huron. North American Journal of Fisheries Management 28, 308-

320.

Riley, S.C., Roseman, E.F., Nichols, S.J., O'Brien, T.P., Kiley, C.S., Schaeffer, J.S., 2008. Deepwater demersal fish

community collapse in Lake Huron. Transactions of the American Fisheries Society 137, 1879-1890.

Robinson, J.G., Frid, C.L.J., 2003. Dynamic ecosystem models and the evaluation of ecosystem effects of fishing: can

we make meaningful predictions. Aquatic Conservation: Marine and Freshwater Ecosystems 13, 5-20.

Roseman, E.F., Riley, S.C., 2009. Biomass of deepwater demersal forage fishes in Lake Huron, 1994-2007:

Implications for offshore predators. Aquatic Ecosystem Health & Management 12, 29-36.

Ruckelshaus, M., Klinger, T., Knowlton, N., DeMaster, D.P., 2008. Marine ecosystem-based management in practice:

Scientific and governance challenges. BioScience 58, 53-63.

Ryder, R.A., Kerr, S.R., Taylor, W.W., Larkin, P.A., 1981. Community consequences of fish stock diversity.

Canadian Journal of Fisheries and Aquatic Sciences 38, 1856-1866.

Smith, A.D.M., Fulton, E.J., Hobday, A.J., Smith, D.C., Shoulder, P., 2007. Scientific tools to suppport the practical

implementation of ecosystem-based fisheries management. ICES Journal of Marine Science 64, 633-639.

Stewart, T.J., Sprules, W.G., 2011. Carbon-based balanced trophic structure and flows in the offshore Lake Ontario

food web before (1987-1991) and after (2001-2005) invasion-induced ecosystem change. Ecological

Modelling 222, 692-708.

Storch, A.J., Schulz, K.L., Caceres, C.E., Smyntek, P.M., Dettmers, J.M., Teece, M.A., 2007. Consumption of two

exotic zooplankton by alewife (Alosa pseudoharengus) and rainbow smelt (Osmerus mordax) in three

Laurentian Great Lakes. Canadian Journal of Fisheries and Aquatic Sciences 64, 1314-1328.

Ulanowicz, R.E., 1986. Growth and development: Ecosystems phenomenology. Springler-Verlag, New York.

Walsh, M.G., O'Gorman, R., Strang, T., Edwards, W.H., Rudstam, L.G., 2008. Fall diets of alewife, rainbow smelt,

and slimy sculpin in the profundal zone of southern Lake Ontario during 1994-2005 with an emphasis on

occurrence of Mysis relicta. Aquatic Ecosystem Health & Management 11, 368-376.

Walters, C.J., Christensen, V., 2007. Adding realism to foraging arena predictions of trophic flow rates in Ecosim

ecosystem models: shared foraging arenas and bout feeding. Ecological Modelling 209, 342-350.

Walters, C.J., Martell, S.J.D., 2004. Fisheries Ecology and Management. Princeton University Press, Princeton.

Walters, C., Martell, S.J., Christensen, V., Mahmoudi, B., 2008. An ecosim model for exploring Gulf of Mexico

ecosystem management options: Implications of including multistanza life-history models for policy

predictions. Bulletin of Marine Science 83, 251-271.

2 - 13

Warner, D.M., Schaeffer, J.S., O'Brien, T.P., 2009. The lake huron pelagic fish community: persistent spatial pattern

along biomass and species composition gradients. Canadian Journal of Fisheries and Aquatic Sciences 66,

1199-1215.

Whipple, S.J., Link, J.S., Garrison, L.P., Fogarty, M.J., 2000. Models of predation and fishing mortality in aquatic

ecosystems. Fish and Fisheries 1, 22-40.

2 - 14

Table 1: Names of species (with scientific names) or groups of species that were included in the Ecopath model, and

their corresponding group numbers. Specific age stanzas are listed for groups modeled with age structure.

Group name Scientific Name Group number

Sea lamprey Petromyzon marinus 1

Lake whitefish Corregonus clupeaformis

age 0, age 1-3, age 4+ 2, 3, 4

Lake trout Salvelinus namaycush

age 0, age 1, age 2-4, age 5+ 5, 6, 7, 8

Chinook salmon Oncorhynchus tshawytscha

age 0, age 0.5, age 1-5, age 6+ 9, 10, 11, 12

Steelhead Oncorhynchus mykiss

age 0, age 1, age 2-5, age 6+ 13, 14, 15, 16

Burbot Lota lota

age 0-3, age 3+ 17, 18

Alewife Alosa pseudoharengus

age 0, age 1+ 19, 20

Rainbow smelt Osmerus mordax

age 0, age 1+ 21, 22

Bloater Corregonus hoyi

age 0, age 1+ 23, 24

Round Goby Neogobius melanostomus 25

Slimy sculpin Cottus cognatus 26

Deepwater sculpin Myoxocephalus thompsoni 27

Ninespine Stickleback Pungitius pungitius 28

Diporeia Diporeia hoyi 29

Mysis Mysis diluviana 30

Benthic invertebrates 31

Dreissenid mussels Dreissena polymorpha

Dreissena bugensis

32

Predatory zooplankton Bythotrephes longimanus 33

Zooplankton 34

Phytoplankton 35

Detritus 36

2 - 15

Table 2: Contributions of prey (rows) to the diet of predators (columns) from the model balanced with the “consumption-based” approach. When

changed, initial contributions are provided in parenthesis and bolded. Numbers in the leading row and column correspond to group numbers

provided in Table 1. Groups not preyed on, or predators that fed outside the system are not shown. Predator group numbers

Pre

y g

rou

p n

um

ber

s

1 2 3 4 6 7 8 10 11 14 15 17 18

2 0.004 0.015

4 0.05

6 0.01 (0.029)

8 0.50

11 0.20

15 0.10

18 0.10

19 0.44 0.54 0.16 0.66 (0.19) 0.37 0.15 (0) 0.16

20 0.28 0.70 0.74 0.50 (0.85) 0.26

21 0.50 0.08 0.19 (0.81) 0.03

22 0.04 0.031 (0.012) 0.26 (0.24) 0.32 (0.09) 0.14 0.35

23 0.015

24 0.05 0.015 0.083 0.03

25 0.001 (0.032)

26 0.03 0.009 0.005

0.005

(0.03)

0.04

(0.115)

0.02

(0.06)

27 0.025 0.018 0.011

0.19

(0.115)

0.10

(0.06)

28 0.005 (0.01) 0.005 (0.048) 0.01 0.01 (0.03) 0.004 0.004

29 0.015 (0.01) 0.074 (0)

30 0.02 0.06 0.003

31 0.11 0.36 0.34 0.15 (0) 0.60 0.03 0.40 0.19

32 0.24 0.56 0.07 0.031

33 0.12 0.02

34 0.87 0.20

35

36

2 - 16

Table 2 cont.

Predator group numbers

Pre

y g

rou

p n

um

ber

s

19 20 21 22 24 25 26 27 28 29 30 31 32 33 34

2

4

6

8

11

15

18

19 0.01

20

21 0.05 (0.18)

22

23 0 (0.064) 0 (0.057) 0 (0.02)

24

25

26 0 (0.09)

27 0 (0.064) 0 (0.057)

28

29 0.035 0.09 0.065 0.20 0.095 0.024 0.72 (0.63) 0.37 (0.32) 0.65 0.03

30 0.21 0.31 0.18 0.56 (0.28) 0.40 0.009 0.17 (0.15) 0.62 (0.55) 0.22

31 0.015 0.02 0.015 0.02 0.004 0.034 0.11 0.011 0.026 0.05

32 0.93

33 0.13 0.05 0.006 0.02 (0.017) 0.005

34 0.61 0.53 0.73 0.14 0.50 0.10 0.67 1.0 0.05

35 0.01 0.30

0.30

(0.51) 0.95

36 0.99 0.95

0.70

(0.49)

2 - 17

Table 3: Parameter values from the Ecopath model balanced with the “consumption-based” approach. Proportional

adjustments from initial parameter estimates that were changed are in parentheses and bolded. Names of group

numbers are provided in Table 1.

Group

number

B (g/m2) P/B (/yr) Q/B (/yr) EE Harvest

(g/m2)

1 0.000470 0.860 16.0 0

2 0.0116 2.00 18.4 0.09

3 0.464 0.355 5.09 0

4 0.96 0.500 2.80 0.19 0.090

5 0.000398 0.00100 13.2 0

6 0.00363 0.410 6.91 0.72

7 0.00782 0.330 4.81 0.30 0.001

8 0.0318 0.600 3.35 0.53 0.006

9 0.00167 0.00100 27.1 0

10 0.0101 3.60 14.2 0

11 0.0500 1.40 6.22 0.29 0.019

12 0.000321 1.40 3.48 0

13 0.00166 0.00100 16.5 (0.75) 0

14 0.0122 0.500 9.16 (0.75) 0

15 0.142 0.106 5.50 (0.75) 0.10 0.001

16 0.475 0.106 4.49 (0.75) 0

17 0.00357 0.745 5.02 0

18 0.0647 0.149 2.00 0.08

19 0.263 4.00 33.7 0.31

20 0.667 1.25 13.6 0.88

21 0.0751 2.64 12.9 0.77

22 0.462 1.17 4.60 0.70

23 0.0147 2.33 31.9 0.02

24 0.309 1.02 8.60 0.06 0.004

25 0.0096 (2) 0.640 4.70 0.44

26 0.0113 (2.5) 1.00 (1.18) 7.50 (0.63) 0.78

27 0.117 0.850 (1.42) 7.50 (0.75) 0.19

28 0.0240 (2) 1.77 12.0 0.66

29 14.7 1.43 25.0 0.25

30 3.53 2.80 25.0 0.80

31 3.92 2.50 8.60 0.40

32 11.0 0.315 8.60 0.60

33 0.610 10.0 86.0 0.33

34 67.8 26.3 110 0.28

35 23.8 (3) 365 (1.31) 0.82

36 147 0.08

2 - 18

Table 4: Parameter values from the Ecopath model balanced with the “production-based” approach. Proportional

adjustments from initial parameter estimates that were changed are in parentheses and bolded. Names of group

numbers are provided in Table 1.

Group

number

B (g/m2) P/B (/yr) Q/B (/yr) EE Harvest

(g/m2)

1 0.000470 0.860 16.0 0

2 0.00886 (0.75) 2.00 13.8 (0.75) 0.06

3 0.348 (0.75) 0.355 3.82 (0.75) 0

4 0.72 (0.75) 0.500 2.10 (0.75) 0.25 0.090

5 0.000398 0.00100 13.2 0

6 0.00363 0.410 6.91 0.72

7 0.00782 0.330 4.81 0.30 0.001

8 0.0318 0.600 3.35 0.53 0.006

9 0.00125 (0.75) 0.00100 20.3 (0.75) 0

10 0.00756 (0.75) 3.60 10.6 (0.75) 0

11 0.0375 (0.75) 1.40 4.66 (0.75) 0.39 0.019

12 0.000241 (0.75) 1.40 2.61 (0.75) 0

13 0.000832 (0.5) 0.00100 16.5 (0.75) 0

14 0.00611 (0.5) 0.500 9.16 (0.75) 0

15 0.0710 (0.5) 0.106 5.50 (0.75) 0.20 0.001

16 0.238 (0.5) 0.106 4.49 (0.75) 0

17 0.00179 (0.5) 0.745 5.02 0

18 0.0324 (0.5) 0.149 2.00 0.16

19 0.263 4.00 33.7 0.17

20 0.667 1.25 13.6 0.63

21 0.0751 2.64 12.9 0.70

22 0.462 1.17 4.60 0.22

23 0.0147 2.33 31.9 0.02

24 0.309 1.02 8.60 0.05 0.004

25 0.00480 0.630 4.70 0.50

26 0.00450 1.00 (1.18) 7.50 (0.63) 0.78

27 0.117 0.850 (1.42) 7.50 (0.75) 0.11

28 0.0120 1.77 12.0 0.90

29 14.7 1.43 25.0 0.24

30 3.53 2.80 25.0 0.79

31 3.92 2.50 8.60 0.32

32 11.0 0.315 8.60 0.34

33 0.610 10.0 86.0 0.31

34 16.8 (0.25) 26.3 100 (0.91) 0.48

35 7.94 278 0.75

36 80.6 0.27

2 - 19

Table 5: Values of index 3 for each scenario. Index 3 is the percentage change in biomass due to balancing, averaged

across all groups, relative to the percentage change in biomass due to other sources, averaged across all groups.

Ranges of values for individual groups are provided in parentheses. Details of vulnerabilities and forcing type are

explained in Fig. 2.

Vulnerabilities Forcing type Index 3 (%) (range)

Low Environ 0.059 (0.00-12.7)

Med Environ 1.40 (0.11-8.65)

High Environ 3.46 (0.19-70.0)

Low Fishing 21.1 (0.00-119)

Med Fishing 23.1 (0.00-203)

High Fishing 41.3 (1.79-170)

2 - 20

Fig. 1: Map of Lake Huron with neighboring countries and lake basins labeled.

Fig. 2: Simplified food-web of the offshore community in the main basin of Lake Huron. “Pacific salmon” represent

groups 9-16 in Table 1, “Main prey fish” groups 19-24, “Other prey fish” groups 25-28, “Main inverts” groups 29-30,

“Other inverts” groups 31-32, and “Plankton” groups 33-35.

Fig. 3: Median proportional differences in biomass between the two balancing approaches taken across groups and 40

years of simulation (index 1) for each of six scenarios. Values above the bars reflect the range of proportional

differences over groups. “Low”, “med”, and “high” represent the magnitudes of trophic interactions (vulnerabilities)

used in each scenario (1.01, 2, and 10, respectively). “Environ” and “Fishing” represent scenarios where

environmental productivity was doubled and fishing mortality was doubled, respectively. The vertical dashed lines

separates scenarios between “Environ” and “Fishing”.

Fig. 4: Median change in relative biomass taken across groups, 40 years of simulation, and balancing approach for

each of six scenarios. Values above the bars reflect the range of proportional differences over groups. Details of each

scenario are explained in Fig. 2.

Fig. 5: Relative biomass for 40 years of simulation from the “consumption-based” (solid line) and “production-based”

(dashed line) balancing approaches for age 4+ lake whitefish, Diporeia, age 1+ alewife, and ninespine stickleback,

under high vulnerabilities and environmental (Environ) and fishing (Fishing) forcing types. Note the different scales.

2 - 21

Figure 1:

2 - 22

Figure 2:

Lake trout

(4 groups)

Plankton

(3 groups)

Sea lamprey

Lake whitefish

(3 groups)

Main prey fish

(6 groups)

Burbot

(2 groups)

Pacific salmon

(8 groups)

Other prey fish

(4 groups)

Other inverts

(2 groups)

Main inverts

(2 groups)

Detritus

2 - 23

Figure 3

2 - 24

Figure 4:

2 - 25

Figure 5:

2 - 26

Appendix A

Table A. 1: Sources of data for biomass (B), production to biomass (P/B) and consumption to biomass (Q/B) ratios, diet, and

when appropriate commercial or recreational harvest harvest. Brody growth coeffecients (K) and age of 50% maturity (Amat) are

given for multistanza groups. When applicable, location of data is listed, LH = lake huron, LM = lake Michigan, LS = lake

superior, LE = lake erie, and LO = lake Ontario. RE = estimates taken from published regression relationships. Group numbers

for pre-stocking stanzas for lake trout (5), Chinook (9), and steelhead (13), as well as age 6+ stanzas for Chinook (12) and

steelhead (16), are not included. Names of group numbers are provided in Tables 1.

Group

#

Parameter Data description Years Source

1 B LH number of spawners

Survival of parasite to spawning = 0.75

Average wet weight (WW) of parasitic

phase

1999 Mike Siefkes (Great Lakes Fishery

Commission, pers. comm.)

Jones et al. (2009)

Bergstedt and Swink (1994)

P/B LM ecopath model estimate Ann Krause (University of Toledo-UT,

pers. comm.)

Q/B Great Lakes bioenergetics study Madenjian et al. (2003)

Diet LM ecopath model estimate Ann Krause (UT, pers. comm.)

2 P/B Assumed value

Diet LH spring-summer sampling 2000, 2003 Nalepa et al. (2009)

3 P/B LH Ontario catch-at-age (SCA) models

LH 1836 treaty waters SCA models

1999

1999

Adam Cottrill (Ontario Ministry of

Natural Resources-OMNR, pers.

comm.)

Ebener et al. (2005), Modeling

subcommittee (2009)

Diet LH spring summer sampling 2000, 2003 Nalepa et al. (2009)

Harvest LH Ontario catch-at-age (SCA) models

LH 1836 treaty waters SCA models

1999

1999

Adam Cottrill (OMNR, pers. comm.)

Ebener et al. (2005), Modeling

subcommittee (2009)

4 B LH Ontario catch-at-age (SCA) models

LH 1836 treaty waters SCA models

1999

1999

Adam Cottrill (OMNR, pers. comm.)

Ebener et al. (2005), Modeling

subcommittee (2009)

P/B LH Ontario catch-at-age (SCA) models

LH 1836 treaty waters SCA models

1999

1999

Adam Cottrill (OMNR, pers. comm.)

Ebener et al. (2005), Modeling

subcommittee (2009)

Q/B Bioenergetics on laboratory fish Madenjian et al. (2006a)

Diet LH spring summer sampling 2000, 2003 Nalepa et al. (2009)

Harvest LH Ontario catch-at-age (SCA) models

LH 1836 treaty waters SCA

1999

1999

Adam Cottrill (OMNR, pers. comm.)

Ebener et al. (2005), Modeling

subcommittee (2009)

K LH Canadian index sampling = 0.229 1999-2008 Adam Cottrill (OMNR, pers. comm.)

Amat LH Canadian index sampling = 6 yrs 1999-2008 Adam Cottrill (OMNR, pers. comm.)

6 P/B LH SCA models 1999 Ji He (Michigan Department of Natural

Resources-MDNR, pers. comm.)

Diet LH bioenergetics model 1984-1998 Dobiesz (2003)

7 P/B LH SCA models 1999 Ji He (MDNR, pers. comm.)

Diet LH offshore spring/summer diet 1998-2003 Madenjian et al. (2006b)

Harvest LH SCA models 1999 Ji He (MDNR, pers. comm.)

8 B LH SCA models 1999 Ji He (MDNR, pers. comm.)

P/B LH SCA models 1999 Ji He (MDNR, pers. comm.)

Q/B LH bioenergetics model 1998 Dobiesz (2003)

Diet LH offshore spring/summer diet 1998-2003 Madenjian et al. (2006b)

Harvest LH SCA models 1999 Ji He (MDNR, pers. comm.)

K LH sampling = 0.303 1975-2007 Ji He (MDNR, pers. comm.)

Amat LH SCA models = 7 yrs 1984-2008 Ji He (MDNR, pers. comm.)

10 P/B LH stocking model 1999 Travis Brenden (Michigan State

University–MSU, pers. comm.)

Diet LM spring-fall diet 1973-1982 Jude et al. (1987)

2 - 27

11 B LH stocking model 1999 Travis Brenden (MSU, pers. comm.)

P/B LH stocking model 1999 Travis Brenden (MSU, pers. comm.)

Q/B LH bioenergetics model 1998 Dobiesz (2003)

Diet LH spring-fall diets 1984-1998 Dobiesz (2003)

Harvest LH US recreational and tribal commercial

harvest estimates

1999 Jim Johnson (MDNR, pers. comm.)

K LH stocking model = 0.37 Travis Brenden (MSU, pers. comm.)

Amat Assumed value = 3 yrs

14 P/B LH stocking model 1999 Travis Brenden (MSU, pers. comm.)

Diet LM spring-fall diet 1973-1982 Jude et al. (1987)

15 B LH stocking model 1999 Travis Brenden (MSU, pers. comm.)

P/B LH stocking model 1999 Travis Brenden (MSU, pers. comm.)

Q/B LM bioenergetics study 1978-1988 Stewart and Ibarra (1991)

Diet LM spring-fall diet 1973-1982 Jude et al. (1987)

Harvest LH US recreational estimates 1999 Jim Johnson (MDNR, pers. comm.)

K LH stocking model = 0.466 Travis Brenden (MSU, pers. comm.)

Amat Assumed value = 3 yrs

17 P/B LH bioenergetics model 1998 Dobiesz (2003)

Diet LH bioenergetics model 1984-1998 Dobiesz (2003)

18 B LH bioenergetics model 1998 Dobiesz (2003)

P/B LH bioenergetics model 1998 Dobiesz (2003)

Q/B LH bioenergetics model 1998 Dobiesz (2003)

Diet LH bioenergetics model 1984-1998 Dobiesz (2003)

K LH sampling = 0.238 1987-2008 Ji He (MDNR, pers. comm.)

Amat Assumed value = 3 yrs

19 P/B LM bioenergetics model 1987 Rand et al. (1995)

Diet LM spring-fall diet study 1998-2004 Pothoven and Madenjian (2008)

20 B LH bottom trawl estimates with fishing

power correction (FPC)

1999 provided by S.R.

P/B LM bioenergetics model 1987 Rand et al. (1995)

Q/B LM bioenergetics model 1987 Rand et al. (1995)

Diet LM spring-fall diet study 1998-2004 Pothoven and Madenjian (2008)

K LH index survey = 0.625 1999-2004 Adam Cottrill (OMNR, pers. comm.)

Amat Assumed value = 2 yrs

21 P/B LH bioenergetics model late 1980s Lantry and Stewart (1993)

Diet LH bioenergetics model late 1980s Lantry and Stewart (1993)

22 B LH bottom trawl estimates with FPC 1999 provided by S.R.

P/B LH bioenergetics model late 1980s Lantry and Stewart (1993)

Q/B LH bioenergetics model late 1980s Lantry and Stewart (1993)

Diet LH bioenergetics model late 1980s Lantry and Stewart (1993)

K LH bioenergetics model = 0.477 late 1980s Lantry and Stewart (1993)

Amat Assumed value = 2 yrs

23 P/B LM bioenergetics model 1987 Rand et al. (1995)

Diet LM fall diet study 1979-1980 Crowder and Crawford (1984)

24 B LH bottom trawl estimates with FPC 1999 provided by S.R.

P/B LM bioenergetics model 1987 Rand et al. (1995)

Q/B LM bioenergetics model 1987 Rand et al. (1995)

Diet LM august sampling 1995-1996 TeWinkel and Fleischer (1999)

Harvest LH fishery harvests 1999 Baldwin et al. (2002)

K LH Canadian index sampling = 0.147 1999-2008 Adam Cottrill (OMNR, pers. comm.)

Amat Assumed value = 2 yrs

25 B LH bottom trawl estimates with FPC 1999 provided by S.R.

P/B RE estimates

Lmax = 11.8 from Detroit River

K = 0.4 from Detroit River

T = 6 from LH SCA model for group 8

1996

1996

Pauly (1980)

MacInnis and Corkum (2000)

Fishbase (www.fishbase.org)

Ji He (MDNR, pers. comm.)

Q/B LE bioenergetics study 2000-2001 Lee and Johnson (2005)

2 - 28

Diet LH fall diet study by number

WW for zooplankton

WW for benthic invertebrates

WW for dreissenids

WW for diporeia

WW for mysis

WW for bythotrephes

2000-2001 Schaeffer et al. (2005)

Hawkins and Evans (1979)

Nalepa and Quigley (1980), Nalepa et

al. (2002)

Mills et al. (1999)

Landrum (1988)

Sell (1982)

Barbiero and Tuchman (2004),

Johannsson et al. (2000)

26 B LH bottom trawl estimates with FPC 1999 provided by S.R.

P/B LS ecopath model Kitchell et al. (2000)

Q/B LS ecopath model Kitchell et al. (2000)

Diet LM diet study in dry weights (DW)

DW:WW for Diporeia = 0.2

DW:WW for Mysis = 0.15

DW:WW for other (chironomids) = 0.14

DW:WW for fish eggs (standard

zoobenthos) = 0.166

2000-2001 Hondorp et al. (2005)

Landrum (1988), Johnson (1988)

Landrum et al. (1992)

Smit et al. (1993)

Jørgensen (1979)

27 B LH bottom trawl estimates with FPC 1999 provided by S.R.

P/B LS ecopath model Kitchell et al. (2000)

Q/B LS ecopath model Kitchell et al. (2000)

Diet LM diet study Hondorp et al. (2005)

28 B LH bottom trawl estimates 1999 provided by S.R.

P/B RE with

Lmax = 7.6 from Canada

K = 1.6 from England

T = 6 from LH SCA model for group 8

Pauly (1980)

Fishbase (www.fishbase.org)

Fishbase (www.fishbase.org)

Ji He (MDNR, pers. comm.)

Q/B Assumed the same as group 26

Diet LS diet study 1968-1969 Griswold and Smith (1973)

29 B LH sampling study in numbers

Average WW per individual

1999 Richard Barbiero (Environmental

Protection Agency-EPA, pers. comm.)

Landrum (1988)

P/B LH sampling study in profundal zone 1980-1982 Johnson (1988)

Q/B LS ecopath model Kitchell et al. (2000)

Diet LM spring-fall sampling 1986-1987 Evans et al. (1990)

30 B LH sampling study 1971 Sell (1982)

P/B LH sampling study 1971 Sell (1982)

Q/B LS ecopath model Kitchell et al. (2000)

Diet LO diet study 1995 Johannsson (2001)

31 B LH numbers

Conversions to WW

2000

1987-1996

Nalepa et al. (2007)

Nalepa et al. (2002)

P/B LO study 1967-1968 Johnson and Brinkhurst (1971)

Q/B LM ecopath model Ann Krause (UT, pers. comm.)

Diet LM ecopath model Ann Krause (UT, pers. comm.)

32 B LH numbers

LO WW

LE conversion to shell free WW

2000

1995

1993-1994

Nalepa et al. (2007)

Mills et al. (1999)

Johannsson et al. (2000)

P/B LO study 1967-1968 Johnson and Brinkhurst (1971)

Q/B LO network model Jaeger (2006)

Diet LM ecopath model Ann Krause (UT, pers. comm.)

33 B LH sampling in DW

DW:WW ratio = 0.1

Average depth = 76 m

1999 Richard Barbiero (EPA, pers. comm.)

Richard Barbiero (EPA, pers. comm.)

Barbiero et al. (2001)

P/B LE RE

LH sampling for average length

LH surface temperature data

1993-1994

1983-1999

Johannsson et al. (2000)

Barbiero and Tuchman (2004)

www.ndbc.noaa.gov

Q/B LM ecopath model Ann Krause (UT, pers. comm.)

Diet LH summer diet study 1988 Vanderploeg et al. (1993)

2 - 29

34 B LH sampling in DW

DW:WW ratio = 0.1

Average depth = 76 m

1999 Richard Barbiero (EPA, pers. comm.)

Richard Barbiero (EPA, pers. comm.)

Barbiero et al. (2001)

P/B LE RE

RE

LH surface temperature data

1993-1994 Johannsson et al. (2000)

Shuter and Ing (1997)

www.ndbc.noaa.gov

Q/B LM ecopath model Ann Krause (UT, pers. comm.)

Diet LM ecopath model Ann Krause (UT, pers. comm.)

35 B LH sampling

Depth of sampling = 20 m

1999

Richard Barbiero (EPA, pers. comm.)

Richard Barbiero (EPA, pers. comm.)

P/B LM ecopath model Ann Krause (UT, pers. comm.)

36 B RE

LH spring euphotic depth = 26

WW:Carbon (C) for phytoplankton = 42

DW:C for detritus = 2.22

DW:WW for detritus = 0.08 (assuming

same as for phytoplankton)

1993-1995

Pauly et al. (1993)

Fahnenstiel et al. (2000)

Cushing (1958)

Jørgensen (1979)

Cushing (1958)

Table A.1 references: Baldwin, N.A., Saalfeld, R.W., Dochoda, M.R., Buettner, H.J., Eshenroder, R.L., 2009. Commercial Fish Production in the Great

Lakes 1867-2006. http://www.glfc.org/databases/commercial/commerc.php.

Barbiero, R.P., Little, R.E., Tuchman, M.L., 2001. Results from the U.S. EPA's biological open water surveillance program of the

Laurentian great lakes: III. crustacean zooplankton. Journal of Great Lakes Research 27, 167-184.

Barbiero, R.P., Tuchman, M.L., 2004. Changes in the crustacean communities of lakes Michigan, Huron, and Erie following the

invasion of the predatory cladoceran Bythotrephes longimanus. Canadian Journal of Fisheries and Aquatic Sciences 61,

2111-2125.

Bergstedt, R.A., Swink, W.D., 1994. Seasonal growth and duration of the parasitic life stage of the landlocked sea lamprey

(Petromyzon marinus). Canadian Journal of Fisheries and Aquatic Sciences 52, 1257-1264.

Crowder, L.B., Crawford, H.L., 1984. Ecological shifts in resource use by bloaters in Lake Michigan. Transactions of the

American Fisheries Society 113, 694-700.

Cushing, D.H., Humphrey, G.F., Banse, K., Laevastu, T., 1958. Report of the committe on terms and equivalents. Rapport et

Proces-Verbaus Des Reunions Conseil Permanent International pour l'Exploration de la Mer 144, 15-16.

Dobiesz, N.E., 2003. An evaluation of the role of top piscivores in the fish community of the main basin of Lake Huron, Ph.D.

Dissertation, Michigan State University, Department of Fisheries and Wildlife, East Lansing, Michigan

Ebener, M.P., Bence, J.R., Newman, K.R., Schneeberger, P.J., 2005. Application of statistical catch-at-age models to assess lake

whitefish stocks in the 1836 treaty-ceded waters of the upper great lakes. In: Mohr, L.C., Nalepa, T.F. (Eds.),

Proceedings of a workshop on the dynamics of lake whitefish (Coregonus clupeaformis) and the amphipod Diporeia spp.

in the Great Lakes. Great Lakes Fishery Commission Technical Report 66, 271-309.

Evans, M.S., Quigley, M.A., Wojcik, J.A., 1990. Comparative ecology of Pontoporeia hoyi populations in southern Lake

Michigan: The profundal regions versus the slope and shelf regions. Journal of Great Lakes Research 16, 27-40.

Fahnenstiel, G.L., Stone, R.A., McCormick, M.J., Schelske, C.L., Lohrenz, S.E., 2000. Spring isothermal mixing in the Great

Lakes: Evidence of nutrient limitation and nutrient-light interactions in a suboptimal light environment. Canadian

Journal of Fisheries and Aquatic Sciences 57, 1901-1910.

Griswold, B.L., Smith Jr., L.L., 1973. The life history and trophic relationship of the ninespine stickleback Pungitius pungitius, in

the Apostle Islands area of Lake Superior. Fishery Bulletin 71, 1039-1060.

Hawkins, B.E., Evans, M.S., 1979. Seasonal cycles of zooplankton biomass in southeastern Lake Michigan. Journal of Great

Lakes Research 5, 256-263.

Hondorp, D.W., Pothoven, S.A., Brandt, S.B., 2005. Influence of Diporeia density on diet conposition, relative abundance, and

energy density of planktivorous fishes in southeast Lake Michigan. Transactions of the American Fisheries Society 134,

588-601.

Jaeger, A.L., 2006. Invasive species impacts on ecosystem structure and function, M.Sc. Thesis, Michigan State University,

Department of Fisheries and Wildlife, East Lansing, Michigan.

Johannsson, O.E., Dermott, R.M., Graham, D.M., Dahl, J.A., Millard, S., Myles, D.D., LeBlanc, J., 2000. Benthic and pelagic

secondary production in Lake Erie after the invasion of Dreissena spp. with implications for fish production. Journal of

Great Lakes Research 26, 31-54.

Johannsson, O.E., Leggett, M.F., Rudstam, L.G., Servos, M.R., Mohammadian, M.A., Gal, G., Dermott, R.M., Hesslein, R.H.,

2001. Diet of Mysis relicta in Lake Ontario as revealed by stable isotope and gut content analysis. Canadian Journal of

Fisheries and Aquatic Sciences 58, 1975-1986.

2 - 30

Johnson, M.G., 1988. Production by the amphipod Pontoporeia hoyi in South Bay, Lake Huron. Canadian Journal of Fisheries

and Aquatic Sciences 45, 617-624.

Johnson, M.G., Brinkhurst, R.O., 1971. Production of benthic macroinvertebrates of Bay of Quinte and Lake Ontario. Journal of

the Fisheries Reserach Board of Canada 28, 1699-1714.

Jones, M.L., Irwin, B.J., Hansen, G.J.A., Dawson, H.A., Treble, A.J., Liu, W., Dai, W., Bence, J.R., 2009. An operating model

for the integrated pest management of Great Lakes sea lampreys. The Open Fish Science Journal 2, 59-73.

Jørgensen, S. E. 1979. Handbook of Environmental Data and Ecological Parameters. Pergamon Press, New York.

Jude, D.J., Tesar, F.J., Deboe, S.F., Miller, T.J., 1987. Diet and selection of major prey species by Lake Michigan salmonines,

1973-1982. Transactions of the American Fisheries Society 116, 677-691.

Kitchell, J.F., Cox, S.P., Harvey, C.J., Johnson, T.B., Mason, D.M., Schoen, K.K., Aydin, K., Bronte, C., Ebener, M.P., Hansen,

M., Hoff, M., Schram, S., Schreiner, D., Walters, C.J., 2000. Sustainability of the Lake Superior fish community:

Interactions in a food web context. Ecosystems 3, 545-560.

Landrum, P.F., 1988. Toxicokinetics of organic zenobiotics in the amphipod, Pontoporeia hoyi: role of physiological and

environmental variables. Aquatic Toxicology 12, 245-271.

Landrum, P.F., Frez, W.A., Simmons, M.S., 1992. Relationship of toxicokinetic parameters to respiration rates in Mysis relicta.

Journal of Great Lakes Research 18, 331-339.

Lantry, B.F., Stewart, D.J., 1993. Ecological energetics of rainbow smelt in the Laurentian great lakes: An interlake comparison.

Transactions of the American Fisheries Society 122, 951-976.

Lee, V.A., Johnson, T.B., 2005. Development of a bioenergetics model for the round goby (Neogobius melanostomus). Journal of

Great Lakes Research 31, 125-134.

MacInnis, A.J., Corkum, L.D., 2000. Age and growth of round goby Neogobius melanostomus in the upper Detroit river.

Transactions of the American Fisheries Society 129, 852-858.

Madenjian, C.P., Cochran, P.A., Bergstedt, R.A., 2003. Seasonal patterns in growth, blood consumption, and effects on hosts by

parasitic-phase sea lampreys in the Great Lakes: An individual-based model approach. Journal of Great Lakes Research

29, 332-346.

Madenjian, C.P., Holuszko, J.D., Desorcie, T.J., 2006a. Spring-summer diet of lake trout on six fathom bank and yankee reef in

Lake Huron. Journal of Great Lakes Research 32, 200-208.

Madenjian, C.P., O'Connor, D.V., Pothoven, S.A., Schneeberger, P.J., Rediske, R.R., O'Keefe, J.P., Bergstedt, R.A., Argyle, R.L.,

Brandt, S.B., 2006b. Evaluation of a Lake Whitefish Bioenergetics Model. Transactions of the American Fisheries

Society 135, 61-75.

Mills, E.L., Chrisman, J.R., Baldwin, B., Owens, R.W., O'Gorman, R., Howell, T., Roseman, E.F., Raths, M.K., 1999. Changes in

the dreissenid community in the lower Great Lakes with emphasis on southern Lake Ontario. Journal of Great Lakes

Research 25, 187-197.

Modeling Subcommittee, Technical Fisheries Committee, 2009. Technical fisheries committee administrative report 2008: Status

of lake trout and lake ehitefish populations in the 1836 treaty-ceded waters of lakes Superior, Huron and Michigan, with

recommended yield and effort levels for 2008. http://www.michigan.gov/documents/dnr/2008-status-

report_288026_7.pdf

Nalepa, T.F., Fanslow, D.L., Lansing, M.B., Lang, G.A., Ford, M., Gostenik, G., Hartson, D.J., 2002. Abundance, biomass, and

species composition of benthic macroinvertebrate populations in Saginaw Bay, Lake Huron, 1987-96. NOAA Technical

Memorandum 122, Great Lakes Environmental Research Laboratory, Ann Arbor.

Nalepa, T.F., Fanslow, D.L., Pothoven, S.A., Foley III, A.J., Lang, G.A., 2007. Long-term trends in benthic macroinvertebrate

populations in Lake Huron over the past four decades. Journal of Great Lakes Research 33, 421-436.

Nalepa, T.F., Pothoven, S.A., Fanslow, D.L., 2009. Recent changes in benthic macroinvertebrate populations in Lake Huron and

impact on the diet of lake whitefish (Coregonus clupeaformis). Aquatic Ecosystem Health & Management 12, 2-10.

Nalepa, T.F., Quigley, M.A., 1980. The macro-and meiobenthos of southeastern Lake Michigan near the mouth of the Grand

River, 1976-1977. NOAA Data Report ERL GLERL 17, Great Lakes Environmental Research Laboratory, Ann Arbor.

Pauly, D., 1980. On the interrelationship between natural mortality, growth parameters, and mean environmental temperature in

175 fish stocks. J Cons int Explor Mer 39, 175-192.

Pauly, D., Soriano-Bartz, M.L., Palomares, M.L.D., 1993. Improved construction, parameterization and interpretation of steady-

state ecosystem models, in: Christensen, V., Pauly, D. (eds.), Trophic models of aquatic ecosystems, vol. 26, 390 p., pp.

1-13.

Pothoven, S.A., Madenjian, C.P., 2008. Changes in consumption by alewives and lake whitefish after dreissenid mussel invasions

in lakes Michigan and Huron. North American Journal of Fisheries Management 28, 308-320.

Rand, P.S., Stewart, D.J., Lantry, B.F., Rudstam, L.G., Johannsson, O.E., Goyke, A.P., Brarndt, S.B., O'Gorman, R., Eck, G.W.,

1995. Effect of lake-wide planktivory by the pelagic prey fish community in Lakes Michigan and Ontario. Canadian

Journal of Fisheries and Aquatic Sciences 52, 1546-1563.

Schaeffer, J.S., Bowen, A., Thomas, M., French III, J.R.P., Curtis, G.L., 2005. Invasion history, proliferation, and offshore diet of

the round goby Neogobius melanostomus in western Lake Huron, USA. Journal of Great Lakes Research 31, 414-425.

2 - 31

Sell, D.W., 1982. Size-frequency estimates of secondary production by Mysis relicta in Lakes Michigan and Huron.

Hydrobiologia 93, 69-78.

Shuter, B.J., Ing, K.K., 1997. Factors affecting the production of zooplankton in lakes. Canadian Journal of Fisheries and Aquatic

Sciences 54, 359-377.

Smit, H., van Heel, E.D., Wiersma, S., 1993. Biovolume as a tool in biomass determination of Oligochaeta and Chironomidae.

Freshwater Biology 29, 37-46.

Stewart, D.J., Ibarra, M., 1991. Predation and production by salmonine fishes in Lake Michigan 1978-88. Canadian Journal of

Fisheries and Aquatic Sciences 48, 909-922.

TeWinkel, L.M., Fleischer, G.W., 1999. Vertical migration and nighttime distribution of adult bloaters in

Lake Michigan. Transactions of the American Fisheries Society 128, 459-474

3 - 1

Appendix 3: Modeling species invasions in an Ecopath with Ecosim model of a Laurentian Great Lake, Lake Huron.1

Introduction:

Non-native species are a continual threat to the maintenance of Great Lakes ecosystems. Approximately 185

non-native species have invaded the Laurentian Great Lakes, and of these 10% have caused considerable ecological

change (Environment Canada and EPA 2009). Not all non-native species cause ecological change; effects range from

negligible, when the species fails to proliferate, to severe, when the species becomes established and alters ecosystem

processes (Williamson and Fitter 1996). Non-native species that become established and alter system processes are

called “invasive species” and will be referred to as such throughout this paper.

Over the past three decades, several non-native species from the Ponto-Caspian region of Eurasia have

invaded the Great Lakes. Those considered to be invasive include the spiny-water flea (Bythotrephes longimanus),

zebra (Dreissena polymorpha) and quagga (Dreissena bugensis) mussels, and round goby (Neogobius melanostomas).

The spiny water flea invaded the Great Lakes during the 1980s (Vanderploeg et al. 2002), and preys heavily on large

zooplankton (Bunnell et al. 2011). Zebra and quagga mussels were first observed in Lake St. Clair in 1988

(Vanderploeg et al. 2002), and have been suggested to contribute to shifts in the zooplankton community (Barbiero et

al. 2009; Bunnell et al. 2011), declines in prey fish abundances (Riley et al. 2008), reductions in the spring

phytoplankton bloom (Barbiero and Tuchman 2004), and declines in abundances of the native benthic amphipod

(Diporeia) (Nalepa et al. 2007). Round goby were first observed in the St. Clair River in 1990 (Vanderploeg et al.

2002) and compete with other benthic fish, predominantly sculpins.

Ecosystem-based approaches to fisheries management are becoming more prevalent in the management of

aquatic resources (Pikitch et al. 2004). Such approaches take into account objectives for multiple species within an

ecosystem, and are useful for exploring possible management scenarios. A popular tool in assessing multi-species

objectives in fisheries management is the Ecopath with Ecosim (EwE) computer software (Robinson and Frid 2003).

Ecopath with Ecosim requires a mass-balance description of a food-web at a single point in time (Ecopath) which is

then used as the basis for running time dynamic simulations of the future (Ecosim; Christensen and Walters 2004).

Published EwE models in the Great Lakes exist for Lake Superior (Kitchell et al., 2000; Cox and Kitchell, 2004) and

Lake Ontario (Halfon and Schito 1993; Halfon et al. 1996; Stewart and Sprules 2011), and models are currently being

developed for the other three lakes.

Ecopath with Ecosim models can help to assess the implications of recent changes in Great Lakes food-webs

for management. Invasive species have played prominent roles in shaping the ecosystems in the Great Lakes and need

to be incorporated into EwE models. Properly simulating the dynamics of invasions is not straightforward. All

modeled species, including invasive species, must have positive biomass when the model is initialized. This presents

a problem for modeling invasive species that have not yet entered the system by the year the model is initialized. One

solution to account for the effect of invasive species on a food-web has been to use Ecopath to construct two models

that describe a system before and after species invasion (Jaeger 2006; Stewart and Sprules 2011). Such an approach

can only be done with Ecopath, and precludes the use of dynamic simulations. A second approach is to initialize an

Ecopath model in a year after all invasive species are present, and therefore have positive biomass values. However,

because EwE fits time-series of data to tune model parameters, initializing Ecopath in a later year would sacrifice

information about species dynamics prior to the invasion that can contribute to the model-fitting process.

Methods for simulating the invasion process in EwE models without reducing the length of time series

available for model fitting have been proposed. These methods initialize invasive species at some positive biomass

value prior to actual invasion, artificially maintain the invasive at negligibly low biomasses until the year in which it

invades, and then afterwards allow the species to proliferate. Pine et al. (2007) simulated the invasion of catfish into

an inland reservoir by artificially increasing its mortality through fishing prior to the time of invasion, and then

reducing fishing mortality. Similarly, Espinosa-Romero et al. (2011) kept sea otter levels low, prior to introduction,

through an artificial cull. Forcing biomass of invasive species (i.e., specifying a time series rather than dynamically

modeling it) has also been successfully employed to model lionfish invasion in the Caribbean (V. Christensen,

Fisheries Centre, University of British Columbia, pers. comm.). In addition to adjusting biomass, Cox and Kitchell

(2004) also adjusted diets and mortalities to properly account for the effect of invasive rainbow smelt in Lake

Superior. Such methods are ad hoc but represent practical attempts to account for the effect of invasive species.

1 Substantial contributions to the ideas in this appendix were provided by Mark Rogers, United States Geological Survey,

Sandusky, OH, and Hongyan Zhang, Cooperative Institute for Limnology and Ecosystem Research, Ann Arbor, MI.

3 - 2

Different methods of accounting for invasive species have not been compared nor summarized. In this paper,

four methods of incorporating invasive species into an EwE model of Lake Huron, one of the Laurentian Great Lakes,

are discussed. These methods included 1) forcing invasive species biomass to observed levels, 2) starting invasive

species biomass at very low levels and allowing them to increase, 3) starting invasive species biomass at recent (high)

levels and artificially removing them until the time at which they invade, and 4) adjusting vulnerabilities of invasive

species over time to match biomass dynamics. We present the advantages and disadvantages of each method, based

on subjective testing and exploration of the models, and provide a quantitative comparison as to which method best

reproduced observed time series of biomasses. Rather than providing a rule that all other EwE modelers should use,

we instead hope to provide guidance about the strengths and weaknesses of our four methods. Future EwE modelers

can then use these strengths and weaknesses to make decisions about how best to include invasive species in EwE

models for their system.

Methods:

Models

Data inputs taken primarily from Lake Huron were used to construct an Ecopath models. The model was

parameterized from data collected around 1981 for 20 unique species or groups of species. When data were available,

multiple age-stanzas were included for biologically important species as well as those targeted by fisheries, which

increased the total number of modeled groups to 36 (Table 1). This model focused on the offshore fish community, so

important nearshore groups were excluded. Five fisheries were included in the model: trap nets and gill nets for the

lake whitefish commercial fisheries in Canadian and US waters; combined trap nets and gill nets for the lake

whitefish commercial fishery in tribal waters; the commercial fishery for bloater (Coregonus hoyi) in Canadian

waters; and the recreational fishery for Pacific salmonines and lake trout (Salvelinus namaycush) throughout the lake.

Sea lamprey (Petromyzon marinus) biomass was forced at observed levels because much of its biological control

occurs outside the model in the form of chemical treatments. Additionally, Chinook and steelhead salmon, and lake

trout are stocked in the model and thus recruitment dynamics were driven by stocking. Bythotrephes were assumed to

enter into Lake Huron in 1984 (Makarewicz 1988), round goby in 1997 (S. Riley, United State Geological Survey,

Ann Arbor, unpublished data), and Dreissenids in 1997. Dreissenids were established prior to 1997, but data were

available beginning in 2000 (Nalepa et al. 2007, French et al. 2009) and thus the same year of “invasion” as round

goby was chosen.

Estimates of biomass (B), production to biomass (P/B), consumption to biomass (Q/B), diet components

(DC), harvest, and biomass accumulation (BA) from the literature and previously published models were used to

define interactions among modeled groups and fisheries within Ecopath (Christensen and Pauly, 1992). Important

parameters governing these interactions which are calculated within Ecopath are search rates (aji) of predator j on

group i, and other mortality (M0i) of group i. Other mortality is the mortality not explained by modeled sources, and is

the difference between total mortality entered into the model (P/B) and the sum of predation mortality, fishing

mortality, and BA. When M0 was negative, the model was unbalanced, and data inputs were adjusted following

recommended practices (Christensen et al. 2005, see also Appendix 2).

After species interactions were defined in Ecopath, Ecosim was used to calculate biomass estimates through

time. Descriptions of the calculations are outlined in Christensen and Walters (2004). Perhaps the most important

parameters governing interactions in Ecosim are vulnerabilities. Vulnerabilities control the strength of the trophic

interaction between a predator and its prey. Although unique vulnerabilities for each predator-prey interaction are

possible, it is recommended that a single vulnerability be used for all prey of a single predator. Vulnerabilities are

difficult to estimate in the field, however general ranges can be specified based on the assumed direction of control in

the system (top-down versus bottom-up) or the observed variability in biomass of a group (where large variability

would suggest high vulnerabilities) (Christensen et al. 2005; Ahrens et al. 2012). Alternatively, Ecosim has a fitting

procedure to estimate vulnerability parameters so that modeled biomass dynamics match time series of user-provided

biomass data as close as possible, and which is recommended (Ahrens et al. 2012). Fitting to time series also provides

a way to estimate the level of relative productivity (measured as the P/B of phytoplankton) in the system for each

year, called “production anomalies”. These anomalies often reduce the deviations in fit by about a third from

estimating vulnerabilities alone (C.J. Walters, Fisheries Centre, University of British Columbia, pers. comm.).

Quantitative criteria for assessing performance

The fits of predicted Ecosim dynamics to observed time series provided objective criteria to compare the four

methods of incorporating invasive species in EwE models. Ecosim calculates residual sum of squares (RSS) between

3 - 3

the Ecosim calculated biomass dynamics and the observed data inputs, calculated on a log scale. Data inputs can be

relative or absolute values. Because our four methods utilized different starting values of biomass for the invasive

species, absolute biomass time series of invasive species were used. Time series of biomasses, fishing mortalities,

stocking levels, growth, and total mortality for some modeled groups were available from published and unpublished

sources and ranged from 1981-2008 (Table 2). Unique vulnerability parameters for every modeled predator were

estimated as well as yearly production anomalies.

Methods of incorporating invasive species:

Overview:

Four methods of incorporating invasive species into EwE models were compared. The methods differed in

terms of whether or not the observed invasive species biomass time series was used to fit the model, assumed initial

values of biomass, and the method used to “release” the invasive species and allow its biomass to increase at the time

of invasion (Table 3). A low initial biomass is arguably more representative of the pre-invasion state, whereas high

initial biomasses facilitate the biomass reaching recent (post-invasion) levels in the later years of the time series.

Forcing the biomass time series for an invasive species (i.e., not requiring the model to predict values that fit the data)

allows model fitting to be focused on non-invasive groups; alternatively the biomass of the invasive species could be

fit along with all other groups. When the modeled invasive species biomass was fit to data, the model was adjusted to

mimic the observed invasion dynamics (low before invasive and increasing after invasion) by artificially increasing

mortality through fishing before invasion, and then removing (“releasing”) fishing after invasion, or by changing the

strength of predator-prey interactions (i.e., vulnerabilities).

Method 1 – Forcing biomass

The simplest way in which invasive species were included in the EwE model was through biomass forcing.

Time series of invasive species biomass were used to overwrite simulated values based on the Ecosim equations.

Time series of all other species groups were used to estimate vulnerability parameters and production anomalies.

Biomass time series of round gobies, Dreissenids, and Bythotrephes were available. Time series for Dreissenids and

Bythotrephes were incomplete, and did not extend all the way to early the years of invasion (Table 2). For years

before data availability, but after species invasion, biomasses were assumed to be zero. Small gaps in the time series

were not a problem because Ecosim estimates biomass values at each time step, and therefore gaps were filled in with

Ecosim estimates.

Using time series of biomass to force dynamics in Ecosim did not solve the issue of initial biomass estimates

in Ecopath, which should be zero because the model was initialized before species invaded. The way in which Ecosim

fills in gaps in the time series was found to be affected by the choice of initial biomass estimate. Low initial biomass

estimates resulted in Ecosim estimating lower biomasses than observed in the time series. On the other hand, high

initial biomass estimates resulted in Ecosim estimates that more closely matched the observed biomasses for years

later in the time series. Thus we chose to use a high initial biomass for this method, as described in detail for method

3.

Method 2 – Starting biomass at low levels

For this method, initial biomass was set at a low value. During an actual invasion, biomass starts at a low

value, and then builds to higher levels. Starting biomass at low levels therefore has the advantage that it more closely

resembles initial biomass values. However, starting with a very low, positive biomass can result in unrealistically high

predation mortality by predators of invasive species. Consequently, diet contributions of invasive species to their

predators must also be low.

Entering biomass or diet at arbitrary low levels could lead to grossly incorrect descriptions of trophic

interactions between groups. To address the concerns of reasonably describing interactions between invasive species

and their prey and predators, we took the following approach when starting invasive species biomass low. We picked

a year, 2002, in the invasive species time series during which the invasive species was well established and which diet

information for its predators were available. The biomass for invasive species at this year was then divided by 1000

and used as an initial biomass value. This scaling factor was large enough to make initial biomass estimates of the

invasive species realistically small, but the factor was also small enough so that diet estimates could be entered into

Ecopath at appropriate precision. Because the invasive species biomass was reduced by a factor of 1000, diet

contributions of the invasive species to its predators had to be reduced as well to maintain search rates at calculated

levels. This assumes that predator search rates for the invasive species do not change over time. To allow all diet

3 - 4

contributions to sum to one, other components of the predators’ diets were proportionally adjusted. Although the

invasive species was also feeding on prey items during the pre-invasions period which in reality do not yet experience

this mortality, the mortality levels were very small initially due to the low initial invasive species biomass, and thus

no corrections were required.

To simulate an increase in biomass during the time at which invasion actually occurred, fishing mortality was

applied to each invasive species initially, and then removed in the year of invasion. The level of fishing mortality was

based on the instantaneous rate of population increase (r) from the time series of data as calculated by B0ert, where B0

is the first biomass estimate in the time series, and t is the number of years after the year for B0. To offset the added

fishing mortality rate, P/B ratios of the invasive species were increased by an amount equivalent to the level of fishing

mortality. This ensured that estimates for M0 were taken only from trophic interactions with other species.

Method 3 – Starting biomass at high levels

Rather than choosing an initial biomass estimate that is much lower than the earliest levels, an alternative is to

initialize the biomass at a future observed level, and then artificially reduce invasive species biomass (by applying

fishing mortality) until the time of invasion, generally around 2002. A biomass estimate was selected from a year in

which diets for predators of invasive species were available, and used as the initial biomass. Diet contributions of

invasive species in that year were also added into their predators’ diets for the initial modeled year. Other diet items

were proportionally adjusted so that total diet proportions summed to one. The initial biomass calculated this way was

substantial for Dreissenids and caused imbalance in groups Dreissenids prey on. To offset the (artificial) predation

mortality by Dreissenids, as well as other invasive species that occurred when the model was initialized, we added a

negative biomass accumulation to prey items that was equivalent to the initial level of consumption of the prey item

by the invasive species. Adding negative biomass accumulation allowed Ecopath to calculate appropriate levels of M0

based on groups actually present in the system in 1981.

Once the invasive species were entered into the initial Ecopath model, it was important to remove their

biomass as quickly as possible during the time dynamic simulations. This was done by a combination of adding large

fishing mortality and removing that mortality with negative BA. Fishing mortality was applied to each invasive group

so that biomass would become zero within the first year. To account for this artificial fishing mortality, biomass

accumulation equivalent to the level of fishing was added to each invasive species so that other mortality estimates

would be based solely on trophic dynamics. This allowed the invasive species to be present in the time dynamic

simulation prior to its actual invasion, but without having any effect on the rest of the food web.

Method 4 – Adjusting trophic interactions by vulnerability mediation

The previous 3 methods required some form of artificial reduction or increase in biomass of invasive groups.

Ecosim controls the strength of interaction between two groups through vulnerability parameters. For the fourth

method, assumptions about changes in the strength of species interactions were used to mediate vulnerabilities, and

therefore influence the response of invasive species. Two forcing functions were used to mediate vulnerabilities: one

to mediate vulnerabilities of invasive species to their predators, and the other to mediate the vulnerabilities of prey to

invasive species. These functions forced vulnerabilities to zero in years prior to the actual invasion, which implies no

effect of the invasive species on either their predators or their prey. Once invasion occurs, prey species were assumed

to become more susceptible to the invasive predator, which was simulated by forcing vulnerabilities to increase to a

peak. Over time, we assumed that prey defense mechanisms would be developed and vulnerabilities would reach

stable levels (i.e. would return to a relative value of 1). Similarly, in the initial stages of an invasion, predators on

invasive species may not have developed a search image, and therefore vulnerabilities of invasive species to its

predators were assumed to be very low. Over time during the invasion, the vulnerabilities of invasive species to their

predators were assumed to increase to stable levels. The shape (Figure 1) of the forcing functions (relative changes to

vulnerabilities over time) was chosen using an ad hoc procedure that reasonably matched the model to the invasive

species time series prior to searching over vulnerabilities and production anomalies. For this method we used the

same approach as for method 3 for setting initial biomasses and diets.

Results:

All four methods performed reasonably in matching basic dynamics of invasive species for the Lake Huron

model (Figure 2). The approach of forcing invasive species or fitting invasive species to biomass time series could

capture increases and decreases in biomass around the same time as the data suggested. Starting biomass of invasive

species at both low and high levels also resulted in increases in biomass in later years.

3 - 5

Fits to non-invasive species ranged from adequate to poor, depending on the species and method employed

(Figure 2). Fits to lake whitefish and Diporeia were poorest among non-invasive groups across all four methods. In

general, each method tended to underestimate lake whitefish biomass and overestimate Diporeia biomass. Dreissenids

are a large proportion of the diet of lake whitefish and thus the way in which this invasive species was modeled may

have contributed to low lake whitefish biomass. Overall, method 2 resulted in the most reasonable fits for invasive

and non-invasive groups (Figure 2).

Quantitative measures of model fit could not be directly compared between methods because the data used to

fit the models differed among methods. All methods except method 1 used invasive time series in the fitting

procedure to estimate vulnerability and environmental productivity parameters. In contrast, method 1 forced (i.e., did

not fit) the invasive time series. Since the goal of estimating vulnerabilities was to match as many of the modeled

groups’ dynamics, which included invasive species, fits to all groups were important.

Among methods where all time series were used to estimate vulnerabilities and production anomalies, method

2 had the lowest sum of squares, followed by method 4 and lastly by method 3 (Table 4). Method 2 performed nearly

as well in terms RSS as did method 1 despite a greater number of data points contributing to the RSS. Method 3

performed substantially poorer than the other methods (Table 4). The RSS for method 3 could actually be improved if

the invasive species time series were not included when searching over vulnerabilities and environmental

productivities, but was then included when calculating RSS. This occurred because the fitting routine worked to

reduce the very high RSS for the invasive species time series at the expense of improving the RSS in other time series

(i.e., the predicted invasive species time series data were very inconsistent with the other groups in the model).

The presence of invasive species in the diet of their predators was a second important attribute that we wanted

the model to capture. In addition to reasonably reproducing biomass time series, estimated diet proportions for

predators of invasive species reproduced observed data for all methods (but are not shown). As expected, the increase

in contributions of invasive species to their predators’ diet matched the increase in abundance.

Discussion:

Methods 1 and 2 had the lowest RSS among the four methods considered (Table 4). However fits were

reasonable for important groups only for method 2 (Figure 2), and is therefore recommended for future EwE

modelers. We also judged method 2 to be the simplest of the three non-forcing methods. Because biomass starts low,

method 2 doesn’t require large changes to predation mortalities of invasive species on their prey, or to diet

contributions of invasive species to their predators.

In addition to quantitative comparisons among methods, another objective of this study was to summarize the

advantages and disadvantages of each method. Below we discuss each method in turn.

Method 1 - Forcing biomass:

The primary advantage for using this method is that forcing biomass of invasive species is the best way to

match observed dynamics of invasive species. Although observed dynamics are not necessarily the dynamics that

actually occurred in the system due to measurement error, they are the best available information on the invasion

dynamics. This method is also simple, in that there is no need for artificial mechanisms to keep calculated biomass

levels low until the time of invasion. The main disadvantage of this method is the requirement for a complete time

series of observed data. Time series for both Bythotrephes and Dreissenids did not extend to the year of actual

invasion. Biomass time series are required for all methods, but having a complete time series is more critical for

method 1. When the time series are incomplete, interpolation between years can be used, but this is only reliable if a

small fraction of years are missing and there are data for the beginning and end of the time series, or the invasion can

be assumed to start later than it actually occurred (as was the case here for Bythotrephes and Dreissenids). Another

important disadvantage of method 1 is that future dynamics for invasive species are not informed by past dynamics

because biomass was forced, rather than having the model estimate vulnerabilities based on the observed patterns of

change in the biomass of the invasive species and their predators/prey.

Method 2 - Starting biomass low:

The primary advantage for starting invasive species low is that this method best mirrors the process of actual

invasion. In addition, a species with very low biomass has very little impact on other species in the model, and thus

few adjustments are needed. With low invasive species biomass, contributions of invasive species to the diets of their

predators are low, resulting in few changes to the contributions of other groups. Similarly, initial predation mortality

by invasive species on prey is minimal, and thus mortalities do not need to be offset. Combined, these attributes

3 - 6

simplify the process of modeling invasive species. The primary disadvantage of this method is that species dynamics

must be maintained by an artificial fishery. The fishery kick-starts the invasion, and thus depends on the time of

release from fishing as well as on the level of fishing. For fishing mortalities that are too low, invasive species

biomass fails to reach observed levels, even with high vulnerabilities. For mortalities that are too high, invasive

species biomass increases very quickly, and vulnerabilities are lower. This method also assumes, simply because

invasive species must increase their biomass greatly during the invasion period, that their vulnerabilities will be high.

Method 3 - Starting biomass high:

This method is the most complicated of all methods. Consequently, the advantages are minimal. The primary

advantage is that invasive species are able to increase very rapidly due to the removal of a very substantial fishing

mortality. This makes this method fairly suitable for species which invade very quickly. The disadvantages are that

because invasive biomass begins at a high level, the impact of invasive groups on other groups is large, and thus must

be accounted for. The high initial biomass must also be removed quickly. The combination of high diet contributions

of invasive species to their predators, and quick removal of invasive biomass, leaves predators without a substantial

proportion of their diet. Although diet proportions are adjusted, substantial time is spent searching for a now non-

existent invasive prey item. Such an effect results in substantial drops in predator biomass and is affected by the value

of vulnerabilities. One potential way to address this effect is to increase the predator search rate on prey other than the

invasive so that once the invasive species biomass becomes zero, the predator spends the same amount of time

searching for existing prey. This effect is most serious for predators with a large proportion of their diets contributed

by invasive species after the invasion has taken place.

Method 4 - Mediating vulnerabilities:

The primary benefit of mediating vulnerabilities is that biomasses of invasive species are altered by biological

interactions rather than artificial fishing. Adjusting vulnerability also provides greater plasticity in biomass dynamics

than does simply removing fishing mortality. Unfortunately, the sensitivity of biomass to the shape of the mediation

function is also the greatest disadvantage of this method. Although the overall shape of the mediation functions we

used made theoretical sense, there was no way to know whether the shape was correct. Although changing the shape

would improve biomass fits, this is not substantially different from forcing the fits. Another disadvantage with this

method is that although artificial fishing mortalities were not used to adjust biomasses, ad hoc changes in diet were

included. Consequently, many of the disadvantages from method 3 exist here as well.

Other considerations:

Of the four methods we examined, one was based on forcing invasive species, and the other three involved

using past data on invasive species to fit a model. The choice between forcing and fitting depends on the objectives

for which the model was developed. We recommend that if there are complete time series of data on invasive species

biomass, and if the objectives of the work are primarily to assess the effects of invasive species on the system, rather

than to predict future interactions, then forcing biomasses may be the best approach. If the objectives of the work are

to account for the effect of invasive species and consider future outcomes of management strategies, then using one of

the fitting methods is preferable.

Sum of squares fitting to data provides an easy metric from which to assess performance between methods

and choose the “best” approach. However, caution with sum of squares data is warranted. When multiple time series

are used, the weighting of such time series to the overall sum of squares can be important. Poorer fits to groups that

are less important to overall objectives may be more acceptable than poor fits to important groups. Similarly, although

scales in Ecosim are relative and residual sum of squares are fit on a log-scale, differences among groups in the

magnitude of change in the biomass time series can bias the fitting procedure toward fitting one group over another.

In general, careful attention should be given to judging which data sources are most informative about food-web

dynamics, and which groups warrant priority in model fitting given the project’s objectives. It is unlikely that giving

all data sets equal weight will result in the best possible model fit.

The methods we examined here emerged from a working group discussion of options for including invasive

species in EwE models. We do not mean to suggest that our list is exhaustive, but we suspect other methods will be

related to one or more of the methods examined here. As the EwE software moves towards more user-developed

plug-ins (Christensen and Lai 2007), additional ways to include invasive species will likely be developed. In addition,

the above recommendations were based on a single model of the Laurentian Great Lakes. Although method 2 was

preferred, it is not possible to say that this method would always out-perform others if tested in different systems. Our

3 - 7

comparison is not meant to be a single recommendation for all modelers, but instead will hopefully elucidate ways to

account for invasive species, and foster new approaches to best account for the interactions of these important groups.

References:

Ahrens, R.N.M., Walters, C.J., Christensen, V., 2012. Foraging arena theory. Fish and Fisheries. 13, 41-59.

Barbiero, R.P., Tuchman, M.L., 2004. Changes in the crustacean communities of lakes Michigan, Huron, and Erie

following the invasion of the predatory cladoceran Bythotrephes longimanus. Can. J. Fish. Aquat. Sci. 61,

2111-2125.

Barbiero, R.P., Balcer, M., Rockwell, D.C., Tuchman, M.L., 2009. Recent shifts in the crustacean zooplankton

community of Lake Huron. Can. J. Fish. Aquat. Sci. 66, 816-828.

Bunnell, D.B., Davis, B.M., Warner, D.M., Chriscinske, M.A., Roseman, E.F., 2011. Planktivory in the changing

Lake Huron zooplankton community: Bythotrephes consumption exceeds that of Mysis and fish. Freshwater

Biol. 56, 1281-1296.

Christensen, V., Pauly, D., 1992. ECOPATH II - a software for balancing steady-state ecosystem models and

calculating network characteristics. Ecological Modelling 61, 169-185.

Christensen, V., Walters, C.J., 2004. Ecopath with Ecosim: methods, capabilities and limitations. Ecol. Model. 172,

109-139.

Christensen, V., Lai, S., 2007. Ecopath with Ecosim 6: the sequel. Sea Around Us Newsletter 43, 4 p.

Christensen, V., Walters, C.J., Pauly, D., 2005. Ecopath with Ecosim: A user's guide. Fisheries Centre, Vancouver.

www.ecopath.org

Cox, S.P., Kitchell, J.F., 2004. Lake Superior ecosystem, 1929-1998: Simulating alternative hypotheses for

recruitment failure of lake herring (Coregonus artedi). B. Mar. Sci. 74, 671-683.

Environment Canada and United States Environmental Protection Agency (EPA), 2009. State of the Great Lakes

2009 Highlights Report. 16 p. available online at http://www.epa.gov/solec/

Espinosa-Romero, M.J., Gregr, E.J., Walters, C., Christensen, V., Chan, K.M.A., 2011. Representing mediating

effects and species reintroductions in Ecopath with Ecosim. Ecol. Model. 222, 1569-1579.

French, J.R.P, III, Schaeffer, J.S., Roseman, E.F., Kiley, C.S., Fouilleroux, A., 2009. Abundance and distribution of

benthic macroinvertebrates in offshore soft sediments in western Lake Huron, 2001-2007. J. Great Lakes Res.

35, 120-127.

Halfon, E., Schito, N., 1993. Lake Ontario food web, an energetic mass balance, in: Christensen, V., Pauly, D. (eds.),

Trophic models of aquatic ecosystems, vol. 26, 390 p., pp. 29-39.

Halfon, E., Schito, N., Ulanowicz, R.E., 1996. Energy flow through the Lake Ontario food web: conceptual model

and an attempt at mass balance. Ecol. Model. 86, 1-36.

Jaeger, A.L., 2006. Invasive species impacts on ecosystem structure and function, M.Sc. Thesis, Michigan State

University, Department of Fisheries and Wildlife, East Lansing, Michigan.

Kitchell, J.F., Cox, S.P., Harvey, C.J., Johnson, T.B., Mason, D.M., Schoen, K.K., Aydin, K., Bronte, C., Ebener,

M.P., Hansen, M., Hoff, M., Schram, S., Schreiner, D., Walters, C.J., 2000. Sustainability of the Lake

Superior fish community: Interactions in a food web context. Ecosystems 3, 545-560.

Makarewicz, J.C., 1988. Phytoplankton and zooplankton in Lake Erie, Lake Huron, and Lake Michigan: 1984. Great

Lakes National Program Office Report No. 3. 274 p.

Nalepa, T.F., Fanslow, D.L., Pothoven, S.A., Foley III, A.J., Lang, G.A., 2007. Long-term trends in benthic

macroinvertebrate populations in Lake Huron over the past four decades. J. Great Lakes Res. 33, 421-436.

Pikitch, E.K., Santora, C., Babcock, E.A., Bakun, A., Bonfil, R., Conover, D.O., Dayton, P., Doukakis, P., Fluharty,

D., Heneman, B., Houde, E.D., Link, J., Livingston, P.A., Mangel, M., McAllister, M.K., Pope, J., Sainsbury,

K.J., 2004. Ecosystem-based fishery management. Science 305, 346-347.

Pine, W.E., III, Kwak, T.J., Rice, J.A., 2007. Modeling management scenarios and the effects of an introduced apex

predator on a coastal riverine fish community. Trans. Am. Fish. Soc. 136, 105-120.

Riley, S.C., Roseman, E.F., Nichols, S.J., O'Brien, T.P., Kiley, C.S., Schaeffer, J.S., 2008. Deepwater demersal fish

community collapse in Lake Huron. Trans. Am. Fish. Soc. 137, 1879-1890.

Robinson, J.G., Frid, C.L.J., 2003. Dynamic ecosystem models and the evaluation of ecosystem effects of fishing: can

we make meaningful predictions. Aquatic Conserv.: Mar. Freshw. Ecosyst. 13, 5-20.

Stewart, T.J., Sprules, W.G., 2011. Carbon-based balanced trophic structure and flows in the offshore Lake Ontario

food web before (1987-1991) and after (2001-2005) invasion-induced ecosystem change. Ecol. Model. 222,

692-708.

3 - 8

Vanderploeg, H.A., Nalepa, T.F., Jude, D.J., Mills, E.L., Holeck, K.T., Liebig, J.R., Grigorovich, I.A., Ojaveer, H.,

2002. Dispersal and emerging ecological impacts of Ponto-Caspian species in the Laurentian Great Lakes.

Can. J. Fish. Aquat. Sci. 59, 1209-1228.

Williamson, M., Fitter, A., 1996. The varying success of invaders. Ecology 77, 1661-1666.

3 - 9

Table 1: Species or groups of species used in the Lake Huron Ecopath model. For multi-stanza groups, the beginning

age (in years) of each stanza is provided.

Group/species name (age stanzas) Scientific name

Sea lamprey Petromyzon marinus

Lake whitefish (0, 1-3, 4+) Coregonus clupeaformis

Lake trout (0, 1, 2-4, 5+) Salvelinus namaycush

Chinook salmon (0, 0.5, 1-5, 6+) Oncorhynchus tshawytscha

Steelhead (0, 1, 2-5, 6+) Oncorhynchus mykiss

Burbot (0-3, 3+) Lota lota

Alewife (0, 1+) Alosa pseudoharengus

Rainbow smelt (0, 1+) Osmerus mordax

Bloater (0, 1+) Coregonus hoyi

Round Goby Neogobius melanostomus

Slimy sculpin Cottus cognatus

Deepwater sculpin Myoxocephalus thompsoni

Ninespine Stickleback Pungitius pungitius

Diporeia Diporeia hoyi

Mysis Mysis diluviana

Benthic invertebrates

Dreissenid mussels Dreissena polymorpha

Dreissena bugensis

Predatory zooplankton Bythotrephes longimanus

Zooplankton

Phytoplankton

Detritus

3 - 10

Table 2: Time series used for comparing performance of methods to incorporate invasive species.

Type of time series # of data points

Fishing mortalities

Age 3 lake whitefish 26

Age 4+ lake whitefish 26

Yearling lake trout 25

Age 2-4 lake trout 25

Age 5+ lake trout 25

Age 1-5 Chinook salmon 28

Stocking

Lake trout 27

Chinook salmon 25

Steelhead salmon 24

Biomass

Sea lamprey 25

Age 3 lake whitefish 26

Age 4+ lake whitefish 26

Age 2-4 lake trout 25

Age 5+ lake trout 25

Age 1-5 Chinook salmon 28

Age 1-5 steelhead salmon 25

Age 1+ alewife 23

Age 1+ rainbow smelt 23

Age 1+ bloater 23

Round goby 9

Slimy sculpin 23

Deepwater Sculpin 23

Ninespine stickleback 12

Diporeia 10

Dreissenid spp. 8

Bythotrephes 8

Zooplankton 8

3 - 11

Table 3: Summary of each method for incorporating invasive species into Ecopath with Ecosim models. Details of

each method are described in the text.

Method Invasive

time series

forced or

fit?

Biomass

high or

low?

Reasoning

1-Forcing Forced High Forcing time series allows fitting routine to

match dynamics to other species, while

invasive species dynamics are fit without

error.

2-Biomass low Fit Low Invasive species begin their invasive at low

biomass levels, and thus should be initialized

as such

3-Biomass high Fit High Starting biomass at levels more similar to

recent years allows invasive species to reach

high levels of biomass

4-Mediating

vulnerabilities

Fit High Biological processes keep invasive species

biomass suppressed until the time in which

they invade

3 - 12

Table 4: Comparison of fits to time series data for each method for incorporating invasive species into Ecopath with

Ecosim models. The number of data point used to fit method 1 differs from those of methods 2-4 and thus cannot be

compared.

Method

Model 1 - Forcing 2 – Biomass low 3 – Biomass high 4 – Mediating

vulnerabilities

Lake Huron 132.3 134.6 3260 167.7

3 - 13

Figure 1: Shapes of vulnerability forcing function for a) prey of invasive species, and b) invasive species to their

predators used in method 4. For vulnerabilities of prey to invasive species, vulnerabilities begin very low for the early

simulation years, and increase to a peak (Y) after the species invades (time period X1), then stabilizes to 1 once the

species begins to become established (time period X2). For vulnerabilities of invasive species to their predators,

vulnerabilities begin very low for the early simulation years, and increase to 1 once the species invades.

X2 X1

1

a)

b)

3 - 14

Figure 2: Fits to key species for each method from the Lake Huron model from 1981-2008. The solid black line

represents model predicted biomass, and the open circles represent observed biomass. Key species include age 4+

lake whitefish (whitefish), age 5+ lake trout (lake trout), age 1+ alewife (alewife), Diporeia (diporeia), round goby

(goby), Dreissenids (dreiss), and Bythotrephes (bytho).

4 - 1

Appendix 4. Evaluation of harvest policies for Lake Huron coldwater commercial fisheries using an Ecopath with

Ecosim model.

Introduction:

Lake trout (Salvelinus namaycush), lake whitefish (Coregonus clupeaformis), bloater (Coregonus hoyi) and

introduced Pacific salmonines comprise the majority of targeted fishing effort within Lake Huron’s cold water fish

community. Lake trout were the primary target of commercial fisheries in the early part of the 1900’s (Baldwin et al.

2002) but increased fishing pressure in combination with the invasive sea lamprey (Petromyzon marinus) resulted in

their near extirpation by the 1960’s. Lake whitefish (Coregonus clupeaformis) harvest has replaced lake trout harvest

since the 1980s (Dobiesz et al. 2005). Current yields of lake whitefish have now surpassed historical levels, and the

lake whitefish commercial fishery on Lake Huron produces the second largest commercial fishery harvest among all

Laurentian Great Lakes in terms of yield (Mohr and Ebener 2005). Bloater harvest has been lower than either lake

trout or lake whitefish, but is not inconsequential (Baldwin et al. 2002). Both bloater and lake whitefish fisheries are

concentrated in Canadian waters, although there are substantial tribal and non-tribal commercial fisheries for lake

whitefish in U.S. waters.

Fishery management goals for a “desirable” fish community have been established for Lake Huron

(DesJardine et al. 1995). These management objectives are called fish community objectives (FCOs) and include

sustainable yields of both lake whitefish and lake trout. In addition, lake trout populations should be self-sustaining,

meaning not dependent on hatchery production. Current lake trout production within the lake is generated almost

entirely through stocking although natural reproduction has increased since it was first observed in 1984 (Riley et al.

2007), providing evidence that management goals are beginning to be realized. In light of the FCO targets, managers

would like to develop a harvest strategy framework that would maintain harvests of coregonines while also increasing

the sustainable production and harvest of lake trout. If this ideal solution is unattainable, then managers wish to

evaluate strategies that represent a compromise between coregonine and lake trout objectives.

Recent large-scale changes in the Lake Huron food-web may have an important influence on interactions

between the coldwater fish community and its fisheries. Abundances of several species of prey fish and an important

benthic amphipod, Diporiea, have declined since the late 1990s (Nalepa et al. 2007, Riley et al. 2008). These species

contribute significantly to the diet of lake trout and lake whitefish (Madenjian et al. 2006, Pothoven and Nalepa

2006). Invasive Dreissenid mussels and round gobies (Neogobius melanostomus) have proliferated in recent years,

further contributing to changes in Lake Huron (Nalepa et al. 2007). Clearly, improved understanding of ecosystem

processes becomes even more important in light of these recent changes, as does appreciation of their implications for

the performance of harvest strategies applied to the multiple fisheries operating in Lake Huron.

We developed a food-web model of Lake Huron’s main basin coldwater fish community using the Ecopath

with Ecosim (EwE) modeling software to explore the effects of alternative harvest policies on management objectives

for lake whitefish and lake trout. Our policy analysis included three types of management options; 1) incremental

adjustments to fishing mortality targets, 2) conversions of the gill net fishery to trap nets, and 3) adjustments to the

seasons in which fishing occurred. The importance of uncertainties surrounding future levels of environmental

productivity, the strength of trophic interactions between predators and prey, and contributions of potentially

important but low frequency prey items were assessed as well. Overall, this analyses should help establish a more

transparent management framework that considers multiple objectives, system uncertainties, and potential tradeoffs

connected to the decision making process.

Methods:

Model

A food-web model of coldwater fish community in the main basin of Lake Huron was constructed using the

Ecopath with Ecosim software package. This model was similar to those described in Appendices 1 and 2 but

contained an additional group, wild lake trout, divided into the same age stanzas as hatchery lake trout in the other

models. The model was parameterized for the year 1981, with data from Lake Huron or as similar of a system as

possible. When data were not available from 1981, data from other time periods were used. Once parameterized,

Ecopath requires the data inputs to balance, meaning the sum of consumption from predators or fisheries on a single

group can not exceed the entered value of production for that group. Given the diversity of data sources used to

parameterize the model, balance rarely occurs without some manipulation of the data inputs. Data inputs were

adjusted following recommended practices (Christensen et al. 2005, but see also Appendix 2).

4 - 2

After balancing was achieved, Ecosim was used to project calculated state variables from Ecopath forward

through time. To assess the accuracy of these projections, Ecosim-estimated biomass dynamics were compared to

observed biomasses. Parameters governing the strength of trophic interaction between predators and their prey, called

vulnerabilities, were adjusted to improve model fits to observed biomasses. Following conventional practice for EwE

models, vulnerabilities to a predator for all prey items of that predator were assumed the same. Vulnerabilities control

the extent to which an increase in predator biomass causes an increase in predation mortality on its prey, with low

vulnerabilities reflecting very little increase and high vulnerabilities reflecting a large increase (Christensen and

Walters 2004). Parameters governing the relative annual level of primary productivity available in the system, called

production anomalies, were also tuned to improve model fits to observed biomasses. Production anomalies are yearly

deviations from the initial productivity entered in Ecopath (Christensen and Walters 2004). These two types of

parameter adjustments were important for matching biomass dynamics estimated in Ecosim to those observed in Lake

Huron (Figure 1).

Policies

After adjusting vulnerability and production anomaly parameters to achieve a reasonable fit to observed

biomasses in 1981-2008, simulations were run to forecast the outcomes of various management policies. Simulations

were run for 50 years, and estimates over the last five years were used to compare policy outcomes. We used a

constant fishing mortality harvest control rule for our simulations, where fishing mortality targets were set for the

primary species harvested in coldwater commercial fisheries, lake trout and lake whitefish. Fishing effort for fisheries

that targeted lake whitefish were then adjusted so that these mortality targets were achieved. The commercial fisheries

included a treaty fishery for lake whitefish, which was a combination of gill nets and trap nets in the 1836 treaty

waters of Lake Huron; a gill net fishery for lake whitefish in non-treaty waters; a trap net fishery for lake whitefish in

non-treaty waters; a Chinook salmon (Oncorhynchus tshawytscha) fishery in treaty waters; and a bloater fishery in

Canadian waters. A recreational fishery for salmonines was also included in the model.

Three types of policy comparisons were used in our simulations. These policies were specified to reflect

earlier discussions with stakeholders (Objective 1). The first policy simply incrementally adjusted lake whitefish

fishing mortality targets above and below the fishing mortality estimated from catch-at-age models in 2006. The

percentage adjustments we considered were -75%, -50%, -25%, 0%, +25%, +50%, and +100 percent. Our purpose

here was to examine how biomass and harvest of both the target species (lake whitefish) and the bycatch species (lake

trout) were affected by changing fishing mortality rates.

The second policy comparison represented a conversion of gill gets to trap nets for non-treaty fisheries only.

The proportion of gill nets converted ranged from no conversion (0%) to complete conversion (100%) in increments

of 25 per cent. Gill nets capture and kill more lake trout than do trap nets (Johnson et al 2004). Consequently, the

purpose of this policy comparison was to explore ways to maintain harvest of lake whitefish while minimizing the

death of lake trout. Harvest of lake whitefish was maintained for each level of conversion; the only change was to the

amount of harvest of lake trout.

The third policy comparison adjusted harvest of lake whitefish or lake trout based on assumed changes to the

seasons in which fishing occurred. We considered two seasonal adjustments to fishing patterns: a) fishing occurred

only in winter; and b) fishing did not occur in summer. Adjustment “a” was considered because stakeholders stated

that lake whitefish prices were greatest in winter, when the supply of lake whitefish was limited. Bycatch of lake trout

in winter was less than the yearly average, and thus fishing only in winter would reduce lake trout harvest for the

same total lake whitefish harvest, or could allow greater lake whitefish harvest while maintaining lake trout harvest at

current levels. We considered both scenarios, referring to the one where the lake whitefish target was achieved and

lake trout were under target as “WF wint”, and the one where the lake trout target was achieved thereby allowing lake

whitefish to be above target as “LT wint”. Adjustment “b” was considered because Johnson et al. (2004) reported

high bycatch of lake trout in gill nets in spring, and high bycatch of lake trout in trap nets in summer. Additionally,

analysis of seasonal observer data in the Canadian commercial fishery suggested bycatch was a greater issue in

summer (Adam Cottrill, unpublished data). As done for adjustment “a”, we considered two scenarios, one where the

lake whitefish target was achieved with lake trout below target (“WF no sum”), and one where the lake trout target

was achieved with lake whitefish above target (“LT no sum”).

For both the gear conversion and seasonal fishing policy comparisons, we allowed for scenarios that were

more extreme than would likely be possible in Lake Huron. We recognize that there are areas of the lake where

complete conversion to trap nets will not be feasible. Likewise it is unrealistic to expect all lake whitefish fishing to

occur in one season (winter). Our rationale for simulating these extreme scenarios was to determine whether direct

4 - 3

and indirect effects of these policies on non-target species (i.e., lake trout) would be great enough to warrant further

discussion.

Uncertainties:

We discussed potential sources of uncertainty with stakeholders during our workshops, and as expected many

candidate areas of uncertainty were identified. For this analysis we decided to focus on three key areas of uncertainty:

(1) future environmental productivity; (2) the strength of trophic interactions between fished groups and their prey

(vulnerabilities): and (3) alternative assumptions about the contributions of particular fish species to the diets of lake

trout and lake whitefish.

Environmental production anomalies played an important role in allowing the Lake Huron food-web model to

reproduce observed dynamics of the system from 1981-2008. The model that best fit observed biomasses included

production anomalies that varied considerably over the time series. In general, anomalies were high during the 1990s,

corresponding to increased biomass of many groups during that time, and declined after 2000 (Figure 2). It remains

very uncertain, however, what future production levels will be. To consider a range of possible futures we used the

first and third quartiles from the estimated time series of past environmental productivities to simulate future

productivity, and compared this with the initial estimate of environmental productivity, that is the initial value in

Ecopath, which was very close to the median of past environmental productivities.

Vulnerabilities are widely recognized as important parameters in Ecosim models (Christensen and Walters

2004), but are difficult to estimate with much precision (Ahrens et al. 2012). Every feeding group in the EwE model

has an estimated vulnerability from the model fitting process. To make an analysis of model sensitivity to

vulnerability uncertainty tractable, we needed to concentrate on a subset of these vulnerabilities. We chose to focus on

vulnerability uncertainty on the oldest age groups of species targeted by the primary commercial and recreational

fisheries in Lake Huron, the rationale being that these age groups are of greatest interest to us, and that vulnerabilities

can influence the level of compensation by fished groups (Ahrens et al. 2012). Species targeted by the primary

fisheries include lake trout (both hatchery and wild), lake whitefish, and Chinook salmon. Vulnerabilities operate, in

effect, on a log scale, ranging from 1 to 109, so order-of-magnitude adjustments to these parameters were appropriate.

Estimated vulnerabilities for lake whitefish and lake trout were very high, and therefore reduced to either 10 or 100;

in contrast estimated vulnerabilities for Chinook salmon were close to 1, so they were increased to 10 or 100. Once

we had adjusted the vulnerabilities for these species, we re-fit production anomalies and vulnerabilities of other

groups to observed biomasses.

Diet contributions of rare prey are difficult to accurately estimate, particularly over long time periods and

large spatial scales. During our workshops, stakeholders argued for the existence of diet contributions that may be

important, but were not supported by our assessment of existing diet data. For example, stakeholders noted that they

sometimes observe lake whitefish in small quantities in lake trout diets, but the diet data we examined (Ji He,

Michigan Department of Natural Resources, unpublished data, Madenjian et al. 2006, Dobiesz 2003; Diana 1990) did

not indicate lake whitefish as a prey item for lake trout. This potential predator-prey interaction could increase the

effect lake trout have on lake whitefish, so to explore this possibility we added 2% of the diet of age 5+ lake trout to

come from age 1-3 lake whitefish. Similarly, stakeholders suggested lake whitefish sometimes show greater levels of

piscivory than suggested by the diet data we used (Pothoven and Madenjian 2008, McNickle et al. 2006, Nalepa et al.

2009). Accordingly, we added small contributions (1% each) of age 1+ alewife and age 1+ smelt to age 4+ lake

whitefish diets. As done when vulnerabilities were adjusted, after making these changes to diets, the model was re-fit

to best match observed biomasses.

Results and discussion:

Lake whitefish and lake trout were the primary species of concern for this project. Critical objectives for these

species included maintenance of acceptable levels of biomass and harvest. As expected, lower fishing mortalities

resulted in increased biomass of both lake trout and lake whitefish while increasing fishing mortality led to consistent

declines in biomass across the range of effort levels considered (Figure 3, left panels). A doubling of fishing mortality

(100% increase) resulted in a 30% decline in lake whitefish biomass and a 17% decline in lake trout biomass.

Harvests increased consistently for both species across the range of fishing mortality targets considered, suggesting

that both species can sustain higher exploitation rates than they currently experience. The relatively large increases in

harvests compared to the corresponding declines in biomasses as fishing mortality increased suggest a high degree of

compensation for both species. Lake trout seem to have even greater compensation than lake whitefish, which may be

simply a consequence of their recruitment being sustained by stocking for the hatchery group. Over the range of

4 - 4

fishing mortalities explored, biomass declined by 54% for lake whitefish and 23% for lake trout while harvest

increased by 271% for lake whitefish and 266% for lake trout (Figure 3, left panels). These findings suggest that the

EwE model is overestimating the steepness of the stock-recruitment relationship for both species at low spawning

stock biomass levels.

Other policies also resulted in expected results. Under gear conversion scenarios, lake whitefish biomass and

harvest remained relatively stable whereas biomass of lake trout increased and harvest decreased with the amount of

gill nets converted (Figure 3, middle panels). Complete conversion of non-treaty gill nets to trap nets resulted in a

15% increase in lake trout biomass. When fishing was limited to winter and total lake whitefish harvests remained at

status quo levels, lake trout biomass was 14% greater (Figure 3, right panels – “WF wint”). In contrast, allowing lake

whitefish targets to increase while maintaining lake trout harvest at status quo levels resulted in a 29% reduction in

lake whitefish biomass, and a 39% increase in harvest (Figure 3, right panels – “LT wint”). Interestingly, under this

scenario the biomass of lake trout increased by 4.4%, most likely due to the decrease in lake whitefish biomass (see

paragraph about assessment of species interactions below). The results were similar but less extreme for the scenarios

where only summer fishing was eliminated (Figure 3, right panels – “WF no sum” and “LT no sum”).

Diet and vulnerability uncertainties had a greater effect on biomass and harvest of lake whitefish than on lake

trout (Figure 4). Under the standard model, variation across all changes to targets was greatest for biomass (Figure 4).

Changes to diet lowered the absolute biomass and harvest for all changes in fishing mortality except 100%. Changes

to vulnerability increased absolute biomass and harvest for lake whitefish, with assumed vulnerabilities of 10 showing

the greatest increase. In contrast, lower vulnerabilities for lake whitefish and lake trout, but higher vulnerabilities for

Chinook salmon, had an opposite, if smaller, effect on lake trout. Lake whitefish biomass was less sensitive to

changes in fishing pressure when diets were changed and vulnerabilities decreased than under the standard model

(Figure 4). Ahrens et al. (2012) stated that greater compensation to fishing pressure when vulnerabilities are low is

expected. Across the range of fishing mortalities explored, lake whitefish biomass declined by 54% under the

standard model; 29% when vulnerabilities for lake trout, lake whitefish, and Chinook salmon were set to 10; 20%

when vulnerabilities were set to 100; and 17% when diets were changed (Figure 4). For lake trout the effect was much

smaller, but in the opposite direction, with biomass being more sensitive when diets were changed and vulnerabilities

decreased (Figure 4). This is possibly due to the effect of changes in lake whitefish biomass on lake trout (see final

paragraph in this section).

Changes to environmental productivities had a far greater effect on biomass and harvest than did changes to

diet or vulnerabilities (Figure 5). Median environmental productivities were the same as initial levels, whereas high

and low environmental productivities represented a 13% increase and 10% decrease from initial levels, respectively.

Lake whitefish responded more to increases in production than did lake trout, presumably because greater production

resulted in greater biomass increases in their prey than the prey of lake trout. Lake whitefish feed at a lower trophic

level than lake trout (3.2 versus 4.2) which implies their prey may be more directly affected by changes in system

productivity. Interestingly, alewife biomass does not recover from 2008 levels unless productivity in the system

increases above initial levels (Figure 6). As expected, alewife biomass increases as fishing on lake trout increases

(Figure 6).

Our goal in building a food-web model was to allow assessment of both direct and indirect interactions

among exploited species, particularly lake trout and lake whitefish. The policies presented above reflect adjustments

to fishing effort that affected both lake trout and lake whitefish. To assess indirect interactions between lake trout and

lake whitefish, resulting from food-web changes, we adjusted fishing mortality on one species without affecting the

fishing mortality of the other. We ran these simulations with the original and modified diets as discussed above. When

the biomass of the exploited species declined due to increased harvest, the biomass of the other species tended to

increase (Figure 7). The only exception was when the original diets were used and lake trout harvest was modified –

this resulted in negligible changes to lake whitefish biomass. More generally, changes in harvest of lake whitefish had

a far greater indirect effect on lake trout biomass than vice versa (Figure 7), probably reflecting the far greater

biomass of lake whitefish than lake trout in the system. Changes to the diet increased the magnitude of the indirect

effect of lake trout harvest on lake whitefish biomass (Figure 7 – “LTH-WF”). In contrast, these same changes

reduced the magnitude of the indirect effect on lake trout biomass of increased lake whitefish harvest (Figure 7 –

“WF-LTH”). When lake whitefish contribute to the diet of lake trout, any competitive release for lake trout that

results from reduced lake whitefish biomass is offset by a reduction in a lake trout prey item.

References:

Ahrens, R.N.M., Walters, C.J., Christensen, V., 2012. Foraging arena theory. Fish and Fisheries. 13, 41-59.

4 - 5

Baldwin, N.A., Saalfeld, R.W., Dochoda, M.R., Buettner, H.J., Eshenroder, R.L., 2002. Commercial Fish Production

in the Great Lakes 1867-2000. (available online at http://www.glfc.org/databases/commercial/commerc.php).

Christensen, V., Walters, C., 2004. Ecopath with Ecosim: methods, capabilities, and limitations. Ecol. Model. 172,

109-139.

Christensen, V, Walters C.J., Pauly, D., 2005. Ecopath with Ecosim: a User’s Guide. Fisheries Centre, University of

British Columbia, Vancouver. November 2005 edition, 154 p. (available online at www.ecopath.org).

DesJardine, R.L., Gorenflo, T.K., Payne, N.R., Schrouder, J.D., 1995. Fish-community objectives for Lake Huron.

Great Lakes Fish. Comm. Spec. Pub. 95-1. 38 p.

Diana, J.S., 1990. Food habits of angler-caught salmonines in western Lake Huron. J. Great Lakes Res. 16, 271-278.

Dobiesz, N.E., 2003. An evaluation of the role of top piscivores in the fish community of the main basin of Lake

Huron, Ph.D. Dissertation, Michigan State University, Department of Fisheries and Wildlife, East Lansing,

Michigan

Dobiesz, N.E., McLeish D.A., Eshenroder R.L., Bence J.R., Mohr L.C., Ebener M.P., Nalepa T.F., Woldt A.P.,

Johnson J.E., Argyle R.L., Makarewicz J.C., 2005. Ecology of the Lake Huron fish community, 1970-1999.

Can. J. Fish. Aquat. Sci. 62, 1432-1451.

Johnson, J.E., Ebener M.P., Gebhardt K., Bergstedt R., 2004. Comparison of catch and lake trout bycatch in

commercial trap nets and gill nets targeting lake whitefish in northern Lake Huron. Mich. Dept. Nat. Res.

Rep. 2071. 25 p.

Madenjian, C.P., Holuszko, J.D., Desorcie, T.J., 2006. Spring-summer diet of lake trout on Six Fathom and Yankee

Reef in Lake Huron. J. Great Lakes Res. 32, 200-208.

McNickle, G.G., Rennie, M.D., Sprules, W.G., 2006. Changes in benthic invertebrate communities of South Bay,

Lake Huron following invasion by zebra mussels (Dreissena polymorpha), and potential effects on lake

whitefish (Coregonus clupeaformis) diet and growth. J. Great Lakes Res. 32, 180-193.

Mohr, L.C., Ebener, M.P., 2005. Description of the fisheries. p. 19-26. In: Ebener, M.P. (ed.) The state of Lake Huron

in 1999. Great Lakes Fish. Comm. Spec. Pub. 05-02. 140 p.

Nalepa, T.F., Pothoven, S.A., Fanslow, D.L., 2009. Recent changes in benthic macroinvertebrate populations in Lake

Huron and impact on the diet of lake whitefish (Coregonus clupeaformis). Aquat. Ecosyst. Health 12, 2-10.

Nalepa, T.F., Fanslow D.L., Pothoven S.A., Foley III A.J., Lang G.A., 2007. Long-term trends in benthic

macroinvertebrate populations in Lake Huron over the past four decades. J. Great Lakes Res. 33, 421-436.

Pothoven, S.A., Nalepa T.F., 2006. Feeding ecology of lake whitefish in Lake Huron. J. Great Lakes Res. 32, 489-

501.

Pothoven, S.A., Madenjian, C.P., 2008. Changes in consumption by alewives and lake whitefish after dreissenid

mussel invasions in lakes Michigan and Huron. North American Journal of Fisheries Management 28, 308-

320.

Riley, S.C., He, J.X., Johnson, J.E., O’Brien, T.P., Schaeffer, J.S., 2007. Evidence of widespread natural reproduction

by lake trout Salvelinus namaycush in the Michgian waters of Lake Huron. J. Great Lakes Res. 33, 917-921.

Riley, S.C., Roseman, E.F., Nichols, S.J., O’Brien, T.P., Kiley, C.S., Schaeffer, J.S., 2008. Deepwater demersal fish

community collapse in Lake Huron. Trans. Am. Fish. Soc. 137, 1879-1890.

4 - 6

Figure 1: Ecosim fits to time series of observed biomasses. In all subpanels, the black line represents estimated

biomasses in Ecosim and the open circles represent observed data.

4 - 7

Figure 2: Relative annual production anomalies during 1981-2008 for the “standard” model. The grey line at 1 is for

reference to the initial primary productivity in the initialized model.

4 - 8

Figure 3: Average biomass and harvest from the last five years of simulation for age 4+ lake whitefish (whitefish) and

age 5+ hatchery lake trout (lake trout) for three policy options in the standard model. For each figure, lake whitefish is

on the primary y-axis and lake trout is on the secondary y-axis.

4 - 9

Figure 4: Average biomass and harvest of age 4+ lake whitefish (whitefish) and age 5+ hatchery lake trout (lake trout)

from the last five years of simulation under uncertainties around diet (diet), and vulnerabilities (vuln 10 and vuln

100).

4 - 10

Figure 5: Average biomass and harvest of age 4+ lake whitefish (whitefish) and age 5+ hatchery lake trout (lake trout)

from the last five years of simulation for three levels of environmental productivities in the standard model.

4 - 11

Figure 6: Average biomass of age 1+ alewife in the last five years of simulation under three different assumptions

about the future level of environmental productivity in the standard model. Biomass is at zero under both low and

medium levels of productivity.

4 - 12

Figure 7: Average biomass in the last five years of simulation of age 4+ lake whitefish at various levels of fishing

mortality on age 5+ hatchery lake trout (LTH-WF) and of age 5+ hatchery lake trout at various levels of adjustment to

fishing mortality on age 4+ lake whitefish (WF-LTH). Biomass values were plotted relative to the biomass when

fishing mortalities were unchanged (0%). The second row of the x-axis reflects the model under standard assumptions

(Standard) or with greater direct and indirect interactions between the two groups (Diet).