NOTES AND

INSIGHTS

Notes and Insights is

zy
zyxwvu
A system dynamics approach to modeling
fisheries management issues: Implications
for spatial dynamics and resolution

zyxwvutsrq
Matthias Ruth

Methods of modern marine fisheries management virtually are no different

a forum for discus- from ancient hunter-gatherer strategies in that the “prey” has to be first spot-
sion and debate
about current issues ted, then caught or collected, and hauled to the home port. The resulting
in the philosophy interactions between the economic and biological systems may be simulated
and application of with well-known predator-prey models. In a past issue of this journal, Swart
the system dynamics
approach, and a
(1990) presented the ability of even the most simple predator-prey models to
marketplace for the yield damped and explosive oscillations as well as stable limit cycles. This
exchange of infor- article expands the standard predator-prey model to capture spatial dynamics
mation about current and draws conclusions for the modeling and management of marine fisheries.
research, policy
The arguments presented can easily be generalized to other mobile renewable

zyxwvutsr
issues, and teaching
experiences. Manu- resources.
scripts treating ma- Much of the real-world fluctuations in the size of fish populations is caused
terial particularly
suited to presen- by unpredictable environmental fluctuations which are poorly understood.

zyxwvuts
tation in short form Reference to those fluctuations is used to justify deviations of model predic-
(less than 2,000 tions from actual population changes, or to stress similarities between seem-
words) should be ingly erratic model predictions and actual population changes (Wilson et al.
sent to Erich Zahn,
Betriebswirtschaft- 1990).
liches Institut, Abt. I contend that fish population dynamics may as well be attributed to econ-
IV,UniversitZit Stutt- omic factors that play an increasing role in the “predator-prey relationship”.
gart, Keplerstr. 17,
D-7000 Stuttgart 1, Many economic forces are poorly understood and difficult to forecast. Further-
Germany or to more, differences between model result and actual system behavior are signifi-
Yaman Barlas, Dept. cantly dependent on the spatial resolution of the predator-prey models in use.
of Industrial Engi-
neering, Bogazici To demonstrate the sensitivity of model results on economic parameters and
University, 80815 spatial resolution the standard analytical description of population dynamics
Bebek, Istanbul, and economic adjustments is combined with a system dynamics approach.
Turkey. While economic models typically concentrate on the analytical investigation
of properties of a system’s equilibrium points, the system dynamics approach
Matthias Ruth is
of this paper admits the richness of experimental investigations to models of
Assistant Professor the dynamic evolution of fishery resources. The paper demonstrates the power
at the Center for of merging system dynamics with economic models of natural resource use to
Energy and Environ- assess the sensitivity of model results to model specifications, initial conditions,
mental Studies and
the Department of and empirical estimates, and to guide data collection, analysis, and ultimately
Geography at Boston policy decision making.
University. His re-
search focuses on
the theory and
modeling of econ- The model

zyxwvuts
zyxwv
omy-environment
interactions. Consider the following model of a marine fishery, following standard assump-
Address: Center for
Energy and Environ- tions of population biology (Brown and Rothery 1993), but subdivided into n
mental Studies and System Dynamics Review Vol. 11, no. 3 (Fall 1995): 233-243 Received June 1994
the Department of 0 1995 by JohnWiley & Sons,Ltd. CCC 0883-7066/95/030233-11 Accepted October 1994

233
 zyxwvut
zyxwvutsr
zyxwvu
234 System Dynamics Review Volume 11 Number 3 Fall 1995

zyxwv
zyxw
zyxwvutsrqp
Geography, Boston
University, 675
Commonwealth Ave-
nue, Boston, M A
02215, U.S.A.
E-mail:
mruth@bu.edu
dt ( zyxwvu
regions. The change in population N,in region i (i = 1:n),measured, e.g., in
tons of biomass per year, is given by
--
dNi
3
- r Ni 1 - - - vEiN,

with r as the natural rate of change. Ki is the carrying capacity of region i,

measured in tons of biomass. The catchability coefficient v depends on tech-
nology and is assumed constant. It is measured in tons of biomass caught per
year per unit effort Ei and biomass Nj in region i. Ei is measured in number of
fishing days. Thus, vEiNi is total catch from region i in a given year. For
simplicity we may assume all regions to be of equal size and quality, i.e., the
carrying capacity of all regions is the same,
Ki= Ki = constant
It is customary in resource economics to express changes in total fishing
effort E as an increasing function of profits (Beltrami 1987; Conrad and Clark

zyxw
1987):
dE
-= CY(PVEN- cE)
dt
with a a constant proportionality factor, P unit price of catch, N population
summed over all n regions, and c unit cost per effort. Total effort is allocated to
region i based on the relative population size di of that region, i.e., El = dlE.
Typically a < 1 owing to the limited ability in the fishing industry to
translate changes in profits fully and immediately into changes in effort (Clark
et al. 1979). For example, if profits increase, fishermen must be hired and
trained and capital equipment must be acquired before fishing effort can be
increased. Similarly,if losses are made, fishermen may not be immediately laid
off and capital equipment such as boats, gear and nets may not be immediately
sold to reduce effort but temporarily remain in use.
Assume for simplicity a four-region model, identical initial conditions for
each region and no migration across regional boundaries. Such a model specifi-
cation is equivalent to a single-region model. Alternative choice of model
parameters and initial conditions yields damped or explosive oscillations
(Figs. l a and Ib).l Neutral stability is found for parameter values between those
that yield damped and explosive oscillations (Fig. lc). For the purpose of the
arguments made in this paper, it suffices only to change a for consecutive
model runs. Lower adjustment coefficients a tend to damp oscillationsbecause
density dependence of the fish population dynamics increasingly dominates
system behavior. For the model a = 2/3 yields neutral stability. The corre-
sponding fish population dynamics are shown in Figure Id.
 zyxwv
zy
Ruth: Modeling Fisheries Management Issues 235

zyxwvutsrq
zyxwvutsr
zyxwvutsrqponm
(a) Damped oscillation (a= 0.60)
54OOO.00, I
(b) Explosive oscillation (a= 0.70)
54ooo.00

E zyxwv E 27000.00

0.00

N N

(c) Neutral stability (a= 0.6666) (d) Neutral stability of fish population dynamics (a= 0.6666)

E dN
ai-
zyxwv
N N

Fig. 1. Single region

model
Of course, real adjustment coefficients depend on the prices of capital of
different vintages, availability and cost of labor, interest rates, institutional
constraints (e.g., tax breaks, subsidies, limitations of access to the fishery,
restrictions on boat size or fishing seasons) and characteristics of fishermen
(e.g., risk aversion, price expectations, experience). Thus, changes of a are very
likely,potentially perturbingthe system’sdevelopmentpath. Even small changes
in a,as shown in Figure 1,result in marked differences of the system behavior.
 zyxwvutsr
zyxw
zyxwvuts
236 System Dynamics Review Volume 11 Number 3 Fall 1995

zyx Little data is actually available to estimate a reliably. The resulting models are
consequently highly sensitive to the assumptions about the actual adjustments
in effort, and it is therefore difficult to attribute population dynamics predomi-
nantly to environmental fluctuations.

Spatial dynamics

The basic model outlined above is used to investigate the impacts of spatial
resolution on model results. A standard predator-prey model with spatially
explicit dynamics is discussed in Hannon and Ruth (1994).Allen and McGlade
(1987)make use of a more elaborate spatial fisheries model, including a specs-
cation of effort adjustments that is similar to the one employed here. Their
model, however, is devoted to the system-wide fishing patterns that result from
adjustments in individual fishermen’s risk perception, and does not address
issues of spatial resolution.
Assume again an equal initial distribution of fish populations across the four
regions. The regions are arranged contiguously in a square, not unlike the

zy
z
statistical areas defined for Atlantic fishing grounds, and migration of fish is

zyxwv
permitted from each region into any other. A fixed proportion of the population
is assumed to migrate, and their destination is determined by random factors.
No fish, however, are leaving the overall system other than through natural or
fishing-induced death. None are added other than through births. A simple
functional specification of net migration between regions i and j (i# f that

zy
fulfills these conditions is

V i, j.
zyxw
Nii is net migration from region i to region j. The parameters m i and mi
are the proportions of migrants from regions i and j, respectively, and pjj and
pii are random numbers ranging between 0 and 1. For simplicity m i = mi,

Assuming that fish move freely in oceans is not at all an unreasonable

assumption. Examples abound of interregional short-term movement and
temporal migration of fish populations, with movement of juveniles frequently
higher than that of older fishes. Atlantic cod, for example, exhibit significant
movement within the spatial confines of Gorges Bank, off the New England
coast (Bowman et al. 1987;Wigley and Serchuk 1992).Similarly, many coastal
reef fish populations exhibit significant movement during their larval and
juvenile stages (Carr and Reed 1992).To the observer (or modeler) the propor-
 zy
Ruth: Model@ Fisheries Management Issues 237

tion of fish moving from one region of the ecosystem to another may seem
higher, the finer the spatial resolution used.
Increase of the spatial resolution from the one-region to a four-region model
results in movement away from neutral stability to explosive oscillations.
With increased migration rates mi,the oscillations become more pronounced.
Figures 2a and 2b show the dynamics in the economic and biological systems

zyxwvutsrq
for the parameter values that yield neutral stability in the absence of migration
(Figure lc). In contrast to Figure lc, the system exhibits explosive oscillations

zyxwvutsrq
zyxwvuts
when migration among regions is allowed.

zyxwvuts
zyxwvutsrqp
Fig. 2. Four-region
model with migra-
tion (a= 0.6666;
mi = mj = 0.90Vi, 1)
Economic and biological adjustments
5m.m -

(b) Fish population dynamics

1500.00

3 -5250.00 zyxwvutsr
- 12ooo.00
1
7 .oo
N
.00 40000.C
 zyxwvutsr
zyxwvuts
238 System Dynamics Review Volume 11 Number 3 Fall 1995

zyxw Noting changes in population sizes is one method of assessing impacts of

human activities on fisheries. Profits accrued over time by fishermen from
harvesting the resource is an economic indicator of the longer-term impacts of
fishing. Given an interest rate q, the cumulative present value of profits can be
calculated as

zy Thus, I1 is the weighted sum of all streams of profits achieved from fishing. The
weights e-qt decrease the further in the future the profits are achieved. In the
spatially explicit model the cumulative present value of profits is lower than
in the single-region model and lower for higher values of mi.Thus, increases
in spatial resolution alter not only the model’s results with respect to the
population dynamics, but, through changes in those dynamics, the economic
assessment of economic activities.

Management strategies

Previous fisheries policies targeted individual species through various restric-

tions on effort or catch (Clark 1991). To be effective, these methods require a
thorough understanding of the natural processes that govern population
changes. In lieu of sufficient understanding of the feedbacks among fish popu-
lations and their physical environment, as well as the corresponding economic
adjustments, management may alternatively be done by setting aside reserves
(Holland 1993).No fishing would be allowed in the reserve but fish could leave
the reserve, thus replenishing stocks in the remainder of the ocean.
Figures 3a and 3b show an example with one of the four regions closed off
from fishing, and no migration across any of the regional boundaries. The result
is a higher total population level. Once the stable oscillation is reached, the
population never drops below the size maintained by the reserve. As a

zy
result, the system moves from its initial oscillation (left part of Figure 3a) to

zyx
neutral stability. Although population size increases, cumulative present
value of profits is lower because part of the fish population is off limits to
fishermen.

zyx
Figures 4a and 4b show the four-region model with one reserve and migra-
tion. It is assumed that the reserve was placed to cover areas with the highest
rate of migration such as spawning areas; thus mResewe > mi = m /(V i,j #
Reserve). Migration leads to abandonment of neutral stability but, unlike in the
absence of the reserve, to damped oscillations. As a result of migration, the
 zy
z
zyxwv
Ruth: Modeling Fisheries Management Issues 239

zyxwvutsrqp
zyxwvuts
zyxwvuts
Fig. 3. Four-region
model with reserve
and without migra-
tion (a= 0.6666;
mj = 0; vi)
-
(a) Economic and biological adjustments
5oooo.00
~-

E 25000.00

0.00

(b) Fish population dynamics

dN
ai
zyxwvutsrq
N

total population size may fall below the maximum that can be sustained in the
reserve. The cumulative present value of profits is typically only marginally
lower than in the one-region model with neutral stability.
Figures 5a and 5b show the four-region, one-reserve model. However, now
the regions are not arranged in a square but a row. All other factors are the
same as in the simulations underlying Figure 4. The result is a faster movement
towards the steady-state and cumulative present values of profit that may
exceed those achieved in the one-region model with stable limit cycle. The
rather erratic initial development is caused by the combined effect of the
 zyxwvutsr
zyxwvutsrq
240 System Dynamics Review Volume 11 Number 3 Fall 1995

zyxwvu
zyxwvut
zyx
zyxwvuts
Fig. 4. Four-region
model with distinct
migration rates for
reserve and remain-
der of ocean
(a) Economic and biological adjustments
- __
~ - -
~~ -- -

(a= 0.6666;
mi = 0.50Vi #
Reserve; mbssrve
= 0.90) E

(b) Fish population dynamics

~ ~ - -

dN zyxwvuts
zyxwvuts
aT

I
-6Ooo.oo
.oo 2oodo.00 ’ 40000.d
N

randomness of the migration direction of fish from the reserve and its im-
mediate neighbor region(s).

Conclusions

A standard predator-prey model was used to model dynamics in fish popula-

tions and the fishing sector of an economy. Small variations in the flexibility
of the fishermen to adjust effort from one year to the next result, either in
 zy
Ruth: Modeling Fisheries Management Issues 241

zyxwvu
zyxwvuts
zyxwvutsrqp
Fig. 5. Regions
arranged in a row
(a= 0.6666; mi
= 0.5V i # Reserve;
m-, = 0.9)
(a) Economic and biological adjustments

(b) Fish population dynamics

N zyxwvut
damped or explosive oscillations or in neutral stability. Although such varia-
tions have not been discussed before in the literature on fisheries management,
they will always be present in real systems. Consequently, erratic fluctuations
in fish populations may as well be attributed to economic factors, rather than
solely to environmental factors, as has been done frequently.
The model was modified to increase spatial resolution and account for move-
ment of fish within the ecosystem. Starting with neutral stability, increases in
the spatial resolution fundamentally change the model’s results towards explo-
sive oscillations.
 zyxwvutsr
zyxwv
242 System Dynamics Review Volume 11 Number 3 Fall 1995

The spatial model was further modified to include a fishery reserve in which
no fishing is allowed, and from which the remainder of the ecosystem may
recruit individuals. Starting with neutral stability, an increase in spatial reso-
lution and an introduction of the reserve leads to fundamentally different
results with regard to long-term population sizes and economic performance.
Surprisingly, the obvious fact that fish populations are typically not stationary
but mobile has not yet found recognition in studies of fisheries management.
To provide even qualitative, let alone quantitative, answers to the dynamics of
fish populations requires clarification of the appropriate spatial resolution of
models and proper delineation of the regions to be modeled. This study offers
a first template and guidance for spatially explicit dynamic models of fisheries
management issues.

Notes

2.
zyxwvu
zyxwvu
1. All simulations are done with the graphical programming language STELLA Il
(High Performance Systems 1994)on a Macintosh IIsi. Each simulation is carried
out for 2000 time steps.
The terms “neutral stability” and “oscillation” are used loosely here since
the randomness of migration will always introduce fluctuations in the time

zy
paths.

References

z
Allen, P. M., and J. M. McGlade. 1987. Modeling Complex Human Systems. Euro-
pean Journal of Opemtional Research 30: 147-167.
Beltrami, E. 1987. Mathematics for Dynamic Modeling. Boston, Mass.: Academic
Press.
Bowman, R. E., T. R. Azarowitz, E. S. Howard, and B. P. Hayden. 1987. Food and
Distribution of Juveniles of Seventeen North west Atlantic Fish Species, 1973-
1976, Wood Hole, Massachusetts, NOAA Technical Memorandum. NMFS-F/
NEC-45, Northeast Fisheries Science Center.
Brown, D., and P. Rothery. 1993. Models in Biology: Mathematics, Statistics and
Computing. New York: John Wiley and Sons.
Cam, M. H., and D. C. Reed. 1992. Conceptual Issues Relevant for Reef Fishery
Management in the US.Southern Atlantic, Miami, NOAA Technical Memoran-
dum, NMFS-SEFC-261.
Clark, C. W. 1991.Economic Biases Against Sustainable Development. In Ecological
Economics: The Science and Management of Sustainability, ed R. Costanza. 319-
343. New York Columbia University Press.
Clark, C. W., F. Clarke, and G. R. Munro. 1979. The Optimal Exploitation of
Renewable Resource Stocks: Problems of Irreversible Investment. Econometrica
47: 25-46.
 Ruth: Modeling Fisheries Management Issues

zyxwv
zyxwv
zyxwv
Conrad, J. M., and C. Clark. 1987.Natuml Resource Economics: Notes and Prob-

zyxwvu
lems. Cambridge: Cambridge University Press.
Hannon, B. and M. Ruth. 1994.Dynamic Modeling. New York Springer Verlag.
High Performance Systems. 1994. STELLA ZZ Technical Documentation. High Per-
formance Systems, 45 Lyme Road, Hanover, NH 0371T,U.I.A.

zyxwv
Holland, D. S. 1993.Managing Fisheries without Restricting Catch or Eflort: The
Use of Marine Reserves for Inshore Fisheries. Masters Thesis, University of
Illinois, Urbana-Champaign.
Swart, J. 1990.A System Dynamics Approach to Predator-Prey Modeling. System
Dynamics Review 6: 94-99.
Wigley, S. E., and F. M. Serchuck. 1992. Spatial and Temporal Distribution of
Juvenile Atlantic Cod Gaudus morhua in the Georges Bank-Southern New
England Region. Fishery Bulletin 90: 599-606.
Wilson, J. A.,R. Townsend, P. Kelban, S.McKay, and J. French. 1990. Managing
Unpredictable Resources: Traditional Policies Applied to Chaotic Populations.
Ocean and Shoreline Management 13: 179-197.