Ecological Modelling 112 (1998) 81 – 84

Modelling ecological and economic systems with STELLA:

Part II

Robert Costanza a,*, Sara Gottlieb b

a
Uni6ersity of Maryland, Institute for Ecological Economics, Box 38, Solomons, MD 20688 -0038, USA
b
Maryland Sea Grant Extension Program, Uni6ersity of Maryland, College Park, MD 20742, USA

Abstract

This special section contains a group of seven modelling studies covering a range of ecological and economic
systems and problems. The models were all developed using STELLA, an icon-based programming language
specifically designed for dynamic systems modelling. Models included in the special section cover: dinoflagellate
(Phiesteria piscicida) dynamics as a result of coastal eutrophication, Menhaden fisheries management in the
Chesapeake Bay, watershed characteristics affecting both brown trout habitat and beaver ponds, the impacts and
economics of riparian buffer strips, the integrated ecology and economics of Brazilian mangroves, and the
macroeconomy including natural resources and a range of critical assumptions about technical change and
substitution. © 1998 Elsevier Science B.V. All rights reserved.

Keywords: Ecological; Economic; STELLA modelling

1. Introduction elling contains a series of models developed by

relatively ‘novice’ modelers who built their models
As interest increases in creating sustainable so- in a relatively short period of time to answer a
lutions to environmental and economic problems wide range of questions about the ecological and/
the need for methods of predicting the outcomes or economic systems they had been studying. It is
of policy decisions becomes more urgent. Where the second in a series of special sections devoted
long term studies or experimental manipulations to this topic1.
are not possible (as is often the case in complex All of the models were initially constructed
ecological-economic systems) representative mod- during a seminar-style course at the University of
els can help to fill in knowledge gaps (Costanza et Maryland at College Park between September and
al., 1993). This special section of Ecological Mod- December, 1996. Participants were first intro-

1
* Corresponding author. E-mail: costza@cbl.umces.edu See Costanza et al., 1998 for the first in the series.

0304-3800/98/$ - see front matter © 1998 Elsevier Science B.V. All rights reserved.
PII S0304-3800(98)00073-8
82 R. Costanza, S. Gottlieb / Ecological Modelling 112 (1998) 81–84

duced to the modelling program STELLA II and performed without an integrated model of natural
then individuals or groups chose a problem that and manmade systems.
each would explore over the rest of the term. The Many of the models presented in this section
remainder of the class was spent in an open will be used as tools in ongoing investigations of
format where students provided one another with the systems they describe. Some may be changed
progress reports and feedback on their projects. radically in structure or parameter sets after more
Participants continued to work on their models empirical information about the system has been
after the course ended. gathered. An important point about modelling is
As one can see in this issue, the problems that that it is an evolving process that sometimes
requires one to discard early work and go back to
concern researchers in the ecological and eco-
the proverbial drawing board. Fortunately, as will
nomic sciences are quite varied. Nonetheless, rela-
be discussed in the next section, the STELLA
tively simple models can provide a great deal of
modelling language makes this part of the process
insight into the questions, ‘what is going on?’ or
very easy.
‘what would happen if...?’ These models should
obviously not be considered to be final answers to
their associated complex questions, but instead as 3. Description of the STELLA modelling
syntheses of existing information and guides or language
maps to direct future work.
STELLA II is an object oriented programming
language that uses an iconographic interface to
2. Modelling approaches facilitate construction of dynamic systems struc-
tures. The essential features of the system are
Scientists may be led to modelling for a number defined in terms of stocks, flows and auxiliary
of reasons other than model creation for its own parameters. The user places the icons for each of
sake. Hall and Day (1977) consider three uses of these features in the modelling area and makes the
models to scientists: understanding, assessing, and appropriate connections between features. The
optimizing. Some of the models in this issue were functional relationships between the features are
developed by participants who simply wished to then defined by the user. These relationships can
gain a conceptual picture of how a system of be mathematical, logical, or graphical function,
interest to them might work. In many cases, these and the program allows quite a bit of flexibility
types of models were generated before any field or here. STELLA II represents stocks, flows and
laboratory studies had been conducted, and their parameters, respectively, with the following icons:
main purpose was to examine what features are
the most critical in determining system behaviour.
At the next level, after empirical measurements
had been taken, models were developed to test Connections between features are defined using
assumptions about the system. For example, in ‘information arrows’ with the following icon:
nutrient-dosing mesocosm experiments, models of
the system can help researchers to determine path-
ways of nutrient flow that result in observed The user generates a complete structure which
conditions over a period of time. Finally, along would resemble the diagram depicted in Fig. 1.
the lines of predicting system behaviour, some Stocks represent a reservoir of material such as
researchers want to know what conditions will population, biomass, nutrients, or money. Mate-
lead to an optimal outcome of some property of rial flows between stocks or into and out of
the system. This type of analysis is essential to undefined sources and sinks (represented by
informed policy decisions and often cannot be ‘clouds’ at the ends of flow structures). Flows are
 R. Costanza, S. Gottlieb / Ecological Modelling 112 (1998) 81–84 83

Anderson (this volume) looks at the effects of

seasonal variability on the germination and verti-
cal transport of a cyst forming dinoflagellate. This
type of dinoflagellate (Phiesteria piscicida) has
been a huge problem in the Chesapeake Bay
recently, causing direct economic losses to the
fishing and tourism industries. Their life cycles are
very complex, and the model explores the links
between environmental factors and movement be-
tween life cycle stages in this class of organism.
This model was an initial scoping of the problem
and will be used to design more complex models
and data sampling regimes. Nevertheless, the re-
sults are intriguing, and indicate the very complex
dynamics that can result from the complex life
cycle stages that these organisms possess.
Gottlieb (this volume) also addresses a problem
in the Chesapeake Bay, namely the implications
Fig. 1. A simple STELLA model showing the use of most of
for management of nutrient removal in age-zero
the symbols.
Atlantic menhaden. She looks at the dynamics of
growth and consumption of age-zero menhaden
affected by auxiliary variables, stocks, and other during a 183 day period when they are abundant
flows through the use of information arrows. in the Bay. The model includes a fishery manage-
Auxiliary variables can take the form of con- ment submodel, and estimates the monetary value
stants, mathematical or graphical functions, and of the fishery, both as a traditional commodity
data sets. Once created, stocks and variables can and as the ecosystem service of nutrient cycling.
be duplicated as ‘ghosts’ and used elsewhere in Jessup (this volume) moves upstream to look at
the model, thus avoiding a jumbled spaghetti of brown trout dynamics and habitat quality in an
information arrows in the model. Portions of a urbanizing watershed. The model links land dis-
larger model can be broken down into sectors turbing activities and the amount of impervious
surface in the watershed to trout habitat quality,
which can be run independently or simulta-
as affected by sediment transport, hydrology and
neously to facilitate debugging. Details about
water temperature. Results indicate that trout
how the user defines initial conditions, functional
populations recover from construction events in
relationships, and parameter values, generates
watersheds with less than 15% imperviousness,
output, and performs sensitivity analyses are pro- but not in more densely developed watersheds.
vided in other publications (Peterson and Rich- Sturtevant (this volume) also investigates an
mond, 1996; Hannon and Ruth, 1994). important watershed issue in the Chesapeake Bay
area; the dynamics of beaver impoundments.
Beavers have made a strong comeback in the
4. Summaries of individual contributions Chesapeake Bay drainage basin, and the model
looks at the spatial and temporal dynamics of
There are seven contributions included in this plant successional changes in a generalized beaver
special section, covering a range of ecological impoundment. Results show how changes in
and economic systems and time and space scales. flooding regimes, sediment accumulation and seed
All of the models address some aspect of the sources (due to changes in watershed conditions)
interlinkage between ecological and economic have dramatic effects on the spatial dynamics of
systems. beaver ponds.
84 R. Costanza, S. Gottlieb / Ecological Modelling 112 (1998) 81–84

Robles-Diaz-de-Leon and Nava-Tudela (this 5. Conclusions

volume) look at an interesting proposed system of
stream buffer management using Asimina triloba The models collected in this special section
(pawpaw), a tree native to the Chesapeake Bay demonstrate the range of ecological and economic
which can provide both a stream buffer and a questions that can be productively addressed with
marketable fruit. The model looks at the inter- easy-to-use dynamic modelling tools. The days of
dynamic modelling as the purview of a few spe-
linked dynamics of growth, production, and pos-
cialists are coming to a close. Dynamic modelling
sible economic gains from pawpaw planting in
is now a tool accessible to researchers in many
riparian buffers. The model shows how the eco- fields, as easily and routinely as statistics. Of
nomic viability of pawpaw planting depends on course, the more powerful the tool, the greater the
labor costs, the market value of the fruit, and the danger of its misuse. It is as important to under-
value of non-marketed sediment control services. stand the range of uses of models and their limits
Grasso (this volume) moves a little further in these uses as it is to understand the technical
south and constructs an integrated ecological- details of the models themselves. This collection
economic model of mangrove systems in coastal of models does an admirable job of using new
Brazil, focusing on the trade-offs between modelling tools in appropriate ways to address
forestry and fishery production. She combines important science and policy questions.
the results of a simulation model with a dynamic
optimization model of the system. The output of
the simulation model was used in the optimiza- References
tion model to find shadow prices for the re- Costanza, R., Wainger, L., Folke, C., Mäler, K.-G., 1993.
sources, which then affected the simulation Modelling complex ecological economic systems: toward an
model. Forest growth rates turned out to be the evolutionary, dynamic understanding of people and nature.
most important variable, since fishery production BioScience 43, 545 – 555.
Costanza, R., Duplisea, D., Kautsky, U., 1998. Modelling
in this area is directly dependent on the man- ecological and economic systems with STELLA. Ecol.
grove forest. Model. 110, 1 – 4.
Finally, Woodwell (this volume) looks at the Hall, C.A.S., Day, J.W., 1977. Systems and models: terms and
basic principles. In: Hall, C.A.S., Day, J.W. (Eds.), Ecosys-
whole ecological economic system, and the fac-
tem Modelling in Theory and Practice: An Introduction
tors that affect the linkages between resource with Case Histories. John Wiley and Sons, New York, pp.
consumption and economic growth. The model 6 – 36.
is used to illustrate the dramatic effects of com- Hannon, B., Ruth, M., 1994. Dynamic Modelling. Springer-
Verlag, New York.
peting world views and assumptions on the re- Peterson, S., Richmond, B., 1996. STELLA Research Technical
sults. Documentation. High Performance Systems, Hanover, NH.