LINKING AND REUSING MODEL COMPONENTS

FOR ENVIRONMENTAL SIMULATION

Jean-Luc de Kok and Guy Engelen

Flemish Institute for Technological Research (VITO), Boeretang 200, Mol, Belgium

Keywords: Environmental Simulation, Coastal Zone Management, Component-Based Simulation, Model Reuse.

Abstract: Simulation modelling is widely accepted as a tool to gain understanding of the behavior of environmental

systems under pressure and support their integrated management. A common practice is that the modelling

process is aimed at identifying the relevant problems and potential management strategies in close

coordination with stakeholders before translating the available domain expertise into integrated simulation

models. The challenges include the need for cross-disciplinary thinking and model linking, incorporation of

social science knowledge, and the communication aspects of projects. Most projects bring along their own

case studies, with simulation models being designed from scratch. Although component-based design is the

common practice in software engineering few examples are found in environmental modelling. Project

deliverables are often not reused by parties different from those that developed them. An open-source

library of well-designed, reusable model building components could pave the way for a more efficient use

of modelling resources and considerably improve the efficiency and flexibility of the modelling process,

thereby creating more room for the communication with end users, scientific discussions, and model

interface design. The experiences of a multi-disciplinary simulation study of coastal systems are used to

discuss the methodological aspects and pitfalls of the design of reusable model components.

1 INTRODUCTION

Environmental systems are under increasing man-

induced pressure and the response to different

external conditions and combinations of

management options depends on the complex

interactions between the social, ecological and

economic processes involved. Different integrated

simulation models have been developed in the past

to support strategic environmental planning.

Unfortunately, these models are often not reused by

research groups other than those that developed the

models. One of the more important reasons is that

the degree of reusability of model constructs is

limited due to the specific requirements of the

modelling framework used (Rizzoli et al., 2008).

Reusability and flexibility for application by third

parties is seldom a priority for domain experts, who

often lack the professional software engineering

skills needed to make their products easily

transferable. Well-designed, reusable components

representing key functionalities could facilitate the

exchange of scientific expertise between

environmenttal modellers across different modelling

domains. In turn this would allow for a more rapid

and efficient design of new and maintenance of

existing simulation models. Multi-disciplinary

simulation studies with a policy-supporting objective

could benefit from a generic library of reusable

components because it reduces the effort required

from generalist modellers for the design of

integrated simulation models (Donatelli et al., 2008).

In turn this will leave more room for the

communication with stakeholders and addressing

scientific issues. The advantages and challenges of a

component-based design process for simulation

modelling is a topic of growing interest (Argent,

2004; Papajorgji, 2005; Rizzoli et al., 2008;

Verbraeck and Valentin, 2008). The aim of a

recently concluded research project SPICOSA

(www.spicosa.eu) was to develop a methodology for

science-policy interfacing in the field of integrated

coastal zone management. The scale and integrated

nature of the project, need to exchange model

fragments, range of different research themes and

component-based modelling platform used made the

project an interesting testing ground for the concept

of reusable model components (De Kok et al., 2010).

103

de Kok J. and Engelen G..

LINKING AND REUSING MODEL COMPONENTS FOR ENVIRONMENTAL SIMULATION .

DOI: 10.5220/0003576001030108

In Proceedings of 1st International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH-2011), pages

103-108

ISBN: 978-989-8425-78-2

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

The project was organized around the different

design phases of integrated simulation models

linking the ecological, economic and social aspects

to the promising management strategies. Regular

consultations with stakeholders and representatives

of coastal planning institutes formed a key aspect of

the project. The modelled themes ranged from

fisheries and shell-fish farming to beach

attractiveness and coastal and inland eutrophication.

ExtendSim® (www.extendsim.com) served as

common simulation platform to allow for easy

exchange of model constructs between the different

project teams. This modelling environment supports

the component-based, hierarchical design of

dynamic simulation models to describe discrete and

continuous systems. Drag-and-drop linking of

reusable model components which are stored in

different libraries makes the design and maintenance

of models relatively easy, even for less-experienced

modellers. In addition, a large number of tools are

available to customize and improve the layout and

control of the model, an advantage in projects

developing policy instruments to be used in

interactive sessions with stakeholders. Simulation

models can be designed and modified by linking

standard components based on graphical icons, or by

means of more powerful coding in ModL, a

programming language similar to C (Fig. 1).

Modellers can also reorganize their components in

custom libraries to support the exchange of model

constructs. The project teams were required to

contribute one or more reusable model building

components to a generic model library, an important

deliverable of the project. Typical design criteria for

model components are reusability, flexibility,

encapsulation, interoperability (Papajorgji, 2005;

Rizzoli et al., 2008; Verbraeck and Valentin, 2008)

as well as framework independence (Rizzoli et al.,

2008). General guidelines for the design of

components were issued in the project (De Kok et

al., 2010) to help the modellers with their task and

ensure a certain degree of reusability of the

delivered components. The practical implementation

of the design guidelines turned out to be less

straightforward than expected, and some interesting

lessons can be drawn from the project. Here we take

a few examples of component linking and reuse to

reflect on the scientific and technical difficulties.

2 DESIGN APPROACH

The majority of environmental modellers are less

familiar with component-based design, because it is

not their first priority. The common practice is that

models expand gradually and iteratively in terms of

the functionalities included while gaining in

complexity. Frequent problems with the early

deliveries to the model library in this project

included a lack of separation of models and data,

excessive and inappropriate use of connectors,

overcomplexity, inadequate documentation and

testing, and functionalities that were too much

dependent on the case study modelled. The

advantages and general principles of component-

based modelling therefore had to be clarified by

means of tutorial examples taken from different

application domains. A literature review of the

design criteria that were mentioned earlier formed

the starting point for a number of more concrete,

project-focused guidelines and templates to

standardize the design process and avoid the

common pitfalls. Domain-independent components

with functionalities related to up- and downscaling,

user dialogues, im- and export of data, encryption,

statistics etc. were developed to support the project

teams with the design and polishing of their

simulation models. For organizational reasons the

majority of the components were designed by means

of a top-down procedure, and derived from the

simulation models at a late stage in the project. The

advantage of this approach is that the modelling

experience and data needed for the calibration of the

components were available, and the priorities were

clearer. On the other hand, the design of the

simulation models could not fully benefit from the

component-based design approach, compared to a

bottom-up approach.

Repeated exchanges to obtain feedback from the

support team and more experienced modellers

helped the project teams with the polishing and

improvement of their components. The approach for

designing the model components addressed the

following aspects:

1. First, it was necessary to identify which

functionalities could be useful to others. Here,

functionality is defined to be the modelled process

with in- and outgoing state variables. An example is

a component which describes the population

dynamics of a fish stock, with the fishing effort as

input and stock size as output variable. The

mathematical and visualization tools available in the

standard ExtendSim® libraries are typical examples

of components with a high degree of reusability, but

pertain to a low level of model detail. More complex

and process oriented components were needed to

reduce the effort required for developing a new

simulation model. On the other hand, components

SIMULTECH 2011 - 1st International Conference on Simulation and Modeling Methodologies, Technologies and

Applications

104

Figure 1: Icon-based (left) and ModL coded (right) component for logistic population growth.

which are too complex and case-dependent can be

expected to be less useful for application in other

modelling studies. Choosing the appropriate level of

detail, also referred to as the granularity (Donatelli

et al., 2008) of a component is essential. Screening

resulted in a tentative list of over 80 candidate

components pertaining to different model domains

and levels of detail. These components were

classified in five subcategories: ecological,

economic, social, physical, and domain-independent

support components. Fig. 2 shows an example of a

phytoplankton dynamics component from the

ecological sublibrary.

2. The technical implementation was relatively

straightforward once the functionality of a

component had been specified, using the supporting

graphical tools available in ExtendSim®. The next

step was then to restructure the component and

remove redundant connectors and model-specific

parameters as much as possible, which would

otherwise impede the connection to components

with complementary functionalities. The use of in-

and outgoing connectors was strictly limited to state

variables, whereas model-specific parameters and

input data were stored in databases or set in user

dialogues. The use of global arrays to store and

exchange model parameters was also discouraged to

in order to reduce the model dependence. Instead the

general recommendation made was to separate the

model code as much as possible from the data, use

generic indexing of database records, and add

transparent user dialogs to pass case-specific model

parameters. Some components were extended with

control options to create or access databases on user

demand, thereby increasing the level of flexibility.

3. After restructuring of each component the design

process focused on the layout and documentation.

Standardized templates were developed to ensure a

certain degree of uniformity and proper description

of the functionality, scientific background,

application domain, and references to the

developers. The use of simplified demonstration

applications with sample input was encouraged in

order to clarify the use of a component to other

modellers.

4. Next, the project teams were asked to verify and

test their components, to ensure the reliability, a

prerequisite for use by others. The components were

extracted from the supportive library and preferably

pasted into a blank model worksheet to test the

functionality with different sets of input variables,

whereas model-specific parameters were restricted

to their physical range. These tests included

component linking if possible and meaningful.

5. The general recommendation was to focus on the

genericity and reusability of the components for

other applications and modellers. Components were

reused within and between simulation models and

frequently exchanged among the modellers. A

proper component should allow for changes in use,

such as the number of fish species. Two other design

criteria were equally important but more difficult to

achieve. Although the interoperability of

components has been addressed in the past (Chiang,

2003) this important design criterion received only

limited attention in the project because most of the

simulation models were used to derive components

rather than the other way around. Moreover, the

majority of the study teams delivered only a few

components related to the processes modelled in

their case study. Nevertheless, the study sites were

asked to limit the number of in- and outgoing

connectors and reserve these for the key state

LINKING AND REUSING MODEL COMPONENTS FOR ENVIRONMENTAL SIMULATION

105

Figure 2: Component for phytoplankton dynamics.

variables to improve the interoperability of their

components. The complementarity of the

components was limited by the modelled domains

and processes, with insufficient contributions from

the social and economic sciences to fully meet the

demands for supporting ICZM. The expectation is

that this imbalance will gradually improve once a

wider community of modellers starts using the

model portal.

3 APPLICATIONS

A total of over fifty reusable components had been

delivered by the end of the project, with overlapping

functionalities and a general focus on physical and

ecological functionalities such as light extinction

and algal development. ExtendSim® can

automatically assign a global simulation order to all

components in a model. This can cause problems

after replacing components passing information at

distinct time steps in a hierarchically organized

model. The problem was solved by setting the

simulation order manually. To avoid these problems

such reusable components should verify the proper

and timely passing and initialization of variables. A

tutorial model of complementary components from

different sublibraries was developed to demonstrate

the advantages of component-based modelling for an

intermediate review of the project (Fig. 3). The

simulation model links plankton dynamics to shell

species growth and the income from harvesting.

Adding and linking components, completing the user

dialogues, and importing the databases takes only a

few minutes for this model. Because the components

are taken from different sublibraries and case studies

this example also demonstrates the complementarity

and connectivity of the components.

Powerful database functionalities are available in

ExtendSim® to drive the simulations, read model

input and store model output. A general

template icon

with connectors

documentation template

list of in- and outconnectors ModL com

onent scri

SIMULTECH 2011 - 1st International Conference on Simulation and Modeling Methodologies, Technologies and

Applications

106

Figure 3: Demonstration model linking components from different sub libraries.

recommendation was to let the users of the model set

the path and label input data files via the user dialog,

rather than defining these in the component script

internally. An example is a user dialog (Fig. 4),

based on resizable tables with economic data and

different options to create new or delete existing

tables, with automatic testing of the size of arrays

being im- or exported. The structure of the database

tables automatically determine the number of

economic sectors.

Figure 4: Tutorial example of linked components for

economic input-output analysis and tourism demand.

4 CONCLUSIONS

Component-based simulation of environmental

systems has considerable advantages: models are

easier to design, restructure and maintain, and model

constructs can be exchanged and combined more

easily. The most important design criteria for

reusable components for integrated environmental

simulation are genericity, functional

complementarity, proper documentation, reliability

and modelling framework independence (Rizzoli et

al., 2008; Verbraeck and Valentin, 2008). A web-

based model portal was set up to support the future

exchange, maintenance and documentation of the

reusable components. This portal will be maintained

after the project and a quality assessment of the

deliveries is currently being made. The majority of

the components delivered by the case studies

pertained to ecological processes with less attention

for social and economic applications. Some of the

delivered components, such as those describing the

population dynamics of fish stocks and shell species,

still are to be recombined into more generic tools. A

following step is then to gain wider support for the

model portal to encourage future contributions, in

particular those of the social sciences. Although the

components developed in this project depend on a

specific simulation platform the ModL coded

components can be transferred to a different

modelling framework, for example using enhanced

interface semantics as proposed by Rizzoli et al.

(2008) instead of the user dialogues and databases

which are currently used. Wide acceptance by a

community of domain experts depends on proper

documentation and concrete tutorial applications of

modelling tools targeting an audience including the

less skilled simulation modellers. Although the

simulation environment used in this project

adequately supports component-based modelling it

does not guarantee the reusability of components.

Modellers should become aware of the

consequences of reuse of their work in a different

model context at an early stage of projects. The

general lesson of the project is that the majority of

researchers are still insufficiently familiar with the

concept of modular design, although the advantages

are generally understood. Tutorial examples and

demonstration projects are not sufficient by

themselves to pave the way for component-based

LINKING AND REUSING MODEL COMPONENTS FOR ENVIRONMENTAL SIMULATION

107

modelling of (environmental) systems.

Organizational and technical resources such as the

web-based model portal will be needed to support

the exchange, quality improvement and management

of reusable model components. Whether the existing

model portal should be generally accessible is a

matter of discussion. The open-source concept

seems to be the best way to ensure sufficient and

adequate feedback on the contents of the model

library, a necessity for quality improvement.

Reclassification in functional categories and

screening of existing simulation models will help

identify missing components. A functional model

library should not only include abstract, generic

objects but also abundant concrete tools at a higher

hierarchical level to ensure general acceptance

among domain experts. Ultimately these efforts

should lead to a gradually expanding library of

reusable model components representing processes

from different domains, including those of the

economic and social sciences, to meet the demands

of all potential users.

REFERENCES

Argent, R. M., 2004. An overview of model integration for

environmental applications – components,

frameworks, and semantics. Environmental modelling

& Software 19, 219-234.

De Kok, J. L., Engelen, G., and Maes, J., Towards model

component reuse for the design of simulation models –

a case study for ICZM, Proc. 5

Biennial Meeting:

International Congress on Environmental Modelling

and Software. Swayne, D. A., Wanhong Yang, W.,

Voinov, A. A., Rizzoli, A., Filatova, T. (Eds.)

International Environmental Modelling and Software

Society, Ottawa, Canada, July 2010, 1215-1222.

Donatelli, M. and Rizzoli A. E., 2008. A Design for

Framework-Independent Model Components of

Biophysical Systems, Proc. 4

Biennial Meeting of

International Environmental Modelling and Software

Society. Sanchez-Marre M., Bejar J., Comas J., Rizzoli

A., and Guariso G. Eds., 727-734.

Papajorgji, P., 2005. A plug and play approach for

developing environmental models – Short

Communication. Environmental Modelling &

Software 20, 1353-1357.

Rizzoli, A., Donatelli, M., Athanasiadis, I. N., Villa, F.,

and Huber, D., 2008. Semantic links in integrated

modelling frameworks. Mathematics and Computers

in Simulation 78, 412-423.

Verbraeck, A. and Valentin, E. C., 2008. Design

Guidelines for Simulation Building Blocks. Proc.

2008 Winter Simulation Conference. Mason S. J., Hill

R. R., Mönch L., Rose, O, Jefferson T. and Fowler J.

W. Eds., 923-932.

Chiang, C. C., 2003. The use of adapters to support

interoperability of components for reusability

Information and Software Technology, Volume 45,

Issue 3(1), 149-156.

SIMULTECH 2011 - 1st International Conference on Simulation and Modeling Methodologies, Technologies and

Applications

108