Integrating Distributed Computational Models as

Dynamic Expressions of Knowledge: The Case for

Evaluating Measures for Urban Ecosystem Sustainability

Steven Kraines

Future Center Initiative, The University of Tokyo

Environmental Science Research Building, Rm 277

5-1-5 Kashiwa-no-ha, Kashiwa-shi, 277-8563, Chiba-ken, Japan

Abstract. In order to conduct simulation studies of highly complex problems,

such as integrated environmental assessment of technologies, policies and other

measures for making urban ecosystems more sustainable, integration of know-

ledge in a dynamic form from a wide range of knowledge domains is essential.

We have proposed that computational models are appropriate representations of

expert knowledge for such integration. Building on previously introduced con-

cepts and software prototypes, we have been designing a dynamic computa-

tional modeling platform for studying the integrated effects of supply side and

demand side technologies and policies to reduce environmental impacts and

consumption of resources caused by activities in urban ecosystems such as

Tokyo, Japan. This platform has been used to evaluate scenarios that include

the introduction of roof-top photovoltaics, a solid oxide fuel cell combined with

a gas turbine topping cycle (SOFC/GT), energy conservation measures in the

residential and commercial building sectors, and waste processing and recy-

cling systems. Here we describe the software engineering issues associated with

the construction of this model integration platform and provide examples of the

techniques we have developed to address those issues.

1 Introduction

Recently, in response to growing realization of the profound effect that urban activi-

ties are having on global sustainability, there is an urgent need for tools and methods

to comprehensively evaluate the potential effects of technologies and policies on

increasing the sustainability of a particular urban ecosystem [1]. Such comprehensive

evaluation of technologies and policies in actual real-world urban systems must in-

clude a wide range of expertise from different domains of knowledge ranging from

economics, to policy science, to engineering, to life cycle analysis [2]. If we are to

rely on information and communication technologies (ICTs) to assist us in integrating

this knowledge, the knowledge must be expressed in a form that can be examined and

manipulated by a computer.

The question “what is knowledge” has an entire field of knowledge devoted to it,

to which we could hardly give justice here. However, at least for expert knowledge in

Kraines S..

Integrating Distributed Computational Models as Dynamic Expressions of Knowledge: The Case for Evaluating Measures for Urban Ecosystem

Sustainability.

DOI: 10.5220/0003700100570066

In Proceedings of the 2nd International Workshop on Software Knowledge (SKY-2011), pages 57-66

ISBN: 978-989-8425-82-9

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

natural sciences, the old saw “data is points, information is a line through the points,

knowledge is knowing what the line means and wisdom is knowing what the line

does not mean” is remarkably insightful. Human expert knowledge is often described

as existing in the form of mental models, which are dynamic, generalized approxima-

tions of the real world phenomenon that is “known”. In particular, these mental mod-

els enable people to consider the meaning of that phenomenon, e.g. by making infe-

rences in relation to other things that are known.

Although printed documents have been traditionally used to express and commu-

nicate knowledge generated through scientific research, we have suggested that a

more appropriate format for representing scientific knowledge might be “computa-

tional models”, which we consider to include all forms of interactive software repre-

sentations of knowledge [2]. For example, computer simulations and other computa-

tional models are important media for expressing specific expertise or knowledge

about the dynamics of particular phenomena in urban ecosystems [3], [4]. In order to

integrate the expertise expressed in different models and to generally increase their

reusability, some way of modularizing and interfacing those models is needed that

preserves the dynamics and generality of that knowledge. New tools and methods

have emerged in disciplines such as software engineering that could help use these

computational models to integrate expert knowledge for studying complex entities

such as urban ecosystems [5], [6], [7], [8]. Here, we present work aimed at applying

some of these technologies to develop a distributed web-based model integration

environment for enabling multiple stakeholders to work together in designing and

evaluating combinations of measures for increasing sustainability in urban ecosys-

tems.

In previous publications, we have described the development and application of a

“virtual urban ecosystem” in the form of a computational model integration environ-

ment on the web [6], [9], [10]. The virtual urban ecosystem simulation platform uses

the DOME (distributed object-based modeling environment) software to make avail-

able a set of computational models that give the boundary conditions of the urban

ecosystem, such as census databases, input-output tables, electric power dispatch

models, transportation models, and atmospheric circulation models. Another set of

interfaces in the urban ecosystem simulation platform link to a variety of models of

technologies and policies that have been designed to make the urban ecosystem more

sustainable, e.g. by reducing environmental impacts or consumption of non-

renewable natural resources. By combining the boundary-condition models with the

technology and policy models, we can more comprehensively assess how the intro-

duction of new technologies and policies might impact on environment, economy and

society. We have described applications of the platform to the evaluation of specific

technologies in Tokyo and other parts of Japan, from introduction of roof-top

mounted photovoltaics to comprehensive plastic separation, collection and reuse

system [9], [10], [11], [12], [13], [14]. In this paper, we focus on issues related to

construction of the platform from a software engineering perspective, and we present

some of the techniques we have used to address those issues.

In the next section, we describe the DOME software that we use as the middle-

ware for constructing the virtual urban ecosystem platform, and we give examples of

two of the modularized computational models that we have integrated by using

DOME. In section 3, we describe our work to improve accessibility of the simulation

platform by embedding it in a web-based collaboration platform. We end by discuss-

ing some ongoing problems related to model interfacing.

2 Supporting Integration of Dynamic Knowledge using DOME

The approach that we have taken for integrating dynamic expressions of expert know-

ledge related to different aspects of urban ecosystems is based on the DOME software

being developed at the CAD laboratory in the Massachusetts Institute of Technology

[8]. The DOME software offers tools that provide a complete environment to devel-

op, interface, integrate and evaluate sets of computational models over the World

Wide Web. DOME tools allow modelers to build, deploy, browse and integrate model

interfaces to different computational models, which have been constructed using

different software applications such as MS Excel and Matlab. The interfaces can be

parametrically interlinked to form DOME integrated model projects. Specifically, the

build tools enable the model developer to create web-accessible objects (e.g. scalar

and matrix objects that have a wide range of attributes for physical units, annotation,

ownership, etc.) that are mapped to the input and output parameters of the model. The

build tool also provides mechanisms for creating simple computational relationships

or “bridge models” between the numerical objects. In fact, the DOME relationship

building environment contains sufficient numerical operators, such as a full range of

matrix operators, to construct complete models of technologies to be linked into a

DOME integrated model project.

In previous work, we have designed a DOME integrated model project as a com-

prehensive simulation platform for evaluating the overall effects of introducing a

particular combination of technologies and policies to a target urban ecosystem [9].

Here, we describe two examples of computational models that have been interfaced

and deployed to the model project. The first is an optimization model for central

power grid planning and dispatch, and the second is a building energy consumption

model that includes roof-top installed photovoltaic solar cells.

2.1 An Example of a Modularized Computational Model for Describing the

Urban Ecosystem: The “Power Planning and Dispatch Model”

The “Power Planning and Dispatch Model” has been described in detail in [10]. The

DOME input objects include a matrix object that contains the representative power

demand curves (e.g. summer weekday, summer weekend, winter weekday, winter

weekend, mid-season weekday, mid-season weekend, and peak power demand day)

in the targeted urban ecosystem, and a set of vector objects containing the installation

costs, running costs, conversion efficiencies, capacity limits, and load following con-

straints for each of the power generation types available in the region (Fig. 1). The

output objects are a matrix object giving the hourly dispatch to each power generation

type for the representative power demand curves, and scalar objects giving the total

emissions and total cost.

In the instance of the “Power Planning and Dispatch Model” used in [10], the

power generation types consisted of hydropower, nuclear power plants, LNG fired

power plants, coal fired power plants, oil fired power plants, gas combined cycle

power plants, energy recovery using pumped water storage, and a novel technology

that combines a solid oxide fuel cell, which uses natural gas, with a natural gas tur-

bine topping cycle (SOFC/GT) [11]. The basic DOME interface for the power plan-

ning and dispatch model showing the input and output parameter objects is shown in

Fig. 1. A custom graphic user interface that we developed for the model is shown in

Fig. 2.

Fig. 1. The default DOME interface for the “Power Planning and Dispatch Model”.

Fig. 2. The custom graphic user interface for the “Power Planning and Dispatch Model”.

2.2 An Example of a Modularized Computational Model for Describing a

Technology: The “Residential Building Energy Conservation Model”

In order to evaluate the potential for reducing electricity consumption in the residen-

tial sector, we have used DOME to modularize the computational model for residen-

tial buildings shown in Fig. 3 [15]. The model, which is implemented in MS Excel

with VBA, uses inputs in the form of a set of demand-side countermeasures selected

by the user and a specified region of Japan. It then outputs the consumption of elec-

tricity, natural gas and other energy sources by all of the residential buildings in that

region from the year 2000 to the year 2050 averaged over five year time periods. The

demand-side countermeasures include both standard energy conservation measures

such as window insulation and efficient electric appliances, which are represented as

DOME Boolean objects, and demand-side power generation using roof-top mounted

photovoltaic (PV) cells, which is represented as a DOME vector object giving how

much area of PV cells will be introduced each year. Outputs are represented as

DOME vector objects giving the demand at each time period for each energy type.

Input data

Region

Countermeasure type

PV introduction rate

Output data

Annual Aggregated

Energy Consumption

between 2000 and 2050

(5 year intervals)

Calculation

Model

(Excel VBA)

Fig. 3. Outline of the “Residential Building Energy Consumption Model”.

3 A Web-based Dynamic Knowledge Integration Platform

We have been developing a Web-based collaboration platform to support communi-

cation of knowledge between different experts towards the goal of evaluating tech-

nologies and policies for mid to long term mitigation of CO

emissions in urban re-

gions in order to achieve Low Carbon Societies (LCSs) [16]. The platform is intended

to form a structured basis for supporting smooth and seamless exchange of know-

ledge from multiple disciplines, knowledge that is related to various measures for

achieving LCSs. A central component of this platform is a dynamic web interface

called the “CO

Tech Table”. The CO

Tech Table enables experts to interactively

manipulate knowledge regarding the characteristics of CO

mitigation options in the

context of specific CO

intensive urban activities; therefore, we can consider it to be a

form of computational model of that knowledge.

The CO

Tech Table has three parts, which are indexed to specific urban ecosys-

tems being studied. On the left, an expandable tree of the major CO

emitting activi-

ties allows users to estimate the contribution of each activity to the total CO

emis-

sions. In the center, the user lists models and data resources for evaluating counter-

measures for CO

emissions from a specific urban activity. On the right, the user

enters the technologies and policies for CO

mitigation being studied, together with

expected reduction of CO

emissions and time scales, etc. The semantics for each of

the parts of the CO

Tech Table are grounded in a formalized ontology that we have

developed to express knowledge in the domain of sustainability science [17].

The CO

Tech Table also includes a database of regional information for urban

ecosystems that are being studied. In the particular application we have implemented

for the study of LCSs in Japan, the target cities are Tokyo, Sapporo, Naha, Kagoshi-

ma, and Utsunomiya. These cities were judged to be representative of the variation in

size, climate conditions, and lifestyles of cities in Japan. Each user is provided with a

set of columns for each urban region or city that is targeted. In addition, the CO

Tech

Table supports multiple user accounts, and it manages the personal information of

each user (such as contact information and publication lists) together with the tables

for classifying and organizing knowledge related to mitigation of CO

from urban

activities.

A view of the CO

Tech Table that is filled out with information for the target city

Tokyo is shown in Fig. 4. The user, Steven Kraines, has entered several urban activi-

ties, classified into general categories, which are logically defined within the ontology

framework. For example, various activities related to electricity consumption by ap-

pliance use and manufacturing processes are grouped under the category “electricity”.

In the center, the user has indicated that a dataset called “central power grid electrici-

ty demand” is necessary for investigating the CO

mitigation potential associated with

electricity consumption in urban regions. In addition, for each of the different types of

electricity consumption, the need for data such as lifestyle parameters and building

floor area is indicated. Finally, in the right column, the user has listed the technolo-

gies and policy options he believes to be effective for realizing CO

mitigation in the

context of each of the types of activities given in the left column.

The collaboration platform also lets members publish a variety of knowledge re-

sources to the web, including computational models. We have used the DOME soft-

ware to create simple web page interfaces that let users dynamically operate the com-

putational models that have been published to the web platform through DOME. By

connecting the model interfaces to the model requirement specifications in the CO

Tech Table, we can relate the interfaced models to the ontology-based knowledge

classification that is expressed there.

Fig. 4. “CO

Tech Table” view showing major urban needs and activities of interest to a par-

ticular user indexed to the types of models and datasets thought to be required for evaluating

that need and the technology and policy options being considered by the user to fulfill the need.

When a user publishes a computational model such as a simulation or database to

the web platform, the user can create input and output fields corresponding to the

DOME objects described by the model interface, as shown in Fig. 5. The input fields

are editable, so that other users may enter values into these fields to express the con-

ditions for model calculation that reflect the particular scenario being studied.

Fig. 5. Web page showing the basic view of one of the DOME models interfaced to the CO

Tech Table collaboration platform.

We have developed a server daemon for the web platform that listens for changes

made to the model parameters and executes the actual computational models through

the DOME software infrastructure. Results are returned by DOME to the server dae-

mon, which then passes the results back to the user interface. A schematic diagram of

the software system is shown in Fig. 6.

In addition to the basic web page view that is provided by the collaboration plat-

form for the models that have been registered to the site, customized layouts includ-

ing dynamic graphing functionality can be created for particular models, so that data

returned from the DOME simulation model servers can be rendered as various kinds

of graphs. We have implemented custom interfaces for several models, including the

“Residential Building Energy Conservation Model” described in this paper.

Fig. 6. Schematic diagram of the software system linking the collaboration platform web site to

the DOME model integration infrastructure.

4 Discussion

We have been using the DOME modeling integration architecture to construct sys-

tems of computational models for studying a number of urban ecosystem processes

[9], [10], [11], [12], [13], [14]. While the model integration approach has enabled us

to draw together a wide range of expert knowledge from researchers in different aca-

demic domains, we have encountered several reoccurring problems when actually

trying to integrate computational models that have been developed by different re-

searchers for different research objectives. Because the actual task of integrating two

computational models consists of connecting their interfaces, interface design is criti-

cal. Of course, multiple interfaces can be created for a single computational model to

satisfy different integration needs. Furthermore, specification of units and dimensions

for the interface parameters together with the ability of DOME to perform unit con-

versions and build simple bridge models between interfaces helps to address many of

the classical issues of model integration resulting from both structural and semantic

gaps between models.

However, one of the most difficult problems that we have had in creating inte-

grated systems of computational models concerns not gaps between interfaces, but

overlap. Particularly in creating a model of a complex system, a modeler often feels

compelled to include all important aspects of the system being modeled, even if many

of those aspects are actually out of the scope of the modeler’s expertise. In many

cases, we have found it necessary to identify what to “cut out” of a model in order to

limit its scope to the part of the system that it is intended to represent. For example,

we had to remove the transformation of electricity use in the “Residential Building

Energy Conservation Model” to CO

emissions in order to integrate that model with

the “Power Planning and Dispatch Model”. More attention by modelers to clearly

define the scope of their computational models would help to increase the general

reusability of the models and facilitate the conceptual tasks of integrating them with

other models.

The afore-mentioned problems have obvious counterparts in classical software

engineering, and it is the author’s hope that this workshop will offer an opportunity

for exploring those relationships and discovering important lessons to be learned.

5 Conclusions

Methods that leverage cutting edge Information and Communication Technologies

are urgently needed to help experts from different research domains to integrate their

respective expert knowledge in order to carry out comprehensive analyses of complex

system problems such as integrated environmental assessments of combinations of

new policies and technologies for urban ecosystem sustainability. The DOME soft-

ware provides a web-based infrastructure for meeting this need. In the paradigm that

we are pursuing, knowledge is expressed in the form of computational models [2].

These computational models are made available over the Internet using the DOME

software, much the same way that web servers make html web pages available on the

World Wide Web [8]. Furthermore, like with hypertext links in the conventional

World Wide Web, computational models distributed on different DOME servers

around the world can be integrated by connecting the model input and output parame-

ters in order to form integrated model projects that can be used for conducting com-

prehensive assessments of complex system behaviors. This paper has discussed some

of the technical issues involved in doing this kind of model integration and the solu-

tions that we have adopted.

Acknowledgements

This research was supported by the Global Environment Research Fund of the Japa-

nese Ministry of the Environment, S-3-3: “Japan Low Carbon Society Scenarios

toward 2050: Scenario development and its implication for policy measures, effects

of introducing countermeasures for carbon dioxide emission reduction in urban areas”.

References

1. Takeuchi, K., Komiyama, H.: Sustainability Science: Building a New Discipline. Sustaina-

bility Science 1(1) (2006) 1–6

2. Kraines, S. B, Batres, R., Koyama, M., Wallace, D. R., Komiyama, H.: Internet-based

Collaboration for Integrated Environmental Assessment in Industrial Ecology – Part 1.

Journal of Industrial Ecology 9 (2005) 31–50

3. Ausubel, J. H.: The Virtual Ecology of Industry. Journal of Industrial Ecology 1 (1) (1997)

10–12

4. Rejeski, D.: Learning Before Doing: Simulation and Modeling in Industrial Ecology. Jour-

nal of Industrial Ecology 2 (4) (1998) 29–44

5. Borland, N., Wallace, D. R.: 1999. Environmentally Conscious Product Design: A Colla-

borative Internet-based Modeling Approach. Journal of Industrial Ecology 3 (2 & 3):33–46

6. Kraines, S. B., Wallace, D. R.: Sustainability Technology Evaluation in a Regional Context

using Distributed Information Technology. Computers, Environment and Urban Systems

27 (2) (2003) 143–161

7. Villa, F., R. Costanza, R.: Design of Multi-Paradigm Integrating Modeling Tools for Eco-

logical Research. Environmental modeling and Software 15 (2000) 169–177

8. Wallace, D. R., Abrahamson, S., Senin, N., Sferro, P.: Integrated Design in a Service Mar-

ketplace. Computer-aided Design 32 (2) (2000) 97–107

9. Kraines, S. B., Wallace, D. R., Iwafune, Y., Yoshida, Y., Aramaki, T., Kato, K., Hanaki, K.,

Ishitani, H., Matsuo, T., Takahashi, H., Yamada, K., Yamaji, K., Yanagisawa, Y., Ko-

miyama, H.: An Integrated Computational Infrastructure for a Virtual Tokyo: Concepts and

Examples. Journal of Industrial Ecology 5 (1) (2001) 35–54

10. Kraines, S. B., Ishida, T., Wallace, D. R.: Integrated Environmental Assessment of Supply-

Side and Demand-Side Measures for Carbon Dioxide Mitigation in Tokyo, Japan. Journal

of Industrial Ecology 14 (5) (2010) 808–825

11. Koyama, M., Kraines, S. B., Tanaka, K., Wallace, D. R., Yamada, K., Komiyama, H.:

Integrated Model Framework for the Evaluation of an SOFC/GT System as a Centralized

Power Source. International Journal of Energy Research, 28 (1) (2004) 13–30

12. Kraines, S. B., Shigeoka, H., Wallace, D. R., Komiyama, H.: Development of an Internet-

based Collaboration Platform and Application to Household Waste Plastic Processing. Intl

Journal of Technology Transfer and Commercialisation, 3(2) (2004) 129–146

13. Kraines, S. B., Koyama, M., Weber, C., Ikaga, T., Chikamoto, T., Wallace, D. R., Ko-

miyama, H.: A Collaborative Platform for Sustainable Building Design Based on Model In-

tegration over the Internet. Intl Journal of Technology Transfer and Commercialisation.

5(2−3) (2005) 135–161

14. Xia, H., Koyama, M., Leyland, G., Kraines, S. B.: A Modularized Framework for Solving

an Economic-Environmental Power Generation Mix Problem. International Journal of

Energy Research, 28 (2004) 769–784

15. Ikaga, T., Miura, S., Tonooka, Y., Shimoda, Y., Koike, K., Fukasawa, O., Mizuishi, T.:

Development of Macro Simulation Method on Household Energy Consumption and CO

Emission by each Administrative Division. AIJ Journal Technol. Des. 22 (2005) 263–268

16. Kraines, S. B.: Using Ontology-Based Semantic Matching for Sharing and Integrating

Expert Knowledge to Evaluate Scenarios for Low Carbon Societies. In preparation

17. Kraines, S. B., Guo, W.: A system for Ontology-Based Sharing of Expert Knowledge in

Sustainability Science. Data Science Journal, 9 (2011) 107–123