Realising the Vision of the Semantic Grid forComputational Land-Use Modelling

P. Edwards∗, E. Pignotti∗, A. Preece∗, N. Gotts†, and G. Polhill†

*Dept. of Computing Science, University of AberdeenAberdeen, AB24 5UE, Scotland

{ pedwards, epignott, apreece}@csd.abdn.ac.uk,WWW home page: http://www.csd.abdn.ac.uk/research/fearg

†The Macaulay InstituteCraigiebuckler, Aberdeen, AB15 8QH, UK{n.gotts, g.polhill}@macaulay.ac.uk

WWW home page: http://www.macaulay.ac.uk/fearlus/

Abstract. In this paper we explore the vision of the Semantic Gridby discussing deployment of a computational land use modelling ser-vice. The FEARLUS-G service architecture is presented which allowslarge scale simulation experiments to be distributed over the Grid. Wealso discuss ontology support for simulation parameters, hypotheses andresults that will facilitate sharing and re-use of such resources amongland-use scientists, together with the creation of scientific arguments.

1 Introduction

Collaborations between large groups of scientists are now seen as essential toenhance the scientific process. While research has always involved collaborationbetween individuals, there is now even greater necessity for tools to supportsharing of knowledge, resources, results and observations. For these reasons re-cent e-science activities [1] have focused on facilitating and promoting collab-oration between scientists using advanced distributed information managementsystems. The vision of e-science is to facilitate large scale science using Grid tech-nologies [2] as a fundamental computing infrastructure to manage distributedcomputational resources and data. However a major gap exists between currenttechnologies and the vision of e-science. The Semantic Grid [3] [4] aims to fa-cilitate ease-of-use of Grid resources and flexible collaboration on a global scale;central to this vision of a ‘smart’ Grid is the use of Semantic Web [5], softwareagent and Web services technologies (amongst others).

The Semantic Grid requires adoption of metadata and ontologies [6] to de-scribe resources, services and data sources in order to promote enhanced formsof collaboration among the scientific community. Ontologies and metadata facil-itate intelligent search mechanisms, one of the key enablers through which suchservices could be realised. The FEARLUS-G system, described in this paper, ex-plores the application of emerging Grid and Semantic Grid technologies within

the social sciences, through deployment of an existing land-use modelling toolinto the (Semantic) Grid context.

The project is one of a number of pilot studies funded by the UK Eco-nomic & Social Research Council (ESRC), under their e-social science1 initia-tive. FEARLUS [7] is an agent-based model of land-use change developed at theMacaulay Institute in Aberdeen. The system contains objects that represent hu-man decision-makers in the real-world (land managers) and takes into accountattributes such as yield from land parcels. Parameters to the modelling environ-ment allow a variety of land-use strategies and their outcomes to be explored.

We chose to deploy FEARLUS as a Grid service for a number of reasons:firstly, the Grid infrastructure would allow large land-use simulation experimentsto be distributed across the Grid; a typical FEARLUS experiment already in-volves running a series of simulations, but on a single machine. Secondly, the useof Semantic Web technologies combined with Grid infrastructure would allow usto create a co-laboratory where land-use scientists can access the FEARLUS-Gservice and share and reuse results and observations.

In the remainder of this paper we briefly discuss the FEARLUS model ofland-use simulation. We then describe the FEARLUS-G service architecturepresenting its various components. We also introduce an ontology to supportdescription of simulations, hypotheses, experiments and a simple form of scien-tific argument structure; a user interface which allows scientists to create andaccess meta-data enriched resources is presented. As a contribution to infrastruc-ture for the Semantic Grid we present a solution which provides storage, queryand retrieval of semantic data to allow a virtual community of land-use scientiststo share and reuse FEARLUS-G experimental results and observations.

2 Simulation Modelling of Land-Use Change

Using computer technologies to study social and economic phenomena is nowcommonplace [8]; such processes are often studied via simulation modelling [9].In such simulations, the model consists of a representation of the structure andbehaviour of a particular real world entity which we wish to study. The aim ofthe simulation is to construct a model where behaviour matches the real entityin at least a significant aspect. By constructing such a model, social scientistsaim to develop conclusions that provide insight into the behaviour of real worldentities or phenomena; such modelling is often exploratory. Alternatively thesimulation may be used to confirm how reliable a predicted behaviour is undercertain key conditions which may or may not be under direct control.

There are often many competing models available, and the problem then ishow to choose between different types of model in a particular problem context.It may also be difficult to interpret the behaviour from a model even when it hasbeen previously validated experimentally. Ideally a model should be structuredin such a way that it is possible to determine if a particular question can beanswered using it.1 http://www.ncess.org/

There has recently been a proliferation of computer models of land-use changeand water management; many of these are spatially explicit models, in whicha set of distinct localities, and spatial relationships between them, are directlyrepresented [10]. As computational resources, and the availability of machine-readable georeferenced data increase, growth in spatially explicit modelling withinthese domains is likely to continue. Attention is now turning to how such modelscan most effectively be used, both in management and policy-related applica-tions, and in social and geographical science.

In FEARLUS Model 0-6-5 [7] parameters to the modelling environment allowa variety of land-use strategies and their outcomes to be explored. FEARLUSaims to improve understanding of land use change, particularly as regards ruralScotland. The agent-based simulation component in FEARLUS is implementedusing the Swarm system developed at the Santa Fe Institute [11]. Swarm providesan environment that can handle large experiments using agent-based simulationmodels. A FEARLUS simulation might involve for example, studying the dy-namics of imitative2 and non-imitative approaches to land use selection changeunder different circumstances, in the context of environments differing in spatialand temporal heterogeneity. This involves an initial set of exploratory studieswhere the model is run in a simulation with different initial parameters and theoutcomes observed. The result of the exploratory study is the formulation of ahypothesis regarding patterns of behaviour of specific aspects of the land usemodel. Figure 1 shows the graphical output of an interactive run of the FEAR-LUS model. It uses graphical output to render different aspects of the model inorder to facilitate exploratory studies. Using experimental validation it is thenpossible to check whether the model does consistently show the patterns of be-haviour suggested by the exploratory studies. Further details of the FEARLUSmodel are available [7].

One of the key justifications for our work is improving the scientific rigourof agent-based modelling. One of the important aspects of science is that workshould be repeatable and verifiable, yet results gathered from possibly hundredsof thousands of simulation runs cannot be reproduced conveniently in a journalpublication. Equally, the source code of the simulation model, and full detailsof the model parameter used are also not journal publication material. We haveidentified four levels of repeatability that are of relevance to such situations:

The first level is being able to access the results data itself, to check that theauthors’ claims (hypotheses) that are based on those results are justifiable. Thesecond level is being able to re-run the experiments to check that they producebroadly the same set of results, i.e. that the researchers have not carefully se-lected runs of their simulation that support their claims and ignored the rest.The third level is being able to manipulate the simulation model parametersand re-run the experiments to check that there is no undue sensitivity of the re-sults to certain parameter settings. The fourth level is being able to modify the

2 Imitation is a social phenomenon, e.g. land managers may be influenced by othermanagers owning a neighbouring land parcel.

source code and/or re-implement the model to check for what might be termed‘algorithmic sensitivity’.

Fig. 1. The FEARLUS Desktop Application.

3 FEARLUS-G Grid Service

We have developed FEARLUS-G - a collection of services to perform land usesimulation in the Grid context based on the existing FEARLUS land use modeland utilising Globus Toolkit 3. We have also developed a Web client applicationthat makes use of the services by uploading and running FEARLUS simulationsand managing scientific resources; the services are also accessible through anextended version of the existing FEARLUS desktop client (Figure 1). We choseto deploy FEARLUS as a Grid service for a number of reasons: firstly, the Gridinfrastructure allows large simulation experiments to be distributed across theGrid to make use of unused processing power; in fact, a typical FEARLUS exper-iment already involves running a series of simulations, but on a single machine.Secondly, the use of Semantic Web technologies with Grid infrastructure sup-ports the co-laboratory structure where land-use scientists and policy makers canaccess the FEARLUS-G service and share and reuse results and observations.

Figure 2 presents an overview of the FEARLUS-G architecture, which is builton top of Globus Toolkit 3. The five core components are as follows:

1. FEARLUS-G Service: The access point for FEARLUS-G simulations andexperiments. This service creates experiment or simulation instances by iden-tifying the components inside an experiment definition.

Base Index Service

Upload Service

FEARLUSG Service

ELDAS Data Access Service

MySQL XML DB

R e a d , W r i t e R

e a d

, W

r i t e

Scientific Objects Repository Service

FGExperiment Instance

Q u e r y , R e t u r n

h a n d l e

U p l o a d

U p l

o a d

JDBC4ELDAS

C r e a t e , R u n R e t r i e

v e S t o r e

Desktop Application

Run Simulation Experiments

FGSimulation Instance

FGSimulation Instance

Fig. 2. The FEARLUS-G Service Architecture.

2. Upload Service: Allows the client application to upload Scientific Objectsto the Grid application. Scientific Objects are defined in RDF [12] using theontologies described in Section 4 of this paper.

3. Scientific Objects Repository Service: Stores and retrieves the modelparameters and experiment definitions acquired from the Upload Service intopersistent storage. This service uses ELDAS [13] to manage data repositoriesfor RDF models; ELDAS was developed by the UK National e-Science CentreeDIKT project3.

4. FEARLUS-G Experiment: Allocates different simulation instances de-pending on the characteristics of the experiment. This service also collectsthe results from the various simulations and stores the results.

5. FEARLUS-G Simulation: Performs the simulation using FEARLUS model0-6-5 and stores the results. It creates FEARLUS runs, and store the resultsin the appropriate simulation instance in the Scientific Objects Repository.

FEARLUS-G uses the Globus base index service to identify available servicesthrough their service descriptors. Experiments and simulations can be allocatedto different nodes depending on the resources available. In our approach, usersprovide specifications of the experiments and the initial parameters for the sim-ulation. This information is stored using the Scientific Objects Repository. Thiscan be done via a client application integrated within the FEARLUS desktopclient (see Figure 1) or through a Web portal designed for FEARLUS-G. TheFEARLUS-G environment contains the middleware to find resources availableand the input data required to perform a specific simulation by making use of aservice descriptor associated with the FEARLUS-G service. The service descrip-tor contains the characteristics of the resources available in the node such as themaximum number of simulations and the simulation instances available.

3 http://www.edikt.org

4 FEARLUS-G Meta-Data Support

As described earlier in this paper, the classical approach to simulation studiesis to create a model of a real entity in order to analyse hypothetical behaviours.Simulations are used to generate numerical results representing future statesof the model under specific conditions. By testing the numerical results withan appropriate method it is possible to gather evidence to support scientificarguments. In this section we propose a conceptual layer through the definition ofan ontology which captures the concepts and relationships important to scientistsin order to improve the scientific rigour of agent-based modelling.

Figure 3 shows the core classes and properties in our ontology. It includesa collection of generic elements that are intended to be applicable to any e-science application, and also elements that are specific to simulation modellingand FEARLUS-G in particular.

Instances generated from the classes in the ontology are of three types: thosewhich need to be created by the user, e.g. the definition of an experiment; thoseinstances created by the Grid service, e.g. the set of simulation run instancesassociated with an experiment, and the experiment result instance summarizingthe results; the third group are instances created to support scientific argumen-tation, sharing and reuse. For example, an hypothesis can be associated withrelated publications and experiments.

The key class in our generic representation is the Hypothesis, which we con-sider to be a scientific concept that has not yet been fully verified. An hypothesisis “a tentative explanation that accounts for a set of facts and can be tested byfurther investigation”4. This implies that a scientific community works to sup-port an hypothesis by contributing publications, experiments and other relatedhypotheses that agree or disagree with the “target” hypothesis. In our vision,the classes HypothesisObject, ExperimentObject, SimulationObject and Publica-tion are subclasses of a generic ScientificObject class.

We define two properties agreesWithHypothesis and disagreesWithHypothesiswhich link any given scientific object to Hypothesis instances. Those propertiesallow us to support scientific argumentation in the FEARLUS context by linkingspecific evidence (FearlusExperiment, Run, FearlusResult) to hypotheses. Thegeneric scientific object representation can, of course, be adapted for differentscientific domains by extending it with more specific subclasses and properties.

We use the Dublin Core5 ontology to provide basic annotation of Scientifi-cObject instances. For example dc:creator defines the creator of an hypothesis,publication or experiment, dc:contributor the contributor(s). Figure 4 showssome examples. We also aim to provide support for refinement of hypotheses;in particular we use dc:replaces to represent that a new hypothesis replacesanother. Another important attribute is describedIn. This is used to recordthat a publication can be used to describe any given scientific object.

4 www.pages.drexel.edu/∼bcb25/scimeth/vocabulary.htm5 http://dublincore.org/

Created by

User

Created by

Grid Services

Support

Scientific Argument

SubClass Of

Relation

Key

ScientificObject

Publication

describedIn*

Hypothesis

agreesWithHypothesis* disagreesWithHypothesis* SimulationObjectHypothesisObject ExperimentObject

describesScientificObject*

ExperimentalHypothesis

hasTailedSignificanceTest Instance* TailedTest

hasStatus Instance KB_836936_Class_8

followsFromTheory Instance* Theory

hasType Instance HypothesisType

hasNullHypothesis Instance* NullHypothesis

NullHypothesis

Output

isOutputOf Instance Run

Run

hasRunDate String

hasPairedRun Instance* Run

hasOutput Instance* Output

hasModel Instance Model

hasParameter Instance* Parameter

...

isOutputOf

Model

hasOutput*

hasPairedRun*

SimulationExperiment

isRunOf hasModel

TwoTailedHypothesis

TailedHypothesis

OneTailedHypothesis

hasRuns*

FearlusExperiment

hasRuns Instance* FearlusRun

hasTerminationYear Instance TerminationYear

hasResult Instance FearlusResult

hasReportingYear Instance ReportingYear

Experiment

hasResult Instance Result

testsHypothesis Instance* ExperimentalHypothesis

hasPredictions Instance* Prediction

Result

isMeasuredBy Instance SignificanceTest

hasResultDate String

isResultOf Instance Experiment

hasResulttestsHypothesis*

isModelOf*

isResultOf

FearlusResult

hasNullHypothesis*

FearlusHypothesis

isResultOfhasResult

isNullOfHypothesis*

Fig. 3. Part of the FEARLUS-G Ontology Illustrating Hypothesis, Publication, Ex-periment & Run Concepts.

An ExperimentObject is intended to be any object that defines a real ex-periment, this includes the experiment definition (class Experiment) and theexperiment results (class Result). We define FearlusExperiment as a subclass ofExperiment to extend our generic ontology to include a FEARLUS experiment.A typical simulation experiment in FEARLUS (class FearlusExperiment) com-pares a number of subpopulations by making them compete in the simulationenvironment. A simulation experiment includes a set of simulation runs (classRun) in a specific simulation model (class Model) and the related output (classOutput). The instances of such classes are generated automatically from theFEARLUS-G Grid services by running a FEARLUS simulation, and annotatingthe system’s output. The annotated output for each run is stored in the Repos-itory Service (see Figure 2). This allows the system to recognize if a specificsimulation run has already been executed by searching the Repository Service.If a pre-existing run and its associated results are detected, the FEARLUS-Gsystem does not need to perform another simulation but only aggregate existingdata.

4.1 Semantic Data Repository Service

In the previous section we introduced a possible ontology for supporting scientificresearch via the Semantic Grid. We use this ontology in the Grid context to makeuse of the potential of both technologies by sharing computational and dataresources across different hosts, with the semantic meta-data providing supportfor management of these shared resources.

Fig. 4. Example of ExperimentalHypothesis and SimulationExperiment Instances.

As part of the FEARLUS-G project we have developed an open sourcereusable semantic data service based on Jena2 [12] and Globus Toolkit 3 [14]to provide semantic data storage, query and retrieval functions. This serviceuses ELDAS [13] to manage data repositories for RDF models; ELDAS was de-veloped by the UK National e-Science Centre eDIKT project6. The advantage ofenabling Jena2 as a Grid service is that semantic resources can be distributed andused in a dynamic environment. Searching a large collection of RDF resourcescan be a computationally intensive task; the Grid could offer the potential fordistributed processing of such queries.

There are issues surrounding deployment of Jena2 as a Grid service as it wasnot designed with this in mind. Jena2 does provide a module which extends theRDF model interface so that it is possible to store and retrieve statements usinga database. Although this module supports different database servers such asMySql, Oracle and PostgreSQL, it needs direct access to the database serverand is not designed to operate in a distributed environment such as the Grid.For these reasons we have developed a bridging solution, jdbc4eldas whichallows connections to different databases supported by the ELDAS data accessservice using standard JDBC library routines. The jdbc4eldas driver7 enablesthe Jena2 database module to store models using ELDAS as a Grid data accessservice. The advantage of using a JDBC bridge driver between Jena and ELDASis that both applications remain untouched. Moreover the jdbc4eldas driver iseasily reusable for any Java application that requires Grid data support.

To allow land-use scientists to create, share and re-use the resources storedwithin the Repository Service we have created a Web-based interface, This hastwo distinct components: My Workspace - a private area controlled by a single

6 http://www.edikt.org7 The drivers and associated documentation are available at

http://www.csd.abdn.ac.uk/research/fearg/links.php

scientist; Public Repository - a shared (public) space allowing access to publishedresources. In the following sections we will describe these interfaces in moredetail.

4.2 FEARLUS-G My Workspace

My Workspace (Figure 5) is a Web-based interface that allows land-use scientiststo manage simulation activities. My Workspace has two main components: theside-bar contains an options box, a list of active projects (My Projects) and alist of active hypotheses (My Hypotheses); the body component is used to renderselected data resources or to render forms to input new instances. Other context-sensitive options appear in the side-bar when a user navigates to different areasof the workspace.

Fig. 5. My Workspace.

Instances contained in My Workspace are grouped by project and by type ofresource (eg. Experiment Set, Hypothesis, Publication). Each resource type hasan associated icon used consistently in the workspace environment to facilitatetheir location. It is possible to browse the properties associated with each in-stance by clicking on its name (rdfs:label). The properties are rendered based ona template which defines how each resource should be rendered in My Workspace.Templates are encoded in XML defined using a set of Java tag-libs which con-tain the code to store, locate and render resources. Templates in My Workspacecould be generated automatically by reasoning from the ontology, but to datewe have not accomplished this; all existing templates are hand-coded. Figure 6

Fig. 6. Form to Create a New Experiment Set (left); Experiment Set RDF Instace(right).

shows an example of an input form and the resulting RDF both generated fromthe defined template.

Certain resources in My Workspace (Experiment Set, Hypothesis, Publica-tion) can be used as a evidence in order to build a scientific argument. When oneof those resources are rendered in the interface, an option is displayed allowingthe user to specifying if the resource agrees or disagrees with an existing hypoth-esis. Figure 7 shows the summary of an hypothesis instance, to which variousresources have been linked as evidence for the argument structure.

Fig. 7. Hypothesis Summary.

Resources appearing in My Workspace can be shared with other users bypublishing them into the FEARLUS-G Public Repository.

4.3 FEARLUS-G Public Repository

FEARLUS-G Public Repository is based on Longwell [15], an existing meta-databrowser developed by the SIMILE project8.

The most important aspect of the public repository is the ‘faceted browsing’user-interface paradigm. A facet is a particular meta-data field that is consid-ered important for the dataset that we are browsing. The system extracts a listof facets, their values and the number of times this facet value occurs in thedataset. This is shown on the right hand side of the Public Repository screen(see Figure 8), where each facet is grouped independently. By clicking on a facetvalue, the browser adds a facet restriction and eliminates all items that do notfulfil that restriction.

Fig. 8. FEARLUS-G Public Repository.

The facet pane on the right-hand side of the screen is also updated, showingthe facets and their values that are available in that sub-set of the data. Clickingon the pane has the effect of ‘zooming in’ on the dataset by locally removing fromthe view anything that is not of interest at that particular time. The browseralso support a ‘free text’ restriction facility; this reduces the browsed dataset toall items that contain the searched strings. By clicking on the icon to the rightof each instance, the user can access the full data associated with that instance.

If the user want to further investigate the selected hypothesis they can addit into their own workspace by clicking on the link (Add to MyWorkspace).

8 http://simile.mit.edu/longwell/

The integration between Longwell and the Semantic Data Repository Servicewas straightforward using the jdbc4eldas bridge. Unfortunately, the perfor-mance of Longwell within the Grid infrastructure is disappointing, as Longwellaccesses the repository frequently to retrieve small chunks of data instead ofusing high level queries. Because of the associated overhead this technique is notoptimal within our infrastructure. Longwell was a valid tool to support creationof an easy to use meta-data browser for the Public Repository. However, thelayout and rendering templates are not designed to be easily customized. Theresult is that the look and feel of our Public Repository differs to that on theMy Workspace interface.

5 Related Work

We have identified a number of activities that share important characteristicswith FEARLUS-G. The MyGrid project [16] [17] enables biologists to performand manage in silico experiments by providing a problem-solving workbenchbased on Semantic Grid technologies. User scientists in MyGrid browse a registryand select a workflow template. The system then asks the user to manuallyselect instances of the services associated with the workflow or automaticallyselects services based on a user profile (user preferences). The MyGrid servicescollect and co-ordinate data inputs and outputs for the experiment, as well asprovenance information about the performance of experiments. MyGrid developsthe use of workflow ontologies to capture Web-based procedures and provenanceontologies [18] to describe how and why results were produced.

Majithia et. al. [19] propose a system which facilitates automated synthe-sis of scientific experiment workflows, based on a high-level goal specification.This approach differs from MyGrid as it uses a composition algorithm to adaptavailable resources on the Grid to high level objectives defined by the user. More-over, the system makes a distinction between different levels of abstraction ofthe workflow in order to allow reuse and sharing.

The two systems described above focus on provenance (MyGrid) and work-flow abstraction (Majithia) to support sharing and reuse of scientific process in-formation. However, in both systems scientific argumentation is not supported;experiments and results are not associated with hypotheses. In our opinion theFEARLUS-G argument structure approach is likely to be more intuitive to userscientists, reflecting as it does the evidence-based approach widely used withinthe scientific community.

SemanticOrganizer [20] is a knowledge management system designed to sup-port distributed NASA projects, including multidisciplinary teams of scientists,engineers and accident investigators. SemanticOrganizer provides a semantically-structured information repository that serves as a common access point for allwork products related to an ongoing project. The repository not only storesfiles but also metadata describing domain concepts. The system allows accidentinvestigators to collect and manage evidence, perform different type of analy-ses that generate derivative evidence, connect the evidence together to support

or refute accident hypotheses. SemanticOrganizer is implemented using stan-dard Web services for both the repository and the user interface. This projectsuccessfully adopts Semantic Web technologies to support construction of anevidence-based argumentation structure. While this approach does have muchin common with the FEARLUS-G solution described in this paper, it does notof course provide support for execution of large-scale experiments.

Other relevant Semantic Grid activities include The Smart Tea Project [21]which aims to provide end to end support for data capture and access for ex-perimental scientists by replacing conventional lab notebooks. The Smart Teaontology supports experimental plans and records, provenance information, andsemantic links between data. The myTea Best Practice [22] project is imple-menting the lessons learned from the Smart Tea project within myGrid.

6 Lessons Learnt & Discussion

Our Grid service provides scientists interested in land-use phenomena with ameans to run much larger-scale experiments than previously possible on stan-dalone PCs, and also gives them a Web-based environment in which to sharesimulation results. The Grid is clearly a valuable tool for managing the com-putation involved in running such large-scale experiments. In fact the Globusenvironment is ideal for simulation modelling because the cost of a simulationrun is less that the cost of the communication necessary to run a distributedexperiment and aggregate the results. However, the Globus environment is notwell-suited for the fine grained exploratory work that the existing FEARLUSdesktop application is able to achieve. Scientists still want to use the patterns ofworkflow related to the desktop application. For this reason, we have extendedthe existing desktop application to be able to upload models to the Grid service,creating a link between the exploratory tool and the Grid experiment platform.

We have defined an initial collection of ontology elements that describe thescientific objects necessary to enable collaboration between members of a com-munity of land-use scientists. Our ontology represents generic scientific conceptssuch as hypotheses and experiments, as well as more domain-specific conceptstailored to the use of FEARLUS. Social scientists are able to run FEARLUSexperiments on the Grid and collect the results; a private interface allows themto manage their activities (experiments, investigations, etc.) while their workis in progress. Scientific resources can also be made public, allowing others toinvestigate them further.

Not all categories of users require access to the same meta-data. For exam-ple, we have exposed FEARLUS model parameters to land-use scientists familiarwith the existing FEARLUS application in such a way that the complexity ofthe Grid services is hidden to them. We provide a Web application which allowsusers to access and manipulate low-level simulation model parameters neededto run a FEARLUS experiment. However, we also need to provide high-leveldata to promote accessibility to the wider community. For this reason we pro-vide support for scientific arguments both in the Public Repository and in My

Workspace where resources are linked to hypotheses by high-level generic rela-tions (agreesWithHypothesis, hasDescription).

Throughout this project we have encouraged computational land-use mod-ellers at the Macaulay Institute to lead the process of engineering the FEARLUSontology. This important activity forced them to think about what conducting anexperiment actually meant. Prior to this project, Perl scripts were used to con-duct experiments with FEARLUS and the experimental design was thus hidden.One positive side-effect of the ontology building exercise is that the experimentalworkflow is now exposed, making its use transparent to other users.

We have engaged a group of “assessors” from the international scientific com-munities interested in agent-based social simulation, land-use and water manage-ment modelling. These assessors have agreed to conduct usability trials using theFEARLUS-G Grid service either in a research capacity (uploading FEARLUSmodels, comparing results from their own models with those from FEARLUS,checking results in FEARLUS publications, or exploring FEARLUS’s capabil-ities and limitations as a stage in the process of designing a new model), oras a teaching tool, or both. Evaluation of the FEARLUS-G service is expectedto provide more general insight into the effectiveness of current Semantic Gridtechnologies and methodologies.

Acknowledgment

The project is supported by the UK Economic & Social Research Council (ESRC)under the “Pilot Projects in E-Social Science” programme (Award Reference:RES-149-25-0011).

References

1. Roure, D.D., Jennings, N., Shadbolt., N.: Research Agenda for the Semantic Grid:A Future e-Science Infrastructure. Technical report, UK e-Science Series UKeS-2002-02, National e-Science Centre, Edinburgh, UK. (2001)

2. Foster, I., Kesselman, C., Tuecke, S.: The Anatomy of the Grid: Enabling ScalableVirtual Organizations. International J. Supercomputer Applications 15(3) (2001)

3. Roure, D.D., Jennings, N., Shadbolt, N.: The Semantic Grid: A Future e-ScienceInfrastructure. In Grid Computing: Making The Global Infrastructure a Reality.Anthony J.G. Hey and Geoffrey Fox. John Wiley & Sons Anthony J.G. Hey andGeoffrey Fox. John Wiley & Sons (2003) 437–470

4. De Roure, D. Jennings, N.S.: Semantic grid: Past, present, and future. In: Pro-ceedings of the IEEE, Volume 93, Issue 3. (2005) 669–681

5. Berners-Lee, T., Hendler, J., Lassila, O.: The Semantic Web. Scientific American284 (2001) 28–37

6. Fensel, D.: Ontologies: A Silver Bullet for Knowledge Management and ElectronicCommerce. Springer-Verlag New York, Inc. (2003)

7. Polhill, J., Gotts, N., Law, A.: Imitative Versus Non-Imitative Strategies in a LandUse Simulation. Cybernetics and Systems 32 (1) (2001) 285–307

8. Doran, J.E.: Simulating Societies using Distributed Artificial Intelligence. In SocialScience Microsimulation (eds. Troitzsch K G, Mueller U, Gilbert G N and DoranJ E ). Springer: Berlin. (1995) 381–393

9. McHaney, R.: Computer Simulation A Practical Perspective. Academic press.(1991)

10. Irwin, E., Geoghegan, J.: Developing Spatially Explicit Economic Models of LandUse Change. Agriculture, Ecosystem and Environment 85 (2001) 7–23

11. Minar, N., Burkhart, R., Langton, C., Askenazi, M.: The Swarm Simulation Sys-tem, A Toolkit for Building Multi-Agent Simulations. SFI Working Paper 96-06-042, Santa Fe Institute. (1996)

12. McBride, B.: Jena: Implementing the RDF Model and Syntax Specification. Tech-nical report, Hewlett Packard Laboratories (Brstol) (2000)

13. Baxter, R., Ecklund, D., Fleming, A., Gray, A., Hilld, B., Rutherford, S., Virdee,D.: Designing for Broadly Available Grid data Access Services. In: UK e-ScienceAll Hands Meeting (CD-ROM). (2003)

14. Foster, I., Kesselman, C.: Globus: A Toolkit-Based Grid Architecture. In: TheGrid: Blueprint for a Future Computing Infrastructure. Morgan-Kaufmann (1998)259–278

15. S. Mazzocchi, S. Garland, R.L.: Simile: Practical metadata for the semantic web.XML.com (2005)

16. Stevens, R., Robinson, A., Goble, C.: myGrid: Personalised Bioinformatics on theInformation Grid. In: proceedings of 11th International Conference on IntelligentSystems in Molecular Biology, 29th June3rd July 2003, Brisbane, Australia. (2003)

17. Lord, P., Bechhofer, S., Wilkinson, M.D., Schiltz, G., Gessler, D., Hull, D., Goble,C., Stein, L.: Applying Semantic Web Services to Bioinformatics: ExperiencesGained, Lessons Learnt. In: proceeding of Third International Semantic Web Con-ference, Hiroshima, Japan. (2004) 350–364

18. Zhao, J., Wroe, C., Goble, C., Stevens, R., Quan, D., Greenwood, M.: UsingSemantic Web Technologies for Representing E-science Provenance. In: proceedingof Third International Semantic Web Conference, Hiroshima, Japan. (2004) 92–106

19. Majithia, S., Walker, D.W., Gray, W.A.: Automating Scientific Experiments on theSemantic Grid. In: proceeding of Third International Semantic Web Conference,Hiroshima, Japan. (2004) 365–379

20. Keller, R.M., Berrios, D.C., Carvalho, R.E., Hall, D.R., Rich, S.J., Sturken, I.B.,Swanson, K.J., Wolfe, S.R.: SemanticOrganizer: A Customizable Semantic Repos-itory for Distributed NASA Project Teams. In: proceeding of Third InternationalSemantic Web Conference, Hiroshima, Japan. (2004) 767–781

21. Hughes, G., Mills, H., Roure, D.D., Frey, J.G., Moreau, L., schraefel, m., Smith, G.,Zaluska, E.: The semantic smart laboratory: A system for supporting the chemicalescientist. Org. Biomol. Chem. (2004) DOI: 10.1039/b410075a.

22. Brostoff, S., Cooke, R., Stevens, R., Gibson, A.: Transparent interaction; dynamicgeneration: context histories for shared science. 1st International Workshop onExploiting Context Histories in Smart Environments (2005)