G E O M A T I C A

SPATIAL DATA UNCERTAINTYIN THE VGI WORLD: GOING FROMCONSUMER TO PRODUCER

Joel Grira, Yvan Bédard and Stéphane RocheCentre de recherche en géomatique, Université Laval, Québec

To date, spatial data quality management has predominantly consisted of documenting processes andmeasuring errors with less concern about meeting users’ varying needs. Despite broad acceptance of the“fitness-for-use” criteria as a key component of geographic information quality assessment, quality infor-mation is still communicated as if it was addressed to a single usage or that requirements are perfectlyknown by the targeted users. Typically, users are disregarded or under-represented from the system designprocess. At best, some users are involved in a working group; at worst, their needs are assumed.

In this paper, we argue that involving a larger group of volunteer end-users in the spatial data uncer-tainty management process contributes to improving the spatial data quality of the designed systems.Accordingly, we discuss the concept of “perceived qualities” as, in a Volunteered Geographic Information(VGI) context, users are usually unskilled and do not have the same understanding of quality as experts do.Hence, a classification framework is proposed of various types of spatial data usage. Then, we addressuncertainty and VGI issues in a context of reshaping the geographic information production process.Furthermore, a set of communication gaps between spatial data producers and consumers are identified anda resulting R&D project introduced.

Introduction

Over the past 20 years, there has been muchresearch conducted on the quality of geographicinformation where spatial data is produced and dis-seminated. Typically, users consumed spatial datapassively, in contrast to the current status in whichusers, even unskilled ones, play an active role in spa-tial data creation and dissemination by publishinggeographic information, editing online maps, anduploading their own observations, georeferencedpictures or geotags. In addition, users commonlygive their opinion about the quality of non-spatialproducts (e.g. Amazon and AppleStore 5-star rat-

ings). Consequently, these new methods of datacreation and quality assessment raise new questionsabout geographic information quality.

Issues dealing with the quality of such user-provided data have been addressed by a number ofrecent studies concerning VGI [Goodchild 2007b;Seeger 2008; Flanagin and Metzger 2008; Haklay2008; Kouandi and Haklay 2009]. In this paper weaddress the issue of spatial data uncertainty asbeing an umbrella term to describe the problemsthat arise as a result of spatial data imperfections

GEOMATICA Vol. 64, No. 1, 2009 pp. 61 to 71

Joel Grirajoel.grira.l@ulaval.ca

Yvan Bédardyvan.bedard@scg.ulaval.ca

Stéphane Rochestephane.roche@

scg.ulaval.ca

Jusqu’à présent, la gestion de la qualité des données spatiales a consisté principalement à documenter lesprocessus et à mesurer les erreurs sans trop se préoccuper de répondre aux besoins propres à chaque util-isateur. Malgré une acceptation générale des critères « d’utilisation » comme un élément clé de l’évaluation dela qualité de l’information géographique, l’information sur la qualité est toujours communiquée comme si elleétait destinée à un seul usage ou comme si les exigences étaient parfaitement connues par les utilisateurs ciblés.Habituellement, les utilisateurs sont ignorés ou sous-représentés dans le processus de conception du système.Au mieux, quelques utilisateurs participent à un groupe de travail; au pire, leurs besoins sont présumés.

Dans cet article, nous faisons valoir que la participation d’un groupe plus important d’utilisateurs finauxvolontaires au processus de gestion de l’incertitude des données spatiales permet d’améliorer la qualité desdonnées spatiales des systèmes conçus. En conséquence, nous discutons du concept de « perception de laqualité » puisque, dans le contexte de l’information géographique volontaire (IGV), les utilisateurs sonthabituellement non spécialisés et ne comprennent pas la qualité de la même façon que les spécialistes. Ainsi,nous proposons un cadre de classification pour les divers types d’utilisation de données spatiales. Puis,nous examinons les questions de l’incertitude et de l’IGV dans un contexte de restructuration du processusde production de l’information géographique. De plus, nous recensons une série de lacunes au niveau de lacommunication entre les producteurs et les consommateurs de données spatiales et nous proposons un projetde recherche et développement qui en découle.

G E O M A T I C A

[Longley et al 2001] and, at the same time, alter andreduce its quality. Managing spatial data uncertaintyrequires the identification of the sources of uncer-tainty in order to better satisfy users’ requirements.In a volunteered context, these requirements areobviously quite impossible to predict given the hugenumber of volunteer users contributing content onthe Web. Furthermore, as quality is not effectivelycommunicated to these users [Comber et al 2006],the latter usually do not share the same understand-ing of the quality of the information provided. Thus,it is obvious that the provided data cannot fit the useof each user. However, our contention is that lever-aging the willingness of a great number of users tospend the time and effort required in order toimprove spatial data quality may overcome the pro-ducers’ lack of resources for monitoring the qualityand validity of sources. Accordingly, the followingsections address the spatial data uncertainty issue bydiscussing how consumers perceive quality.

Based on the quality perception analysis, weidentified the criteria of power and capability uponwhich a framework allowing the classification ofspatial data usages and contributions is built.Leveraging this classification, spatial data uncer-tainty is positioned in the expertise sphere andmanaging this uncertainty within the volunteeredcontext is discussed. A set of communication gapsbetween spatial data consumers and producers isidentified which supports our contention of theneed of volunteers’ involvement into the spatialdata uncertainty management process.

2. Perceived Qualities

Information about data quality is communicatedto both end-users and experts with little considera-tion as to whether it is easily understandable and,whether all spatial data consumers will have thesame understanding of the provided information[Frank 1998; Boin and Hunter 2006; Goodchild2007a]. With spatial data accessible to members ofthe general public who have little formal training inquality issues, the GI science community is facing anew situation that raises questions about the com-municated quality [Boin and Hunter 2007] and itsdifferent users’ perceptions. These perceptions cor-respond to the concept of fitness for use that was firstdefined by Juran with the terms of “subjective con-cept” and described as being something that is per-ceived by the user [Juran et al 1974]. The concept offitness for use was adopted in 1982 by the NCDCDSand promoted by Chrisman who considers that“quality information provides the basis to assess the

fitness of the spatial data to a given purpose”[Chrisman 1983]. More recently, Devillers andJeansoulin used the same concept of fitness for useand considered that it is somewhat unique to everyuse case [Devillers and Jeansoulin 2006a] and thatno single message can be communicated to all users.This concept, often called “external quality,” iscommonly accepted in the GI quality community,and corresponds to the ISO definition of “quality”[Aalders 2002; Dassonville et al 2002; Devillers andJeansoulin 2006b]. Accordingly, reasoning aboutone single quality within a collaborative context isobviously misleading; the increasing number ofusers leads to a wide range of requirements, to dif-ferent assessment processes, and consequently, to avariety of quality perceptions [Brabyn 1996].

Until recently, data quality assessment assumedthat the provider was the ultimate judge of the inher-ent uncertainty of these data [Chrisman 1991]. Thisis no longer the case with respect to a volunteeredcontext where data consumers and individual datacollectors are from a wide range of disciplinarybackgrounds and where their needs for data qualityvary. Basically, organisations providing spatial dataare responsible for the communication of thedatasets, quality. However, on the one hand, dataquality reporting through the use of metadata is notan effective way to inform users about the fitness foruse because “it does not provide full descriptions ofdata uncertainty and allow assessments of data fit-ness” [Comber et al 2006]. On the other hand, spa-tial data providers cannot manage how the commu-nicated information is understood or fits with theuser’s model of the world [Bédard 1986a; Comber etal 2008]. Consequently, spatial data providers arefacing a dilemma; reporting, in accordance withstandards, the spatial data quality while trying tomeet various requirements of a heterogeneousaudience. One may have the temptation to framethe data consumers’ view of quality or to concep-tually represent their various requirements thatlead to a set of perceived qualities; however, itwould be as inefficient as metadata profiles. In fact,creating a metadata profile assumes a perfectknowledge of a specific set of requirements [ISO/TC211 2004]. Hence, taking into consideration the var-ious and different needs of thousands of users isobviously unrealistic, as it creates as many metadataprofiles as there are users’ communities.Alternatively, it would appear more pertinent to helpusers to express their needs in a way that spatial dataproviders could satisfy or suggest other datasets ormashups in a truly collaborative context.

In such a context, the consumers’ intendeduses of the data and quality information vary, and

62

Based onthe qualityperceptionanalysis, weidentified thecriteria ofpower andcapabilityupon whicha frameworkallowing theclassificationof spatialdata usagesand contri-butions isbuilt.

G E O M A T I C A

so do their contributions. It would not be reasonableto think about training thousands of volunteerusers, as it is unrealistic to think that geographicinformation providers will support all spatial dataconsumers. Obviously, a communication gap existsin this regard between the data provider and dataconsumers, as they have different perceptions of thesame represented reality [Boin 2008]. Completelybridging this gap is not possible because of theremaining residual uncertainty [Bédard 1986b;Bédard 1987; Agumya and Hunter 1999; Agumyaand Hunter 2002]. However, it seems possible tobring the consumer’s perception of quality closer tothat of the producer and to reduce the perception gapproperly by using facilitating technologies that couldenable such a reconciliation throughout interactivetools and Web 2.0 interfaces [Seeger 2008; Haklay etal 2008; Scharl and Tochtermann 2007] where usersmay express their opinions about the provided qual-ity. Users are still waiting for such technical facili-ties. Hunter also noted the absence of tools and tech-niques that may help to assess data quality [Hunter2001]. At the same time, data producers may takeinto consideration the different perceived qualitiesand the various possible uses of their datasets.

3. A Classification FrameworkCoined by Goodchild, the concept of

Volunteered Geographic Information (VGI) hasbecome popular [Goodchild 2007c] in the past twoyears. It refers to “geospatial data that are voluntar-ily created by citizens who are untrained in the dis-ciplines of geography, cartography or relatedfields” [Seeger 2008]. Evidence exists in the litera-ture about the need to understand the social aspectsand the environment of VGI in order to compre-hend the phenomenon [Flanagin 2008; Seeger2008; Elwood 2008; Goodchild 2007b]. The socialinteractions among the actors involved in creatingand consuming geographic information is either (1)authoritarian or (2) volunteered:

(1) The first interaction mode corresponds to thecommon practices of the traditional mappingagencies. Traditionally, quality informationwas communicated through a unidirectionalprocess that spatial data producers used to pro-vide metadata. Using this medium of commu-nication may be considered as an implicitexclusion of data consumers from the qualityassessment process. The world of traditionalmapping agencies is described by Goodchildas being a “top-down, authoritarian, centristparadigm that has existed for centuries, in

which professional experts produce, dissemi-nation is radial, and amateurs consume”[Goodchild 2007b]. Although in other casesprofessional experts may have to consume data(e.g. a land surveyor consuming a cadastralmap), an authoritarian relationship still existsbetween producer and consumer actors.Consequently, authority is also anotheraspect—in addition to the social aspect—thathas to be taken into consideration. In fact, ithelps to distinguish between a situation wheredatasets are provided by public or privateagencies and where volunteer users interveneto contribute to existing datasets, as opposed tothe situation that consists in a pure VGI con-text where only volunteered contributions areresponsible for building the resulting datausing mashups and other techniques.

(2) The second interaction mode refers to the VGIworld. In a collaborative context, authority isstill a crucial concern since users have to reactaccording to the fitness for use of a givendataset: consequently, users should have theopportunity to express their needs and indicatehow far the provided quality seems to be fromtheir required quality.

Implementing participatory geogrpahic infor-mation science (GIS) with a role-based authorityrepresents a form of delimiting the actions of vol-unteer users. Access to GIS and data is a key ele-ment to understanding citizen involvement inPPGIS and clarifying the links between data accessand public contributions. In fact, the role of the useris changing the GIS landscape [Budhathoki et al2008; Bishop 2007] considering the varying degreeof access to GIS software and data. Moreover, theaccess to GIS goes beyond the simple connection toa software because it became a matter of skill,expertise, and social interactions [Laituri 2003].Thus, the proposed framework depicts the volun-teered social interaction in opposition with theauthoritarian one (Figure 1). The classificationframework is basically built on two dimensions(Figure 1): the first one is the power relationshipexisting between geospatial data producers andconsumers, and the second one is the capability ofboth of these actors to take the responsibility relat-ed to data manipulation and dissemination.

The power dimension refers to the authoritariantype of social relationships, which is the top level offormal power that actors may have (e.g. cartograph-ic agency producing spatial data), as opposed to thevolunteered one where actors are just deploying vol-unteer efforts with little formal authority on otheractors, procedures, or the resulting outputs (e.g.

63

Access toGIS and

data is a keyelement to

understand-ing citizen

involvementin PPGIS

and clarifyingthe linksbetween

data accessand public

contributions.

G E O M A T I C A

uploading a GPS trace in OpenStreetMap). Thepower dimension refers also to the constraints andstandards that have to be respected by both spatialdata producer and consumer; the authoritarianworld described by Goodchild [Goodchild 2007b]corresponds to a highly constrained world whereasthe volunteered world is constraint free.

The capability dimension refers to the nature ofthe actions that actors are authorised to perform onspatial data and the associated responsibilities relat-ed to these actions. In fact, in both traditional andparticipatory worlds, actors may perform actions onspatial data depending on their attributed capabilities(e.g. approving data, changing data, disseminatingdata, revising data). Thus, a full capability actionrefers, in the proposed classification framework, tothe top level of capability that actors may have andthe usages they are allowed to do whether they arewithin a traditional or a volunteered context (e.g.creating new mashups or overlays from various datasources), contrarily to a limited capability actionwhere actors are limited to a restrained set of usages(e.g. viewing and printing maps).

Crossing these two dimensions—power andcapability—produces a matrix within which it is pos-sible to position the different trends, tools, products,and production processes in the fields of geomatics,geography, cartography, or related fields. This matrixconsists of the following four components:

(1) Volunteered—Full capability(2) Authoritarian—Full capability(3) Volunteered—Limited capability(4) Authoritarian—Limited capability

The first component of the framework corre-sponds to the collaborative context where contribu-tors do not have any control on standardization,production mechanisms, or quality control. Actorsare often simple users, the majority of whom haveneither skills nor background in the disciplines ofgeography, cartography, or related fields.Nevertheless, these actors have the opportunity toupload geographical information, to generategeospatial content, to create mashups, geotags, andmany other types of VGI contributions. Typicalexamples illustrating this component may be thecreation of map mashups (e.g. Google Mapsmashups using Google Maps API)

The second component of the frameworkdescribes situations where highly trained expertsand private/public mapping agencies have full con-trol on production mechanisms, specifications, andquality controls [Goodchild 2009]. Experts andprofessionals have the knowledge and the back-ground to manipulate these institutions’ outputs fol-lowing standards and codes of ethics [Bédard et al2009], so they are included in the same authoritari-an category. A typical example for this dimension

64

Figure 1: A classification framework for geographical information usages.

G E O M A T I C A

consists of an accuracy problem being reported to apublic cartographic agency by an expert (e.g. a landsurveyor for the national cadastral agency). Expertsworking for private firms, such as NAVTEC, mayhave to evaluate and validate a set of POIs (pointsof interest) in terms of temporal and positionalaccuracy based on validation rules, standards, andhuman expertise.

The third component of the framework corre-sponds to the situation where volunteer actors haveno authority on the outputs, production procedures,and specifications. However, they may make deci-sions based on the consultation of provided geo-graphic information. For example, they may print amap from OpenStreetMap or Wikimapia website.Whilst the edition of OpenStreetMap or Wikimapiacontent is technologically constrained (i.e. one has tocreate an account), viewing the maps is totally free.As for the Wikimapia control system, it is describedlater in this paper in more detail.

The fourth component of the frameworkdescribes the situation where authoritarian actors areconsuming data which is used only for consultation.Decisions may be made based on the consumedinformation. The usefulness of this dimension con-sists, for instance, on the usage of a GIS by volun-teers to formulate feedback about a given dataset inorder to improve the outputs of that GIS. The author-itarian actors could ignore the comments, as they arethe only people responsible for the output. However,it is worth noting that the limited/full capability cor-responding to the four components of the frameworkmay result from both technological and knowledgelimitations. Role attribution in GIS software cer-tainly constitutes an access limitation to technologyand consequently, affects the capability of partici-pants [Laituri 2003]. Knowledge limitation refers toa lack of skills or a low educational level in the geo-graphic domain, so that a contributed map may bepositioned indifferently anywhere on Figure 1 (i.e.within one of the four framework components); theonly difference is in the intended usage of that map.

4. Uncertainty in the NewExpertise Sphere

The terms experts and novices belong to a widenumber of disciplines and, according to Goodchild,the separation between the two concepts is drivenby many factors such as the complexity of thedomain concepts and subjects, terminology, andother entry barriers [Goodchild 2009]. On the onehand, expertise refers in general to an extensiveknowledge and ability in a given subject.

Consequently, it consists of skill and knowledgethat distinguish experts from novices. On the otherhand, because of lack of knowledge or skill, anexpert in a given domain may no more be consid-ered as such in another domain. Accordingly, anovice may be considered as an expert in his spaceof familiarity [Goodchild 2009]. Interestingly, onecould use such a concept to analyse the “quality ofusers” with regards to data usages, (i.e. the fitness-for-use of a user’s expertise for the intended use ofgiven data) but this analysis goes beyond the goalof the present paper.

The concepts of activity space [Johnston et al2000] and space of familiarity [Goodchild 2009]bring the notion of expertise from the traditionalworld to the volunteered one. This assumes thatindividuals making entries on websites such asGoogleEarth, OpenStreetMap, or Wikimapia intheir activity spaces are as qualified and efficient inensuring the quality of the contributed data as theprofessional experts are outside of their field ofactivity, which is, obviously, not always the case.

Traditionally, the engagement of experts to fol-low standards and specifications used to ensure ahigh level of data quality. The trend of volunteeredgeographic information and the concerns of spatialdata quality it raises could rely on the collectiveintelligence in ensuring a comparable level of qual-ity. In fact, collective intelligence may contribute toreducing the spatial data uncertainty; however, thereis still no evidence about its effectiveness comparedto the traditional quality controls [Goodchild 2009].Wikimapia, for instance, incorporates a notable qual-ity control mechanism based on two components:

• User levels and permissions—a set of capa-bilities consisting of permissions andrestrictions assigned to users. Users areupgraded to a superior level according tothe overall quality of their contributions

• Voting—a feedback mechanism allowingusers to assess the quality of others’ con-tributions.

The Wikimapia collaborative quality controlsystem seems to interactively communicate qualityinformation to the user through an iterative process.The latter ensures that the consumer-perceivedquality is continuously expressed and that the pro-vided information about quality is continuouslyupdated and disseminated.

Furthermore, and according to the outlinedframework in Figure 1, many GISs are fed withdata coming from volunteer contributions of novicepeople who are uploading data, editing entries, andmaking corrections. As with any observationalprocess which is based on common sense and

65

The trend ofvolunteeredgeographicinformation

and the con-cerns of

spatial dataquality it

raises couldrely on the

collectiveintelligencein ensuringa compara-ble level of

quality.

G E O M A T I C A

human manipulation, these volunteered activitiesand the related contributed data are subject touncertainty. In fact, each collection technique leadsto a level of uncertainty arising from imprecisemeasurements or derived inaccuracy values [Fekpeet al 2009] and the presence of uncertainty relatedto data collection, whether the latter is processed bya professional or an untrained person, goes beyondthe expertise [Couclelis 2003]. Actually, in additionto expertise, it is important to take into account thesources of uncertainty, as these sources are depend-ent on the context, which is itself often heteroge-neous (i.e. neither totally authoritarian nor totallyvolunteered); this means that managing spatial datauncertainty within a VGI context may not be thesame as within an authoritarian one. Accordingly,this legitimizes the asking of questions about howto certify the quality of a dataset resulting from vol-unteered contributions compared to another onethat has been produced by governmental mappingagencies, for instance.

It is worth noting that there are no typical uncer-tainties or distinctive data quality concerns that mayespecially arise in VGI context rather than in anauthoritarian one, because both untrained users andhighly trained professionals may have access toalmost the same range of tools, data, and even equip-ment. For example, data transformations may occurin a volunteered context (e.g. creating mashups) aswell as within a mapping agency (e.g.projection/coordinate transformation while integrat-ing two datasets) and the resulting uncertainties (e.g.positional and temporal errors) can lead to doubtsabout the quality of the analysis and the decisionstaken with that data in both cases [Frank 1998;Gahegan and Ehlers 2000]. However, some types ofuncertainty in a VGI may be more problematic thansome others in an authoritarian world, especiallywhen we get out of the space-time framework. Forinstance, inconsistencies (e.g. topological inconsis-tency resulting from object formation) and incom-pleteness (e.g. omission of a real world entity con-sidered as uninteresting) are more likely to be prob-lematic because they belong to a secondary form oferrors [Hunter and Beard 1992] that are not neces-sarily obvious to untrained users.

5. Uncertainty Managementin a VGI Context

End-users usually face difficulties when itcomes time to assess the suitability of a dataset fora specific application. Accordingly, Comber et al.have suggested that the metadata definition would

be user-focussed [Comber et al 2007a; Comber etal 2007b; Comber et al 2008] rather than producer-centric [Goodchild 2007a]. Users could then focuson their analyses rather than concentrating ondecrypting the technical and hermetic metadata.

In order to bring data quality information clos-er to the users’ perceptions, the capabilities of Web2.0 are worth leveraging [Seeger 2008; Haklay et al2008; Scharl and Tochtermann 2007]; this wouldget users engaged in the process of spatial dataquality assessment and consequently involved inimproving the quality of the manipulated datasets.Providing interactive web interfaces to end-usersthat allow them to create and edit entries about dataquality is valuable; not only will users feel empow-ered to express their requirements [Craglia 2007],but the boundary between the user’s perception ofquality and the producer’s understanding of theuser’s requirements will tend to disappear due tothe bidirectional communication enabled by theweb-based infrastructure [Devillers et al 2007].

Establishing an interactive communicationprocess to manage uncertainties in a VGI contextimplies an iterative mechanism consisting of: (1)visualising spatial data quality, (2) collecting users’entries about quality throughout Web 2.0 forms, and(3) integrating quality information within the model.A similar mechanism based on a feedback loop hasalready been proposed by Boin [Boin 2008] whichconsists of an adaptation of the Peterson’s commu-nication model for communicating the fitness foruse. Nevertheless, the Boin model emphasizesquality visualisation; our focus is on involving end-users in managing spatial data uncertainty to realisea user-centric perspective of quality.

It is worth noting that many studies on visualiz-ing spatial data uncertainty and information aboutquality have been conducted in order to make qual-ity information more understandable to end-users[Clapham and Beard 1991; Buttenfield and Beard1991; Paradis and Beard 1994; Beard andButtenfield 1999; MacEachren et al 2005; Boin andHunter 2006; Boin and Hunter 2007; Boin 2008].Quality information has to be communicated to dataconsumers using simple terms and ergonomic inter-faces in order to ensure their engagement in the col-laborative process. Similar concerns about theappropriate use of dangerous goods by the generalpublic has already led the ISO to adopt “labelingstandards” that define the symbols, form, and con-tent needed to properly warn users in a language theyunderstand [ANSI Z535.4 2007]. In the present work,we assume that users can contribute greatly to iden-tifying, defining, and presenting the risks of inappro-priate uses of geospatial data. We also assume they

66

In order tobring dataquality infor-mation closerto the users’perceptions,the capabili-ties of Web2.0 are worthleveraging…

G E O M A T I C A

are motivated enough to participate voluntarily inquality improvement by fulfilling the Web 2.0 forms.

Furthermore, notable work has been done in theconceptual representation of spatial data quality[Beard 1997; Frank 1998; Fisher 1999; Fekpe et al2009]. Advances in this field are important for thepresent work, as any quality management processhas to be based on a quality model. Thus, althoughwe do not address in this paper any spatial data qual-ity representation issues, we assume that spatial dataquality has been properly presented. The communi-cation of data quality information and how this infor-mation is perceived by end-users constitute animportant factor for their engagement in the processand for the quality of the feedback they will provide;consequently, communicating relevant quality infor-mation to end-users has to take into considerationtheir requirements. Accordingly, when volunteeredcontributions are needed by a data producer or a GISsoftware provider, the displayed quality informationwould be better dynamically generated from a qual-ity model—which takes into consideration the spe-cific needs of the concerned user, instead of beingstatically displayed in hard-coded web forms orpages based on the “One Quality Fits All” assump-tion. As a result, private and public mapping agen-cies may express the need for contributions in waysto valorise public participation; this can be insuredthrough an appropriate, customized, and model-based quality communication. Getting volunteerend-users involved in improving spatial data qualityrequires deploying efforts on designing qualitymeasures and visualizing them, which is, as dis-cussed earlier, beyond the scope of the present work.Nevertheless, bringing volunteers into the qualityimprovement process and asking for their engage-ment and contributions may encounter several obsta-cles consisting of the gaps one can observe through-out the quality management process (Figure 2).Three types of gaps may threaten the success of thevolunteered efforts:

• Normative gap—consists of the differencebetween quality as expected by end users andquality as conceptualized by standards, proce-dures, and normative institutions. Qualityinformation used to be communicated to usersthrough metadata, which is a term commonlyused to refer to a standardized set of metadataelements. The metadata format is compliant toISO 19115 [ISO-TC/211 2003] or similar stan-dards that are generally unknown by untrainedusers who have certain expectations about thequality of the dataset they will manipulate.

• Technological gap—geographic information isusually provided throughout GIS or web map-ping servers that rarely have interactive featuresto allow users to upload feedback and commentsabout data quality. Several existing GISs that areavailable on the Web do not fully take advantageof the potential of this medium; users are stillconsidered as passive data consumers instead ofcontributors [Budhathoki et al 2008]. In a VGIcontext, GIS providers have to reconsider therole of users not only in the design of their webapplications, but also in the design of qualityrequirements and communication.

• Cognitive gap—referring to a user-focusseddefinition of metadata [Comber et al 2008],Comber et al proposed a less static type of meta-data which takes into consideration the variousneeds of users. The different users’ perceptionsof quality are usually very far from the provid-ed quality, especially in a volunteered context(e.g. untrained users are usually confusedbetween accuracy and quality).Replicating a “paper-perspective” of maps on

the Web and providing GIS without either feedbackmechanism or quality management features broad-ens the gaps between spatial data producers andconsumers. The solutions that have been testedover the years adopted a posteriori approaches (i.e.after data have been collected) to help manage

67Figure 2: Spatial data quality communication gaps between providers and consumers.

G E O M A T I C A

quality and improve spatial data usability[Devillers et al 2007]; in fact, data and systemswere typically designed, produced, and distributedwith a priority on delivering a technically workingsystem with limited usability consideration.Following such an a posteriori approach (Figure 3(A)), quality management is not a primary concern,as it is performed in the later stages of the systemdevelopment process. Even if some importantaspects of data quality are sometimes consideredbefore the data collection process begins (e.g.choice of methods or instruments), the point is thatthe model does not reflect these aspects because nospecifications are registered at the design level andbecause metadata are almost the unique reportingapproach where quality information is captured,stored, and communicated. Most concerns aboutdata quality take place once the system is beingused, especially for unintended uses. However, theincreasing access to geospatial data combined withthe growing implications of non-expert users inseveral activities, and the growing number of unin-tended uses, highlights once again the importanceof the problem of quality management; a posterioriapproaches will no more be efficient because userrequirements are varying over space and time, andbecause the volunteered phenomenon is continu-ously expanding. Besides, users need to know moreabout the quality of the provided datasets, whereasgeographic information providers fail in communi-cating that quality mainly because it is impossibleto predict all of the possible usages of the their

datasets and also because of technological limita-tions. Evidence exists about the need to considerquality management issues earlier in the systemdevelopment process [Bédard et al 2009], whichcorresponds to the adoption of an a priori approachto usability improvement, uncertainty reduction, andquality management (Figure 3 (B)). In a volunteeredworld, a large number of users are willing to spendtime on improving the quality of the GIS data theyare interested in: thus, producers have to offer fea-tures allowing end-users to contribute to the qualityimprovement process. Addressing these issues is inprogress within a current research project which isintroduced in the following section.

Figure 3 depicts two situations: the first one(Figure 3 (A)) represents the current state of the art,characterised by a spatial data quality managementperformed within an a posteriori approach becauseuncertainty is dealt with later in the process. Thefirst situation corresponds also to a perspective ofdata and GISs where data quality is neither a col-laborative nor a volunteered issue. On the contrary,the second situation (Figure 3 (B)) represents aproactive vision—an a priori approach—of spatialdata quality where preventive actions occur earlierin the process, namely at the design stage. This sec-ond situation describes how to leverage collabora-tive environments in order to bring end-users closerto the quality decisions’ circle, which was formerlydedicated exclusively to professionals. The currentproject aims at moving from the first situation tothe second one.

68

Figure 3: Positioning systems according to spatial data usability improvement approaches within a volunteered context.

G E O M A T I C A

Conclusions and Future Work

In this article, we have argued that involvingvolunteer users in the spatial data uncertainty man-agement may help improve the resulting quality ofthe designed system. We have presented the con-cept of “perceived qualities” to emphasize theimpossibility of providing one single quality thatfits all users’ purposes; hence the data providersshould take into consideration these different per-ceptions. A possible solution may consist of impli-cating end-users in the design process by gettingthem involved in quality assessment through a col-laborative environment where they can formulatetheir specific needs, give their opinion on therequired data quality compared to the provided one,and ultimately, feed a CASE (Computer AidedSoftware Engineering) tool with contributed quali-ty information. Furthermore, we have proposed aclassification framework that scopes the contribu-tions of volunteers and organizes the various usesof the geographic information around two axes:capability and power. Finally, we discussed the spa-tial data uncertainty issue within the VGI contextwith a particular focus on facilitating volunteeredcontributions to the quality improvement process;we have identified three kinds of gaps between spa-tial data consumers and providers. Finally, we haveillustrated how spatial and non-spatial systems maybe positioned within a volunteered context and rel-ative to a priori and a posteriori approaches ofusability improvement; this highlights the need.

Referring to Figure 3, we endeavour in ourfuture work to reduce spatial data uncertainty withina collaborative context, adopting an a prioriapproach; thus, our contribution will be positioned atthe right side of Figure 3. Believing that involvingend-users in managing spatial data uncertainty leadsto a better quality compared to the traditionalapproach, we attempt to define a collaborativeapproach for designing spatial data quality for agiven system in order to identify potential problemsat the earliest stage of the development cycle.Afterward, will provide a web-based prototype in aconcrete application domain in order to get realmeasurements of our approach. Finally, we willexplore how communicating quality to volunteeredusers may improve spatial data quality.

AknowledgementThe authors wish to acknowledge the contribu-

tion of Canada NCE-GEOIDE projects IV-23 “PublicProtection and Ethical Dissemination of Geospatial

Data” and IV-41 “Participatory Geoweb for Engagingthe Public on Global Environmental Change.”

References

Aalders, H.J.G.L. 2002. The Registration of Quality in aGIS. Spatial Data Quality, W. Shi, P. Fisher andM.F. Goodchild, Eds. London, Taylor & Francis, pp.186–199.

Agumya, A. and G.J. Hunter. 1999. A Risk-based Approachto Assessing the ‘Fitness for Use’ of Spatial Data.URISA Journal, 11(1), 1999, p. 33 and 44.

Agumya, A. and G.J. Hunter. 2002. Responding to theConsequences of Uncertainty in Geographical Data.International Journal of Geographical InformationScience, 16(5), p. 405 and 417.

ANSI Z535.4. 2007. American National Standard forProduct Safety Signs and Labels, p. 49.

Beard, K. 1997. Representations of Data Quality.Geographic Information Research: Bridging theAtlantic. Massimo Craglia and Helen Couclelis,Eds. London, Taylor and Francis, pp. 280-294.

Beard, M.K. and B. Buttenfield. 1999. Detecting andEvaluating Errors by Graphical Methods inGeographic Information Systems: Principles andTechnical Issues. P. Longley, M. Goodchild, D.Maguire, and D. Rhind. Eds. New York, John Wiley,Vol. 1, pp. 219-233.

Bédard, Y. 1986a. A Study of Data Using aCommunication Based Conceptual Framework ofLand Information Systems. Le Géomètre Canadien,40(4), pp. 449-460.

Bédard, Y. 1986b. Comparing Registration SystemsUsing a Communication Based Criterion:Uncertainty Absorption. Proceedings of the XVIIICongress, Fédération internationale des géomètres,Toronto, Ontario, Canada, June 1-11.

Bédard, Y. 1987. Uncertainties in Land InformationSystems Databases. Proceedings of the ACSM-ASPRS Auto-Carto 8 Conference Bethesda, MD -American Congress on Surveying and Mapping, pp.175–184.

Bédard, Y., J. Chandler, R. Devillers and M. Gervais.2009. System Design Methods and Geospatial DataQuality. Association of American Geographers—Professional Ethics Session on GeographicInformation Ethics and GI Science, Las Vegas,USA, 22-27 March.

Bishop, J. 2007. Increasing Participation in OnlineCommunities: A Framework for Human–ComputerInteraction. Computers in Human Behavior, 23(4),pp. 1881-1893.

Boin, A.T. and G.T. Hunter. 2006. Do Spatial DataConsumers Really Understand Data QualityInformation? 7th International Symposium onSpatial Accuracy Assessment in Natural Resourcesand Environmental Sciences, M. Caetano, and M.Painho, Eds., Instituto Geografico Portugues,Lisbon, pp. 215–224.

69

G E O M A T I C A

Boin, A.T. and G.J. Hunter. 2007. What CommunicatesQuality to the Spatial Data Consumer? 5thInternational Symposium on Spatial Data Quality,Enschede, Netherlands, June 13-15, 8 p.

Boin, A.T. 2008. Exposing Uncertainty: CommunicatingSpatial Data Quality via the Internet. VDM VerlagSaarbrücken, Germany.

Brabyn, L. 1996. Landscape Classification Using GIS andNational Digital Databases. Landscape Research,21(3), pp. 277-300.

Budhathoki, N.R., B.C. Bruce and Z. Nedović-Budić.2008. Reconceptualizing the Role of the User ofSpatial Data Infrastructure. GeoJournal, 72(3-4), pp.149–160.

Buttenfield, B.P. and M.K. Beard. 1991. Visualizing theQuality of Spatial Information. Proceedings AutoCarto 10, Baltimore, MD. vol. 6, pp. 423-427.

Chrisman, N.R. 1983. The Role of Quality Information inthe Long Term Functioning of a GeographicalInformation System. Proceedings of the InternationalSymposium on Automated Cartography (Auto Carto6), Ottawa, Canada, pp. 303–321.

Chrisman, N.R. 1991. The Error Component in SpatialData. D.J. Maguire, M.F. Goodchild and D. RhindEds., Geographical Information Systems: Principlesand Applications, John Wiley & Sons, New York,vol. 1, pp. 165-174.

Clapham, S.B. and K. Beard. 1991. The Development ofan Initial Framework for the Visualization of SpatialData Quality. Proceedings ACSM-ASPRS AnnualConvention 2: pp. 73–82.

Comber, A.J., P.F. Fisher and R.A. Wadsworth. 2007a.User-Focused Metadata for Spatial Data,Geographical Information and Data QualityAssessments. 10th AGILE International Conferenceon Geographic Information Science, AalborgUniversity, Denmark.

Comber, A.J., P.F. Fisher and R.A. Wadsworth. 2007b.Approaches for Providing User Relevant Metadataand Data Quality Assessment. GeographicalInformation Science Research Conference, NUIMaynooth, Ireland. April 11-13.

Comber, A.J., P.F. Fisher, F. Harvey, M. Gahegan and R.A.Wadsworth. 2006. Using Metadata to LinkUncertainty and Data Quality Assessments. 12thInternational Symposium on Spatial Data Handling(SDH06), edited by A. Riedl, W. Kainz and G. Elmes,July 12-14, Springer Berlin, Vienna, Austria, pp. 279-292.

Comber, A.J., P.F. Fisher and R.A. Wadsworth. 2008.Semantics, Metadata, Geographical Information andUsers. Transactions in GIS, 12(3): pp. 287–291.

Couclelis, H. 2003. The Certainty of Uncertainty: GIS andthe Limits of Geographic Knowledge. Transactionsin GIS 7(2): pp. 165-175.

Craglia, M. 2007. Volunteered Geographic Informationand Spatial Data Infrastructures: When Do ParallelLines Converge? Retrieved September 5, 2009,http://www.ncgia.ucsb.edu/projects/vgi/partici-pants.html.

Dassonville, L., F. Vauglin, A. Jakobssosn, and C. Luzet,2002. Quality Management, Data Quality and Users,

Metadata for Geographical Information. Spatial DataQuality, W. Shi, P.F. Fisher and M.F. Goodchild Eds.,Taylor & Francis, London, pp. 202–215.

Devillers, R. and R. Jeansoulin. 2006a. IntroductionFundamentals of Spatial Data Quality, R. Devillersand R. Jeansoulin, Eds., ISTE, London, pp. 17-20.

Devillers, R. and R. Jeansoulin. 2006b. Spatial DataQuality: Concepts, Fundamentals of Spatial DataQuality, R. Devillers and R. Jeansoulin, Eds., ISTE,London, pp. 31-42.

Devillers, R., Y. Bédard, M. Gervais, F. Pinet, M.Schneider, L. Bejaoui, M.A. Levesque, M. Salehi,and A. Zargar. 2007. How to Improve GeospatialData Usability: From Metadata to Quality-AwareGIS Community, Spatial Data Usability. A AGILEPre-Conference Workshop, 8 May, Aalborg,Denmark.

Elwood, S. 2008. Volunteered Geographic Information:Key Questions, Concepts and Methods to GuideEmerging Research and Practice. GeoJournal, 72,pp. 133–135.

Fekpe, E.S., T.K. Windholz and K. Beard. 2009.Probabilistic Approach for Modeling and PresentingError in Spatial Data. Journal of SurveyingEngineering, 135(3), pp. 101–112.

Fisher, P.F. 1999. Models of Uncertainty in Spatial Data.Geographical Information Systems: Principles,Techniques, Management and Applications, P.Longley, M. Goodchild, D. Maguire and D. Rhind,Eds., vol 1. Wiley and Sons, New York, pp. 191–205.

Flanagin, A.J. and M.J. Metzger. 2008. The Credibility ofVolunteered Geographic Information. GeoJournal,vol. 72, pp. 137–148.

Frank, A.U. 1998. Metamodels for Data QualityDescription. Data Quality in GeographicInformation: From Error to Uncertainty, R.Jeansoulin and M.F. Goodchild, Eds., Paris, HermèsEditions, pp. 15-19.

Gahegan, M. and M. Ehlers. 2000. A Framework for theModelling of Uncertainty Between Remote Sensingand Geographic Information Systems. ISPRSJournal of Photogrammetry and Remote Sensing,55(3), pp. 176-188.

Goodchild, M.F. 2007a. Beyond Metadata: Towards User-Centric Description of Data Quality. Keynote paper,5th Int. Symposium Spatial Data Quality, ITC,Netherland, 13-15 June.

Goodchild, M.F. 2007b. Citizens as Voluntary Sensors:Spatial Data Infrastructures in the World of Web 2.0.International Journal of Spatial Data InfrastructureResearch, 2, pp. 24–32.

Goodchild, M.F. 2007c. Citizens as Sensors: The World ofVolunteered Geography. GeoJournal, 69, pp.211–221.

Goodchild, M.F. 2009. NeoGeography and the Nature ofGeographic Expertise. Journal of Location BasedServices, 3(2), pp. 82–96.

Haklay, M., A. Singleton and C. Parker. 2008. WebMapping 2.0: The Neogeography of the Geoweb.Geography Compass, 2(6), pp. 2011–2039.

Haklay, M. 2008. How Good is OpenStreetMapInformation? A Comparative Study of

70

G E O M A T I C A

OpenStreetMap and Ordnance Survey Datasets forLondon and the Rest of England. Under Review inEnvironment & Planning B.

Hunter, G.J. and M.K. Beard. 1992. Understanding Error inSpatial Databases. Australian Surveyor, 37(2), pp.108-119.

Hunter, G.J. 2001. Keynote Address: Spatial Data QualityRevisited. Proceedings of the 3rd Brasilian Geo-Information Workshop (Geo 2001), Ocotober 4-5 , Riode Janeiro, Brazil.

ISO-TC/211. 2003. Geographic Information—Metadata19115.

ISO-TC/211. 2004. Geographic Information—Profiles19106.

Johnston, R.J., D. Gregory, G. Pratt, and M. Watts. 2000.The Dictionary of Human Geography. 4th Edition,Blackwell.

Juran, J.M., F.M.J. Gryna and R.S. Bingham. 1974. QualityControl Handbook, New-York, McGraw-Hill.

Kouandi, O. and M. Haklay. 2009. Assessing the Quality ofOpenStreetMap Data. Thesis, University College ofLondon.

Laituri, M. 2003. The Issue of Access: An AssessmentGuide for Evaluating Public Participation.Geographic Information Science Case Studies, Urbanand Regional Information Systems Association(URISA) Journal, 15 APA II: pp. 25–32.

Longley, P.A., M.F. Goodchild, D.J. Maguire and D.W.Rhind. 2001. Geographic Information Systems andScience. John Wiley & Sons, p. 124.

MacEachren, A.M., A. Robinson, S. Hopper, S. Gardner, R.Murray, M. Gahegan and E. Hetzler. 2005.Visualizing Geospatial Information Uncertainty:What We Know and What We Need to Know.Cartography and Geographic Information Science,32(3), pp. 139-160.

Paradis, J. and M.K. Beard. 1994. Visualization ofSpatial Data Quality for the Decision Maker: A DataQuality Filter. URISA Journal, 6(2), pp. 25-34.

Scharl A. and K. Tochtermann. Editors, 2007, TheGeospatial Web: How Geobrowsers, SocialSoftware and the Web 2.0 are Shaping the NetworkSociety. Springer, London.

Seeger, C.J. 2008. The Role of Facilitated VolunteeredGeographic Information in the Landscape Planningand Site Design Process. GeoJournal, 72(3-4), pp.199-213.

Authors

Joel Grira is a Ph.D. candidate in theDepartment of Geomatics Sciences, Faculty ofForestry, Geography and Geomatics, at LavalUniversity, Canada. He obtained a Bachelor degreein computer science in 2004 and a Master ofBusiness Administration (MBA) in 2006 at thesame university. His research interests are in thefield of spatial data quality and GIS. He is also anIT consultant at Fujitsu Consulting.

Dr. Bedard has been professor in SpatialDatabases and Spatial OLAP at Laval Universitysince 1986. He holds an NSERC IndustrialResearch Chair and has a multi-million dollarrecord in R&D. He has contributed to more than100 fully-refereed papers and 300 other papers andconferences. He recently co-founded Intelli3, a pri-vate spin-off merging GIS and BusinessIntelligence solutions.

Stéphane Roche has a surveying engineerdegree (Paris, 1993), a Master of Planning (Angers,France, 1994) and a Ph.D. in geography (Angers,France, 1997). He has been an associate professorfor five years at University of Angers, France. Hejoined the Geomatics Sciences Department at LavalUniversity in 2003, and has been the chair of theDepartment since 2007. He teaches GIS, cartogra-phy, and geographic information society. Hisresearch focuses on spatial data infrastructures,access, and participation—PPGIS, neogeographyand VGI, and wikiGIS. o

71