Towards an Ontologically driven GIS to Characterize Spatial Data Uncertainty

Similar documents
GEO-INFORMATION (LAKE DATA) SERVICE BASED ON ONTOLOGY

MODELING DEM UNCERTAINTY IN GEOMORPHOMETRIC APPLICATIONS WITH MONTE CARLO-SIMULATION

Formalization of GIS functionality

Using Ontologies for Integrated Geographic Information Systems

Canadian Board of Examiners for Professional Surveyors Core Syllabus Item C 5: GEOSPATIAL INFORMATION SYSTEMS

Using Ontologies for Integrated Geographic Information Systems

Twenty Years of Progress: GIScience in Michael F. Goodchild University of California Santa Barbara

Representing Uncertainty: Does it Help People Make Better Decisions?

Deriving Uncertainty of Area Estimates from Satellite Imagery using Fuzzy Land-cover Classification

A General Framework for Conflation

Key Words: geospatial ontologies, formal concept analysis, semantic integration, multi-scale, multi-context.

Cell-based Model For GIS Generalization

WEB-BASED SPATIAL DECISION SUPPORT: TECHNICAL FOUNDATIONS AND APPLICATIONS

Spatial Analysis and Modeling (GIST 4302/5302) Guofeng Cao Department of Geosciences Texas Tech University

POPULAR CARTOGRAPHIC AREAL INTERPOLATION METHODS VIEWED FROM A GEOSTATISTICAL PERSPECTIVE

UNCERTAINTY EVALUATION OF MILITARY TERRAIN ANALYSIS RESULTS BY SIMULATION AND VISUALIZATION

SPATIAL INFORMATION GRID AND ITS APPLICATION IN GEOLOGICAL SURVEY

SVY2001: Lecture 15: Introduction to GIS and Attribute Data

GIS at UCAR. The evolution of NCAR s GIS Initiative. Olga Wilhelmi ESIG-NCAR Unidata Workshop 24 June, 2003

GEOGRAPHIC INFORMATION SYSTEM (GES203)

The Architecture of the Georgia Basin Digital Library: Using geoscientific knowledge in sustainable development

Linking local multimedia models in a spatially-distributed system

Cadcorp Introductory Paper I

Mappings For Cognitive Semantic Interoperability

Geo-spatial Analysis for Prediction of River Floods

Introduction to Geographic Information Systems

USING GIS CARTOGRAPHIC MODELING TO ANALYSIS SPATIAL DISTRIBUTION OF LANDSLIDE SENSITIVE AREAS IN YANGMINGSHAN NATIONAL PARK, TAIWAN

CLICK HERE TO KNOW MORE

KNOWLEDGE-BASED CLASSIFICATION OF LAND COVER FOR THE QUALITY ASSESSEMENT OF GIS DATABASE. Israel -

[Figure 1 about here]

Spatial Webs. Michael F. Goodchild University of California Santa Barbara

SPATIAL-TEMPORAL TECHNIQUES FOR PREDICTION AND COMPRESSION OF SOIL FERTILITY DATA

Error Propagation and Model Uncertainties of Cellular Automata in Urban Simulation with GIS

Toward an automatic real-time mapping system for radiation hazards

REGIONAL SDI DEVELOPMENT

Taxonomies of Building Objects towards Topographic and Thematic Geo-Ontologies

1. Omit Human and Physical Geography electives (6 credits) 2. Add GEOG 677:Internet GIS (3 credits) 3. Add 3 credits to GEOG 797: Final Project

ASPECTS REGARDING THE USEFULNESS OF GEOGRAPHICALLY WEIGHTED REGRESSION (GWR) FOR DIGITAL MAPPING OF SOIL PARAMETERS

Mapping Landscape Change: Space Time Dynamics and Historical Periods.

DEM-based Ecological Rainfall-Runoff Modelling in. Mountainous Area of Hong Kong

Statistical Perspectives on Geographic Information Science. Michael F. Goodchild University of California Santa Barbara

Accuracy and Uncertainty

Lecture 12. Data Standards and Quality & New Developments in GIS

A spatial literacy initiative for undergraduate education at UCSB

A GIS based Decision Support Tool for the Management of Industrial Risk

An Introduction to Geographic Information System

Lecture 1: Geospatial Data Models

GEOGRAPHY (GE) Courses of Instruction

Sensitivity analysis of a Decision Tree classification to input data errors using a general Monte-Carlo error sensitivity model

Fundamentals of Geographic Information System PROF. DR. YUJI MURAYAMA RONALD C. ESTOQUE JUNE 28, 2010

What is GIS? G: Geographic, Geospatial, Geo

Semantic Granularity in Ontology-Driven Geographic Information Systems

Watershed Sciences 4930 & 6920 GEOGRAPHIC INFORMATION SYSTEMS

Visualizing Uncertainty In Environmental Work-flows And Sensor Streams

Introduction to Geographic Information Science. Updates/News. Last Lecture 1/23/2017. Geography 4103 / Spatial Data Representations

What are the five components of a GIS? A typically GIS consists of five elements: - Hardware, Software, Data, People and Procedures (Work Flows)

Course Syllabus. Geospatial Data & Spatial Digital Technologies: Assessing Land Use/Land Cover Change in the Ecuadorian Amazon.

a system for input, storage, manipulation, and output of geographic information. GIS combines software with hardware,

Uncertainty in Geographic Information: House of Cards? Michael F. Goodchild University of California Santa Barbara

A soft computing logic method for agricultural land suitability evaluation

gathered by Geosensor Networks

GRADUATE CERTIFICATE PROGRAM

GIST 4302/5302: Spatial Analysis and Modeling

VCS MODULE VMD0018 METHODS TO DETERMINE STRATIFICATION

Software. People. Data. Network. What is GIS? Procedures. Hardware. Chapter 1

of a landscape to support biodiversity and ecosystem processes and provide ecosystem services in face of various disturbances.

Investigation, Conceptualization and Abstraction in Geographic Information Science: Some Methodological Parallels with Human Geography

Transiogram: A spatial relationship measure for categorical data

Basics of GIS. by Basudeb Bhatta. Computer Aided Design Centre Department of Computer Science and Engineering Jadavpur University

CPSC 695. Future of GIS. Marina L. Gavrilova

Creating A-16 Compliant National Data Theme for Cultural Resources

A 3D GEOVISUALIZATION APPROACH TO CRIME MAPPING

The Quality of Geospatial Context

Spatial Knowledge Acquisition from Addresses

GIS Visualization: A Library s Pursuit Towards Creative and Innovative Research

FUNDAMENTALS OF GEOINFORMATICS PART-II (CLASS: FYBSc SEM- II)

GIST 4302/5302: Spatial Analysis and Modeling

Chapter 5. GIS The Global Information System

An internet based tool for land productivity evaluation in plot-level scale: the D-e-Meter system

GIST 4302/5302: Spatial Analysis and Modeling

Enhancing a Database Management System for GIS with Fuzzy Set Methodologies

Modeling the Uncertainty of Slope and Aspect Estimates Derived from Spatial Databases

An Introduction to Geographic Information Systems (Fall, 2007)

M.Y. Pior Faculty of Real Estate Science, University of Meikai, JAPAN

Part 1: Fundamentals

Spatial Metadata and GIS for Decision Support

From Research Objects to Research Networks: Combining Spatial and Semantic Search

Foundation Geospatial Information to serve National and Global Priorities

A Model of GIS Interoperability Based on JavaRMI

Web Visualization of Geo-Spatial Data using SVG and VRML/X3D

A Small Migrating Herd. Mapping Wildlife Distribution 1. Mapping Wildlife Distribution 2. Conservation & Reserve Management

AS/NZS ISO :2015

Quality Assessment of Geospatial Data

GEOG 377 Introduction to Geographic Information Systems, Spring, 2002 Page: 1/6

Page 1 of 8 of Pontius and Li

USE OF RADIOMETRICS IN SOIL SURVEY

AN OBJECT-ORIENTED DATA MODEL FOR DIGITAL CARTOGRAPHIC OBJECT

GEOGRAPHIC INFORMATION SYSTEMS Session 8

Geographical knowledge and understanding scope and sequence: Foundation to Year 10

GIST 4302/5302: Spatial Analysis and Modeling Lecture 1: Overview

Transcription:

Ontologically driven GIS for Spatial Data Uncertainty 1 Towards an Ontologically driven GIS to Characterize Spatial Data Uncertainty Ashton Shortridge, Joseph Messina, Sarah Hession, & Yasuyo Makido Department of Geography and Center for Global Change and Earth Observations, Geography Building, Michigan State University, East Lansing, MI 48824 1115, USA, email: ashton@msu.edu, jpm@msu.edu, makidoya@msu.edu. Abstract Current data models for representing geospatial data are decades old and well developed, but suffer from two major flaws. First, they employ a onesize fits all approach, in which no connection is made between the characteristics of data and the specific applications that employ the data. Second, they fail to convey adequate information about the gap between the data and the phenomena they represent. All spatial data are approximations of reality, and the errors they contain may have serious implications for geoprocessing activities that employ them. As a consequence of this lack of information, users of spatial data generally have a limited understanding of how errors in data affect their particular applications. This paper reviews extensive work on spatial data uncertainty propagation. It then proposes development of a data producer focused ontologically driven GIS to implement the Monte Carlo based uncertainty propagation paradigm. We contend that this model offers tremendous advantages to the developers and users of spatial information by encapsulating with data appropriate uncertainty models for specific users and applications.

2 1. Introduction The propagation of spatial data uncertainty to application results is not a new idea. Examples from the research literature include a host of implementations, a few examples of which follow to illustrate the point. Canters et al. (2002) employ a Monte Carlo approach to assess the impact of input uncertainty on landscape classification using land cover and terrain data. Aerts et al. (2003) demonstrate its utility for spatial decision support. Ehlschlaeger et al. (1997) investigate the spatial and cost variability of cross country routing due to elevation uncertainty, while Fisher (1991) considers viewshed uncertainty. Variability in soil properties has been examined in numerous studies via this technique, including Heuvelink (2002). In spite of substantial work by the research community, the expectations of early adventurers into spatial data uncertainty propagation (e.g. Openshaw 1989) for widespread adaptation have not been realized. A variety of reasons for this failure exist: complicated models, lack of necessary information, and difficulties with existing GIS, including both the propagation itself, as well as the preprocessing and fusion of data from different sources and spatial resolutions. The following paragraph treats each of these factors in turn. First, uncertainty models are highly sophisticated, typically requiring specialized knowledge of geostatistics to comprehend (Heuvelink 2002) and employ. While software capable of implementing such models is becoming more accessible (e.g. Pebesma and Wesseling 1998; R Development Core Team 2005), the concepts behind these models as well as their parameterization remain forbidding. Second, the information required to parameterize such models, particularly the spatial structure of error, is often unavailable to users (Heuvelink 2002). In general, current metadata standards are inadequate because they fail to convey meaningful information about the fitness for use of data (Fisher 1998; Frank 1998). Approaches to modify the metadata paradigm to address data quality more robustly have been identified. Devillers et al. (2005) propose a framework to structure and communicate data quality about a hierarchy of attributes. Duckham et al. (2001) introduce an ontology of imperfection encompassing error, imprecision, and vagueness, and demonstrate how rough sets might be used to incorporate an assessment of vagueness in retail siting. Third, GIS must be extended to easily handle Monte Carlo

Ontologically driven GIS for Spatial Data Uncertainty 3 simulation. (Karssenberg and de Jong 2005). Methods for incorporating model based uncertainty have also been proposed (Hwang et al. 1998). Methods for integrating model uncertainty and data quality have also been developed (Crosetto and Tarantola 2001). Fourth, the interpretation of results is not straightforward, especially if the desired output scale is not the same as the input (Heuvelink 2002; Heuvelink 1998). We suggest that the overarching problem is the general manner in which spatial data are distributed, and employed, and that its solution requires fundamental changes in this system. The challenge of handling spatial data uncertainty is in fact linked to the challenge of spatial data handling. The following section describes the current state of practice. Section three is concerned with the gap between this current state and what is evidently desirable, and a superior method. Section four integrates this methodology with an ontologically driven geographic information system, and illustrates what such a system might look like. Section five concludes with a discussion of key implementation challenges and expectations. 2. Metadata, Fitness, and Propagation The days when individual researchers digitized data from paper maps are largely behind us. Most digital spatial data many gigabytes per day are developed from primary sources and distributed by providers. Global networking and the continuous development of new application domains have been responsible for these important changes to information dissemination and application processes (Fonseca et al. 2002). Contemporary information systems are becoming increasingly distributed and heterogeneous, while digital libraries are now a significant component of this emerging trend towards knowledge based distributed environments. A significant challenge is how to integrate geographic information of different kinds at different levels of detail (Longley et al. 1999), and to provide spatial data users with appropriate data for their particular applications. Because spatial data producers and users are generally not the same, carefully designed metadata specifications are critical for enabling the fullest possible use of spatial data (Hill et al. 2000). Indeed, the rationale for these specifications is to enhance the sharing of spatial information, to encourage consistency in data generation and use, and to reduce

4 redundancy in data compilation (SDTS 1996). To achieve these admirable goals, metadata reports generally consist of information about the spatial location and extent of the data, the methods by which the data were collected, the storage format, the producer and distributor, and, most relevant for this paper, the quality of the data (Guptill 1999). While the Federal Geospatial Data Committee (FGDC) standard was released in 1994, FGDC development efforts began in 1992 (FGDC 1998) and the standard owes much of its content to work dating from early in the previous decade by the National Committee for Digital Cartographic Data Standards (Chrisman 1984). In fact, the five primary elements of spatial data quality are present in the Chrisman paper, published a decade before the release of the FGDC standard. So, the notion of characterizing data quality in terms of these five components has been around for a long time. The fitness for use objective has been around for at least as long (Chrisman 1984). The highly practical intent of the data quality specifications is clearly to enable data users to determine whether particular data sets are appropriate for their applications. This language is repeated throughout the various map and accuracy standards documentation (FGDC 1998; SDTS 1996; SDTS 1998). Typical data users are not interested in the data set itself, but rather in the phenomenon that the data set imperfectly represents. They need to know how imperfect this representation is, as it relates to their applications. Consider a forester who wishes to use a USGS DEM to help identify promising sites for a new fire tower. The forester has calculated the size of the viewshed for a set of locations, and is interested in determining how closely the calculated viewshed matches the actual viewshed at these sites. That the RMSE for the quadrangle does not exceed 7 meters is not a detail that can easily be used to determine the quality of the viewshed calculations (Fisher 1991). This situation is not unique to the USGS DEM case, but is a general failing of data quality standards (Goodchild 1995). Current standards fail to satisfy their government mandated objective to, "report what data quality information is known", so that users can make informed decisions about the applicability of the data for their applications (USGS 1996). Specifically, currently reported global measures of data quality are inadequate for developing the models that drive the propagation method, since they provide no information about spatial structure (Fisher 1994;

Ontologically driven GIS for Spatial Data Uncertainty 5 Wong and Wu 1996). Indeed, current map and data accuracy standards in general are not sufficient to characterize the spatial structure of uncertainty (Goodchild 1995; Unwin 1995). Without a model characterizing local spatial variation, data users are unable to complete a fitness for use assessment. We suggest that the only general purpose method for achieving the fitness for use criterion is direct propagation of uncertainty from data to the specific application result. This requires a model capable of generating multiple realizations of the phenomenon of interest and a Monte Carlo simulation implementation to apply these realizations to a specific spatial application (see Figure 1). The simulation method is completely general, in the sense that any spatial application may be handled. This appears to be the most adequate way of satisfying the fitness for use requirement (examples and arguments in favor of the approach appear in Heuvelink et al. 1989; Fisher 1991; Lee et al. 1992; Englund 1993; Ehlschlaeger et al. 1997). The implications for modeling uncertainty stochastically are profound: 'fitness for use' mandate for Federal spatial data producers is satisfied multiple sources of data may be combined to improve information content data developers have more control over distribution of information users understand the effects of uncertainty on a case by case basis analysis results are more realistic, more defensible, and more useable DATA Aux. Data Uncertainty Model Realizations GIS Operation Distribution of Results Fig. 1. Propagation of uncertainty from data to results

6 3. Ontology driven GIS In an ontology driven geographic information system (ODGIS), an ontology is a component, such as the database, designed and functioning to fulfill the systems objectives (Guarino 1997). The first step to build an ODGIS is to specify the ontologies that are stored and translated into a formal computing language. The stored ontologies also serve as a library of information about the knowledge embedded within the system. The ontologies are reformed as objects or classes that contain the operations and attributes that constitute the system's functionality (Fonseca and Davis 1999). With this research, special emphasis is given to using the ontological structures for semantic information integration between geographic information systems and geospatial data as both product and producer of multidimensional data. The use of ontologies in GIS has been discussed by Frank (1998, 1997) and Smith (1998). Ontology plays a central role in the construction of next generation GIS through the establishment of correspondences and interrelations among different domains of spatial entities and relations (Smith 1998). Frank (1997) believed that the use of ontologies would contribute to better information systems by avoiding the problems associated with multiple data model representations in existing GIS. Kemp and Vckovski (1998) note that although certain types of geographic phenomena, like discrete objects, have been the subject of ontological study, spatially continuous phenomena have received little attention. The proposed approach provides dynamic and flexible information exchange and allows partial integration of information when security or reliability contraints make absolute confidence impossible. The ODGIS approach is dramatically different than the simple import or data transfer options already in existence. The ODGIS model allows for a common ground in which multiple data models, algorithms, error specifications, and data sets with very different spatial and temporal characteristics can interact (Fonseca et al. 2002). The ODGIS structure has two main components: knowledge generation and knowledge use. Knowledge generation involves the specification of the ontologies, the generation of new ontologies from existing ones, and the translation of the ontologies into software components (Matsuyama and Hwang 1990). The ontologies or definitions are available to the end user

Ontologically driven GIS for Spatial Data Uncertainty 7 and contain metadata. ODGIS are infused with two basic premises: ontologies are explicit before the GIS is developed, and a hierarchical structure. Explicit ontologies are not database schemas, but rather an agreement on the nature of the entities to be generated. The data, e.g., terrain data, must be adapted to fit the different ontological frameworks. User communities create diverse ontologies, but it is our belief that these diverse ontologies can be explicitly specified and reconciled given the finite nature of the data. In an ODGIS implementation it is necessary to assemble and specify ontologies at different scales. An example of this procedure can be found in the document produced by the FAO of the United Nations where high level ontologies were developed for the classification of different soil types (Gregorio and Jansen 1998). The components of the hierarchy of ontologies are classes modeled or defined by their specific characteristics (e.g. parts, functions, attributes, and roles). These components, if completely specified, allow for effective representation with targeted audiences. The ODGIS framework offers a way to explicitly develop geospatial data models that are suitable for specific and different applications. From a use perspective it is a top down approach to conceptualize data. Data uncertainty is one data specific component of the conceptual model, albeit a crucially important one. It is also a bottom up perspective, beginning with data and error specification and rising to match specific applications. Both approaches embrace an object oriented implementation, in which the data model encapsulates appropriate methods to describe, depict, and employ geospatial information. The framework encourages multiple representations, enabling the implementation of truly stochastic spatial data models. 4. Tying Uncertainty Models to ODGIS We propose a prototype model to implement the uncertainty propagation model presented in Figure 1 for geospatial data with a specific goal of incorporating digital terrain data. We envision a system resembling that depicted in Figure 2; this design separates users from the spatial data maintained by the data producer and provides the data producer with substantial security and flexibility in representation. The left side of Figure 2 represents that part controlled by the spatial data producer or maintainer. This includes both a database with all terrain data and information

8 necessary for each specific error model, along with multiple error models. Some of these models are 'better' than others, in that they use higher quality information to produce more accurate and/or higher spatial resolution realizations. The right side of Figure 2 represents different spatial data users or classes of users. Requests for spatial data by different users result in the provision of spatial data realizations produced by different error models. Each user would then subject those realizations to the GIS application, thereby creating a distribution of results, as shown in Figure 1 and discussed earlier. The ODGIS framework described in the previous section provides the basis for linking particular applications or users to appropriate data models. The system would enable the producer to determine an appropriate error model for a particular class of data users. Data users would not have direct access to the original spatial data; in effect, the server delivers appropriate realizations on demand to the user's computer (Shortridge 2000). This model offers security and flexibility to the data maintainer, satisfies the fitness for use mandate, and integrates two decades of GIScience research on uncertainty into the GIS knowledge production process. SPATIAL DATABASE Producer Error Model 1 Error Model 2 Error Model n User 1 Realizations User 2 Realizations User n Realizations User User 1 GIS User 2 GIS User n GIS Fig. 2. A framework for uncertain data distribution 5. Implementation and Evaluation The initial case study for the ODGIS model is set in East Africa. The Climate Land Interaction Project (CLIP), an NSF funded project, supports

Ontologically driven GIS for Spatial Data Uncertainty 9 an interdisciplinary team that is developing regional climate and land use models for this area (CLIP, 2006). The goal is to study interactions between climate change and land use/cover. As climate conditions change, it is anticipated that people will make decisions that affect the cover and use of the landscape. These changes can in turn alter critical surface conditions, with consequent feedback impacts on regional climate. Important input data for these models include land cover and digital elevation models; currently several continental and global data sets (e.g. Africover land cover data and SRTM elevation data) are available for the region. In addition to these readily available datasets, substantial reference data have been gathered in portions of the study region by CLIP researchers and others. An ODGIS will be developed for these inputs with representative models, enabling CLIP scientists to assess the sensitivity of their models to data uncertainty. Preliminary results for the study region illustrate the talk given in Vienna at SDH 2006. Designing a modeling framework is, in many ways, simpler than parameterizing, calibrating, and validating the model. We define calibration as the process of refining the parameter settings to more closely mimic reality. However, parameterization and validation of simulated systems are frequently ignored. Parameterization in this context refers to the process of assigning temporal constants to the structures, procedures, or functions within the model itself, and the process of justifying the existence of those same structures (Messina and Walsh 2001). In processbased models, parameterization occurs via direct measurement or statistical prediction. Both statistical and non statistical inferences are made with the goal of fitting the appearance of the model to reality or a spatial reference, defined by initial conditions, spatial or temporal constraints, or predefined stochastic limits. Parameterization in simulation system models is often confused with calibration through the use of commercial software packages or architectures. In existing software packages, parameters or at least their structures are often hard coded, leaving the user to fit of a list of options. Even in the most liberal of packages the list of options is distinct and predetermined; the user is left with calibration as the primary science tool. Validation is commonly performed with process based models. With system simulation models, validation is often confused with calibration. Historically, models were designed to fit Markovian reality, measured against control or reference images, as the measure of successful validation. A Markov process is a

10 sequence of experiments with a set of properties including a fixed number of states, dependence upon the present, and an outcome defined by movement to another state or maintenance of the same state. However, the most significant limit is that the neighborhood has no impact on the Markov process and successfully meeting a Markovian threshold in no way guarantees a reasonable approximation of reality. However, justifying a simulation result through theoretical interpretations also does not justify the use of the model or allow an expression of confidence in the result. One of the great challenges of simulation modeling is finding a suitable validation technique. In the ODGIS case, validation is possible through the comparison of realizations and the evaluation of (limited) reference data. While it may be argued that this test only quantifies stochastic uncertainty, it serves as a basic framework for the future development of pattern based validations. By evaluating the fit of multiple model runs, both the endpoints of the simulation and the mean terrain or land use and cover surface can be derived, and predictive evaluations accomplished by adding all of the realizations together. A test of validation for these simulations may ultimately involve a test of context. The context here is the nature of the landscapes and terrain represented though multiple realizations for some future time period. 6. Conclusion Large spatial data producers have employed the fitness for use criterion as the objective for metadata quality reporting. However, this paper argues that traditional metadata accuracy reports must change. Those who use spatial data increasingly demand to know how reliable their GIS results are, and standard accuracy statistics are not able to supply answers. This is especially unfortunate since U.S. government spatial data producers (and many others) are mandated to provide adequate measures of spatial data quality to users. Simulation based uncertainty propagation models have been developed for spatial data. These propagation models are uniquely suited to providing general fitness for use assessments, but they remain difficult to understand and utilize for most end users. Because the data producer is in the best

Ontologically driven GIS for Spatial Data Uncertainty 11 position to develop appropriate models, it seems appropriate that the producer should be responsible for model development and documentation. Substantial challenges remain. Currently employed statistical uncertainty models are poor at reproducing key geomorphological neighborhood and global characteristics. Over the coming months we intend to develop a working ODGIS prototype to propagate spatial data uncertainty to specific applications. We will encapsulate and model error for a variety of spatial data types (DEM, satellite image, and land cover) for multiple study sites around the Earth. We will implement error models employing different degrees of information and capable of different spatial resolutions to demonstrate the capacity of the framework to handle different user and application types. The ODGIS framework will be employed on standard and state of the art geoprocessing objectives, such as change detection, spatial environmental modeling, and spatial decision support. We believe the ODGIS approach combined with basic research on innovative stochastic data objects shows great promise for linking geospatial information, models of data uncertainty, and specific applications. References Aerts JCJH, Goodchild MF, Heuvelink GBM (2003) Accounting for Spatial Uncertainty in Optimization with Spatial Decision Support Systems. Transactions in GIS 7(2):211 230. Canters F, Genst WD, Dufourmont, H (2002) Assessing effects of input uncertainty in structural landscape classification. International Journal of Geographic Information Science 16(2):129 149 Chrisman NR (1984) The role of quality information in the long term functioning of a geographic information system. Cartographica 21(2 3):79 87 CLIP (2006) Climate Land Interaction Project. http://clip.msu.edu Crosetto M, Tarantola S (2001) Uncertainty and sensitivity analysis: tools for GISbased model implementation. International Journal of Geographical Information Science 15(5):415 437 Devillers R, Bédard Y, Jeansoulin R (2005) Multidimensional Management of Geospatial Data Quality Information for its Dynamic Use Within GIS. Photogrammetric Engineering & Remote Sensing 71(2):205 215 Duckham M, Mason K, Stell J, Worboys, M (2001) A formal approach to imperfection in geographic information. Computers, Environment, and Urban Systems 25:89 103

12 Ehlschlaeger CR, Shortridge AM, Goodchild MF (1997) Visualizing spatial data uncertainty using animation. Computers & Geosciences 23(4):387 395 Englund EJ (1993) Spatial simulation: environmental applications. In Goodchild MF, Parks BO, Steyaert LT (eds) Environmental modeling with GIS. Oxford Press, New York, pp 432 437 FGDC (1998) Content Standards for Digital Geospatial Metadata (2.0) FGDC STD 001 1998. Federal Geographic Data Committee: Washington, DC. http://www.fgdc.gov/standards/documents/standards/metadata/v2_0698.pdf Fisher PF (1991) First experiments in viewshed uncertainty: the accuracy of the viewshed area. Photogrammetric Engineering and Remote Sensing 57(10):1321 1327 Fisher PF (1994) Sources and consequences of error in spatial data. In Int. Symposium on the Spatial Accuracy of Natural Resource Data Bases: Unlocking the Puzzle. ASPRS: Williamsburg, VA, pp 8 17 Fisher PF (1998) Improved modeling of elevation error with geostatistics. Geoinformatica 2(3):215 233 Fonseca F, Davis C (1999) Using the internet to access geographic information: An OpenGIS prototype. In Goodchild M, Egenhofer M, Fegeas R., and Kottman C (eds) Interoperating Geographic Information Systems. Kluwer: Norwell, MA, pp 313 24 Fonseca F, Egenhofer M, Agouris P, Camara G (2002) Using Ontologies for Integrated Geographic Information Systems. Transactions in GIS 6(3):231 257 Frank A (1997) Spatial ontology. In Stock O (ed) Spatial and Temporal Reasoning. Academic Publishers: Dordrecht, pp 135 53 Frank A (1998) Metamodels for data quality description. In Goodchild M, Jeansoulin R (eds) Data Quality in Geographic Information from Error to Uncertainty, pp 15 29 Goodchild MF (1995) Sharing imperfect data. In Onsrud HJ, Rushton G (eds) Sharing Geographic Information. Rutgers University Press, New Brunswick, NJ, pp 413 425 Gregorio A, Jansen L (1998) Land Cover Classification System: Classification and User Manual. WWW document, http://www.fao.org/waicent/faoinfo/sustdev/eidirect/eire0062.htm. Guarino N (1997) Semantic matching: Formal ontological distinctions for information organization, extraction, and integration. In Pazienza MT (ed) Information Extractions: A Multidisciplinary Approach to an Emerging Information Technology. Springer, Berlin, pp 139 70 Guptill SC (1999) Metadata and data catalogues. In Longley PA, Goodchild MF, Maguire DJ, Rhind DW (eds) Geographical Information Systems, 2 nd Edition. Wiley: New York, pp 677 692 Heuvelink GB, Burrough PA, Stein A (1989) Propagation of errors in spatial modelling with GIS. International Journal of Geographical Information Systems 3(4):303 322

Ontologically driven GIS for Spatial Data Uncertainty 13 Heuvelink GBM (1998) Uncertainty analysis in environmental modelling under a change of spatial scale. Nutrient Cycling in Agroecosystems 50(1 3):255 264 Heuvelink GBM (2002) Analysing Uncertainty Propagation in GIS: Why is it not that Simple?, In Foody GM, Atkinson PM (eds) Uncertainty in Remote Sensing and GIS. John Wiley & Sons, Hoboken, NJ, pp 155 165 Hill LL, Carver L, Larsgaard M, Dolin R, Smith TR, Frew J, Rae M A (2000) Alexandria digital library: user evaluation studies and system design. Journal of the American Society for Information Science 51(3):246 259 Hwang D, Karimi HA, Byun DW (1998) Uncertainty analysis of environmental models within GIS environments. Computers and Geosciences 24(2):119 130 Karssenberg D, De Jong K (2005) Dynamic environmental modelling in GIS: 2. Modelling error propagation. International Journal of Geographical Information Science 19(6):623 637 Kemp K, Vckovski A (1998) Towards an ontology of fields. In Proceedings of the Tenth International Conference on GeoComputation. Bristol, United Kingdom. Lee J, Snyder PK, Fisher PF (1992) Modeling the effect of data errors on feature extraction from digital elevation models. Photogrammetric Engineering & Remote Sensing 58(10):1461 1467 Longley PA, Goodchild MF, Maguire DJ, Rhind DW (1999) Geographical Information Systems (Second edition). Wiley, New York Matsuyama T, Hwang V S (1990) A Knowledge based Aerial Image Understanding System. Plenum, New York Messina JP, Walsh SJ (2001) 2.5D morphogenesis: modeling landuse and land cover dynamics in the Ecuadorian Amazon. Plant Ecology 156 (1): 75 88 Openshaw S (1989) Learning to live with errors in spatial databases. In Goodchild MF, Gopal S (eds) Accuracy in Spatial Databases. Taylor & Francis, London, pp 263 276 Pebesma EJ, Wesseling CG (1998) Gstat: a program for geostatistical modelling, prediction and simulation. Computers and Geosciences 24(1):17 31 R Development Core Team (2005) R: A Language and Environment for Statistical Computing. URL:http://www.R project.org/. R Foundation, Vienna, Austria. Accessed 12/20/2005 SDTS (1996) The Spatial Data Transfer Standard: Guide for Technical Managers. U.S. Dept. Interior SDTS (1998) Spatial Data Transfer Standard ANSI NCITS 320 1998. American National Standards Institute, Washington, DC Shortridge AM (2000) Compact data models for spatially continuous phenomena. First International Conference on Geographic Information Science. Savannah, GA, Oct 28 31 Smith B (1998) An introduction to ontology. In Peuquet D, Smith B, Brogaard B (eds) The Ontology of Fields. National Center for Geographic Information and Analysis, Santa Barbara, CA, 10 4

14 Unwin DJ (1995) Geographical information systems and the problem of error and uncertainty. Progress in Human Geography 19(4):549 558 USGS (1996) DEM/SDTS transfers. In The SDTS Mapping of DEM Elements. U.S. Dept. Interior, U.S. Geological Survey Wong DWS, Wu CV (1996) Spatial metadata and GIS for decision support. 29 th Annual Hawaii Int. Conference on System Sciences. IEEE, 557 566