GEOGRAPHY

REVIGIS at Leicester and CEH Monks Wood
Alexis Comber, Peter Fisher, Richard Wadsworth

REVIGIS is a pan European project looking at the Revision of Geographic Information. Leicester is involved in Work Package 2 which is concerned with identifying reasoning formalisms for multiply defined land classification and for dealing with many-to-many relations between different classification elements. Leicester and CEH are involved in the following REVIGIS work packages:
WP2.2 (Reasoning Formalisms for Multiply Defined Land Classification)
WP2.4 Completing a Many-to-Many Classification
Documentation Contact Details

Work Description
The generic problem we are addressing is how to overcome ontological differences between datasets. Repeat inventories of land resources are rarely executed with the same methodology and set of objectives. We have shown such changes to be due to technological developments and changes in policy, thereby creating new questions to which answers are sought. The example application we are working with is how to determine change from land cover data for the UK produced by CEH in 1990 (LCMGB) and 2000 (LCM2000). These have structural and semantic differences and come with a 'health warning' against direct comparison. A presentation of the problem is here

Approaches

1. Straight forward statistical approaches using tables of mean correspondences are inadequate : tables of correspondence presents the general case they do not give any indication of the spatial distribution of correspondence not any information that relates to the individual object (pixel or parcel).
Documentation is here

2. We have developed semantic statistical approaches using expressions of expert opinion about how the concepts in the LCMGB (1990) and LCM2000 relate to each other. This approaches is generates 2 characterisations of each LCM2000 parcel based on the
a) the number and type of intersecting LCMGB pixels as interpreted through the Expert Semantic relations
b) the parcel spectral heterogeneity attribute interpreted through published information about expected spectral overlap that accompanied the release of LCM2000
Comparing these characterisations in a feature space of Unexpectedness (x) and Expectedness (y) enables a vector to be calculated and visualised. We hypothesised that the parcels with the greatest vectors were candidate areas of change. The vectors illustrate the extent of ontological change between LCMGB and LCM2000.
A full description of this approach is here, an executive summary is here, and a presentation is here.

3. We have validated the approach by visiting parcels identified to have changed. the first set of results revealed considerable artefacts in the spectral attribution. These are described here. The first results allowed a series of filters to be identified which were then applied in the analysis. The second results are much improved, identify change abd reveal some endemic data artefacts. These are described here

4. Ongoing work is incorporate these characterisations into Dempster-Shafer formalisms

WP 2.2 Reasoning formalisms for multiply defined land classification

Work Topic Delivery Date Months work

Scenario definition for the problem of multiply defined land classification 31 December 2001 1

Identification of appropriate formalisms for this problem area with WP 1 (Wilson and Vasseur) and from the literature (classical and novel) 30 April 2002 4

Developing formalisms with WP1 30 September 2002 5

Revisions to formalisms in light of implementation 30 April 2004 18

WP2.4 Completing a many-to-many classification in a land cover to land use experiment

Work Topic	Delivery Date	Months work
Definition for the application of multiply defined land classification (many-to-many classification in the context of a land cover to land use experiment)	31 December 2001	1
Datasets available within project	28 February 2002	2
GeoMedia integration	31 March 2002	1
Finalising Ontologies	31 April 2002	3
Implementing Rough set approach	31 October 2002	6
Implementing Probabilistic approach	30 April 2003	6
Implementing Belief revision approach	31 October 2003	6
Final report preparation and experiments	30 April 2004	6

Documentation

Papers
Comber, A. Fisher, P. and Wadsworth, R., (2003) Actor Network Theory: a suitable framework to understand how land cover mapping projects develop? Land Use Policy, in press.

Comber, A. Fisher, P. and Wadsworth, R., (submitted) A semantic statistical approach for overcoming ontological differences: how to relate similar features mapped in different ways. Submitted to IJGIS March 2003.

Comber, A. Fisher, P. and Wadsworth, R., (Soon to be submitted). What is land cover? Intended journal: Environment and Planning B.

Comber, A. Fisher, P. and Wadsworth, R., (Soon to be submitted). Semantic Statistical Approach, Second Results: Change. Intended journal: Photogrammetric Engineering and Remote Sensing

Trung, P. and Comber, A., (Soon to be submitted). Fusion Of Different Sources By Using Galois Lattices. Intended Journal : undecided

Vasseur, B. and Comber A., (Soon to be submitted). Data Quality revisited: data quality using the LCMGB-LCM2K problem as an example. Intended Journal : undecided

Reviewed chapters in books
Fisher, P.F., Comber, A.J. & Wadsworth, R. (2002). Land use and Land cover: Contradiction or Complement?. In Re-Presenting GIS, edited by David Unwin. Wiley, in press.

Comber AJ, Fisher PF and Wadsworth RA, (2002). Creating Spatial Information: Commissioning the UK Land Cover Map 2000. pp. 351-362 in Advances in Spatial Data, (eds. Dianne Richardson and Peter van Oosterom). Springer-Verlag, Berlin.

Reviewed conferences
Comber, A.J., Fisher, P.F and Wadsworth, R.A, (2002). Land use and land cover: contradiction or complement ?, Association of American Geographers Annual Meeting, Los Angeles, March 2002

Comber, A.J., Fisher, P.F and Wadsworth, R.A, (2002). Alternative Interpretations of Confusion Matrices / Concept Lattices, GIS Research UK Conference, Sheffield, April 2002

Comber, A.J. & Fisher, P.F. & Wadsworth, R. (2002). Creating Spatial Information: Commissioning the UK Land Cover Map 2000, Spatial Data Handling, SDH'02, Ottawa, August 2002.

Fisher, P.F., Comber, A.J., Wadsworth, R.A., (2002). The production of uncertainty in spatial information: the case of land cover mapping. Pp. 60-65 in Accuracy 2002, 5th International Symposium on Spatial Accuracy Assessment in Natural Resources and Environmental Sciences, (eds. Gary Hunter and Kim Lowell), 10-12th July,

Comber, A.J., Fisher, P.F and Wadsworth, R.A, (2003). A semantic statistical approach for identifying change from ontologically divers land cover data. Pp. 123-131 in AGILE 2003, 5th AGILE conference on Geographic Information Science, (eds. Michael Gould, Robert Laurini, Stephane Coulondre), 24th-26th April, PPUR, Lausanne

Comber, AJ, Fisher, PF, Wadsworth, RA., (2003). A Semantic Statistical Approach to Negotiating Heterogeneous Ontologies, International Symposium on Spatial Data Quality, Hong Kong, March 2003, pp349-358 in Proceedings of the 2nd International Symposium on Spatial Data Quality ’03, ed. Wenzhong Shi, Michael F. Goodchild, Peter F. Fisher, HKPU, Hong Kong

Comber, AJ, Fisher, PF, Wadsworth, RA., (2003). Identifying Land Cover Change Using a Semantic Statistical Approach: First Results. 7th Annual conference of Geocomputation 8-10 September 2003, University of Southampton, UK

Comber, AJ, Fisher, PF, Wadsworth, RA., (2003). You know what land cover is, but does anyone else? An investigation into semantic and ontological confusion. Remote sensing and Photogrammetry Society Annual Conference, 10-13th September, university of Nottingham, UK.

For Further information contact
Lex Comber
REVIGIS Project,
Department of Geography,
University of Leicester,
Leicester, LE1 7RH, UK
Tel: +44 (0)116 252 3828
Fax: +44 (0)116 252 3854
email: ajc36@le.ac.uk
http://www.geog.le.ac.uk/staff/ajc36/

Lex Comber
The views expressed in this document are those of the document owner.