EduServ5 - 2007

EduServ5 - Course An Inquiry on the Quality of Geospatial Data and on Related Statistical Concepts

Instructors: Prof. Alfred Stein, Dr. Valentyn Tolpekin (The ITC International Institute for Geoinformation Science and Applied Earth Observation)
Date: 30 April - 11 May 2007

Course Objectives

The course provides an overview of the components of geospatial data quality and measures used for their assessment. It is intended to help the end-users to judge the fitness of use of a geospatial data set and to evaluate the quality of the derived information products. The course combines theoretical material and hands-on exercises that will allow participants to work with provided data sets and software.

Course Outline

Module 1 - Introduction of the various components of the quality of geospatial data
A general introduction on geospatial data quality is given, based on statistical issues of data quality. Examples are included and ways of analyzing and identifying them are provided. Emphasis is on a proper description and definition, as well as on analysis and quantification methods. Linear and circular errors, error propagation and the error band model will be used as statistical tools. The variogram is introduced as a tool to modeling positional data quality.

• Geospatial data quality
• Linear errors
• Circular errors
• Error propagation
• Error bands
• Variogram

Module 2 - Emphasis on positional uncertainty of point, line and area objects and consequences on model calculations
Some possibilities of addressing positional uncertainty and its consequences are discussed. Point, line and area objects are distinguished. Two real life models that work on a GIS are introduced. Consequences in practical case studies will be presented and analyzed.

• Correlated bands and images
• Polygon uncertainty
• Spatial utility values

Module 3 - Thematic uncertainty and thematic completeness
Emphasis will be put on thematic completeness of spatial information. Here, we understand uncertainty that may arise due to a poor definition, incomplete data and inaccurate data. Traditionally, the error matrix is used for that purpose. We will work on it with the Bradley-Terry model, as an extension of the kappa coefficient.

• Error matrix
• Classification and segmentation
• The kappa coefficient
• The Bradley-Terry model