|
The New Age of Topographic Data – Management and Access Boutin Denis Massé François Abstract: The Centre for Topographic Information in Sherbrooke (CTIS) has reviewed the
way it produces (or updates), manages, and disseminates vector
topographic data (NTDB data and others). First of all, the clientele
was broken down into segments in order to better respond to
specific needs (specialized customers, privileged customers,
the general public, and crisis managers). We defined new vector
data specifications taking into account customer requirements
to a greater degree (update cycle, accuracy, accessibility,
change management). International standards (ISO/OGC) were used
to better describe data characteristics and to make use of concepts
familiar to the geomatics community. Landsat 7 and GPS technology
have been used to produce the initial version of this new geospatial
data. The NTDB data (150 Gb) and the upcoming data will be stored
in a database management system (Oracle8I Spatial). The spatial
data will be available in several GIS formats through Web sites
using e-commerce technology. This paper focuses on the characteristics
of this new Geospatial Database (GDB) specifications, with particular
emphasis on spatial relationship expressed by the Dimensionally
Extended Nine-Intersection Model (DE-9IM), update management
(historical data), and data access. 1 Introduction In recent years, the Centre for Topographic Information in Sherbrooke (CTIS) has produced more than 7000 files that comply with the standards of the National Topographic Data Base (NTDB). Many of these files were produced by digitizing topographic maps. The entire Canadian landmass is covered at the 1:250 000 scale, while the data at the 1:50 000 scale covers mainly the southern portion of the country. During the same period, a number of projects were implemented to update these data. While results were qualitatively interesting, they were clearly disappointing quantitatively. Nevertheless, our customers constantly require data that are more accurate and up-to-date. Because the costs for updating NTDB data were prohibitive, we are reviewing the existing specifications. How can we perform faster with fewer resources than before? The number of solutions is quite limited: reduce requirements (fewer topographic entities to produce) or join forces with partners. Both will be used. We will reduce the number of topographic entities with respect to the NTDB, while working to develop long-term work plans with other agencies. Sharing data collection, however, leads to a whole new set of problems stemming from asynchronous data management. Data integration will be more complex because sets will be partially processed. In order to define these new constraints and needs, the specifications for the Geospatial Database (GDB) [1] were drafted. 2 Entities The selection of entities contained in NTDB specifications was strongly influenced by map features. Consequently, the NTDB contains a large number of topographic phenomena. For the time being, it is impossible to keep this information up-to-date within reasonable time frames and at reasonable cost. The new standards must correspond to our capacity to capture and maintain a certain number of phenomena in the real world based on our customers' needs. This capacity will varies with time and opportunities. The first version of the specifications is built around the capacity to extract topographic information from Landsat 7 imagery. The number of entities contained in the GDB could be increased as a result of cooperative agreements. CTIS is working to develop a cooperative network with other agencies and organizations interested in the same or complementary topographic information. While building the NTDB, we set up the data in a synchronous manner. All the layers of information were processed simultaneously. A specific tile was processed as part of a complex integration process responding to specifications governing geometric sharing and connection. Moreover, the information corresponded to a specific image at a given time. Sharing the workload with collaborators opened the door to new ways of doing things. Each partner will be associated with the production of one or more themes, and probably part of the Canadian landmass. As a result, the data must be integrated asynchronously. This new approach has generated a new complex problem in which data inconsistency can crop up more frequently. The challenge is deriving a means for being flexible (to enable us to work in partnership) and coherent. The new GDB specification deals with both of these issues. 3 Spatial Relations 3.1 Flexibility For a number of years, we have been following the development of international standards (ISO TC-211 and the Open GIS Consortium). We are especially interested in the "Simple Features Specification for SQL."[9] The specifications standardize two aspects: geometric representation for access using SQL and the definition of a set of spatial operators. The implementation of these standards should lead to the commercialization of a number of standardized tools. These tools should make it easier to achieve the flexibility that we are looking for. The use of standardized geometric objects and operators will facilitate cooperation between partners. Unfortunately, the market is changing less quickly than hoped for. Nevertheless, the new GDB model uses three types of geometric primitives: Point, Curve, and Surface, in conformity with ISO [6][7] and OGC [9] documents. The spatial operators (or standardized functions) have been built according to the theory of the Dimensionally Extended Nine-Intersection Model (DE-9IM) [2][3][4][5], as described in international standards. The model makes it possible to express a great number of spatial relations (see Table 1: Concepts of boundary, interior and exterior). Using this matrix (or mask) allows to accurately describe the type of spatial relation needed. One of the standardized functions (Relate) can be used to determine the compliance of a relation between two specific objects expressed by a specific matrix. GDB standards express spatial relations using these masks. A total of 27 masks has been used. Using masks allows describing the relations expected between two geometric objects with more detail. There are, in addition to the "Relate" function, other standardized functions that describe specific masks (Equals, Disjoint, Intersects, Touches, Crosses, Within, Contains Overlaps). These functions, however, are inadequate to express the level of detail desired in the GDB for data integrity validation. Despite the fact that most geographic information systems (GISs) have spatial operators, very few, up until now, have offered the full range of standardized operators. Using these matrices to express spatial integrity constraints is inadequate
to achieve all the flexibility sought for asynchronous data integration.
The constraints must also be accompanied by a tolerance. A tolerance is
a zone defined for each object within which the presence of other objects
is checked and justified. The applicable tolerances are set for each relationship
described and each geometric representation.
Table 1: Concepts of boundary, interior and exterior 3.2 Inconsistencies The new operating approach ("asynchronous arrival of data") will inevitably yield data sets containing certain inconsistencies. The question is how to manage these "errors". Checking integrity constraints is a way to detect data inconsistencies. Many factors can account for inconsistencies, such as the fact that the data are coming from variable sources. These inconsistencies fall into three categories: accuracy, time, and continuity. Accuracy inconsistencies refer to the "errors" detected within the accuracies of the two adjacent objects. These inconsistencies can be corrected automatically by setting certain priority rules. Anything found outside the accuracy is considered as a temporal inconsistency
and these can only be corrected by updating the data. Continuity inconsistencies
serve to identify noncontinuous data. This situation is mainly caused by
per-tile acquisition using sources with different accuracies and dates.
The GDB uses three types of geometry to identify inconsistencies. An inconsistency
can be a point, curve, or surface. The geometry is set depending on the
inconsistency detected. For example, a road that extends into water generates
a linear inconsistency.
4 Change Management Until recently, the method used in distributing updated spatial data consisted in delivering all of the data, a particular theme, or all the themes for the specific area. An edition/version number was assigned to the datasets as a whole. Increasing the number of revision cycles, updating on an entity basis, and the amalgamation of data from different sources make it necessary to manage changes over time. Managing change provides users with a means for determining the exact differences in data between two specific dates. A number of projects and papers in the literature [8][10][13] deal with managing updates. They deal with a large segment of the problem (synthetic image generation, phenomenon evolution, the management of differences). CTIS has selected a change management model that is realistic on the short term and offers possibility for expansion. The model used, developed in cooperation with the Centre for Research in Geomatics (CRG) of Laval University [11] [12], focuses solely on objects change management (or computer representation of phenomena). In certain cases, the evolution of phenomena can be deducted by specialized application software; CTIS, however, will not be focusing on this. In other words, we will not monitor the causes of the changes; only their effects on the data. Due to the size of the territory to be covered and available financial resources, the evolution of the phenomenon responsible for changes in the associated computer representations cannot be monitored with an adequate level of confidence. The new process will enable us to monitor object evolution so as to identify any changes between two successive observations or not. Determining if a change has occurred depends on identifying the following effects or differences: addition, descriptive modification, deletion, and confirmation. This is the simplest form of change management. The differences could be sent to data users so that they can integrate them more easily into their systems. This integration process will be facilitated by using unique, non-significant identifiers for all of the data in the database. The following example helps illustrate update management. Figure 1:
Example of an update presents the original data and the new data, respectively.
In terms of geometry, a single road segment (object 6) was added with respect
to the original data. Descriptively, the pavement for the road segment
(object 2) changed from unpaved to paved. Table 2: Updating effects shows
the effects observed after updating.
5 Data Access CTIS's target clientele breaks down into four major groups: privileged customers, crisis managers, specialized customers and the general public. The products and access modes differ from one targeted clientele to another. Privileged customers are groups of agencies with a need for a national or provincial topographic data over a long period of time. Such organizations want to make use of geospatial data without having to worry about managing them. We will work to offer them a connection to our database. Crisis managers constitute a specific group of privileged customers. Specialized customers generally work on a specific territory for a definite period of time. They have specific needs and often carry out complex analyses. They will be provided with continuous access to the data via a Web site. They will be able to implement data and metadata queries in order to download geospatial data in a particular format. Lastly, the general public must be able to access the data over the Internet, locate them, and download a raster representation (gif) of the area of interest (http://toporama.cits.RNCan.gc.ca). 6 Conclusion In the next few years, spatially referenced data will be captured and managed differently than they are today. The new trend is towards collaboration between different stakeholders in order to obtain higher quality data and hold costs to a realistic level. While many technological problems remain, there are promising signs that they can be overcome. The development of standards in the industry (OGC) is a good indicator. Consequently, we will work in the years to come on building and infrastructure that will facilitate the integration of data from different sources. 7 References [1] Centre for Topographic Information – Sherbrooke, Geospatial Database
- Standards and Specifications, 42 pages, June 2000.
Biography Denis Boutin
Denis Boutin has fifteen years of experience in various aspects of geomatics
and in the development of related complex systems. His academic background
includes a master’s degree from the University of Sherbrooke in Computer
Science (management of historical data). Boutin currently works for the
Centre for Topographic Information (CTIS) on the development of the Geospatial
Database under Oracle8I Spatial and on problems related to spatial data
integration (geofusion or conflation).
Francois Massé
François Massé has a degree in Geodetic Science from Laval University (1985). Since obtaining his degree, he has been employed at the Centre for Topographic Information in Sherbrooke (CTIS). His main fields of expertise are spatial databases modelling and data-acquisition production processes design. His expertise was instrumental in implementing the National Topographic Data Base (NTDB), the Mexican National Geographic Information System, and the new Geospatial Database currently being developed. |
|
|||||||||||||||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||||||||||||||
