Juha Tuomaala, Mikko Uimonen - Finland
ABSTRACT
During 1993-1997 the National Land Survey of Finland (NLS) introduced a new GIS-based cadastral information system in which attribute and map data on cadastral unit are stored in the same database. Maintenance of the Cadastre, conducting of cadastral surveys and an efficient information service are now possible using the same information system. The system employs a map user interface and a document user interface which was developed at the NLS. The information system also includes tools for management of long transactions, which are typical in cadastral information systems. The system comprises a centralized 50 GB database, a cache server in each of the NLS District Survey Offices, and 700 workstations supplied with Pentium II processors for cadastral survey production.
RESUME
Le Service de l’arpentage de Finlande a mise en oeuvre en 1993-1997 un nouveau système d’information cadastrale (GIS) où les données cartographiques des propriétés foncières sont enregistrées dans la même base de données. L’entretien du registre cadastral, l’exécution des opérations d’arpentage et le service d’information efficace sont maintenant possibles avec le même système d’information. Le système d’information comprend la base de données concentrée d’environ 50 GB, un "server" à tous les bureaux ainsi que 700 stations de travail (Pentium II processeur) pour la production d’opérations d’arpentage.
ZUSAMMENFASSUNG
Das Landvermessungswesen von Finnland hat im Laufe der Jahre 1993-97 ein neues auf GIS basiertes Katastersystem durchgeführt. In diesem die Daten von Eigenschaft und Karten für Grundstück und Landgut sind in der gleichen Datenbasis gespeichert. Die Instandhaltung von dem Grundregister, anlässlich der Vermessungen geänderte Angaben im Register und intensiver Informationsdienst sind heute mit dem gleichen Datensystem verwendbar. Das System hat eine konzentrierte Datenbasis von etwa 50 GB, Cache-Server in jeder Nebenstelle und 700 Computer mit dem Pentium II -Prozessor für Vermessungsbetrieb.
INTRODUCTION
In 1993, the National Land Survey of Finland (NLS) began a project to renew the cadastral system. The aim was to create an information system where cadastre maintenance, conducting of cadastral surveys and an efficient information service could be carried out as integral parts of a single system. Another aim was to shorten the duration of cadastral surveys. The average duration of such surveys was about ten months in 1996.
Another reason for renewal was that the technical features of the old cadastal system, introduced in the late 1970s, were becoming outdated. In addition, a digital cadastral index map covering the whole of Finland will be completed in 1999, making it possible to offer an efficient information service from the new GIS-based JAKO system using the digital cadastral index map information.
In this paper the terms 'cadastral register' and 'digital cadastral index map' together with their analogue equivalents correspond with the common term 'Cadastre'. The Finnish registers together with the register of titles and mortgages are equivalent to the term 'Land Register'
RESPONSIBILITY FOR THE REGISTRATION OF CADASTRAL UNIT DATA IN FINLAND
The NLS is responsible for conducting cadastral surveys in rural areas. Such surveys can only be carried out by civil servants. In urban areas, the cities conduct cadastral surveys using their own civil servants. The cadastral surveys are used to update the data in the Cadastre. Responsibility for registration rests with the organization which carries out the survey. The NLS and 87 of 452 cities or municipalities have their own cadastral systems.
The Ministry of Justice is responsible for registration of titles and mortgages. Local courts under the authority of the Ministry of Justice examine the preconditions of confirmation of titles and mortgages and carry out their registration. Titles and mortgages can only be granted to cadastral units listed in the Cadastre. (Vahala 1994).
THE OLD SYSTEM
Digitizing the cadastral register
The NLS commenced the digitizing of attribute data in the cadastral register in the late 1970s. The work was completed in 1994. Attribute data were stored in a network database used in HP3000/MPE computers. The databases were held in three District Survey Offices, each of these databases containing a defined part of the total cadastral unit information for the country.
Digital cadastral index map
The NLS began producing a digital cadastral index map in the mid 1980s. GIS software (MAAGIS) developed at the NLS and used in VAX/VMS computers has been applied in the digitizing and updating of the cadastral map in vector form. The GIS software has proved excellent for data collection. However, with the developments in GIS software, customers were transferring to using seamless databases in their own GIS applications. This meant we could no longer guarantee the quality of data using GIS software based on map sheets, because most of the topological errors appear on map sheet borders. In addition, map sheet borders split objects unnecessarily and cause extra revision work for the customer.
By the end of 1997, 83 per cent of the total area of Finland was in digital form, and by 1999 the entire country should be covered. About half of the digital cadastral map in vector form is produced using photogrammetric methods. Approximately half of the boundary points in this data have a spatial accuracy of 0.25 metre, while the rest of the boundary points, which were not marked in photogrammetric surveys, have an accuracy of about four metres.
Figure 1: Digital cadastral index map produced photogrammetrically
The digital cadastral index map for sparsely populated areas is produced by digitizing existing cadastral maps in scale of 1:10 000. The spatial accuracy of this data is 10 metres at most.
Figure 2: Digital cadastral index map produced by digitizing
Problems with the old system
There were many problems with the old system, the most significant being:
JAKO - THE NEW CADASTRAL INFORMATION SYSTEM
Development
Development of the new system was begun with data definition projects carried out in 1990-1992. After a selection process which lasted a year, the commercially available GIS software (Smallworld GIS) was chosen as the development tool in 1994. The production software of the new system was developed during 1995-1997. The development work involved 15-17 NLS application analysts, depending on the time in question.
Development tools
Smallworld GIS offers an object-oriented programming environment in which genuine object-oriented programming language (Magik) is used. In addition, a graphic Case Tool is used in the development environment for designing and implementing the database.
Due to its long history (data covering 470 years) the Finnish cadastral system includes a lot of special features and details particular to different eras which all had to be implemented into the system. For this reason, one of the main criteria in choosing the GIS-software was the availability of efficient development tools which would allow considerable tailoring of the application during development. Other important selection criteria included the behaviour of large databases in performance tests, a well-functioning distribution solution and models and tools for long transaction management offered by the supplier.
Improvements in the information service
The need for efficient and versatile information service functions was given special attention in the design and implementation of the new JAKO system. The new system allows an efficient information service with versatile attribute predicates (name, date, area, owner etc.) as well as spatial predicates on a map. It is also possible to use both of these priedicates. The cadastral database covering the whole of Finland is available for users via an adaptable map user interface, see Figure 3.
Figure 3: Information service via an adaptable map user interface
Functional improvements in maintaining the database
Surveyors employed by the NLS and major cities maintain the cadastre by carrying out cadastral surveys. Survey production at the NLS employs about 400 surveyors and over 600 other persons in 21 District Survey Offices throughout Finland. These offices, in turn, have their own sub-units, the total number of service points thus totalling 42.
Figure 4: There are 21 NLS District Survey Offices and 21 sub-units throughout Finland
The NLS carries out some 25,000 cadastral surveys annually, and data on some 200,000 cadastral units are updated in these cadastral surveys.
The new system was developed to include extensive tools for conducting cadastral surveys, involving everything from processing an application for a survey to its registration. This makes it possible for one person to carry out all the work phases of a simple survey (for example parcelling), including final registration. This new process should speed up the completion of surveys significantly, since there is no longer any need to pass work from one person to another.
In terms of register maintenance, it is significant that attribute and spatial data are processed in the same transaction, as this quarantees the consistency of attribute and spatial data in all circumstances.
Database
Attribute and map data (in vector form) are stored in the same seamless VMDS (Version Managed DataStore) database. The VMDS is a proprietary DBMS, which is integral to Smallworld GIS. Like most other DBMSs it is relational in nature, supporting flat tables in which records are related by data values in fields. The VMDS is used to store geometry and raster data, and may also be used to store attribute data if desired. The primary difference between VMDS and other relational DBMSs is that it supports multiple 'alternatives' of the same database in an economical manner. A user may easily switch between alternatives, or compare alternatives for differences. The ability to support multiple database alternatives enables the VMDS to provide a long transaction concurrency model. This means that many users may update the database simultaneously over long periods, each working on their own stable and self-consistent alternative, without locking out other users doing the same thing. Within each alternative, one or more versions may exist. In, addition a 'checkpoint' facility is provided so that a user may mark a version as one of particular interest and return to it later. (Smallworld 1997)
The database includes data on over 4.2 million cadastral units, of which some 1.5 million have ceased to exist. During a cadastral survey carried out on a cadastral unit, the former parcel identifier is replaced with a new identification code (a number) in the Finnish cadastral system.
The map data includes parcel data on about five million cadastral units formed by closed polylines. There is a database join between each parcel and cadastral unit. Spatial data is stored in the seamless database in a uniform coordinate system. Coordinate calculation must be converted and output presented using Gauss-Krüger projection with five projection zones.
The database also includes raster data of Base Map at scale of 1:20 000 covering the whole of Finland. Also raster maps at other scales are used as overview maps. The total size of the database is estimated to be about 50 GB (loading of old data was commenced at the end of January 1998).
Distribution of the data
All data in the new cadastral system is stored in a database on a Unix server in Helsinki. The distribution is based on a solution developed by Smallworld in which a cache server including a Persistant Cache is installed in all 21 NLS District Survey Offices and their 21 sub-units. The local caches are populated with cadastral data of the municipalities covered by each District Survey Office. Upon initial creation the local cache is ‘empty’ and must be populated. Users can be left to do this during the course of normal usage. Upon first access all data is ‘hauled’ over the wide area network (WAN) connection and then cached. These cache servers have a 2 Mbps WAN connection to the main server. All workstations in the District Survey Offices are connected to the cache servers with a 10 Mbps local area network (LAN) connection. When a user at one of the district offices makes an inquiry regarding cadastral unit information relevant to that NLS district office, the result would normally be received quickly from the cache server via the LAN. Only if the data required has been altered in a cadastral survey, it is necessary to use WAN connection to get the updated data from the main server. This allows an efficient information service at each district office, because most inquiries are made on cadastral information concerning municipalities for which the particular district office is responsible. This solution significantly decreases the communication burden on the critical WAN which would otherwise grow intolerably large because of the map user interface.
All new and updated data generated in cadastral surveys are stored directly to the main server database. At the end of 1997 the NLS acquired 240 portable Windows NT-workstations with 166 Mhz Pentium processors and 460 NT-workstations with 233 Mhz Pentium II processors for cadastral survey production.
Updating the cadastre with cadastral surveys
The data in the cadastral system is maintained and updated with data derived from cadastral surveys. Managing the long transaction forms the most significant element of the updating process. A long transaction means an action in which the updating of database items takes several hours, even weeks or months before the updated data can be registered. Cadastral surveys are a typical example of a long transaction.
Commercial relational databases have been developed to manage short transactions. For instance, bank account transaction maintenance is a typical short transaction: a withdrawal is immediately debited from the account. Many GIS suppliers have sought to solve the problem of the long transaction by using the check-out/check-in procedure. According to this procedure, data on the relevant cadastral unit are copied to a separate work database and the check-out data is locked from other users for the duration of the survey. During this time, other users only have read-only access to the data.
Long transactions are typical in cadastral surveys. These transactions can last up to several years if the land owners in cadastral survey appeal to the supreme court. The average duration of a cadastral survey in Finland is ten months. Due to the relatively long duration of surveys, it was not possible to use the check-out/check-in procedure in our cadastral system because locking out one cadastral unit parcel will, in practice, lock data for the adjoining cadastral unit due to common boundaries, and thus making it impossible to conduct surveys on neighbouring cadastral unit, see Figure 5.
Figure 5: The check-out/check-in procedure also locks data on the neighbouring parcels which is not acceptable.
Smallworld has developed its VMDS relational database to support the long transactions needed in many geographic information applications. The solution is based on a hierarchy of database alternatives. The highest alternative of the hierarchy in the database is called the top alternative. Below this, each user creates a sub-alternative for his or her own use. The sub-alternative is not a copy of the top alternative, but instead includes the changes and additions to the original data made by the user. The user is always able to see all the data which is in the top alternative when the sub-alternative is created. A sub-alternative may in turn have its own sub-alternatives.
Only one user at a time has write access to an alternative; other users have read-only access. Those viewing the top alternative do not see the changes made in the sub-alternatives. The changes cannot be seen until they are posted to the top alternative by using the 'post changes' function.
A user working on a sub-alternative may sometimes need to merge changes made by other users in the top alternative into his or her own sub-alternative. This can be carried out by using the 'merge changes' function.
A conflict situation occurs if some other user has updated the same items as those updated by the user in his or her own sub-alternative. The Smallworld database is able to detect conflicts between items when changes are merged or posted. If a conflict is detected, posting the changes is aborted and the user must sort out the conflicts before he can register the cadastral survey.
The long transaction model provided by Smallworld did not, however, meet our requirements. Using Smallworld's tools, we developed tools especially for presenting and resolving conflicts from the point of view of the end user's needs.
We also developed the possibility to use portable workstations in cadastral surveys. This required development of a check-out/check-in procedure which allows data to be copied into the database in the portable workstation and to be returned automatically when the cadastral survey is finished. In the final version the copied data is not, however, locked, but instead any conflicts are detected, the user resolve them when the updated data is returned. The mechanism can thus be called an optimistic check-out/check-in procedure, because the conflict are rather unusual in real production.
Each portable workstation has a database which in terms of its data model corresponds to the master database. The surveyor then copies the attribute and spatial data on the cadastral unit being surveyed into the workstation database. On site, the surveyor can update the copied cadastral data when carrying out the survey. Returning of the updated data is done automatically when the surveyor reconnect the workstation to the network.
A more detailed description of the structure of the updating procedure in the JAKO information system is presented in the paper How to update a cadastre - a long transaction in a modern cadastral system by Tarja Myllymäki (FIG Commission 3, small group session 2)
Management of historical data
In the Finnish cadastral system, the history of cadastral unit attribute data has been managed right from the time the system was first introduced. History data are available on the formation of each cadastral unit from the date of its original creation back to the first cadastral surveys.
The new cadastral system also includes management of the history of cadastral map data in respect of parcels boundary lines and points which are registered after system implementation. This makes it possible to submit a query which refers to the cadastral map valid at any particular date since introduction of the new system.
The solution to managing the historical map data is as follows: when a cadastral survey is registered, all deleted or altered data are recorded as historical items and the items are stored with an entry and expiry date in the database.
Document user interface
As nearly 1000 users will need to be trained to use the new information system, usability of the system interface and tools will be crucial. The development process produced an innovative way of managing within the GIS system the documents which users have been familiar with.
It is necessary to process a great amount of cadastral unit attribute data during a cadastral survey. Typical tasks include defining an easement (for example the right to use a road via a parcel of another cadastral unit) between cadastral units formed in the cadastral survey, and division of interests in the common area between the new cadastral units formed in the cadastral survey. These data are presented in a survey document which can be browsed in hypertext form on the workstation display. When the user selects hypertext, the application starts up a window needed for editing that particular item and at the same time brings the data into the data fields of the editing tool. The data can then be updated to correspond the decisions made in the survey using that particular tool. The updated data is returned to the survey document automatically, (see Figure 6). This new way of starting up the necessary editing tools from a hypertext document and simultaneously bringing the necessary data to be processed into the editing tool is what we have termed the "Document User Interface".
Figure 6. The Document User Interface based on hypertext
In this way, the surveyor updates the Cadastre at the same time as he or she produces the cadastral survey document. When all items in the survey document have been dealt with, the cadastral survey can be pre-registered. This procedure allows the surveyor to register the survey himself, and there is no need to restore the decisions and matters resolved during the survey at the registration phase.
A more detailed description of the document user interface in the JAKO information system is given in the paper Hypertext user interface in GIS - a new approach to cadastral information by Satu Miikkulainen (FIG Commission 3, small group session 3)
Internet service
A WWW service has also been developed for the system which allows customers outside the NLS to make queries on the new system on the basis of the parcel identifier. At the initial stage, the service is available only to customers with a service agreement, since the NLS has not yet transferred to using 'netcash' on the Internet.
The NLS also has a prototype of a WWW service implemented with JAVA, where the user may select parcels of a cadastral unit or carry out ordinary spatial predicates (on, touches, inside). When a query is made, the user is given a list of cadastral units. The user may then print out a cadastral register extract on the desired cadastral unit from the list.
Loading data from the old system
Developing tools needed for managing attribute data stored in the old system presented an enormous challenge and required plenty of development and testing. The problem was to ensure that attribute data stored at various times was adjusted to the data model in the new database, and to develop the necessary tools for cadastral surveys which would be able to manage such attribute data in survey production.
Loading the data from the old systems was begun in mid-January 1998. Both attribute data and data in vector form were stored in transfer files, a process which took one week, after which all attribute data were loaded into the new database. For loading purposes, the NLS leased 40 Pentium II 233 Mhz workstations, which were connected to the server with a direct 10 Mbps LAN connection. After loading the attribute data, the data of the digital cadastral index map was loaded. In the old system, the map data was stored in map files which each contained data for one map sheet. Data on map borders, however, included quite a lot of topological errors despite the careful border comparisons. A number of new tools had to be developed in the new system to detect and correct topological errors in order to use the data in a seamless database. The loading also revealed clear identification errors in attribute and spatial data when the data was loaded into the same database.
The loading process lasted 2.5 months, most of the time being spent in automatic detection of errors in the vector data and correcting them manually. The correction involved 12 employees during the period of the loading process.
REFERENCES
Smallworld 1997: Smallworld GIS Online Product Documentation, GIS Glossary.
Vahala 1994: Vahala M., 1994, Developing Next Generation Real Estate Information System in Finland, UN-Workshop on Data Processing in Cadastre, Theme III Databank, 6 pages.