The long way to an integrated road network

In the Netherlands the Traffic research centre (TRC), that is a part of the ministry of Public works and transport, is working on a centreline road network based on a 1:10.000 scale. This network is supposed to meet the requirements of all governmental and a lot of non-governmental institutions for an relating all kind of data in a GIS to a road network making analysis and presenting data with a GIS.


THE NEED FOR A CENTRELINE ROAD NETWORK

During the last decade the amount of organisations working with a GIS has increased extremely, also in the Netherlands. Starting to work with a GIS usually starts with some very enthusiastic people, a brainstorm and starting some pilot project, in which all kind of data is gathered. After a successful or promising pilot a lot of organisations decided to 'go for it', bought GIS software and hardware to run it on and started building their own applications. The wheel was invented many times. The data for their GIS they had usually themselves, but the geometry, the basics for the map not, so either they bought it or a second project started to digitise from maps. Application ready, everybody happy or not ?? One big problem was often forgotten is this early GIS adventures: how to keep the geographic data up to date? Initially digitising data is already expensive, keeping them up to date is usually more expensive, specially when the amount of changes is rather high or the required actuality is high.

Situation in the Netherlands

For the coming part of this paper, attention will be focused on the situation concerning a digital map of the roads in the Netherlands. There are several existing digital maps, based on various scales from 1: 250.000 to 1:1000, used by a lot of organisations and not being able to combine their data without a lot of effort. For instance the Transport research centre (TRC) itself has already three digital maps covering (a part of) the road network of the Netherlands, with different administrative data, different data formats, different GIS software, based on slightly different scale, with different actuality en geometric correctness. Outside the TRC, but still within the ministry of public works their are several more digital road networks, not to speak of other ministries or non governmental institutions, industry etceteras.

None of the existing digital road networks is perfect and the costs of updating them are rather high. But after having pioneered with GIS, the main problem seemed to be the amount of time it took to combine date based on different road networks. Often the best solution turned out to be to digitise the events again, but now based on the road network used for analysing or presentation of the combined data.

After a discussion of about a year the decision was reached, that it would be the best solution to make one national road network, based on centrelines, scale 1:10.000, with high update frequency, geometrically correct, to be distributed in a standard data format, accessible for all kind of GIS-systems, and not the easiest, to be distributed as a complete network or as an update, only containing the things changed since the last delivery. This network is called NWB.

How to expand this network, once it exists

The basic NWB had been defined to contain e.g. street names, road numbers, hectometre signs, road ownership etc. It is co-operatively built and maintained by TRC and ministry of defence. It is available to/can be bought and used by second degree partners. They may add data suitable for their specific use, data which is not yet included in the NWB e.g. information on one way-traffic, traffic lights etceteras. Once data is added the expanded NWB may be sold again to third degree partners. Therefore it is essential that all changes in the NWB are kept in file and can be distributed as an update delivery.

THE APPLICATIONS OF TRC USING A CENTRELINE ROAD NETWORK

Before explaining how to realise the network, a short overview of TRC's own applications needing a centreline road network will be given. TRC Has two main tasks: - research on how traffic and transport can be more efficient, with less accidents, less time consuming. - gathering data nation-wide which are needed for e.g. road or traffic statistics and which should be gathered in a way they can be compared all over the country. To give an idea of data available: congestion, road usage, pavement conditions, bus stops, number of lanes on highways, average vehicle speed, road accidents, transportation from all kinds of goods from region/city to other regions/cities, accidents and near accidents with ships. At the moment there is a variety of applications used for data registration. These applications are running on different sizes of computers, they are either GIS oriented or purely administrative, and hardly able to exchange data with each other. It is obvious, that TRC does not only need a road network, but also a railroad network and a waterway network, so both will be added to the NWB. Special should then be given attention to the spots where they contact each other as railway crossings, harbours, bridges etceteras.

HOW TO REACH THE GOAL

Starting

The ideas on how to expand and distribute the NWB arose relatively easy, but now the main problem, how to get a 'first edition'. There where two basic networks possible, of which was chosen the TRC network used for traffic accident registration. The need for the NWB was indeed urgent, once the decision was taken to make it, a lot of institutions wanted it as soon as possible. So TRC took the challenging task to realise the NWB.

The TRC network however did not meet all requirements put on the NWB. In fact two things have to be realised, a NWB available for everybody needing it and all the applications of TRC running on its own NWB in stead of running on various other road networks.

Most important was the decision that NWB should represent the real life situation geometrically correct. Bearing this in mind it was possible to choose for a two-track strategy to reach the ultimate goal.

The first track

The first track is updating/correcting the existing centreline road network by using digital data available already from several sources. The application supporting this track has been developed in 1995/1996 and is fully operational. It is running on TRC's 'old' infrastructure, which is not with Esri-products. It will take nearly two years and is a semi-automatic process of conflation of two digital road networks. Sometimes the TRC network is the best, in other situations the network of ministry of defence is the best. At the moment about 15 people are working on it and this project is planned to be finished early in 1998. Those parts of The Netherlands where both networks are combined and corrected can already be distributed in an ASCCII file format.

The second track

The second track started at the same time (spring 1995) with datamodelling, a pilot implementation in SDE2 and then realisation of the final database. This is easily said but.... - Most of the TRC applications running on the existing databases must be completely rebuilt also. - While rebuilding and creating a new database the existing ones are still in use, they should be maintained also - Once one of the new applications is ready and can be used, the database should be ready also

Overview of projects

After the pilot implementation the order in which all other developments are supposed to take place had been determined and different projects started . - In the first project converting of basic elements of the centreline road network was done (e.g. junctions, road elements, names, road ownership, road numbers, hectometre signs) This project includes software to maintain the SDE database when there will be changes in the original database. - A second project concerned the complete rebuilding of the application used for traffic accident registration. It is a daily used application, about 50 people working with it. - a third project will rebuilt all traffic accident reporting and investigating possibilities in the SDE/Oracle environment. - In another project adding of boundary files will be implemented, province boundaries, municipal boundaries, police district boundaries etceteras. - More details on road usage for especially highways is needed, number of lanes, traffic statistics, average speed, and amount and cause of congestion's. It has to be converted from another database, thus again a new project. - The applications adding and using the above data have to be rebuilt, most likely there will be several projects, one for each application. - As mentioned earlier railroad network en waterways have to be added, in combination with applications using/needing these, so more projects. -Finally a maintenance application and the possibility to deliver not only the complete network, but also updates has to be developed -When the above mentioned is ready, an integrated centreline road network is available, and then extra data not yet related to a network will be added. The first needs are already known, information for radio-broadcasted traffic information to help drivers choose congestion-less alternatives.

WHAT WE HAVE LEARNED SO FAR

1 Datamodelling is extremely important. When this is not done proper, everything else fails. You should use a mix of GIS experts and ordinary datamodelling people to obtain a good result. Also during software development good teamwork is very important, people should continuously discus their ideas with colleges. As a manager you will sometimes have to enforce this. 2 Learn from prototypes, but discard all of it. Do not try to reuse prototype software, it is written quickly, edited a lot of times to if try brilliant ideas are really brilliant. The final result is not easy to maintain afterwards. 3 When it is necessary to convert data from one database to another you need a lot of time for data integrity. Be aware you might encounter situations that in theory can not exist. They are there. 4 If you have the opportunity to do the data conversion twice, then do it. Use the first results for finding all theoretical impossible wrong data, then discard your bran new database, correct the old one and create e really high quality database. 5 When possible, restructure and correct your data in the old database, before converting them into the new database. Especially when you are very familiar with your old system and not yet with your new one. 6 Try to plan different projects in a way that failing or running out of time of one project does not implicate a delay of all others 7 'Just deliver updates of the NWB and keep file of all changes made' was easy said a few years ago. Now it needs a maintenance application very special for TRC, so we have to develop it completely. 8 testing of an application, especially one that updates data , takes a lot of time, count on at least 20 % of design and development time and do not cut down time or expenses for it. 9 If possible use Oracle designer/developer facilities for data integrity in your SDE database in stead of writing your own check procedures in C++ or avenue.


Els Rijnierse
manager software development
Transport Research Centre, Ministry of public works
Postbus 2510
6401 DA Heerlen
The Netherlands
Telephone: + 31 45 5605 333
Fax: + 31 45 5605 224
Email: e.m.rijnierse@avv.rws.minvenw.nl