NZSM Online

Get TurboNote+ desktop sticky notes

Interclue makes your browsing smarter, faster, more informative

SciTech Daily Review

Webcentre Ltd: Web solutions, Smart software, Quality graphics

Feature

Data Dangers

Computerised databases are changing the way in which people manipulate information, but pose real dangers if their limitations are not considered.

By Brian Sheppard

A geographical information system, better known as a GIS, is widely regarded as an essential tool for land management agencies. In the past, GISs have taken the form of maps and plans. They show such things as the urban or rural landscape, locations of underground services, street architecture and natural resource attributes.

With modern technology, these maps can be digitised, superimposed at any scale and reproduced by computer. This saves the need for redraughting. It also provides composite layers of information "at the touch of a button".

Is this a planner's dream or the prelude to a nightmare? The hazards involved in a computer-based GIS are no different from those of its paper-based predecessor. They are, however, magnified by the immediacy of access to, and possible spurious accuracy of, the information.

Room For Error

Take a computer-based GIS depicting a complex urban area. It can display an entire townscape and allow a user to zoom in on a central city area. From there, they could look at a particular road intersection, viewing the locations of street furniture and underground services. Because the program demands little in the way of operating skills, anyone can identify, for example, the location of buried services.

A user may expect all of the buried services to be accurately recorded, probably by direct measurement on the ground with the actual measurements recorded on a survey plan. If pipes and cables run in straight lines between inspection holes, their paths should also be firmly established. It may be, however, that the location of an inspection hole was recorded with reference to a kerb line which was subsequently moved. Alterations such as this should be easily identified and incorporated in a computer-based GIS as its data layers are digitised and assembled. In those contexts limitations are minimal, and easily comprehended.

Scaling Problems

Bigger problems occur when information is digitised from source plans of different scales. If there are two source plans of the same area, one at 1:10,000 (1 mm representing 10 m) and the other at 1:5,000 (1 mm representing 5 m), the biggest scale at which a composite map can meaningfully be reproduced is 1:10,000.

The user may wish to create a composite map using the larger 1:5,000 scale. That would allow them to compare it with existing maps that use the same base. There is unlikely to be any warning in a GIS program to indicate the dangers that result. Magnified lines would not even be shown as thick and fuzzy, as they might be in a photocopier enlargement. The user would remain blissfully unaware of the spurious accuracy of the feature boundaries that had been generated.

Environmental information is likely to have been mapped at an even smaller scale, perhaps 1:50,000 (1 mm representing 50 m). This might be suitable for overlay with the townscape at the same scale but could be totally misleading when viewed at either of the larger scales.

A consequence of this mismatch might be the misdefinition of a height contour used for flood control, or the mismarking of an earthquake fault line. Problems occur when these are incorporated in a plan at a scale larger than that at which they were originally defined. The planner may not be aware of the data limitations and recommend approval for development that could be threatened by the hazard. The naivete of the assumptions might only come to light following a damaging flood or earthquake.

The database that I operate records the locations of nearly 45,000 archaeological sites. These were mostly recorded using NZMS1 maps (1: 63,360 scale) or the later NZMS 260 (1:50,000) equivalents. The precision with which a site can be located from landscape features on these maps is limited to the identification of a 100-yard or 100-metre grid square. Further levels of uncertainty are added when data is converted from one map system to another, particularly where cartographic errors and different but essential generalisations of landscape features are concerned.

Interpretation Hazards

In the archaeological database, no attempt is made to show the size of an individual site, only the recorded grid square. When site locations are replotted on a suitable map base, allowance has to be made for these limitations.

Even then, the data used is from individual site records. A plot of the distribution of those sites would not in itself distinguish areas with no sites from those where site recording is incomplete. Only people who understand these limitations in the archaeological data can interpret their implications. This applies to any geographical or survey data used in a GIS.

A computer-based GIS can offer great research and management efficiencies when used by experts who understand its limitations. Those limitations are rarely fully appreciated, and the computer program itself can issue only general warnings.

The apparent clarity of information offered by a GIS will inevitably lead to its use by people who are unaware of the system's limitations. This will result in misinterpretations which could endanger the very resources that the operating agency exists to manage. Used with the appropriate interpretation by people with expert knowledge, a GIS is a powerful productivity tool, but used naively its potential for disaster is enormous.

Brian Sheppard is the SeniorArchaeologist for the Department of Conservation.