Geographical Information Systems in Practice - Research Case Studies

To illustrate the importance of GIS as an enabling technology various case studies are demonstrated. These illustrate the breadth of opportunity which exists once we can handle the data required to describe the complexities of the world around us. GIS gives us the ability to collect and manage large volumes of complex spatially- referenced data, which can then be used to research some of the problems which currently face the world. Also included is a review of some of the theorectical GIS developments undertaken in Edinburgh, which are helping GIS develop as a technology.

Edinburgh

A larger version of this satellite image of Edinburgh is available by clicking on the picture.

The Department of Geography has a large database relating to the Edinburgh area. It contains, for example, satellite imagery of Edinburgh and its environs, a street network for the city and detailed locations for individual buildings in the city centre.

Satellite Images of the Edinburgh area are available which illustrate the increasing detail which can be detected from space. Two images come from the American Landsat family of satellites; MSS images have a resolution of 80m, and Thematic Mapper (TM) Images down to 30m. The French SPOT satellite has a resolution of 10m. New satellites from France, USA and the CIS have planned resolutions of up to 2m. This level of detail has great potential, but as the resolution increases so does the volume of data, and the time required for processing.

One of many uses for the street network involves the modelling of service distribution networks within the capital. This model can be used as a decision-making tool which could save time and money and provide a better service to customers. For example, a system has been developed to look at the implications of the move by the 'Evening News' printing and distribution centre from North Bridge in central Edinburgh to a new facility on Newhaven Road in the north- east of the city. Timing is all important in the delivery of newspapers to a large number of newsagents and other retail outlets. The aim was to identify improvements in the directing of existing delivery routes and to develop new, and possibly better, alternative routes.

GIS - A Foundation for Rural Land Use Strategies

This system has been used to provide much of the NCCS input to the draft Islay Local Plan/Land Use Strategy, and has enabled the examination of the use of terrain modelling and satellite imagery. Most recently, a user-friendly interface to the database has been constructed. This is designed to function as a teaching tool to introduce the capabilities of GIS and as a system that NCCS staff with minimal GIS or computing training can use to interrogate the data. This interface lies on top of the ARC/INFO software, yet is a large and complex computer program in its own right, being around 8000 lines of code.

Rural Land Use Information System (RLUIS)

However, Fife Region Planning Department, one of the original members of the working party, recently wished to see the data revived, updated and converted for use on their recently acquired Smallworld system. In conjunction with the Department of Geography, selected data sets from the RLUIS pilot project are currently being upgraded. An initial exercise has been to identify the available data, select those of most use and convert these into ARC/INFO format to assess suitability and quality.

Agricultural Census Data for Fife Region

The Geography Department have data for Fife Region for three census years - 1976, 1981 and 1988. This is a fairly complex data set held in a relational database. There are almost 150 census items, some of which have changed between census years. Individual census returns are aggregated and data released as more than 50 parishes.

The analysis system has been designed to be used by someone who knows little about the data. Help and additional information is provided throughout. The user selects the census year and the census item required, and the relevant data are extracted from the database. The options then produce a classification or distribution of the data, if required as a percentage of another item. The data are displayed on a parish map of Fife, either shaded or containing variable sized spots.

Additional options are a group distribution display, where a number of items are shown as a percentage of another to analyse for example landuse, and a parish boundary display, to illustrate changes to boundaries between census years.

Deforestation & Forest Degradation Monitoring

Monitoring the earth from space has proved to be a very useful method in quantifying changes in the rainforests over a period of time. In many cases, the use of satellite imagery is the only practical method of assessing the changes over large areas of inaccessible forest. The amount of data that any satellite provides is substantial. This data set has to be stored on and processed by computers, which are often dedicated to that purpose. For large areas, the amount of data may run into hundreds of megabytes. It is critical that the computer processing and data storage facilities can cope with this amount of data. Parallel processing technology will substantially improve the speed with which these data sets can be processed, and will provide the opportunity to continue this type of work at a regional scale.

Modelling the effects of Global Warming

A computer model of the Greenland Ice Sheet has been developed which simulates how this major Northern Hemisphere ice mass responds to changes in global climate or sea-level. The model is being used to look at the present day sensitivity of the Greenland Ice Sheet and also to investigate how it has behaved since its maximum extent in the last Ice Age.

The model is dynamic over time. From a given starting point in time it recalculates the surface of the ice sheet via two sets of calculations; those that work out the mass balance, that is the net gain or loss of snow at the surface, and those that derive the internal mechanics of the ice from its thickness and surface slope and then allow ice to 'flow' from one area to another. So, over time, the modelled ice is allowed to expand or contract, and thus change its shape and characteristics.

The modelling procedure is computationally intensive. To run this simplistic model of the ice sheet over a suitable period to give meaningful results takes a number of hours. With parallel computation architectures, giving much greater brute processing power, the spatial and temporal resolution of the model can be improved. More realistic modelling of these complex natural processes would then be possible.

The progression of a model run can be appraised using a graphical display program running on a Unix Workstation. The graphics not only show how the model ice sheet has developed at points in time but also recreate the ice dynamics and climate associated with the ice sheet at that time.

This work has recently been extended to the Southern Hemisphere, in particular Patagonia. The ice sheet models have been ported to Digital Alpha UNIX workstations, and also to the Universities' Cray T3E parallel processor (which also makes use of DEC Alpha processors). This example shows the development of the ice-sheet over time. The orange colours represent the height of the land (darker colours indicate higher land) and the blue colours the thickness of the ice sheet (lighter blue indicates thicker ice).

Desert Locust Forecasting

The eradication of locusts remains the single most important factor in the prevention of famine in major areas of the world and the forecasting of plagues is critical in providing information which can be used to support this effort.

In conjunction with the Desert Locust Forecasting Group from the United Nations Food and Agriculture Organisation (FAO) in Rome and Natural Resources Institute (NRI) in Chatham, the Department of Geography is developing a GIS-based system to automate many aspects of the forecasting process. This will include data collation and display, forecasting by analysis of past events which are analagous to the current situation and the production of a monthly bulletin which is transmitted worldwide.

The prototype system, using a Sun workstation, the ARC/INFO GIS and the Ingres database management system, was installed in Rome in early 1993 for testing and evaluation. The results of these trials and enhancements suggested by the forecasters will be incorporated into the production system to be installed by the end of 1993.

In order to provide data for the analysis of past events, whether for forecasting or for the initialisation and validation of models, a historical database of locust sightings was incorporated into the system. This database, which contains a record of sightings from 1930 onwards, was held by NRI on paper maps. It had to be coded, and computerised together with historical data for precipitation, temperature and wind. These data sets are expected to form the basis for many research projects into plagues, their beginnings and the efficacy of methods of pest control.

Environmental Epidemiology

This animation shows the changing pattern of respiratory illness, for part of Scotland, using SMR1 (Scottish Morbidity Records) data as a proxy. SMR1 data relates to all cases of respiratory illness where the patient required hospital treatment. This animation was created as part of an MSc dissertation which explored ways of representing the relationship between opencast coalmines and respiratory morbidity within a spatio-temporal context. The animation reflects one particular area of Scotland, between 1986 and 1995, and is designed to show the potential of temporal visualisation within environmental epidemiology. The Density Ratings represent arbitrary break points, with 9 being areas where the density of clusters was greatest. The prevailing wind for this particular area of Scotland is South-Westerly.

Great care must be taken in using these results, as any relationship could easily be an artefact of the methodology and its limitations. This animation is for demonstration purposes only, and no responsibility is accepted nor relationship implied.

The visualisation was created using Arcview3 (Spatial Analyst), Paint Shop Pro and Alchemy Mindworks GIF Construction Set. It is best viewed using Netscape, which properly handles the inter-frame delays.

GIS Performance and Parallel Processing

Research has involved the detailed analysis of bottlenecks in GIS computing environments and the design of strategies to avoid these. Because of the data-intensive nature of GIS, it has been important to improve data-handling mechanisms before attempting to increase the overall throughput of the system by using advanced parallel processing techniques.

Algorithms for digital map overlay and raster-vector inter-conversion have been developed to utilise the parallel processing capabilities of various major computing resources available within the University (eg. a Cray T3E, and Symmetric Multi-Processing Machines). Work has been sponsored by the DTI, SERC and the EPSRC, with a range of industrial collaborators, including the leading U.K. GIS vendors, notable users (such as British Gas) and data suppliers (such as the Ordnance Survey).

The research carried out in GIS-PAL will allow us to make better use of spatial information (for example, through real-time modelling techniques) and process the ever-increasing volumes of data which are becoming available to the scientific community.

GIS Developments in the Regional Research Laboratory for Scotland

The RRL has acted as a catalyst for several of the research initiatives some of which have been developed elsewhere in this document. Applications-based research is grouped under two main headings; those of Land Use/Conservation and Urban Information Systems. Projects include the examination of educational and medical service provision, census analysis and a major project comparing administrative boundaries used by different agencies. Examples are drawn from Edinburgh District and Lothian Region.

Theoretical research in the computational aspects of GIS algorithms has supported the efforts of GIS-PAL in parallel processing.

Contributors:

B.M. Gittings      N.R.J. Hulton       G.B. Park
C.J. Place         T.W. Rideout        H. Anderson
D.C. Finnegan      A. Douglas