David R Miller, Neil A Brooker and Alistair N R Law
This paper describes the analysis of topographic data for the assessment of space and human perspectives of the landscape. In particular, a technique is presented for calculating a census of visibility for Scotland based upon a digital elevation model. The method uses functions provided within the GRID module. These techniques provide both a basis for assessing historical descriptions of geographic features and as a basis for assessing reactions to changes in landscape.
A calculation of intervisibility was undertaken for each cell and the total area visible was attributed to the cell. Each cell in the elevation model is counted only once, although the calculation could be weighted according to the inverse of the distance from the cell.
Land cover effects may be included in this analysis by adjusting the height of each location for example, approximate heights of build up areas or forests. The equation may be applied to individual locations or to all locations in the area. Vertical angles are calculated making corrections for earth curvature and atmospheric refraction. The analysis considers a complete 360 degree rotation around each location for a radius of up to 50 kilometres and is carried out on the (resampled) 25m resolution raster DEM of source scale 1:50 000 from the Ordnance Survey.
The calculation of census has been carried out for a region of Scotland which was 900 000 cells. On a Sun Sparc 10, this calculation took in excess of 30 days of CPU time. To run the calculation for Scotland, which is 78 000 km2 or 7 800 000 cells would be impractical (7). Therefore, a surrogate approach was adopted.
The surrogate approach was based upon a calculation of the number of points at 5km x 5km intervals which would be visible from every 100m x 100m cell using a 1:250 000 scle DEM. This would provide a scoring of relative visibility across the country. The more enclosed an area the fewer intersects would be visible and, conversely, the more open the land the greater the number of intersects that would be visible.
A file of the co-ordinates of the intersect points was derived for the land area of Scotland and those intersects within 5 kilometres of the coast, a total of 3 500 points. An intervisibility calculation was then undertaken for each 100m x 100m cell, of the intersects and the score attributed for each cell. Figure 1 shows the grid of point for which the visibility calculation was undertaken, the coastline and the buffer within which the points were selected.
Figure 1. Intervisibility observations points for the west coast of Scotland.
The country was segmented into fifteen for processing purposes. This was for two principal reasons:
1. If there was an interruption in processing, due to power failure, exhausted swap space or lack of disk storage space, the results of the entire process were irrevocable. A considerable amount of time was lost due to such interruptions, totaling at least three months work time.
2. There was insufficient memory and disk space to undertake the entire calculation in one operation.
The consequences of segmenting the country for processing was, in itself, a contribution to the length of time taken for processing the data. Each tile required to overlap with its neighbouring tiles by in excess of the 20 km radius of view. Therefore, for every overlap between segments there was up to four intervisibility calculations per point.
The visibility calculation was undertaken within GRID, of ArcInfo Version 7. The same calculation, undertaken in Version 6 was inhibited by a software bug which prevents reliable completion of the processing of certain areas.
The GRID command used was:
Output_File = Visibility ( Input_grid, Input_points, point, frequency )
The time taken for an area of, approximately, 40 x 40 points, was between five and twelve days of CPU time, depending upon additional, external use of the processor.
Test are being carried out to check whether or not the results from ArcInfo Versions 6 and 7 are consistent. there is some evidence to suggest that in correcting the software error in Version 6 another error may have been introduced in Version 7.
Figure 2 shows the completed visibility census for Scotland. The areas with brighter tones are those with the highest visibility. The black tones (which are England and Northern Ireland) are areas for which no visibility calculations were undertaken but are land that may have been visible from Scotland. The central valley, open drainage basins and some coastal areas are those with the hisger visibility levels.
Figure 2. Left: Digital Elevation Model For Scotland (brighter tone relates to higher altitude). Right: Visibility census for Scotland (brighter tone relates to higher visibility).
Figure 3. Visibility census for Central Scotland (brighter tone relates to higher visibility).
Figure 3 shows the resultant visibility census for Central Scotland. In the area of Central Scotland, the central valley and the larger river basins display a higher level of visibility. The valleys which are `wide' and `open' have views to a large area of hill tops and sides, which provide barriers to visibility. These are sufficiently distant that areas of high visibility represent `openness' in the landscape.
Enclosed land is that with valley sides that are closer to the observer, narrower valley bottoms and thus a lower area of visible land. This description applies to most of the uplands of Scotland, north and south of the Central Valley.
There are some areas of the uplands which are relatively open. Coastal areas are one such landscape type. The second type is that of large expanses of peatlands, such as Rannoch Moor. In the latter example, the levels of visibility of the valley sides are greater than those of the plain. Therefore, the nature of the description of an area as `open' is related to scale. The level of low lying relief is sufficient to reduce total visibility, but the land is `enclosed' by hillsides of high visibility which are visible from much of the lower land.
Further work is required to assess the nature of the openness and enclosure of land for different types of geomorphological features and geographic areas. In particular, the integration of distance between the location of the observer and the barrier to view with that of the total view (10).
In an environment where multiple users have equal rights of access to disk space and CPU time and when circumstances prohibit a dedicated facility, an arrangement had to be negotiated whereby processing of segments could be run on different processors but that a longer total time for the project is accepted by each interested party. In fact, the principal applies equally to single user facilities.
If the priority of processing this task had substantially increased, a different set of agreements would have been required, specifically, with respect to access to CPU time with competing processes.
It is probable that circumstances such as those described in this paper are very common. The task, its length and the total time (CPU, process and personnel time) involved may vary considerably but the need for careful resource management and negotiation of agreed solutions has clearly been required in many organizations.
Alternative solutions are often an option. Such solutions may include an entirely different approach to the algorithm used (9), its implementation and a rigorous questioning of whether the task, as proposed, is actually necessary. In most GIS related processing there are trade-offs between scale, resolution, accuracy, data volume and processing time according to the task objective. Equally, continued increases in computing resources moves forward the limits for each of the parameters mentioned but probably also increases the expectations of the user. Such user expectations are likely to vary by discipline (and individual).
The experience of this project, wher considerable personnel time was lost due to the requirements of closely monitoring unreliable software (the `Visibility' command within Grid, Version 6) points to a requirement for a non-interactive processing facility which can manage processes and data such that system interruptions have a minimal effect on the time to obtain the end result.
A further outcome of the experience of the processing of a visibility census for Scotland, complementing other research experiences, is an acknowledgement that the need for adequately describing the production and value of a product includes an appraisal of the resources available (REF). Specifically, such a description requires appraisal with the manager of the GIS facility. Finally, the negotiation of the processing priorities of the product involves placing it in context with other, competing tasks.
One example of current work is in studying the origins of geographical place names with respect to `openness.' In the Scot's language, and thus used a place names and on maps, the geographical feature type of a valley is referred to as either a `Glen' or a `Strath.' The definition of a Glen is of narrow, lower land lying between adjacent hillsides whereas a Strath is a wide, open valley often with a river running through.
The hypothesis which is being is that the naming of Glens and Straths was that the broadest, identifiable valleys were named with the prefix Strath and that all other valleys were named Glens. he initial referencing of a valley broad enough to be called a Strath by local peoples would probably have been independent of first hand experience of other such features. Standardization of terms may have occurred but this was not the remit of the survey teams whose responsibility was to collect evidence of all names that were used for any particular feature and the selection of the name for use in mapping and a national gazateer of place names. To test the hypothesis an analysis of the levels of visibility for each Strath is being carried out and compared to a range of sizes of Glens. The understanding of the extent of these features is an issue in itself.
The linkage of the analysis of views with the visibility census provides for analysis of elements of landscape with respect to impacts of land use on the observer. In particular, the assessment provides an audit or inventory of national intervisibility against which land cover, land use and other activities and resources may be compared. This type of wide area assessment highlights areas of greatest importance across the country and is of value for strategic planning and management. The dataset presented here may be used in visual impact assessments which are sensitive to the geographic context within which they occur and provide a role for GIS in data synthesis, analysis and evaluation as part of visual impact assessment.
Progressively, the means to link geometric and cognitive aspects of landscape are being built coupled with spatial analysis. The geometric measures can be used in studies of visual and metaphoric space as described by individuals. Thus space representation by imagery and language can be one coupling of the individual's perception of landscape with Euclidean and non-Euclidean measures of landscape. Future research will develop this theme, linking geographic information to spatial reasoning (1 and 11).
2. Kliskey, A D, Hoogsteden, C C and Morgan, R K (1994), The application of spatial-perceptual wilderness mapping to protected areas Management in New Zealand, Journal of Environmental Planning and Management, 37(4), pp 431 - 445.
3. McLaren, R. A. and Kennie, T. J. M. (1989) Visualisation of digital terrain models: techniques and applications. In: J. Raper (ed.), Three dimensional applications in geographic information systems, pp. 79-98. Taylor and Francis: London.
4. Kliskey, A D and Kearsley, G W (1993), Mapping multiple perceptions of wilderness in southern New Zealand, Applied Geography, 13, pp 203 - 223.
5. Coeterier, J F (1994), Cues for the perception of the size of space in landscapes, Journal of Environmental Management, 42(4), pp. 333 - 348.
6. Land Use Consultants (1990) Landscape Assessment: principles and practice. A report to the Countryside Commission for Scotland.
7. Miller, D R, Morrice, J G, Whitworth, P L and Aspinall, R J (1994), The use of GIS for the analysis of scenery in the Cairngorm Mountains, GIS in Mountainous Regions, (Eds. I Heywood and M Price), Taylor and Francis.
8. Lee, J (1991), Visibility dominance and topographic features on digital elevation models, In: Proceedings of the 5th International Symposium on Spatial Data Handling, Charleston, SC, USA. pp 622 - 631.
9. Franklin, W R and Ray, C K (1994), Higher isn't necessarily better: visibility algorithms and experiments, In: Proceedings of the 6th International Symposium on Spatial Data Handling, Edinburgh, UK. pp 751 - 770.
10. Fisher, P F (1994) Stretching the viewshed. In: Proceedings of the 6th International Symposium on Spatial Data Handling. Edinburgh, UK. pp. 725 - 73
11. Hetherington, A, Daniel, T C and Brown, T C (1993), Is motion more important than it sounds? The medium of presentation in environmental perception research, Journal of Environmental Psychology, 13(4). pp 283 - 291.