1In the United Kingdom county coverage by metric aerial photography is now a routine decennial practice with flights occurring in the census years of 1971, 1981, 1991. The newly flown photographs are often used intensively for the first few years and then the next flight is eagerly awaited. Once this new photography becomes available, the old photographs are usually archived and often forgotten. Old photography, however, can be of great potential use for site assessment as they allow identification of the location and the appraisal of change to forms and features that may be no longer visible in the landscape. Besides this purely visual assessment, old aerial photography can be used as a data source to quantify the extent and rate of any changes to the area covered.
2While the benefits of using aerial photography are relatively well understood, the use of historical aerial photography is less developed. Tharp, Conway and Wade (1994) used sequential historical aerial photographs to identify potential waste disposal sites. Eblen (1989) used historical aerial photography to identify an old landslide thought to be associated with ground instability under the pier of a bridge. Henry (1989), of Ove Arup and Partners, drew attention to the uses that consulting engineers find for archived air photographs. Examples he cited included using archived photography to identify old ponds that caused subsidence after land reclamation and building construction. Merin (1990) illustrated that historical aerial photography could complement other data sources, such as ground penetrating radar images, to identify previously unrecognised waste pits.
3If the use of historical aerial photography is, as yet, poorly developed. The work discussed in this paper illustrates the use of this type of photography and shows how it can make an important contribution to environmental studies.
4The papers cited above generally applied qualitative interpretation techniques to historical aerial photography. Quantitative analysis was limited to relatively simple planimetric measurement. In part, this is because all aerial photographs contain distortions due to height variations and any tilt in the camera at the time of exposure. Traditionally, it has been difficult to remove those distortions without access to expensive photogrammetric equipment and skilled operators. However, it is now possible to extract good quality quantitative data from such photographs should the need arise using simple techniques. However, for more complex manipulation of photography, an understanding of the nature of the errors inherent in the photographs is required, together with an understanding of the techniques available for image correction and translation and the alternate models that can represent the data as a continuous surface. This paper outlines some general principles in undertaking this more complex manipulation of historical aerial photography.
5A project intended to identify the nature and extent of former waste disposal sites in West Yorkshire serves to illustrate the potential benefits and problems of using historical aerial photography. Location of the former waste disposal sites was important as they were thought to be the source of groundwater contamination. In this particular case the tipping had ceased in the mid 1950s and the area had been subsequently developed. No evidence of the former sites could be identified on more recent photography.
6Using photography from the Royal Commission on the Historical Monuments of England's archive, flown by the RAF in the 1940s (Figure 1), it was possible to identify the landfill sites and to mark their positions on modern maps. It was discovered that several buildings, including a garage, had been built on the former waste disposal sites that contained the toxic waste.
7The identification of these former waste disposal sites was carried out using a mirror stereoscope to give a three-dimensional view of the area for conventional photo-interpretation techniques. Prior to their use for waste disposal, the sites had been quarries. Individual sites could be identified and, by studying the tone and texture of the bottom of the quarries, it was possible to conclude which sites were already being used for tipping. The original intention was to use an analogue photogrammetric plotting instrument to measure the surface in X, Y and Z coordinates, create digital terrain models of the surface and subsequently to determine the volumes of the holes. Unfortunately, during the period after the original photographs had been taken, the photography had distorted, stretching more in one direction than in the other. In addition, the geometry of the original air survey camera (calibration data) was not known with any precision. This made it impossible to create the accurately scaled photogrammetric model within the instrument so a high precision method was precluded.
8As a result, the approach adopted was to use approximate techniques. The depths of the quarries were measured using a large number of parallax measurements. This technique involves the measurement of the parallax differences between objects measured on the two photographs of the stereo pair. It is a labourious process and does not yield the same level of accuracy as a photogrammetric plotter but was sufficiently precise for the desired volume calculations. The errors in depth measurement would typically be +5%. The planimetric positions of the holes were determined by using a stereoscopic radial line plotter to overlay an image of the aerial photograph on a recent map of the site. While this was not an ideal solution, it was adequate for the purpose and provided data that was not otherwise available.
9Many modern GIS packages, such as MapInfo, contain software which permits an approximate rectification of aerial (or terrestrial) photographs. Such solutions are often adequate for many environmental purposes.
10Where the photographic base material has not been stretched, it is possible to recreate the three-dimensional geometry of the original photography and use modern photogrammetric plotting machines in the measurement and mapping process. To do this, however, it is still necessary to determine the focal length of the camera used and the size of any errors to due imperfections in the lens. Camera calibration of modern survey cameras would provide this information. For historical photography it is not normally possible to calibrate the camera as, in most cases, the camera no longer exists. What is needed, therefore, is a way to obtain the necessary information by taking measurements from the historical photographs themselves. The technique is known as a "self-calibrating bundle adjustment", so called because it involves attempting to reconstruct the bundle of light rays that passed through the lens to create the photographic image at the moment of exposure.
11To carry out a bundle adjustment it is necessary to identify a large number of specific points of known coordinates. The more points the better but a minimum of 9 points should be collected if a reasonable level of accuracy is sought from the final data. If the photograph covers an area which has changed very little since the photography was taken, it is often possible to fix the positions of points identified on the photograph using other sources. Position fixing can be derived from Ordnance Survey maps, conventional land survey techniques, high precision Global Positioning System (GPS), or even fixed from more recent photography. The X and Y coordinates of these points are then measured on the historical photography itself (normally to a few microns) using a photogrammetric plotter. This creates a data set which, when compared with the known coordinates, allows the computation of the image distortion caused by the camera lens and its focal length. Software packages such as CUBA (Short, 1997) are employed to carry out this stage of the process by reconstructing the bundle of light rays and subsequently providing the calibrated focal length of the camera lens and image distortion data. This is a complex process and is likely to be expensive. However, if highly accurate data is needed to fix a benchmark in historical time, perhaps forty to fifty years ago, it may be the only way of collecting the information.
12Once camera calibration data is available it is possible to use the photography to capture reliable X, Y and Z values for the ground surface. This is achieved by viewing the stereoscopic model in the photogrammetric plotter and using a "floating mark" (a small dot of light seen by the operator as moving in three-dimensional space) to measure the positions of points within the stereo model.
13Over the past fifteen years or so, analytical photogrammetric techniques have been the typical method employed to carry out this task. Such methods have replaced the use of comparatively simple mechanical, or so called analogue photogrammetric plotters. Though analogue machines could produce acceptable data from high quality photography, they were considerably less effective than the analytical machines when used with distorted and degraded historical aerial photographs. More recently, new digital techniques have become available which bring even greater flexibility than the analytical methods.
14In analytical photogrammetric techniques, the relationship between the object and image is established through a mathematical solution (Ghosh, 1988). A pair of overlapping photographs is placed in the analytical plotter. By taking a series of measurements in X and Y in the plane of the photographs it is possible to reconstruct the bundles of rays that originally projected onto the photographic negatives. This allows the creation of an object-model within which all measurements in X, Y and Z can be made.
15The advantages of the analytical technique are considerable. The same points on any surface can be revisited automatically on photography of different dates and hence changes through time can be monitored. The relationships between height characteristics of surface points and non-morphological attributes can be observed and assessed. Different image formats can be accommodated including oblique aerial photographs, and the data captured is in digital form and can therefore be easily transferred to a Geographical Information System and surface modelling packages.
16One of the main reasons why this approach has not been more widely applied is that the technique requires expensive, highly specialised equipment and skilled personnel to operate it. The mathematical model used in the analytical solution is, however, fundamental to the greater freedom of choice in type of photography which may be used.
17In digital photogrammetry the image is no longer held and used in "hard copy" form. Instead of photographic prints, scanned copies of photographs or digital data from CCD cameras are used (Wrobel, 1991). These digital data are displayed on a computer screen for both image viewing and measurement. To do this it is necessary to display two images simultaneously in such a way that a three dimensional image of the surface can be created in the mind of the observer. Different approaches have been adopted to this problem, usually involving the use of polarising filters. The method of model creation is essentially the same as for analytical photogrammetry but it does offer certain advantages over analytical techniques. The equipment costs are significantly lower than analytical photogrammetry. The main cost is associated with the need for stereo-viewing. The photography can be corrected to remove distortions due to stretching before carrying out the photogrammetric process. Image quality can be enhanced by using image processing techniques on the scanned images. Digital techniques lend themselves to being semi-automated, leading to cost reductions and orthophotographs may be generated as a by-product of the digital process.
18The application of digital techniques depends upon access to high resolution scanning facilities and the capacity to hold very large data sets in digital form. Though problems of digital data storage are reducing rapidly the problem of archiving scanned photographic images for use in future decades remains.
19One characteristic of the earlier studies discussed above was that changes that had obscured subsurface differences were the direct result of human activity. There are other reasons for such differences to be obscured which are either the result of natural processes or only an indirect consequence of human activity. In these areas aerial photography may provide the only source of reliable historical data. This is particularly important in coastal environments where map records have been shown to be unreliable (Collier et al., 1995). The work carried out in Langstone and Chichester Harbours illustrates the nature of the problem and shows how historical aerial photography can help.
20The Department of Geography has studied Langstone Harbour and Chichester Harbour over a period of more than fifteen years using air photography. The earliest studies (Coulson et al. 1980 ; Budd, 1980 ; 1981 ; 1982 and 1983) were carried out on behalf of Southern Water to determine whether the discharge from Budds Farm sewage treatment plant was having any effect on the vegetation in Langstone Harbour. A more recent study has been carried out for the Royal Society for the Protection of Birds (RSPB) to look at the erosion of island nesting sites in Langstone Harbour. This work included the production of a high resolution digital terrain model (DTM) from specially commissioned large scale (1 :4,000 scale) aerial photography. Aerial photography has also been a primary data source for the mapping of vegetation in both harbours for the National Rivers Authority (NRA) in the light of growing concerns over eutrophication (Collier and Fontana, 1995 and Collier, Fontana and Pye, 1996).
21In common with a number of other inter-tidal areas in Southern Britain, Langstone and Chichester Harbours were colonised during the late Nineteenth Century by Spartina. This vegetation traps suspended sediments and forms beds that continue to grow in height until the top of the bed is near the normal high water level. Spartina beds are considered one of the best natural forms of coastal defence as the vegetation absorbs much of the wave energy. Unfortunately, however, there seems to be a natural cycle of Spartina growth and 'dieback'. When dieback occurs, the Spartina beds begin to erode, leading to the remobilisation of the formerly trapped sediments. In some cases the released sediments will be removed from the area by currents and tidal forces. However, in semi-enclosed systems, like Langstone and Chichester Harbours, the remobilisation leads to a redistribution of the freed sediments within the harbour complexes. Much of this sediment is deposited in existing channels where the scouring action of the tide is insufficient to keep the channel open.
22As part of the studies which were focussed on the vegetation changes and erosion within Langstone and Chichester Harbours, it was noted that a number of the channels which had been mapped as major features in earlier surveys were no longer visible. This was noted in the report (Collier and Fontana, 1995) but the significance of the observation for other disciplines was not fully appreciated at the time. In early 1996 the Department received a request for a copy of the DTM of Langstone Harbour from consultant engineers working on a scheme to lay a pipeline between the sewage treatment plants at Budds Farm and Eastney (Map 1). As part of the feasibility study the consultant engineers were using a range of geophysical techniques to locate subsurface variations in the sediments of the harbour. The DTM and a digital base map were required for use with the accurately positioned geophysical data.
23Among the features of interest were any former channels that might be filled with unconsolidated sediments. Conventional geophysical techniques, including sub-bottom profiling would reveal the presence of such features and, when linked to modern GPS, would permit the accurate mapping of the features. However, the costs of carrying out such work are very high and add considerably to the overall cost of the scheme. For a relatively small additional outlay the historical aerial photographs give a useful insight into the changes that have taken place within the Harbour over the past 60 years and give valuable information on the location of former channels.
24Figure 2 shows the channels in the area of the Milton landfill site in Langstone Harbour which have been derived from photography taken in 1996 and 1961. The changes in this part of the Harbour are of fairly recent origin and are probably the result of local changes in the circulation pattern caused by the landfill operations on Portsea Island. Elsewhere in the Harbour there is evidence that the changes have been taking place over a longer timescale.
25Figure 3 shows photography taken of the landfill area 1961. Figure 4 shows the same area in 1996. It can be seen that some of the smaller channels which were present in 1961 have disappeared. It seems that the channels in 1961 were incised into resistant sediments and that these have now been partially or totally infilled by soft sediments following the changes in the circulation pattern.
26As noted above, the use of historical photography is not, however, limited to photographic interpretation. Using modern photogrammetric techniques it was possible to obtain accurate planimetric data on the channels (see Figure 2) and volumetric data from historical photography.
27The accuracy of the maps and DTMs derived from the historical photography was, in theory, only limited by the scale of the original photography. That is, the maps and DTMs derived from historical photography are potentially as accurate as maps and DTMs derived from new photography of the same scale.
28Terrain modelling software, in addition to providing a good visual representation of the surfaces (Figures 5 and 6), allows volumetric and planar area calculations to be carried out between models of the same area that were captured for different periods of time. This permits further insight into the evolution of a surface as, for example, it undergoes erosion or deposition events. It is possible to create contour maps and surfaces of change between two data sets by subtracting one surface from another (Figure 7). Additionally, the software allows the overlay of different layers, thus permitting direct visual comparison of a surface as it evolves through time. It is possible to either stack contour maps, surface plots, surface change plots or a combination of all three in order to compare different surfaces.
29In addition to using the DTMs to study surface change, they can also be used with other data sets. In the case of the Milton site it was possible to overlay the DTM with a map of methane levels and to see whether there was a relationship between depth of fill and methane levels. In the event, no clear relationship was established.
30The work on Langstone Harbour illustrates a useful extension in the application of a well known and understood technique. By using historical photography it is possible to extend studies of environmental change back in time. New software tools for photogrammetric analysis and GIS and making the process of using historical photography both easy and economical, thus opening up this valuable data source to a much wider user community.
Map 1 : Locations in Langstone Harbour
Fig.1 : Landfill site before the start of tipping
Fig.2 : Channels in the area of the Milton landfill site, 1961 and 1996
Fig.3 : Milton landfill site, 1961
Fig.4 : Milton landfill site, 1996
Fig.5 : DTM of the Milton landfill site, 1961
Fig.6 : DTM of the Milton landfill site, 1996
Fig 7 : DTM of change on the Milton landfill site 1961-1996
© cybergeo