Case study 15: Forest of Dean
- 1st January, 2008 |
- 12:00 pm #
The Forest of Dean had been subject to standard aerial survey techniques as part of the National Mapping Programme (NMP) for Gloucestershire. Using historic aerial photographs taken over a number of years it had been possible to record some features that were visible during periodic phases of felling, but given the nature of the Forest, with a high proportion of land covered by dense woodland, there were large areas where very few archaeological features were recorded. The Aerial Survey team at English Heritage have been examining lidar data with a view to assessing its suitability since 2000. It was therefore seen as an excellent site on which to test the capabilities of lidar.
Introduction
The Aerial Survey team at English Heritage have been examining lidar data with a view to assessing its suitability since 2000. A fresh aspect of lidar and new potential was recognised in 2003 when the possibility of using last pulse data was pointed out in a presentation by Simmons Aerofilms. It was realised that the ability to penetrate tree canopies could be extremely useful in revealing features in areas where traditional aerial survey was unsuitable. The Forest of Dean had been subject to standard aerial survey techniques as part of the National Mapping Programme (NMP) for Gloucestershire. Using historic aerial photographs taken over a number of years it had been possible to record some features that were visible during periodic phases of felling, but given the nature of the Forest, with a high proportion of land covered by dense woodland, there were large areas where very few archaeological features were recorded. The Aerial Survey team became involved in a project with the Cambridge University Unit for Landscape Modelling (ULM), Forest Research at the Forestry Commission and Gloucestershire County Council Archaeology Service to look at the Iron Age hillfort at Welshbury. The site had previously been recorded on the ground by the Royal Commission on the Historical Monuments of England (RCHME), but when the area was covered by the NMP project very little detail was recoverable due to the density of the vegetation. It was therefore seen as an excellent site on which to test the capabilities of lidar.
Instruments and software
An airborne lidar survey of the site was carried out in February 2004 using the ULM’s Optech ALTM 3033 system. The details of their project are recorded in Devereux et al (2005) from which the following technical specifications are also taken. Ground GPS support was provided by a dual frequency, Novatel receiver located at an Ordnance Survey passive recording station. The maximum distance from the base station to the most extreme point on the survey site was 28km whilst the shortest distance was 1.2km. Two separate surveys of the site were conducted to generate approximate point densities of 4 per square metre and 1 per square metre. The size of the laser footprint was set to a nominal 0.8m for the 4 points per metre survey and 1.25m for the 1 point per metre survey. By flying the surveys during winter, the deciduous canopy was devoid of leaf cover and the understorey vegetation was at a minimum, thus ensuring maximum laser penetration to the ground surface. The survey point cloud data were converted to a 0.25m and 1m grid (for the high and low resolutions surveys, respectively) by assigning cells with the point value of the laser observation that falls within the cell. Where more than one laser observation was found in a cell the last one encountered in the point cloud was used. Empty cells were filled by smoothing their neighbours. Images were collected for first pulse, last pulse and intensity (the overall strength of the laser return). Staff at ULM wrote a vegetation-removal algorithm to create a digital elevation model (DEM) of the topography of the site under the forest canopy (Devereux et al, 2005). As there is no mathematical expertise within the Aerial Survey team for the writing of algorithms it was felt important to analyse the potential of using just raw last pulse data to see what information could be gained that was not available from the first pulse. The data was provided to English Heritage in the form of ASCII tables recording the x,y,z and intensity data for the first and last pulse in a single table. This was separated into tables that could be read into ArcGIS 8.3 where the 3D Analyst module was used to interpolate to raster using Inverse Distance Weighting. The results were very positive in that though they did not remove all traces of vegetation, as was achieved by the algorithms, they did reveal a large amount of previously unseen detail.
Why was scanning selected?
The potential of lidar to penetrate the canopy and allow the recording of features that were invisible to standard aerial photographic techniques made it ideal to test in such an environment that is also difficult to survey on the ground.
What problems were encountered?
The Forest of Dean was the first survey area where the Aerial Survey team had direct access to the lidar data and this led to a very steep learning curve in how best to use the data. The large file sizes also created practical difficulties in terms of the processing power of the teams PCs. On a more technical note, whilst the processed algorithm left a bare earth DEM the raw last pulse data left a large amount of “stumps” representing either the actual trunk of the tree or particularly dense areas of foliage that could not be penetrated. This was particularly noticeable in the area of conifer plantation where even the last pulse data was unable to penetrate the canopy due to the density of foliage. This data is useful for more detailed analysis as it provides information to aid location for follow-up fieldwork and on the condition of the archaeological features.
What were the final deliverables?
Whilst the ULM algorithm produced a true DTM the raw last pulse data produces something between a DTM and DSM as it removes the bulk of the vegetation, but not all of it. This was illuminated from various elevations and azimuths to reveal variations in the surface that might relate to archaeological features. Because the lidar coverage extended beyond the edges of the woodland it was possible to compare the results with those from the conventional NMP survey and confirm the presence of known features.