Knowledge Centre- Subject Index     Main Help page

The technical side of the TrenchMapper project.

There were significant problems to overcome to show a large number of maps and aerial photos on a website.

1. Scanning. The maps were scanned using a Colortrac Smart LF scanner set to 300dpi. The resulting scans were stored in uncompressed .TIFF format to preserve the image quality. Care was taken with colour matching, the colours recorded will reproduce very close to the original. As a rule of thumb, 300dpi scans of lithographic maps will print at a quality that is indistinguishable from the original to the human eye. The scans were then processed, the WFA and IWM icons added etc. and stored in .JPG format at a compression level that produced compression artefacts that are hardly visible. It is these .JPG files that are available for downloading. When printed at 100% on a suitable large format printer the results will be very close to the original, especially if a heavy grade cream coloured paper is used.

2. Web presentation. The web version of the maps uses a very large number of small map tiles that form the screen image on demand, each zoom level has its own set of map tiles, a system commonly used to present web based maps. Zoom levels are limited to what is practical and to match the screen resolution to that of the original maps. A typical screen will have a resolution of about 72dpi, a 4K screen will be double that. There is little point in zooming in close enough to see individual pixels on screen, especially as storage of the huge number of high zoom level tiles becomes excessive.

3. Trench Map References. TrenchMapper uses sophisticated mathematical techniques to convert trench map references to latitude and longitude values suitable for modern maps and GPS receivers, phones etc. These latitude and longitude coordinates conform to the common World Geodetic System, known as WGS84, a coordinate reference system standardised in 1984 and used worldwide.

To make the conversion, several problems had to be overcome. Latitude and longitude coordinates are marked on the corners of some of the 1:40,000 series maps and it would seem simple to use these to calculate positions anywhere on the map. Sadly, that does not work because those coordinates did not use the WGS84 standard. To many people it comes as surprised that for any given point on earth, there are many different latitude & longitude values that define that single point, sometimes differing by a kilometre or more. Different countries used their own Coordinate Reference System, CRS. In the past a CRS was sometimes referred to as a datum. Many would assume that the Greenwich meridian is the only one used as the basis for mapping but this is far from the truth. Although it was ratified in 1884 by the International Meridian Conference, various countries continued to use their own meridians, the French used one through the Paris Observatory and the Belgians used one through the Brussels observatory. Even in the UK there is not one meridian, the Ordnance Survey still use the Bradley meridian, this is 19 feet west of the one that everyone sees when they visit Greenwich, the so called Airy meridian; both were named after the Astronomer Royal who set the up. Some of the maps in the TrenchMapper collection reflect this difference in meridians, longitude values measured from Paris or Brussels instead of Greenwich.

To cap all that, the meridian used by WGS84 is different again. It runs about 109 metres east of the Greenwich (Airy) meridian. So if you visit the Greenwich observatory there is evidence of several previous meridians but three are in current use. For small scale mapping this is not significant but if you want to use a GPS to stand where grandfather stood, it is very significant.

The historical geodetic problem

When the British went to war in 1914 there were no adequate maps. The war was expected to be one of rapid movement and of short duration. However, as it became clearer this was not to be the case, emergency measures had to be taken.

The Royal Engineers and the Ordnance Survey had similar CRS problems in 1914 but did not have the benefit of computers, they had to rely on pre calculated tables often in the form of logarithms. These tables were supplied by the French and sadly contained some errors so further compromises had to be made to allow for them.

The best solution would have been to resurvey the whole of the British front but this is a slow, painstaking undertaking. The first triangulation of Britain, a geodetic survey of great accuracy published in 1858, had taken 70 years to complete. The same level of care and accuracy to cover the British front could not be achieved in the time available even when it became apparent the trench lines were almost static. This meant that faster yet less accurate methods had to be employed but this resurvey was not complete at the time of the Battle of Loos in September 1915 and is reflected in the maps that cover the area; the mistakes in the French tables were in places 150 yards out. This can be seen on the TrenchMapper website on sheet 36C maps of 1915, using the opacity slider it is clear the roads do not match those on the modern map very well.

We live on an approximately spherical planet; an oblate spheroid. That means that any map drawn on a flat sheet is distorted. If an accurate map is required this distortion must be known and measurable. This is why map projections are used, to present the curved world on a flat map.

There is an additional problem for this project, the Bonne projection was worked out and drawn using the meridian running through the Brussels Observatory and the earth radius used, the so called Figure of the Earth, was different from that used in France and different again from the values used in the UK. What makes that worse is that the values used were not recorded explicitly so to make the calculations for this project, they had to be assumed, i.e. the radius of the earth at the poles and at the equator. These differ by a fraction of about 1 in 300, a measure of how far the earth is from being a sphere.

The normal way to transform coordinates from one CRS into another is to use published equations to reduce the latitude and longitude to x, y and z coordinates measured through the meridian, then apply a shift to each of these three values. These updated coordinates are then used in the reverse equations using the meridian of the target CRS. This should give the correct WGS84 values but it doesn't in this case because of the uncertainty in the Figure of the Earth values and the unavailability of the three shift values. After a great deal of research these difficulties were overcome and accurately known Great War coordinates could be converted to WGS84 values.

The cartographic problem

There are various projection types used, the one the British used for trench maps was inherited from Belgian maps of 1:40,000 scale, the Bonne projection. This has the property of preserving area, areas drawn on the map are directly scalable to reality on the ground.

That was not good news for the artillery because the Bonne projection does not preserve bearings, a line drawn on the map from a gun battery to the target will not give the correct bearing.

Much later in the war, the British planned to use the same projection as the French, one devised by Johann Heinrich Lambert, an 18th-century Swiss mathematician. This does preserve bearings and is eminently suitable for artillery. A very small number of British maps show the Lambert grid, the old French maps used Bonne but they had adopted the Lambert grid before the war.

The Georeferencing problem

Georeferencing is the assignment of the absolute location of data points on an old map to the same points on earth as shown on a modern map. The modern maps used on the web are a spherical normal (equatorial) variant of the Mercator projection. With sufficient points chosen on the old map, a good fit to the modern map can be achieved, the resulting errors of this process can be hard to spot or at least, acceptable and are generally far smaller than those caused by the dimensional instability of the original paper maps.

The TrenchMapper team is able to use a variety of georeferencing techniques. The quickest works well for regular map sheets and involves cropping to the mapped extents then using calculated coordinates of the four corners, this calculation coming from the extensive research described above. Another method is slower, it involves finding matching Ground Control Points, GCPs on both old and modern maps. This takes some judgement as suitable GCPs are not always obvious, roads are generally in the same place now as then but not always, especially in heavily shelled areas such as in the Hooge locality. This works for non-regular sheets. Another method is to use grid points on a map and use the calculated equivalent WGS84 points. The georeference some items such as distorted maps or some difficult aerial photographs, it is sometimes required to georeference to a previously georeferenced photograph or trench maps. This increase the errors but can be the only option available.

Howard Anderson, Bill Frost. March 2022

 

 

 

Knowledge Version 2 1.1