History Workshop at SotmBaltics

On the Sunday morning of the State of the Maps Baltics Susanna Ånäs had organised a workshop on mapping historical geographies.

Getting the History mapping workshop underway
Senate Room, University Main Building, Tartu

Susanna is a wikipedian from Helsinki who has been recently very active in working with GLAMs (Galleries, Libraries, Archives and Museums) to build resources in the wikimedia commons.

Tartu City plan dated 1941 with war damage overprint.
Note the streets in square H9, these were never built.
source Estonian National Archives via Wikimedia

We were divided into small groups and provided with a nice set of historical maps of Tartu (mainly from by the Estonian national archives) to frame our discussions. I was glad to see that everyone was interested in these historical maps. The summary of the workshop is to be found here. There were many good ideas. I particularly liked zkir's desire for a map of Paris in the time of D'Artagnan; and the use of historical vector maps and routing to illuminate aspects of historical geography).

The rest of this post is really here to add some more flesh to a few things which I mentioned in the workshop.

• Collection of Snapshots or an integrated temporospatial database. Vvoovv mentioned in discussion an expectation that any data would be stored as a set of discrete snapshots with some kind of as-at date I realised that I have been assuming that what we want to achieve is some kind of integration across times, allowing objects to be used across time. Obviously the latter is likely to present some difficult technical problems. My blog posts on persistence (here and here) and using OSM history to look at how we mapped Berlin were all partially concerned with one of these technical issues: potential database size. They were also motivated by the potential to reuse current spatial data from OSM for historical purposes.
  
  The problem I foresee with a collection of snapshots is one of maintaining consistency between them. If someone refines an elaborate building outline (such as the old cathedral in Tartu) using current data for instance from a national survey), then ideally we could instantly reuse this data for earlier periods. I don't quite know how we do this, perhaps by replicating OSM changes to a historical database with some default time-stamp. More sophisticated methods would involve using some kind of persistent object identifier (this is a technology we could do with OSM itself, as I've argued before), which would allow much more selective replication.
  
  There is a suitable half-way house. This is to start with snapshots, but ensure the temporal tagging can be generalised so that data shared between two snapshots can be merged (essentially a temporal equivalent of Ctrl-J in JOSM). This is the sort of thing we should be trying out on the HOSM sandbox.
  
• Improving the pre-processing of maps before vectorisation. This was an issue which Susanna raised and which is already the focus of extensive work. As I put an example on Flickr using a map from 1835 about 18 months ago with a detailed breakdown of my processing steps, I merely refer the interested reader to the description of the image below.
  
  

  

  Sites of Old Osier (willows used for basket-making) at Attenborough, Notts in 1835.
  A plantation of willows covering 4 acres (1.6 ha) is known to have existed in this area in 1086,
  as it is referred to in Domesday Book. Most likely this would be close to the pair of shaded areas shown top-right.
  Copyright is complex. Underlying map is from a publication jointly copyrighted by Nottinghamshire & Derbyshire County Councils; modern map is Ordnance Survey OpenData StreetView released under the OSGB Open Government Licence. Location of Oak Trees is my copyright  and database right.

  This image demonstrates the use-case of making use of historical data, and especially maps, to understand current ecological and environmental patterns (i.e, which areas are likely to have had consistent vegetation over the past few centuries). Such information can be very useful to inform conservation strategies.
  
  
• Date stamp issues. These are of two types. The first just reflect imprecision in available sources. I imagine we need from and to dates expressed with the following types of precision markers: no_later_than, no_earlier_than, at_least_from (=already_existing_at), at_least_to (still_existing_at), existing_around, about. In addition precision may also exist in the date tags (e.g., 20C, 1914, 1914-08, 1914-08-04, etc). Furthermore even apparently accurate source date and times may have discrepancies (see the account of signal book times at the Battle of Jutland in Andrew Gordon's The Rules of the Game) let alone complications of local time prior to the railways. The second issue is probably more intractable. The example was Gaudi's basilica in Barcelona, Sagrida Familia. This structure started being built in the 1920s and is still under construction now. What do we choose for a start date? what do we mean when we allocate a start date? how many different start dates do we need? Large churches provide many such examples: the cathedrals of Ulm, Koeln, Liverpool and Orleans were all built over long time periods mainly to a consistent architectural plan. I can think of several "Bridges to Nowhere" which remained in an uncompleted state for many years (a recent example is the one just finished in Glasgow).
  
• What do we believe? It would be very foolish to take a Soviet era map as an accurate representation of what was on the ground during the period. However, any map will selectively map features according to the needs and biases of the cartographers, their users and their times. A hundred years from now a historian would look in vain on official maps for traces of places like Kibera in Nairobi. Similarly early US automobile maps deliberately excised railways. The collection of historical maps of Pittsburgh has relatively little information on the Steel industry in the period when it was the biggest steel town in the world. All of these examples how that we need to interpret the information in a map with care, and preferably use multiple maps in conjunction with other resources. This is not too bad : for one thing the detective work involved is fun.

Tartu seems to be a very useful exemplar for a whole range of issues which we know about with dealing with historical maps. It has a long history certainly extending before 1000 AD, good documentary resources, a distinguished university and the National Archives on hand. It is also small enough for trying out different approaches. Available open data from the National Archives (e.g., photos, featured later in very interesting lightning talk about crowd sourcing geolocation for historical images) and the National Mapping agency, are also very valuable resources for Tartu.

Other things which either struck me or were mentioned by others:

• Changes in street names, numerous changes between German, Estonian and Russian, plus, no doubt, Soviet-era names which have now gone.

• Several buildings with a complex history of use and re-use. The example quoted in the workshop was a church on , but on visiting the old Cathedral this struck me as even more vivid example. The Old Cathedral fell into disrepair after the reformation, loosing the spires to its twin towers. Later the chancel was restored and used for nearly 200 years as the University Library, now it is the University Museum. This summer an open-air theatre is situated in the nave. Thus this single building covers both historical changes in fabric and usage.

• Interpretation Issues. On several of the maps features exist which were never built or features which existed were never shown (particularly on Soviet era maps). The Attenborough map above illustrates the former well as the railway line shown was actually built further S (just about visible on the current map). The Tartu map from 1941 also shows some planned streets which were never built. (Until quite recently the British Ordnance Survey did not show certain high-security installations, such as Aldermaston AWE, Burghfield ROF, GCHQ, and Fylingdales, even though they were often very prominent features on the ground). Soviet-era mas may also be deliberately generalised or distorted. Similarly NPE maps used by many of us a few years ago for OSM mapping in England and Wales were not reproduced in a way which guaranteed high accuracy at the 100m level (not to mention all the artefacts introduced by scanning and rectification).
  
• Networks Effects. Vectorised historical maps will enable the exploration of topics which were otherwise inaccessible: particularly how changes in the transport network impacts on the delveoping geopgraphy of an area. The example cited was the destruction of Tartu's bridges by the advancing Soviet army in 1944. As the bridges were not replaced for several years this affected how the city developed after the war. Obviously the arrival of railways and motorways would also have profound effects. On a minor scale small foor ferries disappeared along the River Trent sometime around the middle of the 20th century destroying once quite intimate links between riverside villages.

Many thanks to Susanna for running this workshop, which proved to be a great start to a very stimulating series of presentations on Sunday morning.