Georeferencing

Do you want to browse old maps with your smartphone? Experience the history of where you live from a new perspective? Discover fascinating facts about geographical points along the way? You can do all this and more through georeferencing, with which spatial information is added to digital media.

Terms like georeferencing, spatial data or geoinformation all sound pretty technical. They tend to be associated with specific branches of science, such as geography, geoinformatics, environmental sciences and spatial planning, but this has changed over the last 15 years. Google Maps opened the door to web-based maps and geodata to an ever-broader public.

The ETH Library also makes use of georeferencing methods, allowing it to index more facets of its holdings and thus offer users new, intuitive access through interactive map interfaces. Map collection holdings are particularly well-suited for georeferencing, but other media such as old prints, photographs, periodicals or herbaria also have spatial references that can be recorded and further processed. These newly recorded references help users to answer questions quickly and precisely, to find specific media, to establish links between holdings and knowledge, and to acquire new knowledge. However, they also allow you to simply browse maps and images with your smartphone. This translates into enormous added value and benefits not only students and researchers but broad swathes of the interested public as well.

Georeferencing is when certain media are linked with spatial information, with a distinction being made between three different methods:

• Point referencing (also referred to as geotagging)
• Surface referencing (2-D georeferencing)
• 3-D georeferencing

In point referencing, coordinates are assigned to a single object. This could be an object in a photograph or a place mentioned in a text, for example. The coordinates define a specific point on the globe through the specification of latitude and longitude. Localities or lakes can also be referenced by a single point, with the geometric centre of gravity of the respective area often determining the precise location.

An example of the use of point referencing at the ETH Library is ETHorama. This platform aspires to visually link places and areas in Switzerland with appropriate contemporary and historical content.

An interactive map (based on Google Maps) shows locations to which individual digital documents are linked: images, texts and historical maps from the various source systems of the ETH Library (E-Pics, e-rara, e-manuscripta, E-Periodica and Research Collection). 1 A distinction is made between two types of markers: points of interest (POI) and areas of interest (AOI). A POI is used for defined points, such as sights and tourist attractions, individual buildings or bridges, while an AOI marks a specific area, such as a village, lake or valley. However, an AOI is also recorded as single point coordinate.

The content on ETHorama is selected and point-referenced, that is, assigned to a POI or an AOI, by ETH Library staff members. This means that journal articles and digital copies of old prints can be indexed by location down to the page level. For example, users can follow the entire route on a map while reading a historical travel report. New additions to ETHorama can be found in the thematic collections, which display content in a thematically consolidated manner (e.g. only documents on the subject of castles and palaces, or only aerial photographs by Walter Mittelholzer). ETHorama now contains more than 45,000 images, texts and historical maps at over 5,600 locations.

Georeferencing of documents in ETHorama is also used for the research platform swisscovery. For resources with a location reference, a “Geographical reference” tab appears in the list of results. This tab lists the locations to which the document was linked on ETHorama.

Using point referencing, spatial attributes can thus also be assigned to media like texts or objects, making it easier to find search objects, such as the location of a specific plant.

Surface referencing assigns coordinates to each individual point (pixel or dot) of a digital image. In this way, an image or a map can be precisely located on a virtual globe. Libraries have been using an intermediate form of point and surface referencing for some time. A national map or an aerial photograph taken vertically is recorded using its four boundary coordinates. This creates what is referred to as a bounding box, which represents the corresponding surface in the form of a cuboid.

The ETH Library has been recording bounding boxes since it has had the Map Department, that is, for almost 50 years. It is the only institution in Europe to do this for each of the 400,000 maps currently in its holdings. Over the years, the technology has continuously evolved: from manually recording coordinates in the library system, to the use of geographic information systems (GIS) that provide ready-made boundary coordinates, through to web-based applications, the tools for surface referencing have become better and more efficient over time.

Surface-referenced data is also used outside the library system, such as on Kartenportal.CH, the specialist portal for maps in Swiss libraries and archives. The advantage of the recorded boundary coordinates becomes apparent here: the user can see at a glance whether a map covers the part of the world that interests them.

Georeferencer shows how it works to surface-reference a complete image – that is, not just the frame. This tool converts digitized maps – such as city maps, along with geological or other subject-based maps – into a virtual globe using reference or ground control points. Each point or pixel on a map is assigned its own coordinate and can be combined or compared with other maps on this basis. Such data allow high-quality analyses – such as in the field of spatial sciences – or browsing of maps with your smartphone or GPS device. In order to be able to carry out the extensive work required for this, the ETH Library relies on crowdsourcing. Interested persons can georeference maps themselves, with ETH Library staff members performing quality checks.

Swiss cantons and federal offices such as swisstopo are also offering a growing number of georeferenced maps and spatial data – in part due to legal requirements. This often leads to cooperation with libraries and archives, such as with the ETH Library or the Staatsarchiv Zürich state archive.

The ETH Library also makes this official data available to members of ETH Zurich and other universities. For easier access, it operates the GeoVITe portal together with the Institute of Cartography and Geoinformation (IKG) at ETH Zurich.

3-D georeferencing draws the eye into the depth of the space. If surfaces are supplemented by height values, the third dimension can also be referenced. This is particularly interesting in the case of photographs or aerial images taken at an angle, as they can then be correctly located in the space.

The Laboratoire of SIG, School of Management and Engineering Vaud (HEIG-VD), recognised the potential of 3-D georeferencing for historical images and developed sMapshot in 2017. This platform is “a participatory time machine”, according to its own statement, because “in the past, there was no GPS”. A virtual, web-based globe forms the central element, enabling both 3-D georeferencing by volunteers and visualisation and searching of georeferenced images. 2

When georeferencing an image, volunteers first record the camera position, the direction in which the shot was taken and the height of the camera. After that, the sMapshot algorithm carries out calculations based on this data.

In a second step, at least four points on the globe and in the image must be identified and reconciled, that is, matched up with each other, so that the algorithm can produce an initial calculation for the accuracy of the georeferencing. Finally, the image can be georeferenced if at least six points are provided.

3-D georeferencing of an image is also quite valuable for another reason: it enables precise identification and indexing of places that are visible in the image and that cannot be recorded in such detail with the usual textual metadata. The conventional metadata in the E-Pics Image Archive Online image database is supplemented with the more precise geoinformation from sMapshot. The entire image is recorded as an information unit and users are able to search more precisely.

The ETH Library Image Archive has been working with sMapshot since January 2018. More than 210 volunteers with local knowledge have positioned and thus georeferenced more than 110,000 photographs – primarily aerial photographs so far – on the virtual 3-D globe. Those who register can manage their georeferenced images themselves, but quality checks are carried out by ETH Library staff.

The three methods of georeferencing build on tasks performed by scientific institutes and the ETH Library. They lead to the establishment of specialist portals and open up new access to digitized media, thus adding value.

Research also benefits greatly from this. More and more students and researchers require different types of geodata, and efficient access to spatial data makes their work easier.

Georeferencing makes the ETH Library’s documents a gold mine for science. The spatial data can be used to combine content from the holdings with other resources, allowing completely new research questions to be raised and answered. There are also points of contact with the non-natural sciences, such as:

• Linguistics – in regard to research into how place names originated and changed over time
• Political sciences – in regard to defining and changing geopolitical boundaries
• Economics and tourism – in regard to management of supply and demand
• Journalism – in regard to preparing maps, infographics and factual data, for example in the case of referendums or the very topical subject of epidemics
• Data sciences – in regard to linking of different data sources using spatial specifications

The methods of georeferencing presented here are essentially based on the manual recognition and specification of the location of reference points by library staff members or volunteers. Of course, now there are also automated procedures that are being used and tested at the ETH Library.

One way to recognise geographical names is using OCR (optical character recognition) and NER (named entity recognition). With these procedures, algorithms are used to make text passages such as words, book pages or place names on map sheets machine-readable, and then recognise them as names of persons, institutions or places. These can be linked to geographical information that is already available.

Georeferencing gives users access to better library catalogues and specialised portals for fascinating collections. But the added value that institutions like the ETH Library create by indexing their holdings with spatial information is even greater, because linking different media using spatial references creates the basis for new knowledge. In addition to researchers and students, it is of use to everyone interested in finding answers to exciting (research) questions. Georeferencing is thus opening science to inquisitive laypersons and enabling citizen science.

The ETH Library has played an active role in this area for many years. It has fostered the digitization and georeferencing of its own holdings, making the latter available in enhanced form to both experts and the general public.