We can easily imagine the whole picture, if a particular problem is sketched on a sheet of paper, or when the observed area is illustrated with a painting, in conjunction with understandable symbols. A map is therefore a visual representation of particular portion of space by means of symbolic delineation of properties and relations between the mapped elements, in our case in a geographic space.
A cartographer selects by a creative abstraction of reality only the information that is essential to fulfil the purpose of the map, and is suitable for the target scale.A map production has a long history with roots in visual representation of geographic space extending back at least 30,000 years.
It appears that the demand for land surveying arose from the need to measure the land for agricultural purposes beginning some 5000 years ago. These are the foundations of cartography as a science and technology. Looking traditionally, the maps are actually a combination of science, technology and art. Those that comprise enough realism and aesthetics can serve as a comprehensive communication instrument between people.
Getting Information from the Old Maps
The map’s interpretation depends on a cognitive perception of the Earth by both, cartographers and users. Until the recent past, examination of the maps have been possible only visually (Gašperič 2007), by analysing a single, or comparing two or more datasets lying adjacent to each other. It is based on a human interpretation of a spatial location and orientation, size, shape, colour, texture, pattern, and contextual resources in the maps.
This kind non-automatic interpretation is used, along the semiautomatic, as a common technique for data acquisition, i.e. converting the analogue maps into a digital form. The acquired data is usually vectorised to the topologically adjusted datasets, founded on object-based model as points, lines and polygons, with corresponding attributes.
A geographic information system (GIS) allows many possibilities of manipulations with those datasets, whereas spatial analysis allows a combination with the other maps or spatial datasets by means of overlaying principles. The GIS also has the tools for transforming the digitised maps into a common coordinate system.
The old (historical, archival) maps merit respectfully consideration with reliable reconstruction of the different kinds of ancient issues. Thus these maps may get a valuable applicability potential through the following standard workflow: digitalisation (1), georeferencing (2), and thematic acquisition (3). The old maps biases to different inaccuracies that are interwoven through the workflow; hence spatial data error handling (4) is required.
Digitalisation Problem – 1
The access to archival maps is increasing, but at the same time there are an increasing number of scanned maps available. However, digital media lifetime is much shorter than the analogous data sources and the established paradigms of archiving were not adapted to them. The quality of the scanned maps is often pure considering their resolution, contrast and many kinds of sources distortions, such as instability of the paper on which the old maps were printed.
Georeferencing Problem – 2
The georeferencing or rectification is a problem that considers the nature of mapping and error patterns of the old maps. This complexity is usually solved with planimetric procedures in a 2D space.
Technical equipment and technology. Geometrically and topologically correct presentations have been demanded in the topographic maps. However, different techniques of measurements with various surveying equipment are succumbed to various random, systematic and gross errors.
For example the 1st military survey was carried out using plane table and busole, without precise (geodetic) surveying measurements; mountains areas where especially problematic, the cartographers even mapped by eyeballing (Podobnikar 2009a).
Mapping standards. Maps or map sheets for diverse areas produced under different conditions that allow no common and objective approach for georeferencing.
The subsequent problems have been: applying different mapping techniques; different projections, coordinate systems (with geodetic datums), scales, etc.; inhomogeneous metrical quality of the maps; different or not defined generalization methods; administration (countries) changes; and uncommon map legends keys (object catalogues) (Fig. 1).
Figure 1 – Common object catalogue regarding legend keys of maps Josephine 1st Military Topography, Franciscan cadastral maps, and Italian and Yugoslav topographic maps.
Skills of the surveyors and cartographers, who might have had a high level experience and technical skills for their time, but not for nowadays. Considerable variability (“styles”) between different operators may appear, too. In general, the applicability is lower when the map is older.
Landscape changes through the time, where they have been anthropogenic, e.g. felling, urbanization, and natural, e.g. flood, erosion, landslide, earthquake, fire (Podobnikar et al. 2009). If these changes appear on the map, these features cannot be used as reference (Fig. 2).
Figure 2 – Changes of the Mura River with orthophoto in background (below) according old maps (above), for the NE Slovenia (4 by 3.5 km). Wider surrounding of the river cannot serve for georeferencing.
Purpose of mapping and policy. Maps have been produced for a specific purposes, e.g. for military use, tax collection, routing or other. Their content and accuracy was intentionally adapted and limited with their purposes.
Reference datasets of a higher quality than the old maps can be used, usually in a modern coordinate system. It is not possible to get an absolutely ground truth reference.
The following datasets can be used: (contemporary) topographic maps in considerably greater scale than the target old maps; DEMs; orthophotos; field survey data; old photographs; textual sources; and various other materials, e.g. archaeological data.
Ground control (identical, reference) points (GCPs). Locations have to be chosen carefully paying attention to the: reliability of the reference dataset; ancient measurement techniques; generalisation issues; changes of the natural and built-up environment; possible misunderstanding of the toponyms due to change their names or translation to the other languages (e.g. German to Slovenian). Apart to the GCP’s, ground control lines or aerial features may be used, e.g. peaks, ridges, valleys, lakes (Molnár 2010).
According to our experiences, the most relevant locations of the GCPs for most of old maps have been churches, towers, bridges, road crossroads, railways, estate borders (land cadastre), distinguishable peak points or river confluences.
Mosaicking of singular map sheets to a seamless map. This is an optional procedure in a case when a system of map sheets is to be reconstructed by stitching into a cohesive map (Fig. 3). It is a kind of relative georeferencing that relates singular map sheets to dependable positions. The technical solution considers resembling the map sheets original shape and simultaneously in fine detail stitches the map edges.
An additional option is colour balancing among the map sheets as a process of inconsistent colour cast homogenisation. It can improve the maps interpretation and is important for an efficient visualisation. The failures of colour casts can be a result of imprecise scanning process or, more commonly, an outcome of the original print where the inhomogeneity may increase over the centuries due to aging. Since different cartographers may have produced the maps, their overall colour cast differed originally.
Figure 3 – Left: distortions at the map sheet borders of the single map sheets and a mosaicking afterwards. Right: prevention of distortions using rubber sheeting technique (Neubert and Walz 2002).
Transformation to the target coordinate system is the main goal of old maps georeferencing, where previously described problems might be carefully considered and adapted. The robust and as much as universal solutions are being a matter of the scientific development.
Two basic methodological groups may be distinguished: (i) transformation on the base of known projection parameters, and (ii) transformation based upon GCPs.
(i) The first group requires uncovering of the old maps reference system, where projection parameters, and in the best case geodetic datums, should both be known. These parameters are seldom available. This is a standard procedure for cartographic data that is necessary for transformation between different coordinate systems, e.g. between national and international ones. Since the GCPs are not used, the results may yield to mapping inhomogeneity errors.
(ii) The second methodological group is established upon the determination of appropriate GCPs. The georeferencing is carried out using reference datasets (Fig. 4).
The transformation methods to a target coordinate system bases on GCPs with unknown parameters. This principle is also used in cases when the first group of methods was applied, but a higher accuracy of results is required – with additional GCPs many local distortions can be eliminated. The common methods are polynomial transformations and rubber sheeting.
The polynomial order 1 (linear transformation) is appropriate for transformation between two projections where great distortion is not presented. This was useful especially for the relatively contemporary maps in similar projections transformed to the target coordinate system. The polynomial order 2 is appropriate for transformation between geographical coordinates (sphere) and projected data, or for locally systematically distorted datasets in larger areas.
The rubber sheeting method is based on triangulation, where the positions of the identical points are kept after the transformation. This method could be useful for areas with high distortion. It requires a high number of identical points where their deficiency can cause unpredictable distortions.
Figure 4 – Georeferencing principle: original old map, reference contemporary map, and georeferenced old map.
Thematic Acquisition Problem – 3
A thematic acquisition is considerably interwoven with the georeferencing problem. The goal is to get a spatial dataset as a structured information founded on an object-based model of discrete entities (points, lines, polygons), or seldom on a field-based representation (grid). The acquisition to the target dataset, e.g. to a categorised land use, requires fine interpretation of the old maps (Csaplovics 2009).
To get homogeneously classified attributes of the target dataset, the map legends keys are needed to be converted to standardised object catalogues. In the case when time-series datasets are the target application, which reflects specific discrete moments, a backward editing (reverse order by editing) method as kind of reverse engineering can be applied.
This method bases on a map-by-map overlay interpretation and adaptation to the momentary dataset, acquired in the previous step, starting with the most contemporary dataset (Fig. 5). The most typical problems of the thematic data acquisition are low level of details in the old maps, and inhomogeneous symbology in the different old maps.
Figure 5 – Backward editing of land use data (below), with assisting of contemporary orthophoto, TK and the last, JMT, for the area of Lake Bohinj, NW Slovenia (4.2 by 3 km).
Error Handling Problem – 4
Statistical and visual methods of quality assessment are needed to obtain in all parts of the entire process (Drecki 2002). Positional, thematical, temporal errors together with completeness and logical consistency are to be assessed. The positional errors are simply assessed through the transformation processes on the basis of various polynomial transformations.
Although the positional errors in old maps can be normally distributed, their character often contains admixtures of poorly explained heterogeneities and other uncertainties that are not simply described. They may be classified as random, locally systematic, and systematic (with the limits at the gross) error distributions (Podobnikar 2009b). The visual methods may employ various cartometric analyses (Jenny 2009) of positional, rotation and scale displacements, and on simulations of the errors (Fig. 6). Both approaches can assist in understanding the nature of errors, and may be used for improving the techniques of georeferencing in further iterations.
Figure 6 – Left: rectangular grid squares of 200 m with assigned land use after a Monte Carlo simulation, for Josephine 1st Military Topography error model and area of Lake Bohinj (7.5 by 12 km). Right: cartometric presentation of positional error (exaggeration factor 10), for 3rd Military Topography and area of Murska Sobota (12 by 19 km).
Integration of the Old Maps with Other Datasets
Datasets acquired from old maps can be associated together with many others like already listed datasets for reference datasets, in order to semantically enrich them and gain more efficient interoperability or even integration. Significantly relevant and independent properties of any quality datasets may support better results (Podobnikar and Vidiček 2010). Since there is not a real homogeneous reference data as ground truth data information, cross validation of many different data can drop hints to their impact. For example, the land use datasets may be improved by using other maps and even (historical) DEMs.
Spatial datasets acquired from the old maps can be involved in numerous applications that include fields like geography, geodesy, geophysics, history, archaeology, ecology, climatology, biology, environmental studies, or even toponomastics. We can study a landscape dynamics history, with datasets of relief, palaeo-hydrography (Fig. 7), cultural landscape, aerial or satellite images.
It is possible to reconstruct fragments or all historical courses of border changes influences, urbanisation roles according to the natural and cultural lawfulness, natural disasters and risk management. There are even interesting applications on visualisation a landscape history on a contemporary way, e.g. according to modern topographic maps or thematic cartography principles to produce stability maps or land use changes trajectories…
Figure 7 – Application: changes of the river course the Ljubljanica River (Ljubljana, Slovenia) in comparison between georeferenced Josephine 1st Military Topography (~1780) with blue line as contemporary river course, and DEM visualised by bipolar differentiation (2005) where traces of the ancient river course can be identified, for the area of 6 by 4 km.
Conclusions
Concerning historical cartography, amongst the most prominent tools and technologies is GIS with its all capabilities for numerous locational data manipulation and analysis. The valuable information on the base of the old maps can be attained by digitalisation, carefully georeferencing and attribution, and semiotical understanding of their backgrounds, which may further lead to beneficial applications.
These modern tools are enabling the realisation of many more challenges and ideas than ever before and in considerably shorter time. Not only this, the input of novel technologies allowing us faster discoveries and gratification of our curiosity. Moreover, the semantically enriched interdisciplinary corpuses of spatial data, together with the old maps, can lead to very accurate analysis.
These issues can apparently help our learning from past mistakes. The overall good practice may serve as support for better decisions and, hopefully, more sustainable development of our society.
Acknowledgments
The maps used in this research were provided with the permission and kindness of many institutions and people. The originals of the historical maps were digitally photographed. The main map sources were provided from: Scientific Research Centre of the Slovenian Academy of sciences and Arts (Anton Melik Geographical Institute), Vincenc Rajšp, Gábor Timár and Romeo Varga. The contemporary topographic data was provided by Surveying and Mapping Authority of the Republic of Slovenia.
—————————————————————
Tomaž Podobnikar – Scientific Research Centre of the Slovenian Academy for Sciences and Arts, and Faculty of Civil and Geodetic Engineering, University of Ljubljana.
More Information: (www.geospace.si, [email protected])
References
Csaplovics, E., 2009. Transnational Ecological Networks. Moravske toplice, http://web.rra-mura.com/prenosi/predstavitev_06.pdf
Drecki, I., 2002. Visualisation of Uncertainty in Geographic Data. Spatial Data Quality.
Gašperič, P., 2007.Cartographic images of Slovenia through time. Acta geographica Slovenica 47/2, http://giam.zrc-sazu.si/index.php?q=en/node/118
Jenny, B., 2009. Automated cartographic techniques for terrain representation, map distortion analysis, projection design, and interactive mapping. PhD dissertation, ETH Zurich. http://jenny.cartography.ch/phd/phd.html
Molnár, G., 2010. Making a georeferenced mosaic of historical map series using constrained polynomial fit. Acta geod. geophys. Hung., 45/1, http://www.akademiai.com/content/n7n7r760x551114r
Neubert, M., Walz, U., 2002. Auswertung historischer Kartenwerke für ein Landschaftsmonitoring. Angewandte Geographische Informationsverarbeitung, Heidelberg, http://www2.ioer.de/recherche/pdf/2002_neubert_walz_agit.pdf
Podobnikar, T., Schöner, M., Jansa, J., Pfeifer, N., 2009. Spatial analysis of anthropogenic impact on karst geomorphology (Slovenia). Environ. geol., 58/2, http://dx.doi.org/10.1007/s00254-008-1607-3
Podobnikar, T., 2009a. Georeferencing and quality assessment of Josephine survey maps for the mountainous region in the Triglav National Park. Acta geod. geophys. Hung., 44/1, http://www.akademiai.com/content/q8hm1230m260
Podobnikar, T., 2009b. Methods for visual quality assessment of a digital terrain model. S.A.P.I.EN.S., 2/2, http://sapiens.revues.org/index738.html
Podobnikar, T., Vidiček, M., 2010. Positional and thematic digitisation error assessment in the practices. Accuracy 2010, Leicester.
