Cartographic Heritage into the Digital

Georeferencing and Cartoheritage on the Web

6111.2 - Accuracy Assessment and Georeferencing Historic Sanborn Fire Insurance Maps

Thursday, July 6
8:50 AM - 9:10 AM
Location: Hoover

Historical maps were some of the most important scientific documents of their time. They systematically encode unique and reliable information about past time-periods for which there is no contemporary substitute (Rumsey and Williams 2002). Large scale maps like those used in the fire insurance industry of the late nineteenth and early twentieth centuries, were produced from intensive field surveys that are thought to have been highly accurate in their spatial representation of historical features. Fire insurance maps continue to be used by a diverse set of people today including scientists performing environmental site reviews, historians, geneologists and others. A primary challenge faced by contemporary collectors and curators of historical fire insurance maps is to unlock the wealth of information that they contain for the benefit of a diverse set of potential users.
Two preparatory activities are required to unlock the information contained on historical maps: accuracy assessment and georeferencing. For small scale “one-off” maps of ancient origin best practices for accuracy assessment and georeferencing have been established and are widely seen as unproblematic given the often low-quality nature of the geographic information that they contain (Favretto 2012; Grosso 2010). In contrast, voluminous large-scale map sets (i.e. that contain many map sheets all produced in the same way), of more recent origin likely present a different set of accuracy assessment and georeferencing challenges that have received comparatively little attention in the published literature (Balletti 2006; Fuse 1998). The research objectives of the current project were to: 1) Develop and demonstrate methods to assess the geometric accuracy of approximately 31,000 Sanborn fire insurance maps of Pennsylvania (PA), USA that are housed in the Penn State University Libraries map collection. 2) Empirically derive a recommended number of ground control points (GCPs) for georeferencing PA Sanborn maps that minimizes labor while also preserving the native accuracy of print maps in digital geospatial surrogates. 3) Develop a set of best-practices for production-scale accuracy assessment and georeferencing large-scale print maps.
A probability-based, spatially-balanced random sampling design was used to both select 202 individual sheets from the entire Sanborn map set as well as to identify individual GCPs to assess the geometric accuracy of individual map sheet. A probability-based sampling design ensured inference to the entire map set while examining only a small portion of the entire map set, thereby minimizing labor. To develop a recommendation for the number of GCPs required to georeference sheets, we examined the change in total root mean squared error (RMSE) as GCPs were iteratively added to each sheet. Non-linear regression models were fit to RMSE observations to identify the inflection point where total sheet RMSE stabilized and this became the recommendation for the minimum number of GCPs required to georeference PA Sanborn maps.
Lessons learned from this project can be summarized into a set of best practices for assessing the accuracy and georeferencing large scale historical map series. Best practices include: 1) Use reference base maps of the best known positional accuracy available. 2) Use a probability-based spatially-balanced random sampling design to statistically assess the geometric accuracy of map series and determine a minimum number of GCPs required prior to initiating production-scale georeferencing efforts. 3) Report map accuracy and methods of accuracy assessment in the metadata of digital geospatial surrogates. Following these best practices will ensure that the digital geospatial data produced from georeferencing large scale historical print maps will meet the needs of a diverse set of potential data users from hobby enthusiasts to quantitative researchers.

Nathan B. Piekielek

Geospatial Services Librarian
The Pennsylvania State University

Nathan Piekielek, (Ph.D.) is the Geospatial Services Librarian at Penn State where he supports the spatial research and teaching activities of all disciplines and units at the university. In addition to providing research consultation services upon request, he builds, maintains, and manages an extensive print map and digital geospatial data collection. He has a largely quantitative background in geographic information systems, remote sensing, and conservation ecology and is interested in how this expertise can be applied to contemporary challenges faced by academic libraries, like the large-scale conversion of print maps and historic aerial photographs to web-accessible, open source, high-quality digital geospatial data products.


Send Email for Nathan Piekielek

Angeliki Tsorlini

ETH Zurich, Institute of Cartography and Geoinformation

Angeliki Tsorlini is a postdoctoral researcher in Cartography at the Institute of Cartography and Geoinformation, ETH Zurich, Switzerland, since 2012. She graduated in Rural and Surveying Engineering at the Aristotle University of Thessaloniki (AUTH), Greece in 2004. In 2005, she received a M.Sc. degree in Cartographic Production and Geographic Analysis at the same University and in 2011, the Doctor of Engineering degree in Digital Cartography. She has worked on different projects and her research interest is focused mainly on the digital analysis of historical maps and the information which can be extracted from them.


Send Email for Angeliki Tsorlini


6111.2 - Accuracy Assessment and Georeferencing Historic Sanborn Fire Insurance Maps

Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for Accuracy Assessment and Georeferencing Historic Sanborn Fire Insurance Maps