Open Source Spotlight: Old Insurance Maps


7 min read

Tell Us About This Project and Why You Started It

I wanted to create a collaborative space for georeferencing historical maps, especially multi-page atlases, to turn them into publicly accessible mosaic layers. My idea wasn't new – check out MapWarper, Georeferencer, or the development of AllMaps, for example – but I focused specifically on the Library of Congress' digital Sanborn Map Collection, a fascinating collection of historical maps. Not only does crowdsourcing distribute the work of georeferencing historical maps, but it also provides a new way for people to engage with them. This historical archive covers over 8,000 cities, towns, and villages across North America, so I felt that building a project around it would create something many people could connect to.

Sanborn Maps

The Sanborn Map Company created these maps between the 1860s and 1960s to help fire insurance underwriters balance their risk. They show building footprints, uses, and construction materials, focusing on commercial and industrial areas. The Library of Congress has the most extensive public archive of these maps. Through a massive digitization effort, there are now 35,000 digitized editions available online, and a JSON API makes it easy to access this content. It is an unparalleled, large-scale, and enthralling visual record of urban and architectural development around the turn of the century.

Researchers use these maps all the time, but to perform meaningful spatial analysis with them, each sheet must be georeferenced–rotated, warped, and embedded with geospatial coordinates–turning a scanned document into a Cloud Optimized GeoTIFF (COG). As the project evolved, I focused on creating an easy-to-use browser interface and exposing these historical maps to the public at every stage with a robust, iterative workflow.

  • New Orleans, LA

    From February through May last year, I facilitated a public participation period including only maps of Louisiana, where I live. A few dozen people participated (with a small proportion doing the vast majority of the work, not uncommon in crowdsourcing), and they georeferenced over 1,500 layers from 270 different atlas editions covering 140 cities and towns. Of course, we worked on the largest city in the state, New Orleans. Link to interactive viewer.

  • Plaquemine, LA

    But people also added and georeferenced small towns they were interested in seeing through time, like Plaquemine, where, over the years, the Mississippi River has slowly carved away a few blocks of the downtown. Link to this exact view.

Example of the georeferencing interface. After three control points have been created, a live preview (rendered by TiTiler) shows the georeferenced results.

Once georeferenced, each layer can be accessed and downloaded in multiple ways. An XYZ or WMS endpoint generated by TiTiler can be used in any common geospatial applications.

The “MultiMask” interface eases the creation of contiguous cutlines for individual sheets across a volume.

What we're doing.