Unbinding the Atlas: Working with Digital Maps
NYPL has now scanned nearly all of its public domain New York City atlases (a collection of now more than 10,000 maps, the wonderfully graphical title page at left is from a recently scanned Sanborn atlas of Staten Island) and built a web tool (blog post announcement) where users both inside and outside the Library can virtually stretch old maps onto a digital model of the world à la Google Maps or OpenStreetMap, thus creating a new copy that is not only aligned with spatial coordinates on the Earth, but normalized across the entire archive of old maps. And once we’ve done that, we can walk this digital spatial object through a workflow, adding useful information and context with each step. All of this is done collaboratively, through the piecemeal efforts of staff, volunteers, and interns, a group of roughly 1,500 participants worldwide.
It’s helpful to think of this type of work in terms of moving or changing contexts. A book is a linear intellectual object that has a particularly bibliographic context… a start, a sequence of pages, and an end. An atlas is a type of book within this framework that also has geographic and spatial contexts, that is, its pages have a frame of reference (earth) and therefore relate to one another spatially, through adjacency, proximity, scale, and coverage area. In the process of aligning old maps, we’re rearranging the atlas, transforming it from a linear, bibliographic object into one that is more true to its spatial nature. Below is a sketch outline of the processes we perform on maps to recontextualize them as spatial digital objects.
Imagine we’ve taken a photograph of a map. By simply looking at that map, we can access our own geographic knowledge base, through visually scanning its shapes, curves, and contours, by reading the title, or by decoding the various symbols that appear throughout. When we make a digital image of a map, there is nothing inherently geographic about it. It, like a photograph of a person or building, is a mass of pixels of varying color. A computer cannot presently look at pixels on a map and divine its geographic locale. This is where map warping comes into the picture. Using the tool we’ve created at maps.nypl.org, we enable users to geolocate pixels on a map. By inserting a virtual pushpin on the pixels that represent a location (e.g. Fifth Avenue and 42nd Street) and then inserting a virtual pushpin on that same geographic location, but on a digital map (e.g. Google Maps), we create a relationship between pixel space of the image of a map and the geographic space of a digital model of the earth. Once we’ve compiled a series of these pushpin equivalencies (three at the very least), we can create a new derivative work from the original digital photograph of a map, one that is spatially aware from the perspective of a computer. The image at right is John Montrésor's 1775 A Plan of the city of New-York, originally surveyed 1766, in the foment of pre-revolutionary America, undergoing the warping process, with color coded virtual pushpins tying pixel to geographical space.
This makes it possible to create really cool visualizations and map overlays (see Montrésor's map, at left) in applications like Google Earth (Montrésor's map for those who have GE installed) and more importantly, to create visual spatial histories of your locale of interest. We can also easily compare maps of the same place that might not be at the same scale, might not have the same title, or might be from atlases where the page schemas are completely at odds with one another. All of these issues, bound so to speak in earlier bookish iterations (as atlases), are mitigated through this process.
Map warping further allows us to create new mosaic-like copies of larger works, like this fire insurance atlas published by G. W. Bromley in 1909, seen at right with its 47 map sheets as a unified whole. Before creating a composite image we need to crop away all of the extraneous marginalia, that is, the non-cartographic information that usually frames a map. Maps can be cropped individually and, if part of an atlas, viewed together with their adjacent neighbor maps. Think of cropping as like applying a map-coverage-area shaped cookie cutter to rolled out map dough, then resizing and assembling those pieces into a unified whole. Here's a handy guide to help walk you through Map Warping and Cropping.
Once we’ve warped a map or warped and cropped an atlas, we can proceed to tracing. This is the final step necessary in transforming printed cartographic materials, maps, and atlases, into machine-readable data. For those whose eyes just glazed over, here it is put another way. This is the final step necessary in creating the means to build a time machine. Remember in the Matrix, how everything we know and see is a simulation created from digital information? Well, what if these maps and all the information on them are the building blocks of an arguably less insidious, historical and spatial research environment to be woven together with documents from the past? We believe they are, but we need to first harvest all of their information through tracing. The map at left is a composite image taken from William Perris' 1852 Maps of the City of New York and is the first to comprehensively detail the built environment of Manhattan. This particular section shows where the Brooklyn Bridge would eventually attach to Manhattan and those buildings, both their spatial geometries and their attributes (e.g. street names and addresses), that would be removed for the bridge construction. Here's a guide to help walk you through Map Tracing.
Implications and Ramifications
So, what does this all mean? If we have documents related to past times and past places (old maps), then we can create data to “rebuild” those past times and past places. And if we “rebuild” old places in virtual space, we can then organize a universe of other information around those old places. Wouldn’t it be great to have yelp.com and menupages.com, but for old restaurants and with old menus and prices? Or to have at least a smattering of old photos in a historical street view? Or to search the National Newspaper Digitization Project using a map interface? At the core of all of these dream-like research futures is geographic information, in machine-readable format. And to get there, we need to warp, crop, mosaic, and trace our old maps. That’s why we’re doing what we’re doing. And as a positive byproduct, the maps just so happen to become more useful at each step along the way.
So if you’re interested in participating in this long-term, collaborative research project, one that accretes little bits of new information to a collectively held historical knowledge base, free and open to all… let us know, or start in on it now by creating an account at maps.nypl.org. Here again are the instructions for Map Warping and Cropping and Map Tracing. Stay tuned for the next post, where we'll talk about some of the projects that use these tools, including the NYPL's own New York City Historical GIS Project, a grant funded by The National Endowment for the Humanities.