Georeferencing the Past

We cannot know where to go next if we do not know where we are coming from.

Fig. 1

Tacoma, Washington, Circa 1891

Maps have been in existence for thousands of years, most commonly recored on paper. Computer-based Geographic Information Systems have grown only out of the last half a century.

Comparing modern day spatial data with non-digitized maps from the past deepens our understanding of how time has shaped a space or place. A richer knowldge of a place's history can lead to better informaed decisions about future planning, as well as deeper connections to place.

A method of digitaizing old maps (or handrawn maps) is Georeferencing. This method utilizes control points, or reference points to rotate, stretch or alter the original map so that it lines up most closely with modern projection.

The map used in this geofreferencing example is a map of Tacoma, WA from 1891 (Fig. 1) that has been fold and warpped and will be lined up with a current projection of Tacoma to see where old landmarks are in modern day Tacoma.

Step 1: Import reference vectors

A Tacoma boundary shapefile was imported into a working feature dataset within a geodatabse, along with Tacoma streets, water bodies and railroads (Fig. 2). Having additional features (aside from just a city boundary) can help find more common feautres to "match" with the 1891 map.

The vectors used have been symbolized with bright colors to be easily distinguishable when set over the 1891 map.

Fig. 2

Tacoma, Washington

Circa 2013

Step 2: Import rasters

The 1891 map used is a jpeg image (a scanned image). The raster was imported into the same geodatabase as the vector files and brought into ArcMap.

Fig. 3

Step 3: Begin Georeferencing

With the vector files and raster file visible in ArcMap, the Georeferencing toolbar is added, and with the toolbar, the 1891 map is selected (Fig. 3).

The raster is manipulated, stretched and rotated so it is most closely in line with the vectors. Control points (x/y coordinates) are then added. These first set of points (First order polynomial transfermation furthur stretch the map as a sheet of rubber would be stretched out. A good place to start is to find a common feature in both the vector and raster, such as a body of water, a prominant street intersection or landmark. As part of a 1st Order Polynomial tranformation in which the initial

manipulation occurs, control points were added in the downtown Tacoma area where most of the streets in 1891 are that same in 2013 (amazing!). A resulting RMS error number indicates how accurate the manipulation is. A higher number indicates greater residual error. A 2nd Order Polynomial transormation will likely be needed.

City Park (today known as Wright Park) is another good place to start adding control points (Fig.4). Note how the original map and bright green lines representing 2013 streets are not quite lined up, but are close. The 1891 map also seems to be warpped, which means additional manipulation will be needed.

Fig. 4

Step 2: Feature Digitization

Once a low RMS error has been reached, and after updating or saving the control points (a table of control points, essentially a table of x/y coordinates, can be exported for others to utilize), landmarks and other features of interest can be digitized. For this example, Narrow Gauge Railroads (the equivalent of a commuter rail line) were digitized. A number of landmarks from 1891 were digitized, as well (Fig. 5). These digitzed features will be part of a new vector that includes modern streets and rail lines.

Fig. 5

Fig. 6

To digitaze features, empty feature classes were created. In the "Editor" toolbar, lines, points and polygons can be drawn, and in doing so the featureclass's attribute table is filled. Figure 6 is a sample of the newly created attribute table for the Narrow Gauge Rail Line polygon feature class.

Digitized 1891 landmarks and transporation features are added to the 2013 features to create a final output map, below.

Tacoma Transportation and Landmarks

1891 and 2013

Lauren

Marie

McKenna