CNSI and its subcontractors adopted a vector approach for its GIS-based site selection process. When the process began in 1990, far less geographic data was available in digital form than it is today. Most of the necessary data was available only as paper maps, which had to be converted to digital form using one of two digitizing procedures, manual digitizing and scanning. In one of its interim reports, CNSI described two digitizing procedures used. Here's how it described "digitizing:"
In the digitizing process, a GIS operator uses a hand-held device, known as a cursor, to trace the boundaries of selected disqualifying features while the source map is attached to a digitizing table. The digitizing table contains a fine grid of sensitive wire imbedded within the table top. This grid allows the attached computer to detect the position of the cursor so that the system can build an electronic map during the tracing. In this project, source maps and GIS-produced maps were compared to ensure that the information was transferred accurately. (Chem Nuclear Systems, 1993, p. 8).
One aspect overlooked in the CNSI description is that operators must encode the attributes of features as well as their locations. Tablet digitizing (illustrated in the photo below left) is an extraordinarily tedious task.
Compared to the drudgery of tablet digitizing, electronically scanning paper maps seems simple and efficient. Here's how CNSI describes it:
The scanning process is more automated than the digitizing process. Scanning is similar to photocopying, but instead of making a paper copy, the scanning device creates an electronic copy of the source map and stores the information in a computer record. This computer record contains a complete electronic picture (image) of the map and includes shading, symbols, boundary lines, and text. A GIS operator can select the appropriate feature boundaries from such a record. Scanning is useful when maps have very complex boundaries lines that cannot be easily traced. (Chem Nuclear Systems, Inc., 1993, p. 8)
Notice that the CNSI's description glosses over the distinction between raster and vector data. If scanning is really as easy as they suggest, why would anyone ever tablet-digitize anything? In fact, it is not quite so simple to "select the appropriate feature boundaries" from a raster file, which is analogous to a remotely sensed image. The scanned maps had to be transformed from pixels to vector features using a semi-automated procedure called raster to vector conversion, otherwise known as "vectorization." Time-consuming manual editing is required to eliminate unwanted features (like vectorized text), correct digital features that were erroneously attached or combined, and to identify the features by encoding their attributes in a database.
In either the vector or raster case, if the coordinate system, projection, and datums of the original paper map were not well defined, the content of the map first had to be redrawn, by hand, onto another map whose characteristics were known.