The power of a GIS is in the data analysis. Without data, all the bells and whistles of GIS are just that.
The effort to bring stable, accurate data is an enormous one for any GIS. Data development and maintenance is the most costly, labor intensive part of developing a GIS.
There are several ways in which to bring spatial data into a GIS. This article provides a brief overview of some of the more common methods.
The paper map is attached by tape to a digitizing table (or tablet as the smaller digitizers are known). Usually between 4-6 initial points of which the coordinates are known are logged. Optimally these points are such locations as the intersections of graticule lines.
In the absence of an overlying grid system, points are taken from identifiable locations such as street intersections or landmarks.
Using the Puck to Digitize Paper Maps
The data is then digitized by tracing the features of interest with a mouse-like hand-held device called a puck. Once all the features are traced, the newly acquired data is transformed from table units (the coordinates of the digitizing table) to real world units using an algorithm.
This algorithm takes the known table coordinates of the initial points and warps the data to match the real world coordinates assigned to those points.
The error in the adjustment from the table units to real world coordinates is called the RMS error. Results are reported as root-mean-square (RMS) error and average error. The RMS value reflects the range of the error; the precision of the digitized data.
Factors contributing to this error can be human error, shrinkage or physical alteration of the paper map and projection differences.
Heads up digitizing
With the proliferation of low cost sources of digital imagery and large format scanners, heads up digitizing is becoming a popular method of digital conversion. Also know as on-screen digitizing, this method involves digitizing directly on top of an orthorectified image such as a satellite image or an aerial photograph on a computer.
The features of interest are traced from the image. The benefit of this over manual digitizing is that no transformation is needed to convert the data into the needed projection. In addition, the level of accuracy of the derived dataset is taken from the initial accuracy of the digital image.
Coordinate Geometry (COGO)
Coordinate geometry is a keyboard-based method of spatial data entry. This method is most commonly used to enter cadastral or land record data.
This method is highly precise as entering the actual survey measurements of the property lines creates the database. Distances and bearings are entered into the GIS from the original surveyor plats. The GIS software builds the digital vector file from these values.
Geocoding is another keyboard-based method. Geocoding uses addresses from a flat file (such as a .dbf file, MS Access database or excel spreadsheet) to create x,y coordinate locations interpolated from a geocodable spatial database. These spatial databases are most commonly street centerline files but can be other types. The resultant geocoded database is a point file. For more information read my article called “Geocoding 101.”
Global Positioning Systems (GPS)
GPS is a way to gather accurate linear and point location data. Originally devised in the 1970s by the Department of Defense for military purposes, the current GPS consists of 28 satellites that orbit the earth, transmitting navigational signals.
Through interpolation, these signals received by a data logger can pinpoint the holder’s location. Depending on the unit, the locational accuracy can reach to the millimeter.
Combined with attribute data entered at the time of collection, GPS is a rapid and acccurate method of data collection.
Geodatasets can be derived from digital imagery. Most commonly satellite imagery is utilized in a process called supervised classification in which a user selected a sampling of pixels for which the user knows the type (vegetation species, land use, etc).
Using a classification algorithm, remote sensing software such as ERDAS or ENVI classifies a digital image into these named categories based on the sample pixels. In contrast to the other methods discussed, supervised classification results in a raster dataset.