Introduction:
Historic maps are an excellent source of data for archaeological geomatics research. This lab will explore how we can work with historic cartographic products to extract the information they contain and put them into the GIS context for further analysis. This is the same process for integrating remote sensing data.
Lab Part 1
Using ArcGIS you will learn the process of map georectification using the georeferencing tools available in ArcGIS.
Start ArcGIS, choose 'open existing map', and 'browse' to open the GIS database located in your new data CD
and open (double click on) the project document maps.mxd
Look at the various data layers. This gis database is part of the large GIS project I am currently doing for NCDOT to create both prehistoric and historic predictive models of North Carolina. See our project website. For this project we have scanned over 400 historic maps of NC, dating back to the 1730's. We had to georeference them to extract historic features such as roads and structures from them in order to build models of historic settlement in the state. These are propriatary data, so please do not make copies or distribute the data.
We have here some local raster data from Orange county, NC., including Digital Raster Graphics (DRGs), which are scanned 1:24k topo maps (.jpg files), Digital Ortho Quarter Quads (DOQQs), which are scanned aerial photos, NC county boundaries (CB100), and several historic soils maps, including one from 1918 (Soil_or_1918.jpg).
We will see how we georeference the DRG and historic map so that it 'lines up' with the other data in the GIS. We must do this before we can extract the historic features. The same process of georeferencing is also used on remote sensing data as well.
In order georeference the historic soils map, we first need to access the ArcGIS georeferencing tools. You need to go to 'View', then 'toolbars', and click on 'georeferencing'. This will open the toolbar with the required tools. You can 'park' the toolbar on the top of your arc screen by moving it there, or leave it to float.
Georeferencing allows us to 'warp' one raster map so that it will properly overlay other GIS data. Obviously, historic maps are not as precise as modern data, so they must be spatially corrected. Again, this is the same process we use with remote sensing data.
Be sure that you have chosen the correct map to georeference in the area next to the georeference toolbar. The map we will work with is Soil_or_1918.jpg. Put this map on the top of the list at left, so you can see it. Right click on the map name at left and click on 'zoom to map' to see the entire map. You can see it was scanned 'on its side'. Click on the 'rotate' icon on the georeference toolbar (a circle with arrow) and rotate the map 90 degrees to the right, so that North is up.
Now zoom in on a corner of the county soil map, REALLY zoom in! Next, click on the georef icon on the georeferencing toolbar and "pick a point" at the corner of the county. Now, on the list of files at left, click on CB100 (county boundaries at 1:100,000 scale), and move it up the list to make it visible. Right click and use the 'Zoom to Layer' option, and then zoom into the same corner area where you picked the point. Click again on the georeference icon and click on the same corner point. You have now told the computer that these are to be the same location in the GIS.
Now look at the soils map again. You can see that the soil map is now tied in to this point on the screen, but the other corners are not.
Click on the right-most icon on the georeference toolbar and look at the points in the Link File. These are in NAD'83 feet. You will see the two sets of XY coordinates, the first for the soils map, the second for the other map. You can keep this open or get rid of it and look at it again. You can save this link file and refer to it again.
Now do the other 3 corners of the county. Go to the opposite corner of the county soil map (upper right to lower left), and do the same process. You will need to do a minimum of 4 points in order to get an RMSE (root mean square error) number, which is an indication of how well you have matched the points on the two files.
Overlay the historic map on the DRG quad files, how do they line up?
A definition of RMSE is:
"RMSE--Root Mean Square Error. The RMSE statistic is used to describe accuracy encompassing both random and systematic errors. The square of the difference between a true test point and an interpolated test point divided by the total number of test points in the arithmetic mean. The square root of this value is the root mean square error.
Try doing some additional points using the DRG (digital raster graphic-or scanned topo maps) instead of the county boundaries.
A perfect transformation would produce an RMS error of 0, but this is unrealistic. The Rule-of-thumb for georeferencing raster image files is that the RMS Error be less than or equal to one half the map unit size of a cell (pixel).
You can also make a map partially transparent to see how things are lining up. Right click on the name of the soils map on the list at left, and go to properties, then display, then set the transparency at 50%. Now you can look at both layers at the same time. Reset it if you like.
The person with the smallest RMS wins! The rectified map already in the database (recsoil...) has an RMS of 172 feet. Can you do better? Check to be sure that features are lined up.
When you have getten an acceptible RMS and things look like they are lined up well (they are NOT always the same!), you complete the georeference process of the map by going to the 'georeference' toolbar and click on 'rectify'. This will ask you to name the file. You should also be sure to save the link file. The original map file is not altered in any way.
Important hint-do not click on 'Rectify' or 'Update georeference' until you are completed, or ArcGIS will not be happy, and we want ArcGIS to be happy....
Fun!
Overview of the Georeferencing process in ArcGIS 9
* Some materials for this lab were taken from http://www.columbia.edu/acis/eds/gis/georef.html
Georeferencing is the process of aligning an image file such as an historical map, satellite image, or aerial photograph to spatial data (layers that are shape files: polygons, points, etc.). This document describes the basic steps for georeferencing an image using ArcGIS.
Preparation* Prepare the Software
Add the Georeferencing toolbar if it is not already showing by selecting:
View > Toolbars > Georeferencing
* Prepare all the spatial layers
o Open all spatial layers you will use to georeference the image.
o Zoom in to the approximate area that is covered by the image to which these layers will be added. This does not have to be exact as it is done just to provide you with an easier workspace.
o Adjust the color and shape of any points so that they will be visible when displayed on the image.
o Polygons will be easier to use if you make them "hollow" (hollow is a choice in the list of fill colors for symbols). Also choose a color and width for the line around the shape that can be seen against the image layer.
* Prepare the Image Layer
o Open the image layer.
o Rescale the image to fit within the window; select from the Georeferencing tool bar:
Georeferencing > Fit to DisplayAdding control points
This is the process of matching points on the image layer to corresponding points on the shape layers. From the Georeferencing toolbar select the
Add Control Points icon
This tool allows you to georeference the image:
* First, using your mouse, left click on a known point on the image. This will place a cross mark on that location
* Next, left click on the matching control point in another layer. This will 'move' the image and better aligning the control points
* repeat this step with each control pointIf you would like to input x and y coordinates as an alternative to the method mentioned above, after placing a cross mark on a location right click on it and enter in the coordinates.
Note: it is a good idea to zoom in on your image when adding control points for better accuracy.
For every set of control points you create, an entry is created in a table that records the original coordinates, the control point coordinates, and the residual error. Access the table by choosing the View Link Table icon from the Georeferencing toolbar
Entries in this table can be deleted one at a time (highlight the entry in the table and click the delete icon) thus making corrections is easy.
Adjusting the Final Product
Residual error is the measure of the fit between the true locations and the transformed locations of the output control points. If there is a link with a high level of error, consider deleting it by highlighting the point in the link table and clicking the delete icon.
Depending on the number of control points you have, from the Georeferencing table you can perform either a 1st, 2nd, or 3rd order transformation. The transformations compare the coordinates of the source image with the control points creating two least-square-fit equations to translate the image coordinates into map coordinates.
A 1st order transformation shifts the image up, down, right, or left, stretches the image larger or smaller, or rotates the entire image
The 2nd and 3rd order transformations fit higher order polynomial equations to the data, allowing points to be shifted in a non-uniform manner
Most of the time either a 1st or 2nd order will suffice, try all three and choose the one that works best
Saving changes
When you are satisfied with the georeferencing process, three options for saving changes are offered:
* As a text file - you can load the text file when you need to have the image georefenced again:
From the Link Table select Save
* As an .aux file - this saves the changes in a file readable by ESRI products
From the Georeferencing toolbar select Update Georeferencing
* As a World File - this creates a new image (either TIFF, ESRI Grid, or ERDAS Imagine) with the coordinates stored. Use this option if you plan on using the image with other software
From the Georeferencing toolbar select Rectify
Lab Part 2 (You can do these later, before the next class, if you prefer)
Try goereferencing one of the other historic maps. What rms can you get?
Lab Part 3
Review the historic map work done by Scott Madry in France here and here. I will have several of these maps in the lab for you to review.
Digital Mapping Web Resources
The David Rumsey Historical Map Collection contains to date over 8,000 maps online and focuses on rare 18th and 19th century North and South America cartographic history materials. Check out their 3-D GIS viewer. Here is a Wired Magazine article about David Rumsey and his Maplibraries.com project. This is some of the best online historic map GIS work available right now.
The Library of Congress Map Collection
The National Park Service Historic Preservation Service
UC Berkeley Digital Map Collection
The Yale Map Collection consist of over 200,000 map sheets, 3,000 atlases, and 900 reference books. Many of the Collection's historically significant maps are now available online.
NOAA's Office of Coast Survey's Historical Map & Chart Collection contains over 20,000 maps and charts from the late 1700s to present day.
The Perry-Castañeda Library Map Collection, University of Texas
The Bodleian Library Map Collection, Oxford University
The Portsmouth Harbor, UK Project
The William P. Cumming Map Society (NC map society)
USGS Historic and out-of print maps
The Hargrett Rare Book and Manuscript Library at the University of Georgia maintains a collection of more than 800 historic maps spanning nearly 500 years
The UNC Ancient World Mapping Center
Feel free to contact me if you have any questions. See you next week.
~ Main page ~ Schedule ~ Web Resources ~ Bibliography ~
