ArcGIS Pro Lesson 2: Georeferencing Historical Maps

This lesson is undergoing open peer review. Please use the comment section to review the lesson and highlight any areas that require more clarity or could otherwise use improvement. Contact Jim Clifford if you have any further questions about the process: jim.clifford@usask.ca

by Bailey Clark and Joshua MacFadyen, University of Prince Edward Island

October 7, 2020

This lesson will teach you how to georeference in ArcGIS Pro. 

Georeferencing is a digital process whereby a digital image of a map (usually acquired through scanning a physical map) is given real-world coordinates. This is done by aligning the geographic features of a digital image of a map with those same geographic features on a digital map. Known as a reference map, the latter has accurate coordinates. Because of this, aligning the two maps will assign those accurate coordinates to the digital image of a map as well. 

A digital image of a map that has accurate coordinates added to it through georeferencing is known as a raster. A raster is a type of data in ArcGIS. A second type is vector data, which comprises data that one uncovers by analyzing the raster. Because the vector data is based on a georeferenced raster, users must first turn a digital image of a map into a raster before they may gain any vector data from it. 

Here is an example of georeferencing: if we had a scanned image of an 1880 map of Lot 46 in Prince Edward Island and a current map with accurate coordinates of Lot 46 in Prince Edward Island, we could assign accurate coordinates to the former by digitally aligning its geographic features with those of the latter. Once this georeferencing process was complete, users could analyze the resulting raster; any data they gained from the raster, such as the number of houses in Lot 46 in 1880, would be considered vector data. 

The remainder of this guide will walk you through the process of georeferencing.

Table of Contents

Before We Begin: Downloading the Lesson Files
Step 1: Saving a New Project File in ArcGIS Pro
Step 2: Adding Reference Data
Step 3: Adding Digital Images of Maps (AKA Rasters)
Step 4: Georeferencing
Step 5: Finalizing the Georeferenced Raster Image

Before We Begin: Downloading the Lesson Files

An 1880 map of Lot 46 in Prince Edward Island will be used throughout this guide as an example. If you would like to follow along with the steps in this guide and georeference this lot as well, please download this zipped folder. It contains the entire suite of folders and files used in this guide, including the digital image of the 1880 map of Lot 46. By downloading this zipped file, you will have access to the same files that are used in this guide, and your folder structure will be identical to the one contained in this guide. All of this should allow you to follow along with this guide more easily. Note: once downloaded, make sure to unzip this folder before proceeding.

Step 1: Saving a New Project File in ArcGIS Pro

  • Open ArcGIS Pro

At the first screen that appears, let’s create a new project file. 

  • Under the New heading, click Map
  • In the “Create a New Project” dialogue box that appears, provide a suitable name for your project, save your project in the ProjectFiles folder, and then click OK.

Step 2: Adding Reference Data

ArcGIS already provides us with a base reference map, which it calls “World Topographic Map.” We will use this map in conjunction with other reference data, namely layers that show the lot boundaries and road network of Prince Edward Island.

Adding PEI Lot and Road Network Layers

  • Under the Map tab and in the Layer group, click Add Data. This will bring up a file explorer, which you may use to navigate to the location in which you have your reference data saved. In our case, we will navigate to Lesson2_ArcGISPro\ReferenceData.
  • While holding the Ctrl key on your keyboard, select the files called Lot_Township_Lot46clip.shp and National_Road_Network_PEI_Lot46clip.shp. 
  • Click OK.

Note: for the full dataset containing all of the PEI townships go to the province’s Open Data portal.

Symbolizing Layers

Once the reference data has been added, we can change their look in order to make them easier to work with. To change the look of the Lot_Township layer, 

  • Click the pink square directly underneath the name of the layer in the Table of Contents. This will open the Symbology pane to the right of your screen.
  • In the Symbology pane, you have multiple options to choose from; for this demonstration, we will select the Black Outline (2 pts).
  • Once you select the Black Outline (2 pts), click the Properties tab.
  • In the Properties tab, click the Outline color dropdown menu and select Electron Gold
  • Click Apply at the bottom of the pane.

To change the look of the modern roads layer, 

  • Click the blue line directly underneath the name of the layer in the Table of Contents. As before, this will open the Symbology pane to the right of your screen. 
  • In the Symbology pane, click the Properties tab. 
  • In the Properties tab, click the Color dropdown menu and select Anemone Violet.
  • Click Apply at the bottom of the pane.

Step 3: Adding Digital Images of Maps (AKA Rasters)

Once the reference data has been added to your project and symbolized, you can add the digital images of maps, which are also known as rasters. The latter will be aligned with the reference data in subsequent steps.

We will add to our ArcGIS project the raster of the map of Lot 46 in Prince Edward Island from Meacham’s 1880 Atlas. We will add the raster in its original format, which in this case is .JPG. We will follow similar steps as in Section 2 to do so:  

  • Under the Map tab and in the Layer group, click Add Data. This will bring up a file explorer, which you may use to navigate to the location in which you have your raster saved. In our case, we will navigate to Lesson2_ArcGISPro\Rasters.  
  • Double-click the file called 1880_Lot46.jpg to add it to your project.

You will now see the raster show up as its own layer in the Table of Contents pane to the left-hand side of the ArcGIS window.

Step 4: Georeferencing

Preparing to Georeference

There are a couple things that we need to do before georeferencing the raster we just added. 

  • In the Contents pane, select the raster layer that you have just added (so that it is highlighted blue). Also, ensure that the raster layer is “on” (i.e., there should be a check mark visible). 

At this point, your raster will still not be visible on your screen, even if it is “on.” This is because you have not yet assigned any coordinates to the raster layer, so ArcGIS does not know where to place it in relation to your reference map. The process of adding those coordinates is central to the georeferencing process, and it is what we will do next.

Before we add specific coordinates to the raster, we first need to navigate to the geographic area of the reference map that roughly lines up with the geographic area of the raster we wish to georeference. In our case, we want to navigate to the area on the reference map around Lot 46. To do this, 

  • Pan to the geographic area on the reference map that corresponds to the geographic area of your raster (Lot 46 for us). (To do this, ensure that the Explore option is selected within the Map tab’s Navigate sub-group. With Explore selected, your cursor will be shaped like a hand and you may pan around the reference map.)    
  • Click the Imagery tab.  
  • Within the Imagery tab, click Georeference, which can be found in the Alignment sub-group.

Once you click Georeference, a new tab called “Georeference” appears to the left of the “Map” tab. 

  • Within the Georeference tab, click Fit to Display

Now that we have clicked Fit to Display, the raster is visible. The raster will not line up exactly with the reference map at this point, as we have not added any control points, but its position roughly corresponds with the geographic area of the reference map.

To further optimize the location of the raster in relation to its reference map, we can also use three more options under Prepare in the Georeference tab: Move, Scale, and Rotate. The Move tool can be used to move the raster image around your display to better fit your reference map. Scale can be used to resize your raster in a similar way to resizing an image in Google Docs. The user can manipulate the 4 corners of the raster to adjust the size of the raster to appropriately fit its reference map. Finally, the Rotate tool can be used to rotate the image around its center axis to better align with its reference map. Keep in mind that these tools are only to be used to prepare the raster for georeferencing—they are not a substitute for the georeferencing process itself. There are still inaccuracies inherent in such things as the cartography of the raster to the point that it is not geographically accurate, which georeferencing helps rectify.

Adding Control Points

Now that we have roughly aligned the raster with the reference map, we will add control points. Adding control points is the means by which users precisely align the raster map with the reference map—it is the crucial step by which a raster is georeferenced. 

A “control point” is actually made up of two sub-points—one that users place on the raster map and one that they place on the reference map. Doing this tells ArcGIS that these two points represent the same geographic location. Once ArcGIS understands this, it will automatically and immediately re-align the raster image so that the control sub-point you placed on it is positioned directly above the corresponding control sub-point you placed on the reference map. After the user has added a few control points, the raster and reference maps should align very accurately.

Before Adding Control Points

To make things easier when inserting control sub-points, we can utilise a feature called “Snapping.”

  • In the Edit tab and the Snapping group, ensure that the Snapping option is on.
  • Click the small “List By Snapping” icon in the Table of Contents. 
  • Ensure the National_Road_Network and Lot_Township layers are “on” (i.e., ensure that they have a checkmark beside them). 
  • Return to the default Table of Contents view by clicking on the “List By Drawing Order” button, which is found to the left of the “List By Snapping” icon.

Adding Control Points

Now we can begin to insert control sub-points.

  • First, you need to visually identify the geographic features that are clearly common between the raster data and the reference data. 

These features are often political borders and boundaries. Sometimes, as in our case, additional reference layers besides the reference map need to be used in order to identify commonalities between the raster and the reference data. In our case, we have used a reference layer that has mapped the roads of Prince Edward Island. We can identify where the placement of the intersection of roads is consistent between the raster map and the reference layer containing the modern roads.  

I have decided that a point in the south-east corner of the raster at which a road meets the border of the township is a good place to put a control point.

Once this decision is made, you can begin to add control points. To do so, 

  • In the Georeference tab and the Adjust sub-group, click Add Control Points.

Note: After clicking “Add Control Points,” your cursor will now have changed from a hand to a crosshairs. This change in cursor means that you can no longer pan around the map by dragging your cursor across the screen, as doing so will now insert a control sub-point instead. Rather, once your cursor has changed to a crosshairs, you must now click and hold your mouse’s scroll wheel in order to pan around the map.

  • Align your cursor—the crosshairs—with the location on the raster of the historical map at which you would like to place a control sub-point. Click to place this first control sub-point. 
    • Note: always place the first control sub-point on the raster of the historical map.
    • Suggestions: try to place your control sub-points on the political boundaries or borders of your historical maps, if possible. Try also to place your control points at the intersection of two roads or the intersection of a road and a political boundary, like we are doing here. Moreover, it is advisable to place your control points as close to the corners of the raster of the historical map as possible. For instance, if you are able to place a control point on each of the four corners of the historical image, your historical image will quickly acquire an accurate orientation in relation to the reference map.

After you place your first control sub-point, a red square will appear. 

Before placing the second control sub-point, it is sometimes—but not always—necessary to temporarily turn “off” the raster of the historical map (i.e., ensure there is no checkmark beside it) so that it does not obscure the reference data. To turn “off” the raster of the historical data, 

  • Clear the checkmark beside the name of the historical map raster in the Table of Contents. 

You are now ready to place the second (of two) control sub-points. The second sub-point can be placed on the reference data—either the reference map layer (i.e., the World Topographic Map) or the modern-roads layer. In this demonstration, we will place the second control point on the modern-roads layer. To place the second control sub-point, 

  • Click the location on the modern-roads layer that corresponds to the location at which you placed your first control sub-point on the raster of the historical map.
    • Caution: be aware of places that have changed between the time your historical map was created and the time your reference map was created.

Once the second control sub-point has been placed, the location of the control point will be denoted by a fixed, encircled crosshairs, or, sometimes, a “B.” This symbol, whether a crosshairs or a “B,” will have two colours to it: the bottom colour will be green and correspond to the first control sub-point you placed, while the top colour will be red and correspond to the second control sub-point you placed. 

As mentioned before, as soon as the second control sub-point has been placed, ArcGIS automatically and immediately re-aligns the raster image so that the control sub-point you placed on it is positioned directly above the corresponding control sub-point you placed on the reference map. This automatic re-alignment is what happened when we placed the first set of control points:   

If you had to turn “off” the historical raster map to place the second control sub-point, you should now turn that layer back “on.” To do so, 

  • Click the now-empty box beside the name of the historical map raster in the Table of Contents. A checkmark will appear in this box.

With the historical raster map once again visible, you can now assess the first control point you placed. If you decide that it is unsatisfactory, you can remove the point and try again. To remove a placed control point, 

  • In the Georeference tab and the Review group, click Select
    • This will allow you to select any unsatisfactory control points. 
  • Select the unsatisfactory control point. 
    • After it is clicked, it will turn into a green square.
  • With the control point selected, click Delete in the Georeference tab and the Review group.

You can now repeat the steps in the “Adding Control Points” section until your raster is completely georeferenced. Generally speaking, in order to completely georeference a raster of a historical map, you should add a minimum of four and a maximum of ten control points. Four control points—especially if placed in the four corners of the historical map—can quickly orient the raster into the correct position in relation to the reference map. Although you may place more than ten control points, after you have placed ten, any further points you place will have an increasingly diminished effect on the orientation of the historical map in relation to the reference map. Lot 46 on Prince Edward Island can be georeferenced fairly accurately with four control points, one on each of the four corners:

If, during the process of adding control points, your work goes totally askew, you can erase all of your control points and restart. To do so, 

  • In the Georeference tab and the Adjust group, click Reset
  • In the Georeference tab and the Prepare group, click Fit to Display.

Assessing the Accuracy of Control Points

Once you have added a sufficient number of control points, it is time to assess their accuracy. There are two main ways of assessing the accuracy of control points: the residual error and the root mean square (RMS) error. The former regards the accuracy of individual control points while the latter concerns the accuracy of the entire collection of control points.

To view the residual error

  • In the Georeference tab and the Review group, click Control Point Table. The residual error is listed in the column farthest to the right.

We can use the residual error to assess how close the control sub-point on the raster is to its corresponding control sub-point on the reference data. Ideally, we want those two sub-points to line up as closely as possible because that means that our raster and reference data are lined up well. The residual error is expressed as a number; the higher the number, the greater the disparity between the location of the sub-point on the raster and the location of the sub-point on the reference data. 

To view the root mean square (RMS) error, look at the georeferencing-status box that is placed over the top of your view of your reference and raster maps; you will see Total RMS Errors. The RMS error with which we are concerned the most is the Forward error. The RMS forward error measures the overall error of all of your control sub-points combined. In our case, it is expressed in metres. Thus, if our RMS error is ~1000, we know that our control sub-points on the raster are, on average, about a kilometre away from their corresponding control sub-points on the reference data. When assessing your RMS error, it is important to keep scale in mind. If you are dealing with maps of large areas (like a province or country), then the large scale of your project renders a 1-km RMS error acceptable. But if you are dealing with maps of small areas (such as a city or neighbourhood), then the small scale of your project demands a lower RMS error value for the sake of accuracy.

Transformation and Control Points

Transformation is the equation by which ArcGIS “transforms” the raster to its real map coordinates. Every time you georeference a raster, it goes through a transformation to reach its real-world coordinates. We can use the various transformation options to increase or decrease the residual and RMS error values.

There are several types of transformations that ArcGIS uses to place the raster dataset accurately onto its reference map. The most common type of transformation used is a polynomial transformation, which uses a polynomial equation to transform raster data. Here is a general run-down of the transformations you will have available in ArcGIS:

  1. Polynomial Transformations: Used to derive georeferencing formulas under a “least-squares-fitting” (LSF) algorithm. The LSF method essentially applies a formula to all control points. This formula will cause local distortion (i.e., a higher residual error between your control sub-points), but it will increase the overall accuracy of the raster’s placement. Thus, these transformations are useful for georeferencing for global accuracy, but your raster may not align perfectly at smaller, localized scales. There are several “orders” of formulas derived by ArcGIS Pro, based on what distortion your raster needs relative to its reference image. Zero- and First-Order polynomials will stretch, rotate, move, and scale your image; straight-lines remain straight but the angles between these lines will change. At higher orders, those straight lines will begin to bend and curve.

    There are several types of polynomial transformations. Zero-Order Polynomial uses only 1 control point. Typically, this transformation is used when the raster is already georeferenced but needs a slight shift to more accurately align with its reference map. First-Order Polynomial (Affine): after selecting this transformation, any straight lines in the raster will remain straight after georeferencing, but the angles between them will change. Most commonly, this transformation is used in georeferencing images. It requires 3 control points. Beyond 3 control points, residual error begins to increase. Although adding more control points will increase your residual error, the overall accuracy of the transformation will increase as well. Second-Order Polynomial requires 6 control points and introduces distortion of straight lines, while Third-Order Polynomial requires 10 control points and increases the distortion of straight lines. Similarity is a first-order transformation that requires 3 control points. Its algorithm attempts to preserve the general shape of the image. Similarity will produce a higher residual error, as it prioritizes the shape of the image rather than its alignment.
  2. Adjust: Requires 3 control points. Adjust optimizes between global and local accuracy. Its algorithm uses a polynomial transformation, and then further optimizes the raster using “triangulated-irregular-network” interpolation. 
  3. Projective: Requires 4 control points. Projective transformation is useful for the georeferencing of maps and imagery where there is a need for accurate local data without distortion of straight lines. Projective formulas may warp bent lines to remain straight, but in doing so removing their parallelism. 
  4. Spline: Requires 10 control points. Spline optimizes local accuracy, and will have 0 RMS error as control points are transformed exactly to their targets. However, parts of the map farther from the control points may become distorted and be less accurate.

For more information about transformations and how they work, check out these two articles on the ArcGIS website:
Overview of georeferencing—ArcGIS Pro | Documentation
About spatial adjustment transformations—Help | ArcGIS for Desktop 

As you can see, the number of control points we place determines the transformation we can use. The more control points added, the more complex the transformation that can be applied. However, more control points will not necessarily equal more accuracy. Control points should also be evenly spread throughout the raster, as concentrating control points in one area would heavily bias the transformation. We placed four control points in Lot 46—one on each corner. This number of control points will allow us to use most of the transformation options, and we placed them well enough apart to avoid the transformation being biased toward one area.

Now that you know more about the types of transformations, let’s apply one to our project:

  • In the Georeference tab and the Adjust group, click Transformation

From the dropdown menu that appears, try selecting different transformation options to see which one results in the most accurate georeference of Lot 46. See if you can notice the effect that each transformation option has on the residual and RMS errors.

Step 5: Finalizing the Georeferenced Raster Image

Saving Control Points

Once you have completed adding all of your control points, you can save them. This is important because, should you misplace your finalized raster of your historical map, you could still import the original raster and import the corresponding saved control points to create another finalized raster. To save your control points, 

  • In the Georeference tab and the Save group, click Export Control Points
  • Navigate to the ControlPoints folder, name the file, and click Save


Saving Raster as Rectified Image

With the control points saved, we will now save the raster of the historical map as a “rectified” image. This creates a new version of the historical map raster file. This new, rectified version of the historical map file contains the coordinate data that we just added through the placement of control points. This is very useful because it means that you can import the rectified image into any ArcGIS Pro project, and, provided that it has the same Spatial Reference (see below), ArcGIS Pro will know exactly where to place the image. That is, the rectified image will automatically be placed in its correct geographic location on the reference map. To save the georeferenced raster image as a new rectified image file, 

  • In the Georeference tab and the Save group, click Save as New. A pane called Export Raster will appear to the right of the screen.

In the Export Raster Pane, 

  • Next to the Output Raster Dataset field, click Browse… and locate the “Rectified” folder. You can then provide an appropriate name for your file and click Save.
  • Next to the Coordinate System field, click the Spatial Reference icon. In the window that appears, type 2954 in the search box and press Enter on your computer’s keyboard. In the results list of this search, select NAD 1983 CSRS Prince Edward Island and click OK. 
  • Please note that you may also change the Output Format and the Compression Type of the rectified image in the Export Raster pane. For this demonstration we will change the format to TIFF and change the Compression Type to LZW
  • At the bottom of the Export Raster Pane, click Export

The exporting process may take a moment to complete; its status will be shown at the bottom of the Export Raster Pane. Once the exporting process is complete, the rectified image will be automatically and immediately added to your Table of Contents. (If the rectified image does not appear automatically after the exporting process is done, you can add the rectified image to your project file in the same manner as you originally added the raster image. That is, by clicking Add Data in the Layer group of the Map tab.)

Once the rectified image appears in your Table of Contents, you can rename it in order to provide it a suitable name, and you can also drag it to a suitable location in the Table of Contents (such as to a location within a parent folder). Also, you may wish to remove the original raster, as we just replaced it with the rectified image. To remove the original raster, right-click it in the Table of Contents and then click Remove

Completing the Georeferencing Process 

  • Click the Save icon in the Quick Access Toolbar at the top of your screen. Alternatively, you may press Ctrl + S on your computer’s keyboard. 

And that completes the georeferencing process.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: