• news
  • events
  • admissions
  • academic programs
  • professional development
  • people
  • research & publications
  • inside the gsd
  • home
  [an error occurred while processing this directive]   Computer Resources GIS Manual  

Harnessing Projections in ArcMap: A tutorial

Geometric datasets record relative positions of entitites and their vertices. A coordinate referencing system may be completely arbitrary, as is often the case with images and CAD files. Many data sources, however, use geodetic and projected coordinate systems whos referenced locations may be associated with specific locations on the earth (in the atmosphere, in space, or on other planets.) These coordinate referencing systems are one of the most important aspects of Geographic Information Systems that allow representations in different datasets to be associated with one another.

This tutorial is designed to provide hands-on experience with the various issues of working with projections in ArcMap. It is important that educated users of GIS understand how to control the projection of their maps for a couple of reasons: First, accepting the default display projection that ArcMap provides will often result in maps that are ridiculously distorted at any scale. Second, when bringing together data from different sources, it is sometimes impossible to get layers to align without first defining projection properties for one or more layers.

This tutorial is concerned with GIS layers that use non-arbitrary geographic or projected coordinate referencing systems. The problems of georeferencing images and CAD files that use arbitrary coordinate systems are covered in another tutorial.

This tutorial builds on the ideas and skills and data resources explored in the followng pages in the GSD GIS Manual.

For more in-depth reading on these issues, download the following software use guides for ArcGIS:

Sample Dataset

Right-Click Here to download the sample datsaset. Expand the zip archive to your personal folder in the directory c:\temp\yourname.

A Preview of the Major Issues to Understand

  • We will always encounter data files that record geometry with a variety of different coordinate systems.
  • GIS tools are capable of transforming the geometry of data layers of various coordinate referencing systems (CRS) on-the-fly.
  • What coordinate framework is used to display data on a map is a critical decision based on the location and scale of your study. You, the map maker must know how to choose an apropriate projection.
  • The projection is a critical aspect of any portrayal and should be explicitly specified on the map!
  • Sometimes datasets are not associated with information that identifies the proerties of the coordinate referencing system inherent in its geometry. Software is not able to project thezse layers until we add information about the inherent CRS inherent in the data.

Understand the Coordinate System Properties of the ArcMap Data Frame

One of the great capabilities of a GIS like ArcMap is the abbility to associate locations referenced in independent datasets -- even if these datasets use different coordinate referencing systems. This is accomplished by transforming references gibven in one system, to aanother corrdinate refereincing system on-the-fly. However, ArcMap does not always do a good job of choosing what coordinate referencing system to use for a given location and purpose. This is why we, the users, need to understand how all of this works. First is to explore the properties of the ArcMap dataframe to see where the target coordinate system can be set. And then, to observe how ArcMap sets this target coordinate system -- sometimes counter to our best interests!


The Dataframe inherits the CRS of the first dataset opened

  1. Start ArcMap with a new empty document
  2. Check the coordinate system properties of the dataframe. Undefined. Right?
  3. Now open the datasets from the ESRI folder of our tutorial data compilation.
  4. Now look at the CRS properties of your dataframe. It is now set to GCS (Geographic Coordinate System) with the earth model assumotions of the North American Datum of 1983. Where did this come from?
  5. The ArcMap dataframe displays data using the CRS of the first layer that is added.
  6. Check the properties of some of the datasets. Aha! Now you see how ArcMap gets its CRS.

Observe on-the-fly Transformation

So ArcMap can read the CRS properties of a Layer and adopt these for its display. Great. Now lets open another layer that happens to use a different CRS for its cooridinate references. We will see that ArcMap will transform these coordinates to align with other data.

  1. Open the Coastline.shp dataset from the rigis folder of our tutorial dataset.
  2. Observe that it seems to align in a logical fashion with our streets.
  3. Check the Source -> Coordinate System Properties of the coastlines dataset.
  4. So the CRS of the coastline and the Local Streets are different, but ArcGIS is making them align!
  5. Go to the data frame properties, and see that the map is still being displayed in a Geographic Coordinate System.
  6. Now, for fun, Clear the coordinate system property of the data frame and, in the General tab, set the Units to Unknown.
  7. Now note that the layers don;t line up any more. Use Zoom to Layer to go to the Coastline dataset, and look at the coordinates in the lower right corner of the map window. Do the same thing on your Streets layer.
  8. Now, set the projection back to GCS.

Choose an Appropriate Coordinate System for Displaying your Data

As explained in Concepts of Geographic Coordinate Referencing Systems, GCS use latitude and longitude as regular cartesian coordinates, which severely misrepresents distance and size relationships and is very embarrassing to the cartographer whether he/she knows it or not. So let's over-ride ArcMap's choice of GCS for our map display and choose another projection. The shape files, utm_zns and st_plane_zns in your sample dataset will tell you what local projection cases will work best for our study area. For the sake of this tutorial, we will project our map to the special case of the Transverse Mercator Projection, identified as Universal Transverse Mercatoe Zone 19.


Choose an Appropriate Projection

  1. Zoom out to the entire united states.
  2. Note that Geographic Coordinate System (GCS) used in this displ;ay uses latitude and longitude as cartesian coordinates. Whats wrong with that?
  3. Observe that the USA looks all stretched out. COmpare the area of Montana with that of texas. Is this a true representation?
  4. Zoom to your local streets layer.
  5. Use your Drawing Tools to create a circle on the map, centered on a point on the shore line, and double-click the circle and make it transparent.
  6. Measure it with the measure tool. It makes no sense to measure length in degrees (at least almost nobody can imagine what this means in terms of actual length on the ground.) Change the units of your measure tool to kilometers and measure it again.
  7. When is a circle not a circle? When your map is distorted in a very significant way!!!
  8. Lets choose a projection that is portrays the united states for true area comparison. Reproject your dataframe for a projected coordinate system for continental north america that preserves area.
  9. Zoom out to the world country boundaries.
  10. Note that this projection does not work well for China.
  11. Zoom to your local streets layer and note that streets that should be running north south are not represented well, nor is anything else.
  12. Project your data into the official coordinate system of rhode island: Rhode Island State Plane Coordinates. Observe the various properties we have discussed -- Shape, Distance, Direction. This projection seems to do a good job with shape direction and distance. Right?
  13. Now project your data into the apropriate UTM zone. You can use the maps of UTM zones and State Plane ZOnes to figure out what the right zone is. Not bad, either, right? Take a look at the projection details to see what the actual projection parameters that make up the particular case of the Transverse Mercator projection that we call UTM 19.

Observe a Problem with Automatic Transformation

Norbert Wiener once observed that a good strategy for understanding the way a system works is to study the ways that it fails. So here we go. We will open the aerial photo from the Rhode Island GIS. Although we got this image from the State GIS Agency of Rhode Island, It does not have the special sort of metadata that arcMap can understand. This is why when it opens we get a message that says Coordinate System Not Defined. This problem is common, not only with georeferenced images but also with vector datasets, because the idea of embedding CRS properties with a dataset are very new (since about 2002.) Datasets that don't originate from ArcMap version 8.2 or later need to have their inherent coordinate systems identified using ArcCatalog before they are subject to on-the-fly transformation.

Trouble-Shooting Mis-Aligned Layers

  1. Add the image file 3426.sid to your map document.
  2. You may get an error reflecting that this layer does not have its coordinate system defined.
  3. In fact, the image does add, but it dosen't seem to align with the other data on our map. This is a problem that you will become comfortable if you don't completely give up on GIS.
  4. Where is the image? Right-Click on it and Zoom to Layer. There it is!
  5. the Image
  6. Check the Source Properties of the image layer, and take a look at its projection properties. Undefined!
  7. While we have image properties open, also observe the Extent in the Source Properties of the dataset.

Guessing a Coordinate Referencing System

The problem with our new layer is that it does not have its coordinate system identified in a way that ArcGIS recognizes. In fact, if we look at the metadata that was provided by Rhode Island GIS, the CRS is identified there, but lets pretend for a minute that we don't have any information other than what we can observe and infer from our understanding of projection systems.


Educated Guesses about a Layer's CRS

The critical aspects of a CRS that need to be correctly identified are discussed in Understanding Geographic Coordinate Systems. They are:

  • The Projection Method, or unprojected Geographic (lat/Lon) Coordinate System
  • the earth model
  • The Case of the projection
  • The Coordinate Units

The CRS used in a particular datsaet is chosen by the creators of the data based on the location, the extent, the purpose and the political context of the data collection activity. It is always better to find the official metadata for a dataset to ascertain the particulars of its CRS, but in this primitive state of data sharing in which we find ourselves today, this metadata is often lacking and we have to guess. The following logic can be helpful for guessing the coordinate system of a map:

  • Images and CAD files may have a completely arbitrary coordinate system. A non-georeferenced image will have its orign at the upper left hand corner, and its units will be one-per-pixel. CAD datsest often have their origin at the bottom left corner of the page and may have units of page inches (silly!). In this case, you should go find the documentation on georeferencing images and CAD data.
  • Geographic datasets that cover a very broad area (larger than a state) are likely to use unprojected latitude and longitude (GCS) for their coordinates. In this case, the units are almost always Decimal Degrees.
  • If a dataset is worldwide and created after 1990 or so, the earth model is likely to be the World Geodetic Spheroid of 1984.
  • If a dataset is national or continental in scope, the earth model is likely to be a local datum, such as the North American Datum of 1983 (NAD83) or rarely, NAD27. Other countries have their own national earth models.
  • Datasets that are collected by state GIS agencies or local governments within a state typically use that state's state plane coordinate systyem. State Plane systems have units of either Feet, or Meters.
  • Datasets collected as part of a national or international map series of a large scale (detailed) are likely to use the apropriate case defined in the Universal Transverse Mercator system. UTM coordinates are almost always meters.

Explore the Inherent CRS of an Undeclared Layer

  • Zoom to layer on the image.
  • Move your mouse to the upper left corner and observe the coordinates at the bottom right of your map. This is a georeferenced image, otherwise the coordinates at the upper left corner would be 0,0.
  • Zoom in to a residential street, and measure its width with the measure tool. As we already noted, even though arcmap says that these units are meters, it actually doesn't know this, and is, in fact confused. A residential street is between 30 and 60 feet wide, and 30-60 meters does not make sense. These units are feet.
  • Since the units are feet, this means that the data are most likely to be using projection method and case defined by the the local state plane coordinate system.
  • This image was collected in 1995, therfore we presume the earth model assumptions are those of the North American Datum of 1983.
  • To confirm this guess, lets have ArcMap transform our other layers to Rhode Island State Plane Feet on NAD83.
  • Since this gives an apparent match, we can rest assured (somewhat) that our logic is correct, though it is nice that we have good metadata in this case.

Correctly Assigning the CRS Properties of A Dataset

So we have figured out that if all of our other layers are transformed to Rhode Island State Plane Feet then our image file will associate logically with them. This is good, but if we properly inform ArcMap of this image's CRS propertiues, then ArcMap can display our data together in any CRS we want. This principle is the same with other sorts of datasets like shape files. This is not necessary merely for display purpises, but also many of the associative capabilities of Arcmap will fail if the arcmap CRS properties are not set up correctly. m So now we will use arccatalog to update the properties of our dataset.


There is One Right Answer!

One of the most common misunderdstandings of the issue of identifying a datasets inherent CRS so that ArcMap can transform it for display is that people often enter the wrong CRS properties for the dataset and the result not only causes the layer to fly into hyperspace, but is also very confusing to people who encounter it later. Correctly identifying the properties of the coordinates that are inherent in the vertices and pixels in a dataset is not a matter of stating what you want them to be! There is one right answer to this question. This is why we performed the experiment in the example above to confirm our guess before we actually update the properties of the data.

  1. Open the Toolbox panel by clicking the little red toolbox.
  2. Go to Data Management Tools and open Projections and Transformations.
  3. Open the Define Projection Tool, and use it to set the projection properties of the 3426.sid image.
  4. Take a look at the files in the rgis folder. Note that there is a new xml file in there. This is the machine-readable metadata that has been created for your image.
  5. You may have to remove the image from the map and reopen it to see the effect of this change.
  6. The real test is to reproject the dataframe to a CRS that is not the native one for your layer.
  7. Reproject your dataset to a projected coordinate system for North America. Experiment!


Now that you understand how ArcGIS deals with Coordinate Referencing Systems, you should be able to responsibly choose an appropriate coordinate system for your maps and to troubleshoot problems with datasets you find that do not have ArcMap-compatible metadata for their inherent CRS properties.