Differential GPS (DGPS) is required to obtain accurate GPS positions. Without utilizing differential techniques, either in real-time or post-processing, GPS is generally considered accurate to only within about 100 meters. This is due to the combined effect of many error sources, however, through differential processing, most of these errors can be removed. This month we shall examine GPS error sources to address which of these errors can be removed with differential techniques and which cannot.

"...we have so much in common!"
In order to obtain GPS differential corrections you always require at least one GPS receiver that is stationary at a location of known coordinates. This receiver is often referred to as a base station or a reference receiver. Because this receiver is at a known location, it is able to compute the error associated with every satellite that it tracks. These corrections can then be applied to any near-by roving GPS receivers.
     The underlying premise of differential GPS is that any two receivers that are relatively close together will experience similar errors. By having one receiver at the known location, all errors that are common to both receivers can be entirely removed. For the most part, this works pretty well. The errors that are common to both receivers are removed entirely. The problem is that there are a few error sources that are not common between both receivers and therefore are not removed by differential correction. Let's look at the most common and most significant errors and determine which errors are correctable.

What are the error sources? Strictly for my own convenience, I will divide the GPS error sources into three categories.
• Trivial Errors
• Shared Errors
• Unique Errors
      The first category, "Trivial Errors," are those that are typically not significantly large in the context of GIS data capture. This assumes that the users are working in a relatively limited geographic area, say within a hundred kilometers of the GPS reference receiver. The other assumption is that the roving receiver(s) are being used to collect GPS positions with a spatial accuracy ranging from sub-meter to perhaps as much as 5 meters. In this situation, error sources that typically contribute only a few centimeters of error will usually not have a significant impact on the results.
      The second category, "Shared Errors," are those which are common to all GPS receivers in a limited geographic region. These errors are generally not a problem as they are entirely removed by differential correction.
      The third category, "Unique Errors," are those which are unique to an individual receiver. These are often induced by some local ambient condition such as nearby tall buildings or surrounding foliage that reflects and/or obstructs the satellite signal to the receiver. In this case the resulting error is unique to that receiver and therefore cannot be removed by differential correction techniques. Because these errors are more difficult to correct, the more detailed discussions that follow will focus on how to avoid and/or correct these errors.
      The table on the following page lists the most common GPS error sources along with a categorization and description. It may be interesting to note that Selective Availability (S/A) appears twice in the table below. This is because S/A can be induced in two ways. The U.S. Department of Defense (DoD) can either degrade the satellite clock, or they can degrade the satellite orbital parameters. The degradation of the satellite clock is referred to as dither, and the degradation of the satellite orbital parameters is referred to as epsilon.
      It is possible to implement both types of S/A (dither and epsilon) at the same time, and it is also possible to alter the severity of either on a frequent basis. At any given time, the severity and type of S/A that is implemented is not public information.
      Historically, and as of this writing, S/A is usually limited to clock dither. The epsilon aspect of S/A appears to be implemented only occasionally. This is fortunate in that differential techniques can easily remove errors induced by clock dither, but not necessarily for errors induced by epsilon. Be warned, however, that the U.S. DoD can, and does, implement and vary both forms of S/A, on a continual basis.

Dealing with epsilon
There is no effective way to recognize the presence of epsilon while collecting data in the field. S/A is on continuously, and without differential correction, your position will continuously wander within about 100 meters of truth. However, although it's very clear that S/A is turned on, you have no way to tell whether it is due to epsilon or clock dither. The nature of epsilon is that it will have little impact on data that was collected when the base and rover receivers were very close together. The greater the distance between the base and the rover(s), the greater the potential for error.
     Again, I want to stress that historically the epsilon aspect of S/A has rarely been implemented by the U.S. DoD.
     A single base station cannot be used to remove the effects of epsilon. However, a widely distributed network of base stations can be used to generate corrections that remove the epsilon factor. The use of several GPS reference stations to generate correction data that are accurate over a wide geographic region is known as wide area DGPS. There are several companies that offer wide area DGPS services. These services generally use real-time DGPS to provide accurate navigation in areas such as the North Sea.

Dealing with multipath
Multipath is probably one of the most common GPS error sources. Multipath errors are due to the GPS receiver receiving a reflected signal in addition to the same signal that travels a direct path to the GPS antenna. This interference can result in position degradation that ranges from a mild wandering of a few meters to wild position jumps of dozens of meters.
     Multipath is generally a highly localized phenomena. Typically it is the roving GPS receiver that is carried in and out of areas where multipath signals are present. This results in the rover receiving erroneous signals that the base station does not receive. Since these errors are not common to both the base and rover receivers multipath is virtually never removed by differential correction.
      Fortunately several things can be done to avoid multipath problems. The most obvious solution is simple avoidance. When possible, avoid collection GPS data near large, reflective surfaces. A few of the most common multipath sources include large steel and glass buildings, automobiles, and large bodies of water.
      Of course, it is not always possible to just avoid the areas that have multipath potential. Receivers vary widely in their resistance to accepting multipath signals. The best advice for the urban GPS user is to test any receiver in your own environment before you purchase. Some receivers are equipped with software/firmware that can recognize when multipath is present (and in the more sophisticated models, reduce or eliminate those signals). A hardware feature that is available in other models are antennas that are equipped with a ground plane to reduce the incidence of multipath on the antenna.
      From the user's end there are a couple of things that the typical user can do to minimize the impact of multipath signals on the data. A well designed GPS system will allow the user to set the minimally acceptable satellite elevation and signal strength. My recommendation is that a roving receiver should never be allowed to use satellites that are less than 15 degrees above the horizon. Such low elevation satellites are much more prone to reflected signals. In addition, the signal from a low elevation satellite is also more subject to propagation errors since the signal travels through more atmosphere on the way to Earth compared to a satellite overhead.
      When a GPS receiver receives multipath signals the signal strength is usually significantly lower than a comparable direct signal. A well designed receiver will also allow you to set the minimally acceptable signal strength. When the user sets a minimum signal strength mask, the weaker signals are automatically rejected as opposed to accepting all signals equally.
     I hope these thoughts help in your selection of the best GPS system for your application.

About the author:
Chuck Gilbert has over a decade of experience as a GPS user. He has been employed as an applications engineer for Trimble Navigation since 1989. If you have a suggestion or request for a future article, please drop a line to Chuck care of Earth Observation Magazine, 13791 E. Rice Place, Suite 204, Aurora, CO 80015, or fax to (303) 690-2522.

Back