The GPS Consumer Series is a monthly column that explores the issues associated with GPS data collection. This column explores the benefits provided by various GPS receiver features on today's market. Issues commonly encountered in differential GPS data capture are examined from the user's perspective.

Introduction
A single GPS receiver, run autonomously, is subjected to several sources of error. (Autonomously means: a solitary receiver collecting data with no contact or input from other receivers collecting data at the same time.) All GPS receivers will be impacted by the combination of these errors, resulting in accuracy that ranges from dozens to hundreds of meters. This is true for all receivers, regardless of manufacturer or receiver type.
      Only by obtaining data from two receivers, and by using a process known as differential correction (DGPS), can much of this error be corrected. The two most basic requirements to perform differential correction are that both receivers must collect data simultaneously, and that one of the receivers must be stationary at a known location. Utilizing differential GPS can improve your accuracy from the 100 meters typical of uncorrected data to better than a centimeter, depending on the type of equipment used. DGPS is based upon the principle that at a receiver at a known location can compute the errors associated with each satellite. These errors are converted into correction terms and applied to a roving receiver(s).
      This principle is true for all error sources that are common to both receivers. Fortunately the largest errors are very similar between any two receivers within a few hundred kilometers of each other.

Are there alternatives to DGPS?
Can higher accuracy, say 1 meter or better, be achieved without DGPS? The short answer is "No." Last month this column examined the averaging of GPS positions to overcome these errors. The result is that hundreds of hours of data must be averaged to reliably achieve even 2 meter accuracy. In addition, the rate at which averaging improves accuracy is not very predictable. Therefore, the majority of users collecting GPS data for the population or update of a GIS will require a source of base data so that they can perform a differential correction.
      There are two ways to obtain GPS base data for differential correction. Either you obtain your base station data from an outside source, such as a commercial vendor or a government agency, or you supply your own base data by owning multiple GPS receivers. Either way, this is generally not a significant problem for GPS users in many urban areas. A wide variety of commercial and government differential correction sources are available in many cities of the world. Even if an urban user chooses to purchase a base station of their own, it will usually be easy to have the new base station surveyed in and tied to local geodetic control.

Out in the Boondocks
In the process of speaking with GPS users I have encountered users deep in the outback of Australia, in the jungles of Congo, tracking snow leopards in the deserts of China, and in many parts of the Amazon. These are but a few of the areas where it will be difficult to find local geodetic control. When working internationally, it may sometimes be difficult to obtain local control simply due to language barriers and from being unfamiliar with the structure of local governmental agencies.
      GPS users in a remote region may therefore have a much more difficult task in fully achieving the joys of DGPS. First, it is unlikely that any commercial or government source of DGPS correction data will be available in a remote area. (The desert and forest creatures of the world have not yet proven themselves to be a viable DGPS consumer base.) Although there are some limited regions of Earth that can receive DGPS corrections via satellite, these wide-area DGPS services are more expensive and are not yet world-wide in geographic extent. Finally, even if you purchase your own GPS receiver to serve as a base station, you still have to obtain geodetic control so that you can place your base at a location of known coordinates.
     This last option above, setting up your own base station, is usually the most practical solution for the user in remote field areas. The biggest difficulty is obtaining the accurate coordinates of the base station antenna. In these situations, when accurate coordinates are not known, the most important factor in selecting the base location should be "obtaining a clear view of the sky." Assuming that power supply and protection from the weather are not issues, an experienced user will select some location that can be conveniently reoccupied each day. As long as the location has an unobstructed view of the sky, the only remaining problem is to obtain "accurate" coordinates for our fence post. The question is, "How accurate do these base station coordinates have to be?" The answer is, once again, ... It depends. Your application will determine your accuracy requirements.

How good is good enough? It is possible to simply estimate the coordinates of the base station antenna. It is critical, however, that the user has a sound understanding of the impact that the inaccurate, "estimated" reference coordinates will have on the accuracy of the "differentially corrected" rover data. All of the error in your base station reference position will be added directly to your rover data.
      Note that DGPS provides the user with relative accuracy. That is, in differential GPS, the position of the roving GPS receiver is computed relative to the base station. Thus a small error in the base station reference coordinates will result in only a small additional error at the rover(s). In fact, if the estimated base station coordinates are within a few hundred meters of truth, the rover data will probably appear to have a simple homogenous shift in the same direction of the base error. For example, if the base coordinates are estimated incorrectly by 15 meters to the east, the entire rover data set will appear to have corrected perfectly, (relative to the base station) and a closer examination against truth would reveal a consistent 15 meter offset to the east of truth.
      For some applications, such relative accuracy is more than sufficient. In fact, as long as the same base location and same estimated reference position are used, the user will be able to consistently repeat the results on subsequent visits even months or years later. The rover positions will have excellent internal consistency from second to second, and can be matched to other data sets that were corrected by a base station at the same location using the same reference coordinates.
      Therefore, if your estimated reference position for the base station is incorrect by only 5 meters, your rover data will have a uniform 5 meter shift (in addition to the typical error level of your receiver). If your estimate is incorrect by 250 meters, your rover data will be quite consistent internally, however, there will be about a 250 meter shift for the entire data set when compared to local "truth."
      The risk you incur by taking a guess at the base station reference position is that your data may not align accurately when compared to other data sources. In this scenario, if matching to local maps is important, you may be able to take advantage of the fact that you are "simply guessing." Suppose for instance, that your initial guess at the reference coordinates produces rover files that consistently mis-match the local maps by, say, 25 meters to the southwest. You are free to consider changing your reference position by about 25 meters northeast to remove the mis-match. You are, after all, simply guessing, and one guess is as good as another! Do not, however, represent your data as having any known "absolute" accuracy. Remember that a sub-meter receiver, differentially corrected by a base that has a 30 meter error, is accurate to only 31 meters at best.
      If you require confidence in your GPS differential accuracy, and you require an absolute accuracy of better than 5 -10 meters, you are advised to contact a local professional surveyor and obtain a different sort of estimate.

More unknowns?
Again, I want to urge caution when averaging uncorrected position data. The accuracy indicated in this article is a direct function of the severity of Selective Availability (S/A) on the day of observation. The Department of Defense has the ability to alter the severity of Selective Availability. Therefore, your results could vary significantly depending upon daily variations in S/A.
      Software is another unknown to consider before you get all trussed up and march off into the jungles of Irian Jaya. How will your specific differential GPS software handle an incorrect reference position? (Don't forget to consider both embedded code for real-time differential and PC software for post processed differential.)
      Some manufacturers may detect that the reference coordinates are suspect and handle the situation in varying ways. There are some differential processors that will refuse to correct data if the reference coordinates are suspect, other packages may simply apply the correction without any regard as to data integrity, others still may simply burst into flame (just kidding) when confronted with inaccurate coordinates.
      At a more basic level, if you plan to get an estimate by averaging, check that your system has a provision for computing an average of large data sets. You don't want to have to average everything by hand. In any case, as always, you should test/examine these things at home before you buy the plane tickets.

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.

Back