The Joy of DGPS
Any individual GPS receiver, used by itself, is only accurate to about 100 meters. However the data from two receivers can be combined to eliminate many sources of error, resulting in accuracies ranging from a few meters to less than a centimeter. This process is termed differential GPS (or DGPS).
      Differential GPS is used by virtually all who collect GPS data for GIS. This is understandable since the 100 meter accuracy of non-differential GPS is not sufficient for most GIS data collection applications. Differential corrections can be performed either in real-time via special radio links, or as a post-processing procedure on a computer. In general, both techniques are equally accurate. However, in reality, the accuracy of post-processed DGPS can vary dramatically depending on the quality of your differential software. A good post-processing program can be more accurate than real-time corrections. A poor post-processing program can be less accurate than real-time corrections.

Real-time DGPS
Despite the variability in software quality, presently more people are utilizing post-processed DGPS than real-time DGPS. Perhaps because, for most applications, the expense, limitations, and maintenance of the telemetry link between base and rovers (to perform real-time DGPS) far outweigh the benefit of collecting accurate data in real-time. The primary benefit of having an accurate, real-time position is for navigation. The accurate position can be used to navigate to a precisely known location that is not readily visible or otherwise identifiable.Post-processed DGPS, on the other hand, can be used for virtually any data collection exercise.

Post-processed DGPS
The elimination of the real-time telemetry link provides a great deal of flexibility in where roving field crews can collect information. With post-processed DGPS the maximum range of differential correction is extended to as much as several hundred kilometers from the base. Additionally, there is no line of sight requirement between the base and the rovers; a common limitation of many telemetry systems. Later, back at the office, the data is typically downloaded to a PC for evaluation and then transferred to the GIS. Since most users download their data to a PC anyway, and because differential post-processing can improve the position accuracy from 100 meters to well under one meter, the extra one or two minutes required to select one additional menu item (differential correction) is well worth the effort.

How far can you go?
It is important to consider the distance between your field work area and the base station location. There are limitations as to how far from the base you can work effectively. In general, the greater the base/rover distance the less the probability of obtaining successful corrections and the less accurate those corrections will be if they are successful. Two factors control how far from the base you can collect correctable GPS rover data; first, which satellites are visible to both the base and the rover, and second, the quality of your differential correction software.

Which satellites are visible?
The first factor is because the base and rover GPS receivers must be close enough together to track a common set of satellites. The differential process removes only errors that are common to both the base and the rover. A requirement to compute these individual satellite errors is that the satellite in question must be tracked at both the base and the rover locations. (This is the reason that I cannot use the base station available in Christchurch, New Zealand to apply corrections to my rover data collected in London.)
      Fortunately the extremely high altitude of the satellites (about 20,600 kilometers) results in each satellite being visible to a very large portion of the Earth's surface. Nearly all users are able to successfully correct rover data collected 500 kilometers (300 miles) from the base station. Some users successfully apply corrections to rovers that are as far as several thousand kilometers from the base station. See the sidebar entitled "How far is far?" for a little perspective on the potential range of a base station.
      Presently there are very few people who utilize GPS base stations to the limit of their effective range. This is due in part to the fact that for most users, typically there are multiple base stations relatively nearby. When users are forced to use a particularly distant base station they must deal not only with the risk of lower accuracy, but also with the risk of some positions in their rover file(s) not correcting. Understandably, most users simply select either the nearest base station, or more likely, the one with the most convenient data access.

How much corrected data is enough?
Studies have been made wherein data has been differentially processed at a range of over 4,000 kilometers from Miami, Fla. to British Columbia, Canada (Biacs and Bronson, 1995). However, as the base/rover distance is increased to such extremes, it is likely that a lower percentage of rover positions will be correctable.
      Usually the only people who must work at such large base/rover distances are those in extremely remote regions such as the deserts of China or the Amazon Basin. For example, a group of astronomers collected data in the rainforests of northern Peru during an eclipse in April 1995. Their rover data was successfully corrected using a base station in Miami (a baseline of 3,350 kilometers).
      Note that working at such extremely long baselines might require a slight modification of the receiver configuration. A parameter known as the elevation mask is usually used to configure receivers so that the base will always track satellites before and after the rover receivers. An elevation mask limits a receiver so that it will use only satellites at elevations above the horizon greater than that specified by the mask. The appropriate configuration for long baseline work is to set the roving receivers to a higher elevation mask than the base station. This results in the base acquiring satellites prior to the rover when the satellites rise, and results in the base continuing to track satellites longer when they set. A few typical examples of reasonable elevation masks are:
      Typically for every degree of difference between the base and rover elevation masks, the base and rover can be separated by about 100 kilometers with little risk of differential correction failure.

Spatial Decorrelation
Spatial decorrelation is the other important issue to consider when preparing to indulge in long baseline, post-processed DGPS. Spatial decorrelation can be described as a gradual increase in error as the base/rover distance increases. This error is due to differences in the path of the radio signal from the satellite to the base and rover receivers. When the signal paths to base and rover are not identical, the resulting differences show up as position errors that cannot be removed by differential correction (because this error is not common to base and rover). Usually such differences are the result of variation in the Earth's ionosphere and troposphere.
      There are several factors that control the severity of this degradation. The intensity of ionospheric activity is one factor, the local variations in troposphere (that means "weather") are another. Even the manner in which Selective Availability is implemented can contribute to long baseline errors. However, one of the largest factors, and the only one that the user can control, is the quality of the differential software being used to compute the differential corrections.
      This type of error is often measured in parts per million (ppm) of the baseline length. That is, if the base and the rover are separated by 100 kilometers and the spatial decorrelation error is specified as 10 ppm; then the user can expect an additional error of up to 10 millions of 100 kilometers, or 1 meter (100 kilometers divided by 10 millionths equals 1 meter). Under the current S/A and atmospheric conditions, software on the market today provide results ranging from as little as 1 ppm to more than 20 ppm. The table below illustrates the impact of these errors as a function of baseline lengths.
      Note, however, that the decorrelation errors above are in addition to any other error sources. For example, if your receiver is rated at 2 meters plus 10 ppm; then at 200 kilometers from your base you can expect as much as 4 meters of error (2 + 2 = 4) under ideal conditions. If your receiver is rated at 0.5 meters plus 1 ppm; then at 200 kilometers from your base you can expect as much as 0.7 meters of error (0.5 + 0.2 = 0.7) under ideal conditions.

Summary
Even today, as real-time differential services are becoming more common, post-processed DGPS is the technique utilized by most GPS users. Base data is available for all but the most remote locations on Earth. This allows the collection of accurately corrected data nearly anywhere on Earth, assuming that your differential correction software is up to the task. As always, there is no substitute for testing the system yourself before you buy. Decide in advance from where you will obtain base station data or whether you will buy your own base station system. Armed with this knowledge you will be able to more effectively test the post-processed DGPS quality.

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