GPS Solutions

, Volume 10, Issue 1, pp 12–20

Online GPS processing services: an initial study

Authors

  • Reza Ghoddousi-Fard
    • Department of Geodesy and Geomatics Engineering, Geodetic Research LaboratoryUniversity of New Brunswick
    • Department of Geodesy and Geomatics Engineering, Geodetic Research LaboratoryUniversity of New Brunswick
Original Article

DOI: 10.1007/s10291-005-0147-5

Cite this article as:
Ghoddousi-Fard, R. & Dare, P. GPS Solut (2006) 10: 12. doi:10.1007/s10291-005-0147-5

Abstract

There are a number of online Global Positioning System (GPS) processing services that provide GPS processing results to the user free of charge and with unlimited access. These services provide solutions for a user-submitted Receiver Independent Exchange Format (RINEX) file based on differential methods using reference stations or precise point positioning using precise GPS orbit and clock data. Different data sets varying in time and location were submitted to the online services and their results compared. Although the quality of results depends on many factors, in most cases the users can expect reliable online processing results for a 10-h data set made by a geodetic dual frequency receiver anywhere in the world.

Introduction

Over the last few years a number of organizations have developed online Global Positioning System (GPS) processing services. These services provide GPS processing results to the user free of charge and with unlimited access. The user sends a Receiver Independent Exchange Format (RINEX) file to the service and within a short period of time, the estimated position of the receiver used to collect the RINEX data is sent back to the user. Organizations that provide these free services include: Geohazards Division of Geoscience Australia, the Geodetic Survey Division (GSD) in Canada, the United States’ National Geodetic Survey (NGS), Scripps Orbit and Permanent Array Center (SOPAC) at the University of California and the Jet Propulsion Laboratory (JPL) at National Aeronautics and Space Administration (NASA).

The objective of this paper is to evaluate these online services and compare their position results with expected values. A comparison has also been made between the results obtained using data sets with varying observation time intervals. Furthermore, the results of the services are evaluated using data collected in different parts of the world.

Online services: an overview

Each of the above-mentioned organizations provides their own free online GPS processing service. The basic requirements that the user needs to take advantage of these different services are almost the same: access to the Internet and a valid email address. The following sections will give a brief description of each service.

AUSPOS

The Geoscience Australia [formerly the Australian Surveying and Land Information Group’s (AUSLIG)] Online GPS Processing Service (AUSPOS) was officially launched in late 2000 (Dawson et al. 2004), and has been in continuous operation since then processing data for dual frequency geodetic GPS receivers located anywhere on earth. The AUSPOS positioning is by differential GPS to the nearest three International GNSS Service (IGS) stations and uses the IGS precise orbit information. This service is accessible via the Geoscience Australia website at: http://www.ga.gov.au.

SCOUT

The Scripps Coordinate Update Tool (SCOUT) was developed by the Scripps Orbit and Permanent Array Center (SOPAC). This service also uses by default the three nearest IGS stations. However, this service allows the user to choose up to four different reference stations. The SCOUT uses the GAMIT processing software. This service is accessible from the SOPAC website at: http://sopac.ucsd.edu.

OPUS

The United States’ National Geodetic Survey developed the Online Positioning User Service (OPUS). This service generates coordinate results by using data from three Continuously Operating Reference Stations (CORS). The CORS sites are chosen not according to closest proximity but picked according to compatibility between the user’s data and the CORS site (OPUS Team 2004). There is also an option that allows the user to choose the CORS stations to be used. The service can be found at: http://www.ngs.noaa.gov.

Auto-GIPSY

Auto-GIPSY is an e-mail/FTP interface to the GPS Inferred Positioning System (GIPSY) developed by JPL. This service performs single point positioning, and is therefore not dependent on the proximity or availability of CORS/IGS data (Macdonald 2002). The FTP address of user’s data should be submitted by email to: ag@cobra.jpl.nasa.gov.

PPP

The Geodetic Survey Division (GSD), Canada, developed the Canadian Spatial Reference System (CSRS) Precise Point Positioning (PPP) service. Single point positioning is provided for users operating in static or kinematic modes using precise GPS orbits and clocks (GSD 2004). This service is available via the GSD website at: http://www.geod.nrcan.gc.ca.

Online services: a preliminary assessment

As well as the proximity of the online service’s coordinate results to the expected receiver position values (the topic of the following sections of this paper) there are other general factors to consider in the evaluation of services; these include: the method of sending and receiving the data, the time delay in receiving the results; available options and limitations. An overall assessment of each service (summarizing the above-mentioned aspects) can be seen in Table 1. As can be seen in Table 1, the services use either uploading of the data or an FTP site in order to access the user’s RINEX file. However, all of the services send an e-mail to the user either including the results or the FTP address of where the results can be obtained. Time delay on receiving the results depends on several factors including the traffic on the Internet and the number of users accessing the service at the same time. The displayed times in the Fourth column of Table 1 are only a rough indicator in order to compare the speed of each service and were obtained by submitting the same 6-h data set to each service.
Table 1

An overall assessment of online GPS processing services

Name of service

Data transfer method

Available options

Elapsed time to receive results (min)

Restrictions on length of GPS data set

Limitations

AUSPOS

Uploading

Antenna height

>25

Minimum of 1 h

Dual frequency

Via anonymous FTP

Antenna type

Static

No. of RINEX files (maximum 7)

SCOUT

Via anonymous FTP

Antenna height

>15

Minimum of 1 h

Dual frequency

Upload the file to Scripps FTP site

Antenna type

Static

Selection of reference stations

PPP* (*An “expert” version is also available with more options)

Uploading

Mode of processing (static or kinematic)

<3

No minimum

 

Reference system (NAD 83 or ITRF)

Maximum 6-day long providing uncompressed RINEX file is less than 100 MB (GSD 2004)

OPUS

Uploading

Antenna height

>4

Minimum of 2 h (recommended by the service)

Dual frequency

Antenna type

24 h maximum

Static

Additional options: selection of state plane and base stations, extended output, set user profile

Only available for use in Central and North America

Auto-GIPSY

Via sending the anonymous FTP address

None

<3

At least an hour, preferably more (Zumberge 1999)

 

Data within 15 h of GPS noon of obs. day will be analyzed

Results validation

One results validation method is to process observed GPS data at known points and compare the resultant coordinates with the known position values. Also evaluated was the relationship between the accuracy of the results and the observation time span.

Analysis of solutions for UNB1 RINEX data

In order to evaluate the accuracy of the online services, data collected at the UNB1 station on April 27, 2004 was submitted to the online services. UNB1 (see Fig. 1) is an IGS station at the University of New Brunswick (UNB) managed by the Department of Geodesy and Geomatics Engineering at UNB. UNB1 uses a continuously operating Javad Legacy GPS/GLONASS receiver located on the UNB Fredericton campus. The JPL has processed the UNB1 data and the latest coordinate of this station in the ITRF2000 reference frame can be found at: http://sideshow.jpl.nasa.gov/mbh/all/table2.txt. It is this JPL estimation that has been taken as the known value for UNB1 in this investigation.
https://static-content.springer.com/image/art%3A10.1007%2Fs10291-005-0147-5/MediaObjects/10291_2005_147_Fig1_HTML.jpg
Fig. 1

The location of UNB1 (after SOPAC 2004)

A single 24-h RINEX file was decimated into 1-h length, 2-h length, and then every 2 h up to 24 h. The 24-h file and the decimated files were submitted to the five online services. The OPUS does not process the RINEX files that contain GLONASS observations while other services remove the GLONASS data from the RINEX file before processing. Therefore GLONASS observations were excluded from UNB1 RINEX data in order that they could be submitted to OPUS. As mentioned before, OPUS and SCOUT also allow the user to choose the reference stations. In order to see whether the user’s selection of reference points can provide better results than default selections, OPUS was also tested using 3 user-selected reference points (OPUS3) and SCOUT using 3 and 4 user selected reference points (SCOUT3 and SCOUT4) with different geometric configurations than the default. The OPUS used by default BARN, BRU1 and PNB1 as reference stations while SCOUT used BARN, PNB1 and WES2. In the OPUS3 and SCOUT3 scenarios ALGO, STJO and PNB1 were selected as reference stations while BARN was added to these stations for the SCOUT4 scenario. It is worth mentioning that AUSPOS used NRC1, WES2 and ALGO. The location and type of reference stations can be seen in Fig. 2.
https://static-content.springer.com/image/art%3A10.1007%2Fs10291-005-0147-5/MediaObjects/10291_2005_147_Fig2_HTML.jpg
Fig. 2

The location of reference stations for OPUS, SCOUT and AUSPOS runs (after SOPAC 2004)

Figures 3, 4 and 5 show the differences between services’ resultant coordinates and the known values as a function of data set length. Figures 6, 7 and 8 give a more detailed look at the differences over the first 10 h. As might be expected the results in ellipsoid height show more variation over the time period (Fig. 5). It can be inferred from the figures that after almost 8–10 h observation the latitude and longitude have converged to within a centimeter of the known value. The height solution for each service continues to show variations at the centimeter level after 8–10 h, but with a variation of 7 cm between the services. Submitting less than 6 h of data to the services resulted (in most cases) in a few centimeters disagreement with the expected values. Auto-GIPSY did not provide proper results for the data set of 1-h length, therefore Auto-GIPSY results start from 2 h in the figures.
https://static-content.springer.com/image/art%3A10.1007%2Fs10291-005-0147-5/MediaObjects/10291_2005_147_Fig3_HTML.jpg
Fig. 3

Latitude differences between known value and online services result versus length of data set for UNB1, April 27, 2004

https://static-content.springer.com/image/art%3A10.1007%2Fs10291-005-0147-5/MediaObjects/10291_2005_147_Fig4_HTML.jpg
Fig. 4

Longitude differences between known value and online services result versus length of data set for UNB1, April 27, 2004

https://static-content.springer.com/image/art%3A10.1007%2Fs10291-005-0147-5/MediaObjects/10291_2005_147_Fig5_HTML.jpg
Fig. 5

Ellipsoid height differences between known value and online services result versus length of data set for UNB1, April 27, 2004

https://static-content.springer.com/image/art%3A10.1007%2Fs10291-005-0147-5/MediaObjects/10291_2005_147_Fig6_HTML.jpg
Fig. 6

Ten hour length plot of Fig. 2 (latitude differences)

https://static-content.springer.com/image/art%3A10.1007%2Fs10291-005-0147-5/MediaObjects/10291_2005_147_Fig7_HTML.jpg
Fig. 7

Ten hour plot of Fig. 3 (longitude differences)

https://static-content.springer.com/image/art%3A10.1007%2Fs10291-005-0147-5/MediaObjects/10291_2005_147_Fig8_HTML.jpg
Fig. 8

Ten hour plot of Fig. 4 (height differences)

As it can be seen in Figs. 6, 7 and 8 OPUS3, SCOUT3 and SCOUT4 provided closer latitude and longitude results to the known values than OPUS and SCOUT for data sets up to 6 h. After 6 h no significant difference can be seen. For the height results, however, SCOUT3 and SCOUT4 provided closer results to the known value than SCOUT for up to 8 h. In the case for OPUS3 the height results with respect to OPUS were improved for up to 2 h.

Results in different parts of the world

Except for OPUS, which is limited to Central and North America, all of the online services provide GPS processing results for observations made anywhere in the world. The PPP and Auto-GIPSY processing are based on precise GPS orbit and clocks products that are global in nature while SCOUT and AUSPOS use differential methods to the nearest three or four reference stations. However, these reference stations are not uniformly distributed in the world. In order to investigate the effect of reference station proximity on online services results two further tests have been done, as explained in the following sections.

Analysis of solutions from commercial RINEX data

A further investigation was carried out to compare the results of processing GPS data that were collected in Ethiopia on April 2, 2002 using commercial Trimble GPS equipment. The assumed coordinates of the point DODOLA (see Fig. 9) were obtained from data analysis using commercial software and UNB’S DIPOP scientific software (Dare and Baglole 2003). Different data set time intervals of the DODOLA GPS observation were submitted to the online services, and the differences in the coordinates with the assumed values can be seen in Figs. 10, 11 and 12. A warning message appeared in AUSPOS results indicating some modeling problems in the 24-h data set, so this data set was excluded from the figures.
https://static-content.springer.com/image/art%3A10.1007%2Fs10291-005-0147-5/MediaObjects/10291_2005_147_Fig9_HTML.jpg
Fig. 9

Location of investigated GPS points (after University of Alabama 2004)

https://static-content.springer.com/image/art%3A10.1007%2Fs10291-005-0147-5/MediaObjects/10291_2005_147_Fig10_HTML.jpg
Fig. 10

Latitude differences between assumed value and services result versus data set time (DODOLA, April 2, 2002)

https://static-content.springer.com/image/art%3A10.1007%2Fs10291-005-0147-5/MediaObjects/10291_2005_147_Fig11_HTML.jpg
Fig. 11

Longitude differences between assumed value and services result versus data set time (DODOLA, April 2, 2002)

https://static-content.springer.com/image/art%3A10.1007%2Fs10291-005-0147-5/MediaObjects/10291_2005_147_Fig12_HTML.jpg
Fig. 12

Ellipsoid height differences between assumed value and services result versus data set time (DODOLA, April 2, 2002)

Processing of observations from an IGS point in Africa

Three 24-h data sets (the first 3 days of 2004) of point MALI were submitted to the online services. The MALI is an IGS point located in Malindi, Kenya (see Fig. 9). An Ashtech Z-XII receiver is operating at this station. The differences between service results and expected values (JPL estimation) are presented in Figs. 13, 14 and 15. The PPP results did not converge on January 1, 2004, so these results are excluded from Fig. 13. Further investigations carried out on the PPP results will be described in the following section.
https://static-content.springer.com/image/art%3A10.1007%2Fs10291-005-0147-5/MediaObjects/10291_2005_147_Fig13_HTML.jpg
Fig. 13

The difference between online services results and expected values (January 1, 2004)

https://static-content.springer.com/image/art%3A10.1007%2Fs10291-005-0147-5/MediaObjects/10291_2005_147_Fig14_HTML.jpg
Fig. 14

The difference between online services results and expected values (January 2, 2004)

https://static-content.springer.com/image/art%3A10.1007%2Fs10291-005-0147-5/MediaObjects/10291_2005_147_Fig15_HTML.jpg
Fig. 15

The difference between online services results and expected values (January 3, 2004)

Further investigation on PPP results

As mentioned in the previous section, the results of PPP did not converge for point MALI on January 1, 2004. To investigate this further, IGS points were selected in different locations and their data for January 1, 2004 were processed by PPP. These points are: ALGO (Ontario, Canada), UNB1 (New Brunswick, Canada), STJO (Newfoundland, Canada), STR1 (Australia), BAHR (Bahrain) and RIOG (Argentina). The location of these points is indicated in Fig. 9.

The differences between the PPP results and expected values (JPL estimation) can be seen in Fig. 16. The PPP results did not converge for STR1 (the same problem that occurred for MALI) and this point is excluded from Fig. 16.
https://static-content.springer.com/image/art%3A10.1007%2Fs10291-005-0147-5/MediaObjects/10291_2005_147_Fig16_HTML.jpg
Fig. 16

The PPP results versus expected values on January 1, 2004

Analysis of the results

On average, Auto-GIPSY was found to produce the closest horizontal and vertical coordinates at the investigated points. Furthermore, the quality of this service and the PPP service are independent of site location due to using precise GPS orbit and clock data in point positioning mode. These two services are also the fastest to return the results.

Ellipsoid height results of SCOUT and Auto-GIPSY show some unusual changes as can be seen in Fig. 5 at hour 4 and at hour 20 for SCOUT and at hour 6 for Auto-GIPSY. However, the SCOUT3 and SCOUT4 results do not show such unusual changes, even though the average baselines length in SCOUT3 and SCOUT4 are more than SCOUT.

For short data sets, user selected reference stations in SCOUT3/4 and OPUS3 scenarios provided closer results to the known values than SCOUT and OPUS. This may be due to reference station quality and geometric configuration. A significant change in vertical accuracy of SCOUT can be seen at point MALI, where on the first day the vertical accuracy was about 3 cm but on the next 2 days was more than 10 cm (Figs. 13, 14 and 15).

At DODOLA, the latitude converges as it did for UNB1. The longitude solutions from SCOUT and PPP, however, continue to show variations at approximately 2–3 cm and the convergence pattern is not as clear as UNB1. The AUSPOS produced identical results for the last 8 h and Auto-GIPSY for the last 12 h. In the solution for height (Fig. 12), Auto-GIPSY provided the closest results to the assumed value while it had a systematic difference of 12 cm with other services. After 10 h, the height solutions for the services (ignoring Auto-GIPSY) vary by about 4 cm.

The PPP results did not converge for points MALI and STR1 on January 1, 2004 (both of these points located in the southern hemisphere). On January 1, 2004 PRN 23 experienced failure in its atomic frequency standard (Sigmond 2004). However it does not seem that the two mentioned reasons caused the failure of the PPP results in MALI and STR1 because although the point RIOG is also in the southern hemisphere accurate results were provided by the PPP (Fig. 16). Furthermore, other services provided reliable results on the same day for point MALI (Fig. 13).

Conclusions

Online GPS processing services can help GPS users all over the world to take advantage of precise point positioning or differential methods with one single receiver, and without requiring detailed knowledge of processing software. Solution quality depends on the availability, proximity and quality of base station data, and the availability of precise satellite orbits and clock corrections. Performed tests in this paper indicate that users can expect reliable results from online services, although some problems have occurred, such as those mentioned for PPP.

The resultant coordinates converged after almost 10 h of observations using default-processing parameters. This shows that users can expect almost the same results for a 10-h data set as for a 24-h data set. With user-selected reference stations for the examples used, the data set length could be reduced by a few hours.

Acknowledgements

Preliminary work on this research was carried out by two undergraduate students at UNB. Their work has been published in Leslie (2004) and Hatch (2003). We thank Duncan Moss and Neil Stuart of the University of Edinburgh, Scotland, for providing the GPS data they collected at Dodola, Ethiopia. Paul Jamason is thanked for providing comments that improved the quality of this paper. We also acknowledge Canada’s ‘Natural Science and Engineering Research Council’ (NSERC) for providing funds to enable this research to be carried out.

Copyright information

© Springer-Verlag 2005