Data From the First Week
Without Selective Availability

GPS Fluctuations Over Time on May 2, 2000

Source: GPS Support Center, Air Force Space Command

This is a plot of GPS navigational errors through the SA transition prepared by Rob Conley of Overlook Systems for the GPS Support Center in Colorado Springs, Colorado. The GPS errors can be seen diminishing significantly around 0405 UTC (shortly after midnight EDT). The data indicates a circular error of only 2.8 meters and a spherical error of 4.6 meters during the first few hours of SA-free operation. The data was measured using a Trimble SV6 receiver.

GPS Accuracy Before and After SA Removal

Click either image for full size view

Source: NOAA National Geodetic Survey

The images compare the accuracy of GPS with and without selective availability (SA). Each plot shows the positional scatter of 24 hours of data (0000 to 2359 UTC) taken at one of the Continuously Operating Reference Stations (CORS) operated by the NCAD Corp. at Erlanger, Kentucky. On May 2, 2000, SA was set to zero. The plots show that SA causes 95% of the points to fall within a radius of 45.0 meters. Without SA, 95% of the points fall within a radius of 6.3 meters.

As illustration, consider a football stadium. With SA activated, you really only know if you are on the field or in the stands at that football stadium; with SA switched off, you know which yard marker you are standing on.

For additional information:

Dr. Dennis G. Milbert
Chief Geodesist
National Geodetic Survey, NOAA
301-713-3222

Fine Print: Data taken at the Erlanger National CORS station, National Geodetic Survey, NOAA. Data with SA were taken from 0000 to 2359 UTC on May 1, 2000. Data without SA were taken from 0000 to 2359 UTC on May 3, 2000. Both data sets were taken at 30 second intervals. Instrumentation was an Ashtech Z-12 receiver. GPS data were dual-frequency pseudorange (both L1 and L2) incorporating ionospheric correction. Data were processed in accordance with the GPS Interface Control Document ICD-GPS-200C, using the broadcast orbit parameters in the World Geodetic System WGS 84 (G873) reference system.

The May 2 Transition as Observed From Around the World

Select "24" to view 24-hour data set, "D" for detailed view

ID Location N. Lat. E. Long. Ellip.Ht. Agency Data
AMMN Amman, Jordan 32.0300 35.8800 1055.8300 RJGC 24 | D
CAS1 Casey, Antarctica -66.2834 110.5197 22.5500 AUSLIG 24 | D
CENA Central, Alaska 65.4982 -144.6776 283.7740 NOAA/FSL 24 | D
ERLA Erlanger, Kentucky 39.0232 -84.6059 253.8970 NCAD 24 | D
FORT Fortaleza, Brazil -3.8774 -38.4256 20.4918 NOAA 24 | D
KOK1 Kokee Park, Hawaii 22.1263 -159.6649 1167.3680 NASA/JPL 24 | D
KSTU Krasnoyarsk, Russia 55.9900 92.7900 210.0000 GFZ 24 | D
MALI Malindi, Kenya -2.9959 40.1944 22.7200 ESA 24 | D
NYAL Ny-alesund, Norway 78.9300 11.8700 82.0000 NMA 24 | D
RIOG Rio Grande, Argentina -53.7900 -67.7500 32.0000 GFZ 24 | D
WUHN Wuhan City, P.R. China 30.5317 114.3573 93.4000 WTUS 24 | D

The images display the simultaneous termination of GPS selective availability (SA) at GPS reference stations located around the world. Each plot shows the ellipsoidal height scatter of 24 hours of data (0000 to 2359 UTC) taken on May 2, 2000. At approximately 0405 UTC (14700 sec.), all GPS satellites stopped introducing the intentional SA error. The data show this occurring over an interval of a minute or two. Detailed plots are provided to show the first moments when SA was discontinued.

The parabolic patterns evident in the plots with SA off are most pronounced at very high and very low latitude stations. These stations have poorer satellite coverage. Error sources such as broadcast orbit and troposphere error can more readily influence those coordinates. Of interest is the amount of vertical scatter in these cases, since ionospheric error was removed through dual frequency pseudorange data.

Our thanks go to the International GPS Service (IGS) for organizing GPS tracking networks around the world, and for insuring that GPS data are available online for all types of studies. We also thank the various agencies who contributed the data used in this study.

For more information on the GPS tracking networks:

Data taken at numerous sites from 0000 to 2359 UTC on May 2, 2000. GPS data were dual-frequency pseudorange (both L1 and L2) incorporating ionospheric correction. Data were processed in accordance with the GPS Interface Control Document ICD-GPS-200C, using the broadcast orbit parameters in the World Geodetic System WGS 84 (G873) reference system. Tropospheric corrections have not been applied.