JPL has been performing GPS orbit and clock predictions (and disseminating them over the internet) since the early 1990s. As enduring leaders and innovators in modeling satellite dynamics and in orbit determination, we bring unparalleled expertise to deriving these predictions.

The GDGPS predicted orbits and clocks states extend 7 days into the future, and are produced six times daily. They are based on the GDGPS real-time orbit and clock solutions so as to minimize the extrapolation period.

We incorporate into our prediction every known future event (announced in the NANUs) that might compromise the integrity of the predictions. While our predicted orbits and clocks are highly accurate (see figure below), prediction is fundamentally risky because certain aspects of the satellite and clock dynamics are difficult to model, and NANU announcements may not be completely accurate.

Our prediction scheme incorporates state of the art satellite dynamics models, the most recent tracking data, and sophisticated continuously-adaptive prediction algorithms. As an additional service to our customers we provide continuous, near-real-time quality control of the predicted products. These enable our customers to determine whether or not a new prediction needs to be downloaded to replace a previous file with perhaps a problematic satellite. This capability saves precious wireless bandwidth.

The key features of the GDGPS predicted orbit and clock product are:

• 7 day prediction, 6 times daily, 15 minutes temporal resolution
• Accuracy estimates are included for every epoch
• Timed NANU-based warning flags are included
• Continuous quality control using the GDGPS real-time orbit and clock states available with 4 hours latency
• Multiple file formats are available, including SP3c, RINEX Navigation, and internal GIPSY
• The prediction files are available through sFTP from multiple redundant servers

Any prediction involves a fundamnetal level of uncertainty, and must be evaluated in a statistical sense. While we cannot guarantee any level of accuracy for future predictions, we can report statistics of past performance. The figure below represent the typical median errors for the satellite clock, 3D position, and the User Range Error (URE - the key metric for user positioning, incoporating both orbit and clock errors). Note that this is the median: half the satellites will perform better, and half will perform worse.

We have found that the roughly linear relationship between the median URE and the prediction length for 1-3 days predictions is quite typical. The quadratic behavior for longer term prediction is also quite typical. Except for the first one or two prediction days, when the URE can be smaller than the clock error, the URE is strongly dominated by the clock prediction error.

As with any other GDGPS product, we continue to actively research methods and means to improve the quality of the prediction products.