Conclusions

We have presented results from two observations in which we were able to compare S/X derived ionospheric corrections with corrections based on GPS data. We come to the following conclusions:

• If you want to remove the ionosphere accurately, you should observe strong sources and design the experiment with enough spread in frequency to measure the ionosphere. Dual band S/X observations are the traditional way to do this, but other pairs (requiring new hardware), and even just multiple IF's at L band and below, could be used. Such a correction scheme cannot be used on weak sources, which eliminates its use for a majority of phase referencing observations.
• The GPS models provide useful ionospheric corrections. Phase offsets due to the ionosphere, which dominate at low frequencies, can be reduced by a factor of roughly 2 to 5 using such models, but cannot be corrected perfectly.
• Ionospheric modeling is in a sufficiently primitive state that differences between models from different analysis centers are commonly 20-50%. Model errors at this level are consistent with the quality of phase improvements seen in the VLBI data.
• The SATLOC models do not appear to have any advantage in accuracy over the global models despite the finer grid and higher time density. The global models are easier to obtain and cover all stations and so are better for our purposes.

Based on the above conclusions, we recommend:

• We should NOT make the gathering and archiving of SATLOC models an observatory priority. It may be worth keeping the receiver functional for any users who might wish to gather the data. But this can be on an as-needed basis.
• Ionospheric corrections based on the global models should become a common part of VLBI calibration, especially at low frequencies and perhaps at all frequencies. Since these models are only provided on time scales long compared with typical coherence times, they should not degrade data, and they are likely to help.

Gathering the models from CDDIS and applying them is rather easy -- an approximately 10 minute job if it is not necessary to edit together multiple files. Therefore, providing the data as a service to the users is not a high priority, but would be a convenience. We need to discuss whether to apply the models in the correlator, pass them in calibration transfer, or do nothing at the correlator. Given that the user might wish to use a different model from whatever is passed, we recommend the calibration transfer option. The model would be applied as part of post-processing. This option may require a new AIPS table to hold the model because currently the information is only included in CL tables which are often destroyed and rebuilt for VLBI data (even version 1). In any case, one small but high priority task is eliminate the need to edit IONEX files for observations near the day boundary.

It is likely that the GPS ionospheric models will improve with time. One significant improvement for our application is likely to come from 3D modeling. We need to watch for this to become available and be prepared to have our software accept such models. Related to this, TECOR currently assumes an infinitely thin ionosphere. A possible future development would be to use a more realistic profile. This might allow better results to be derived even from the current 2D models, let alone from future 3D models.

We have not studied the use of data from the individual receivers at each site. Such data would have the advantage that it is subject to the same height effects as the VLBI data and so could potentially provide better corrections. But there are problems with the data, or at least our use of it. Our first look had results clearly in error by an order of magnitude. The problems may lie in the area of corrections for the offsets between the channels on the satellites and in the receivers, but that is not certain. It could be something as simple as a format conversion problem. Considerable effort will be required to learn how to use such data effectively. That is beyond the scope of this memo but is an approprate project for future work.