The
interface is pretty much straight forward. On the website you see
satellites listed with
the available predictions from different providers and days as well as
the predicted time bias value for the current point in time. There are four groups where you can find the individual satellites listed which can be
found at the top of the page and which are:

- Watch List

13.3333px;” align=”left”>- LEO Targets

- Geodetic Targets

13.3333px;” align=”left”>- Debris Targets

align=”left”>So if you
are
about to track a satellite, find the group, search for the satellite in the list, check the currently predicted time bias value and use

that as an a priori start value durning acquisition. In brackets first you see style=”font-size:
13.3333px;”>the
RMS of all time bias values/passes with respect to the
applied fit and second the number of time bias values/passes used for
that fit. From that you can get a first idea about the reliability of a predicted value. If

you click on a linked provider name you will see a graphical
representation of the time bias values of all available passes and the
predicted one. This allows you to evaluate the quality of the predicted
time
bias value but also of the prediction itself by checking the trend and
the scatter. It is
interesting
to compare predictions from the different providers even with satellites
like Lageos 1 where you do not really expect a time bias. However, when you take
a
look at the HTS prediction over multiple days you can see modeling
errors periodically adding up which is not the case with the JAX
predictions. In
some cases you might also see that a predicted time bias value does not
look too good with respect to the trend. In such cases just do an
”extrapolation on the screen” for a better estimate since we are
currently experimenting with data weights which do not always work well
with sparsely
distributed data.

Let

us know if
you encounter issues, if you are missing information, satellites,
providers, anything else or if you have suggestions. The contacts are
given on the
site and are Jens and me. Please remember that this is a beta test, so the service might encounter outages or


other issues. 

As we have shown in our Riga presentation,
the time bias value prediction becomes better with passes being
processed
and submitted to EDC quick (e.g. after a calibration and not after a
shift) as well as with overlaps of concurrent predictions. Overlaps of
e.g. 6
hours help to predict time bias values from data already collected with a
past prediction instead of starting with a new prediction without any data
13.3333px;”>.

13.3333px;” align=”left”>

Thanks and best

Sven