improve interpolation in IERS data

[mhvk]

Currently, UT1-UTC is interpolated linearly in the IERS data, ignoring the effects of tides (which are not included in the tables)..While this is fine for almost all purposes, the recommended procedure is to use a Lagrangian interpolation scheme and include the tides [1]. This is a reminder to implement this in IERS, following the ideas in the fortran code mentioned in [1], after checking what should be done for jumps in UT1-UTC due to leap seconds.

[1] http://maia.usno.navy.mil/iers-gaz13

[mhvk]

Mostly so there is a record: from an e-mail exchange with Daniel Gambis (IERS), the recommended procedure is:

1. Transform UT1-UTC into UT1-TAI by removing leap seconds. You get a continous time scale
2. Transform UT1-TAI into UT1R-TAI by correcting for the luni-solar zonal tides
3. Interpolate by applying for instance interp.f
4. Add back the zonal tides to get UT1-TAI
5. Transform back to UT1-UTC by adding back the leap second

The interpolation/tide routine to follow would be ftp://hpiers.obspm.fr/iers/models/interp.f

[mhvk]

Cleared milestone as I won't be able to do this for 1.0. It should be done for ultimate precision and is not in itself difficult.

[mcbeth]

I find myself in a position of needing this rather badly. I've made some progress on getting interp.f ported, but have run in to troubles with the IERS_A code apparently always reporting the IERS_B columns of the finals2000A.all, and completely missing the bit about needing to back out the zonal tides before I put them back in.

#!/usr/bin/python3
from astropy.utils import iers
from astropy.time import Time
t = Time('2017-08-20',format='isot')
iers_a = iers.IERS_A.open()
dX,dY = iers_a.dcip_xy(t)
XP,YP = iers_a.pm_xy(t)
dUT1UTC = iers_a.ut1_utc(t)
# 17 820 57985.00 I  0.228614 0.000009  0.378523 0.000024  I 0.3403042 0.0000064  0.2552 0.0047  I     0.122\
    0.134     0.179    0.208  0.228598  0.378522  0.3402929     0.150     0.190
print(XP)
print(YP)
print(dX)
print(dY)
print(dUT1UTC)

[mhvk]

@mcbeth - yes, this has long been on my list! I'm a bit confused about what problem you have with the columns, though it is certainly the case that the various methods use IERS_B results in preference to those of IERS_A (since IERS_B results are presumably the "final" ones).

But perhaps more importantly, about the backing out of the zonal tides: I looked back at the e-mail exchange I had with Daniel Gambis of the Observatoire de Paris, and he wrote that the corrections that are in the table include the zonal tides already (unlike what the text accompanying interp.f suggests), and one would first take those out (presumably using the RAY routine), then interpolate, and finally add them back again.

I must say that I have partially held back from implementing all this because I wasn't sure how to test that whatever I did was in fact correct... Do you have any way to verify?