Check OCC/ROPP, Invert/ROPP test height output scales

[Huw Lewis]

Tests IT-PP-02, IT-PP-03, IT-PP-04 and IT-PP-05 present comparisons of height scales produced by GRAS NRT data (IT-PP-02) or OCC (IT-PP-03, IT-PP-04, IT-PP-05). The ropp_test codes suggest that results are plotted as %s. (ropp-occ)/ropp

For IT-PP-03, IT-PP-04, IT-PP-05, the height differences are constant with height. Why is this?

For IT-PP-02, the height differences are up to 4% at the surface. Why is this?

See also #157.

[Ian Culverwell]

it_pp_02_plot.gif

[Ian Culverwell]

it_pp_03_plot.gif

[Ian Culverwell]

it_pp_04_plot.gif

[Ian Culverwell]

it_pp_05_plot.gif

[Ian Culverwell]

All these tests have problems.

IT-PP-02 (Invert(?) vs ROPP on GRAS SAF NRT data) has decent bangle profiles, but an almost constant offset between the geopotential heights (of around 20m for profile 1):

IT-PP-03 (OCC vs ROPP on CHAMP data) has decent bangle profiles, but this time the offset in GPH increases linearly with altitude (at around -1m/km for profile 1):

IT-PP-04 (OCC vs ROPP on COSMIC data) has decent bangle profiles, but again the offset in GPH increases linearly with altitude (at around 0.2m/km for profile 1):

IT-PP-05 (OCC vs ROPP on GRAS RS data) has very poor (~5-10% different) bangle profiles, but, for variety, the GPHs agree pretty closely:

Obviously these different problems are likely to have different solutions. For the worst one, IT-PP-05, it's pretty obviously a configuration/metadata issue. For the others, it looks more subtle.

[Ian Culverwell]

it_pp_05_plot_new.gif

[Ian Culverwell]

Kjartan Kinch (DMI) has investigated the sensitivity of the IT-PP-05 results to the co-ordinate system used in OCC. By changing from ECI to ECEF 4 out of the 5 profiles changed and gave an almost perfect match to the ROPP bangles:

The 1/5 profile where changing the co-ordinate system has no effect, and therefore no improvement ion before, is the bottom on above. I've tried changing the input parameters to ROPP, to no avail. Kjartan wonders whether the source (phase delay) data might be different in the two cases, and I think this is likely the case. We need to compare.

Quite why the co-ordinate system might have this strong an effect in OCC is unclear, and Michael Gorbunov is looking at it. For our purposes, it suffices to show that the two codes can agree in a large number of cases.

[Ian Culverwell]

It was agreed at the ROM-SAF Steering Group Meeting on 13th Nov 2012 that we could abandon OCC as the reference for ROPP, and use ROPP6.0 instead. Eventually, this will be superseded by a later release, but the reference doesn't always need to be the latest release. Closing the ticket as a result.