Altimeter data analysis
The main issue in altimeter data harmonic analysis is the aliased frequencies and subsequent separation periods. The extent of this issue depends on the considered mission time sampling. The correction of ocean bottom displacements (due to earth tides and ocean loading) is also affecting the ocean tide determination.
After 20 years of duration, most of the alias issue have vanished in T/P and Jason nominal mission (later referred as T/P). This is clearly not the case for the T/P interleaved mission (equivalent to about 5 years of continuous measuremenst). The case of GFO and ERS/EnviSat missions is even worth as not only the data quality is inferior to the one of T/P mission, but also their time sampling is disfavourable to tidal analysis.
In the context of FES2012 project, it is advised to:
- use only the best times series, i.e. discards diminished series (i.e. series with a significant loss of data compared to mission nominal acquisitions)
- filter internal tide surface signature (along-track filtering, shelf-edge exclusion as filtering will be spoiled by tidal elevation sudden change above the continental edges)
- freely use T/P data compliant with above criteria
- use ERS/EnviSat and GFO data compliant with above criteria at crossovers only, and only in case of lack of T/P data (high latitudes, shelves, ...)
Action: investigate loading correction accuracy.
Altimetry-detidor is based on the harmonic analysis method with admittance capabilities in case of separation issue. A tidal prediction is first removed (FES2004, GOT4.8, ...), then the harmonic analysis is performed on residuals. Harmonic constants come with error bars that are based on (tidal peaks) surrounding spectral energy (publication...).
Admittance pivots are:
Semi-diurnal: N2, M2, K2 (S2 being contaminated by radiational tides)
Diurnal: O1, Q1, K1
Long periods: Mtm, Mf, Mm
Trap: use of admittance on residuals needs to be extremely cautious with the prior detiding, so that residuals still fit the "credo of smoothness" (Tides in the North Atlantic, Cartwright et ???)
Multi-threading has been achived in altimetry-detidor, scaling nearly optimal.
Documentation: alias and separation tables for T/P and others
T/P: no particular issues
- separation issues: etc...
- Sun-synchronous tides unavailable (S1, S2)
- K1 aliased on annual period, contamination from ocean (large) variability at this period. Should we remove a circulation model (Mercator...) annual signal first?
After the first tests on the CLS altimeter time series, it seems quite clear that T/P interleaved, ERS/EnviSat and GFO along-track analysis will be difficult to use in data assimilation because of their poor level of accuracy. I strongly recommend to use cross-overs analysis only, and only in case of specific necessity.
Action: prepare alias and separation tables, and actual analysed spectra for each mission
Harmonic analysis examination
- TPN/J1N Xovers
As mentioned in the above-paragraph, it would be very useful to eliminate low frequency sea level variability to improve the accuracy of some tidal components aliased on monthly to annual periods (typically K1 in ERS/EnviSat mission), especially in the high latitudes where T/P data are not available. We need to assess the feasability to take profit of Mercator re-analysis (GLORYS) for this matter.
access to data
In addition, it would be profitable to assess meso-scale signal energy in specific frequency bands to verify harmonic analysis error bars and/or mask regions where energy is significant in data assimilation editing (see following paragraph).
GLORYS-v2 1992-2009 re-analysis has been acquired (cartesian grid courtesy of CLS). First tests are in progress.
Semi-annual signal (close to 173 days K1 alias) from GLORYS-v2 analysis
Semi-annual signal (close to 173 days K1 alias) from GLORYS-v2 analysis, low-pass filtered
Major constituents such as M2, S2 and marginally K1 and O1 do generate internal tides that have a significant surface signature coherent with barotropic tides (phase-locked signature). Along-track filtering will be used to eliminate the internal tide contamination in harmonic constants prepared for data assimilation. Filtering window will be tuned from internal tide wavelength (at least for the first modes) but also to avoid barotropic tide erosion. A special effort has been done to determine true barotropic tide wavelenght (or more precisely typical lenght scale).
tidal wavelength (based on M2, S2, N2, K1,and O1 tides)
Data assimilation capabilities
Data assimilation of contaminated data would be safe if error bars were known with enough accuracy. It is rarely the case, so a preliminary editing (such as assessing the actual volume of data with signal-to-noise ratio better than 90%) might be required.
In case the volume of data is insufficient for a minor astronomical constituent, it is our advise to left it aside from assimilation and derive it from admittances methods.
Action: verify this point as soon as possible, as it can ease the ensemble generation.