AMSR2 collated or uncollated?

[stef]

Am a beginner in data assimilation and a question came up relating assimilation of microwave SST data, specifically from AMSR2.

In Lecture 6 of the 4D-VAR tutorial, it says that the sampling frequency for AMSR2 and WindSat is daily (Table 21.1). According to [1], AMSR2 is near-polar orbiting and samples about twice daily. Does this mean that for assimilation in MARACOOS, L3C (Level 3 collated, observations combined from a single instrument) are used?

In other words, the ascending and descending passes have already been merged together before feeding it to the ROMS assimilation (so this would have to be an estimate of foundation SST with diurnal cycle removed)? Or are you using the individual passes (although that would result in a sampling frequency of half a day)?

Assuming that you do use collated/merged data once per day, may I ask where you get the collated data from? I'm looking through the datasets on

https://podaac.jpl.nasa.gov

and I can find the REMSS L2P (satellite projection) REMSS L3U (regular spatial grid, but uncollated), and also the L4 MW_IR optimal interpolation, but I can't find L3C.

Or are you collating the L2P/L3U datasets yourself? If yes, this involves estimating the foundation SST from the subskin SST? It sounds difficult.

[1] https://www.remss.com/missions/amsr/

[wilkin]

In the real-time Doppio analysis system we run for MARACOOS/IOOS, we currently access AMSR2 L2 from dataset ID AMSR2-REMSS-L2P-v8.2 at PO.DAAC.

I'd have to check, but I believe the granules that intersect with the model domain bounding box are identified using the PO.DAAC subsetter.

[stef]

Thanks for the information! Wow, that's cool! So it probably means you assimilate the ascending and descending passes individually, and maybe the subskin temperature directly as opposed to converting it to a foundation temperature.

[wilkin]

The AMSR file format includes a sses_bias correction term that a user can apply to obtain the best estimate of foundation temperature. See Petrenko et al. https://doi.org/10.1175/JTECH-D-15-0166.1

[stef]

Thanks, great resource!!

Hmm, I've skimmed through the paper and there is a sentence:

PWR SST is consistent with in situ data during both day and night and thus may be viewed as an estimate of ‘depth’ in situ SST. The PWR SST is expected to be a useful input into L4 SST analyses, especially for foundation SST products, such as the CMC L4.

Then later is a sentence:

Since many L4 products position themselves as ‘foundation’ (defined at the depth below which no diurnal warming is present; Donlon et al. 2007) or ‘depth’ (i.e., representative of in situ SSTs), our primary objective was achieving a measurable improvement in estimation of ACSPO SST biases with respect to in situ SST (and thus minimize the need for an empirical bias correction as an initial step of L4 processing)

Then I found [1], where they talk about the ACSPO algorithm, and it says

ACSPO reports ‘subskin’ SST (in the ‘sea_surface_temperature’ layer), from which ‘depth’ SST can be derived by subtracting the sensor specific error statistics (SSES) bias (in the text below, the ‘depth’ SST is also referred to as ‘debiased’ SST). The ‘subskin’ SST is the sea temperature at depth of ~1 mm, and ‘depth’ SST is a proxy for SST at depth typically sampled by drifting buoys (~20 cm).

I get the impression that when the paper you linked to talks about "depth"-SST (representative of in-situ SST), they mean the same as in [1], i.e. at 20cm. Getting the foundation temperature from that would be yet another processing step, right? Because the diurnal stratification layer can extend deeper than 20cm, right? When they say "useful as input to L4", they mean it's useful in the sense that the diurnal signal between "subskin" (1mm below interface) and "depth" (20cm below interface) has been reduced, but they don't attempt to calculate SSTfnd.

Am I on the wrong track?


[1] Jonasson O, Ignatov A, Pryamitsyn V, Petrenko B, Kihai Y. JPSS VIIRS SST Reanalysis Version 3. Remote Sensing. 2022; 14(14):3476. https://doi.org/10.3390/rs14143476

https://www.mdpi.com/2072-4292/14/14/3476

[wilkin]

When I read these papers some time ago I concluded the quantity most representative of ROMS surface later (k = N) temperature would be to compute sea_surface_temperature minus sses_bias . That's what we do in operational Doppio.

[stef]

Thanks for clarifying!

Makes sense to me, because you are probably resolving some diurnal signal in the top layer.

Thanks again for all the help, it is all much clearer now.