Now that the scripts have been written for descending passes, doing the same for ascending passes only requires some minor changes, so the computer will do most of the work.
Yet, downloading the files will take some time and there is a holiday next week, so I may not be able to provide the new plots before the second half of the week.
Sylvain,
I perfectly remember what you told me before that downloading the files takes some significant time.
And I am pleased to have the confirmation of what I imagined that dealing with ascending passes only requires minor changes in the script.
We are not at all in a hurry so take the time you want, the end of next week will be perfect for me.
Now, from a pure "software-engineer" point of view, do you think there is some interest for you to write something for the SWOT community? I mean I already discussed with "SWOT specialists" about what I wanted to do (so about what you did) that is, in some sense, computing a refined geoid from the data available from the web site and they did not believe in what I proposed, so in what you did. In case you plan to write something, please give me the reference.
In case you do not plan to write anything, I think I will have enough results from your SWOT results to efficiently complete what I already wrote and submit kind of manuscript to an oceanographic journal. Let me know if you are interested in reading what I will write... hopefully accepting to co-sign?
I am not a researcher and my field of expertise is in building processing systems and visualisation tools, so I don't plan to write about this, my opinion would not be relevant. Maybe there are altimetry experts outside the SWOT community who would be interested in investigating the physics aspects of this phenomenon with you? Or maybe the SWOT specialists would show interest now if the plots we prepare clearly confirm your hypothesis?
In any case, I would definitely be happy to read what you will write about this.
Hi Sylvain,
I perfectly understand but your help has been so important to get what I will write!
Anyway, if we are "very lucky", the relatively important and purely oceanic lowering sampled south and east of Maupiti during the descending SWOT path on Dec. 30 will be sampled again (more to the west) during the ascending paths. Obviously, feel free to appreciate whether to consider such a path (eventually several) in the mean or not.
Enjoy your holidays.
Cheers,
Claude
Finally, I regrouped the small scale anomaly plots for the three passes (444, 472 and 541). One row corresponds to a SWOT cycle and columns are ordered from West to East:
Hi Sylvain,
I think that we are getting results (about the minimum SLA SWOT values) that are of general interest.
Please, frankly let me know what you think.
Cheers,
Claude
I guess the next step would be to check that similar results are obtained when switching to the sla_unfiltered variable, as we used only the ssha_filtered variable so far. It would highlight:
potential effects of the experimental algorithm used for filtering data on the structures contour. Note that there is a disclaimer regarding the filter: "Experimental algorithm. This parameter is the noise-mitigated counterpart of the ‘ssha_unfiltered’ variable. The noise was mitigated through a machine-learning algorithm (Treboutte et al., 2024). Caution is advised because the algorithm validation is still ongoing: some of the ocean features less than 50km in wavelength may also be affected by the de-noising algorithm"
how the filter discards potentially useful data: as the ssha_unfiltered variable is said to be equivalent to the nadir data where you managed to spot the lowering even on areas where SSHA is masked on my plots, I guess the masked areas in pass 541 on 2024-01-02 and 2024-01-23 would be narrower and more measurements of the lowering would be revealed.
Also please note that I heard about an upcoming new version of SWOT L3 products (at least for products at 2km resolution), so it is important to mention that the plots have been generated using v2_0_1 data. This new version should notably include some corrections/changes related to the geoid and MSS (see https://youtu.be/bsjv4Z2duYk?feature=shared&t=851).
As expected there is a lot more noise, and contrary to my assumption the areas masked in the ssha_filtered variable are also masked in ssha_unfiltered.
Now the areas are not masked anymore, and I think the lowering can still been seen quite clearly among the noise.
Finally, I regrouped plots for the filtered, unfiltered and unedited ssha and added a plot showing ssha_unedited only in the the areas that are masked in the other variables, to show what potentially useful information has been discarded: https://ftp.odl.bzh/odl/sherleda/maupiti...discarded/
I am sorry but I do not have the competencies to think about "your next step". Please remember I am a 77-year physical oceanographer using satellite data only to better understand in situ ones.
I decided to write a few-page note reporting the experiences of a standard user of SLA data made available from the ESA website.
I will emphasize:
1) The interest of daily interpolated nadir data to make statistics and follow a given oceanic structure.
2) The accuracy of the 250-m SLA SWOT data that allows evidencing, in particular, a small scale geodesic variability that should have been removed
3) The efficiency of computing mean data to specify this small scale variability and the interest to remove it from original data to get the oceanic variability
4) The evidence that many SLA data actually rejected because too large do correspond to purely oceanic structures, hence should not be rejected.
As a general comment about your last computations, I would say that I prefer by far dealing with the data as we did (in your last series of plots, the ones in the left are much more pleasant to deal with).
Finally:
Would you accept reading and correcting my note?
Could you propose a journal where to submit it?
YesterdayT07:40 AM (This post was last modified: YesterdayT07:41 AM by Sylvain Herlédan.)
Hi Claude,
The goal of my "next step" was
to show that the removed data actually contained measurements of the lowering. If the plots ended up showing that there was only noise in these areas, or even worse, a positive anomaly, then it would weaken the claim that the algorithm responsible for rejecting data is too heavy-handed. So it felt safer to check.
to identify "when" these data have been removed during the process which generates the nice-looknig SSHA. The SWOT files contain several variables for SSHA, corresponding to different steps of this process, ssha_unedited -> ssha_unfiltered -> ssha_filtered.
As can be seen in the last set of plots I shared (https://ftp.odl.bzh/odl/sherleda/maupiti...discarded/), the culling happened when transitioning from ssha_unedited to ssha_unfiltered, so the filter qualified as "experimental" was not the culprit. And in the plots of the last column, one can see that the removed pixels actually contain measurements of a negative anomaly.
I would of course be glad to read, and, if I feel confident, proof-read your notes.
Regarding the choice of a journal, I don't publish papers so I cannot really help with that.
(YesterdayT07:40 AM)Sylvain Herlédan Wrote: Hi Sylvain,
I do not want taking too much of your time but, for me, on the basis of pass 541 on Jan 2 and 23, considering your "to show that the removed data actually contained measurements of the lowering", I would say that nearly ALL removed data ARE measurements (of the lowering).
I am not competent to soundly discuss your comments but, since you accept reading, eventually proof-reading my note, I propose that you add comments "on my name"...
I will send you my note before sending it to Pascal Bonnefond and Gérald Dibarboure to have their advice and suggestions.
