C2RCC_more_information

Is it possible to see the equations that the C2RCC uses? Is there any other documentation for this tool except the paper mentioned in the snap help?

I am interested in chlorophyll-a retrieval .

Thank you in advance

1 Like

There is only one additional document. It will be included in the help with the next release.
It is specific for MERIS, but it explains also the general concept: CRCC_MERIS_ATBD_4Reproc_20150319.pdf
You can have a look at the source code if you like: GitHub Repository
However, the C2RCC processor uses neural nets and therefore there are no equations available.

Thanks for the answer. In some point of the document says

For each simulation run (case) a random value of the following parameters is selected from a
uniform distribution:
sun zenith angle: 0 -75 deg
view zenith angles: 0 – 60 deg
view azimuth angle: 0 -180 deg
surface pressure: 800 – 10140 hPa
aerosol optical thickness (aot at 550 nm): 0.0 – 0.8
angstrom coefficient: 0.0 – 2.5
wind speed: 0 – 10 m s-1
water temperature: 0 – 36 deg C
salinity: 0 – 43
For the corresponding sun and viewing angle and temperature and salinity the water leaving
irradiance reflectance is computed using the forward NN.

So when you define the salinity, temperture,pressure in the parameters are they used?

My point is to define the exact values of this parameters and see if they will have any effect on the chlorophyll and surface reflectances.

Yes, the parameters are used but the influence on the results is not that big.
However, it would be interesting what your comparison reveals.

dear marpet,
I have a doubt in turn of the processing parameters options in the C2RCC processor. There is the posibility to change the TSM factor and Chla factor. However I do not know how to find the correct factors for my study area (Beagle Channel). Furthermore, we have in situ samples to compare the results and I am very excited about them
Thanks in advance

Xime

1 Like

Hi Ximena!

These factors are the conversion from IOPs to concentrations on the bases of empirical relationships. For OLCI you have:

chl = 21.0 ∗ (a_pig443)^1.04
tsm= 1.06 * (bp + bw)^0.942

The factors and exponents were derived with NOMAD data mainly (and some coastal data). If you have your own data measured in situ, you could try to derive this relationship by yourself and change the factors conveniently to your study area. That should improved the conversion to fit ranges closer to your expectations.

2 Likes

Thanks Ana
I will take a look in Werdel 2005 where the NOMAD dataset and algorithms are described. However I see in the map with the datasets that many points near my study area are considered.
Maybe it is not necessary to change those factors in my case, I don’t know.

I really appreciate your help today

1 Like

Hi Ximena

If you compare your in-situ data with the results of the C2RCC computation you will see if they match already quite well. As you said, maybe no adaptation is needed.

Thank you Marco
I am trying to apply de C2RCC processor to a resampled and subseted (both at 10 meters) product level L1C, but it takes so long
My computer is a i5-5 with 8Gb RAM and SSD 480Gb.
Does it should be faster or is normal?

Yes, at a resolution of 10m it takes some time, especially when only having 8GB. What do you mean by “it takes long”. Is it hours or days?
Do speed up you can disable the computation of uncertainty bands. This takes a good part of the computing time.

It takes a few hours but I will be processing several products for different dates in the future, so I need to make this a little more dynamic.

I’m going to try these two options

  • lower the spatial resolution
  • ignore the uncertainty bands

Thank you very much, Marco

If you are processing several scenes, use GPT. It helps reducing processing time.

Wow! now it flies! less than a minute using a subset of 10m resolution (B2 band) and resampling with 60m (B1 band).
however the chla_conc and tsm_conc looks pixelated while the resampled bands look perfectly fine.
I will continue testing with the 10m resolution but ignoring the uncertity bands.

Yes, the results can be a bit noise. This is a known issue.