Transparency of the water column (Zsd and Kd)
#############################################

Transparency products are based on optical satellite data from the Sentinel-2 constellation. It is generated in near real-time (NRT) once a week for the entire France and overseas area and derived from the Remote Sensing Reflectances (Rrs) products. 
The products are of two kinds : Secchi-disk depth estimation Zsd (m) and its optical counterpart the diffuse attenuation coefficient Kd (m-1).

Retrieval methodology
*********************

Products are derived from a single Sentinel-2 L2A that has been corrected for atmospheric contribution (see section about :ref:`rrs`). 
Two different algorithms are then applied to retrieve the estimations of Kd and Zsd from bottom of atmosphere Rrs derived from the atmospheric corrections.
The diffuse attenuation coefficient (Kd) is obtained with the chain described in  from the article (Lee 2013, Pitarch 2021), 
whilst the Secchi depth (Zsd) is obtained from a chaine produced by Magellium and the ECLA centre.


Product limitations
*******************
As Level-2A optical Rrs estimates are the input data used to derive transparency products, uncertainties in Rrs are consequently propagated to Level-2B data. 
Moreover, no data can be produced when clouds are present in the atmosphere. Clouds can be present over the scenes but the masks are not applied directly over the level 2B produced. 
Therefore, the user needs to specify (mask==0) to eliminate land and clouds to obtain usable transparency data.
The Kd RGB is supposed to be robust as it derives from the QAA-rgb version (Pitarch and Vanhellemont, 2021) that adapted the estimations of Inherent Optical Properties (IOPs) by the QAA algorithm (Lee et al., 2002) to the RGB channels. 
We used QAA rgb outputs within the Lee et al., (2013)’s estimate of Kd to finally derive Kd rgb.
For inland waters and in complex waters, IOPs are hardly resolved correctly which may alter the Kd estimates on certain sites.
Finally, Secchi-disk Depth Zsd rgb is then derived from Kd rgb and QAA rgb estimates as described above and therefore relies on the good estimation of Kd rgb.