Temperature of the surface of water or skin-temperature (Tskin)
###############################################################

The surface water temperature product (Tskin) is derived from Thermal Infrared (TIR) satellite data collected by the TIR sensors onboard the Landsat 8 and 9 missions. 
It is generated in near real-time (NRT), once a week, for the entire France mainland and overseas regions. 
The product is pixel-wise and has a spatial resolution of 30 m x 30 m. Products derived from thermal bands have a resolution of 100 m but are resampled to 30 m.

Retrieval methodology
*********************
The products are derived from the two thermal bands of Landsat 8 and 9 at level L1C. 
Skin temperature is determined using a split window algorithm (Jimenez-Munoz et al. 2014). 
The use of a split-window algorithm allows the temperature to be retrieved directly from the L1C, 
avoiding the atmospheric correction step by using the signal difference between the two thermal bands. 
In addition to satellite images, the algorithm uses the water vapour content that we retrieve from the CAMS data produced by the ECMWF [#f1]_ , 
and the emissivity values of water (which are considered constant in a given wavelength) in both infrared bands of 0.998 at 10.5µm and 0.992 at 12µm (Qui et al. 2006). 
For Landsat-8/9-OLI, the L2A cloudmasks produced by USGS are used.



Product limitations
*******************
The determination of water surface temperature from Landsat thermal data is subject to several limitations. 
The spatial resolution of the thermal bands (100 m for TIRS, commonly resampled to 30 m) limits the ability to capture fine-scale thermal variability, 
especially in heterogeneous landscapes, as well as on narrower objects such as rivers, ponds and elongated waterbodies.
The relatively long revisit time (16 days for a single satellite, 8 days combined for Landsat 8 and 9) restricts the monitoring of temperature changes and transient events.
Additionally, the noise equivalent delta temperature (NEΔT) introduces uncertainty, particularly for small temperature differences or low-contrast surfaces. 
This characteristic is specific to the instrument and induces a fairly high degree of uncertainty (Jimenez-Munoz et al. 2014), which imposes an uncertainty that cannot be reduced by modifying the algorithm.
Finally, edge effects can occur at boundaries between land cover types or at scene edges, leading to mixed pixels and reduced accuracy in temperature estimation.
It is also important to note that this refers to the surface temperature – that is, the top few millimetres of the water column – which may differ significantly from the temperature at depth.

.. rubric:: Footnotes
.. [#f1] European Centre for Medium-Range Weather Forecasts (https://atmosphere.copernicus.eu/)