I’m currently using band S8 to create thermal maps but the BT values I get don’t make sense, specially when compared to Landsat 8 values for the same day (see image bellow). The evalscript and request I use in Python are the following:
I’ve been having a look into this issue for you and have been coming up with similar results until I have spotted what might be causing this!
Due to the fast return time of Sentinel 3, you can get multiple acquisitions in one day. For example, my AOI had a Landsat 8 acquisition at 10:01am and a Sentinel 3 acquisition at 10:37am. I was also seeing large difference between the land pixels of each acquisition but a much smaller difference between the ocean pixels.
However, looking further into this I found that there was an additional Sentinel 3 acquisition at 10pm that evening too. This explains the large error in land pixels but a much smaller error in the ocean pixels. The land cools much more at night while the ocean retains its heat and varies far less day to day
By default the SentinelHub will always return the most recent image even if your time range is only one day. Therefore, by adding the “mosaickingOrder”: “leastRecent” parameter to my request, I forced the API to return the 10:37am Sentinel 3 acquisition which got rid of the large difference in thermal response between the two sensors.
I would suggest trying this and seeing if this solves your problem