Generating 10-meter Land Surface Temperature with Sentinel-2, SRTM, and Landsat in Earth Engine

Hi geospatial enthusiast!

Just want to share that I have been experimenting on. I make a script that could help you generate higher resolution land surface temperature (LST) using open source data. Previously, the highest LST data we have is the 100-meter or 30-meter is resamples LST from Landsat TIRS instrument.

Script: https://code.earthengine.google.com/a...

Now, in the script with the help of Landsat LST as label, we could predict what it will look like in 10-meter resolution. Using Sentinel-2 band, indices and its derivatives also Digital Elevation Model (DEM), we extract the raster together with LST.

This extracted value sample then will be assessed to decide which features or bands or variables is the most important (with high R^2 and low correlated with other top features). For this script, only 3 variable are used. Which is B4, dryness, and elevation. The result show a good 0.88 R^2 and Mean Absolute Error of 1.4 Celcius.

This assessment is very limited of course, we don't have a field data but if you want to advanced, continue!

I hope this video can help you on your research, project, or work! Have a good day!