Scaling satellite imagery before using it in GEE

Question

I want to conduct an analysis on satellite imagery from the Sentinel-2 mission using Google Earth Engine, and after some research I made, I saw that the data always have to be scaled first before making any calculations, according to the scale-factor mentioned in the Collection's metadata.

This is a bit strange, as a lot of tutorials on the internet are not applying any pre-processing on the imagery before using them.

For example, I followed this tutorial, to create a composite at my area of interest, and then calculate the NDVI values, using the Landsat 7 & 8 Surface Reflectance collections.

Does that mean that my NDVI values might not be correctly calculated? Or because NDVI is an index, therefore a fraction of the NIR and RED bands, is still calculated correctly even with the unscaled values?

Even if that is the case, should I always scale by data before applying any type of analysis?

Answer 1

It is correct that because the normalized difference is a ratio, then the scaling "cancels" out (if both bands have the same scaling factor). In general, you should use the scaling factor.

The reason for the scaling factor could be a technical one, e.g. the data is stored as an unsigned int and thus occupies less storage than in float format.

The tutorials might be using images where the data is already scaled, they might be doing calculations where the scaling factor "cancels" out, or they might be visualizing the images with min/max parameters accounting for the scaling factor. Either way, in my opinion it is a better practice for you to always check the band information and apply the scaling factors.