This might be a “harmonization” question, but nonetheless it seems like for the most part when I compare NDVI values from Sentinel-2 and Landsat 8/9 from SH that the values mostly match, except that Landsat 8/9 values are slightly biased low. When I compare level-2 products (surface reflectance) directly from AWS buckets for both sources, Landsat is way different (NDVI values are much lower).
Do you guys apply some harmonization processes to try to get them closer? Just curious why your values appear closer while the AWS bucket ones are not close. Thanks.
Naturally there is some differences between the NDVI values of Sentinel-2 and Landsat 8/9 as they are both different sensors. In addition, they collect data at different spatial resolutions and the spectral range of the Red and NIR bands is also subtly different.
We don’t apply any harmonization processes to get them closer in the separate collections. This is not to be confused with the data harmonization for Sentinel-2 data collections, documented here. Perhaps, the differences you are observing might be in relation to the processing level? You didn’t mention if you are using the L1C or L2A Sentinel-2 data collection, for instance.
And finally, in case you didn’t already know, we also offer a Harmonized Landsat Sentinel data collection provided by USGS. If you are looking for harmonized data, then this is the best one to use.