It’s great to see HLS being available on SentinelHub! I’m interested in tracking NDVI for an AOI and I was wondering how comparable the results are with NDVI from Sentinel-2 L2A.
The inputs to HLS are Sentinel-2 L1C and Landsat 8/9 (both of top-of-atmosphere reflectance), whereas Sentinel-2 L2A is of bottom-of-atmosphere reflectance. According to HLS documentation, atmospheric correlation is applied to both inputs:
A series of radiometric and geometric corrections are applied to convert data to surface reflectance, adjust for BRDF differences, and adjust for spectral bandpass differences.
“convert to surface reflectance” → Does it mean that HLS is in fact of bottom-of-atmosphere reflectance?
Apologies in advance if this question is more suitable to HLS authors (USGS) but I thought I could get a quicker response here
It is probably worth confirming with the HLS authors, but referring to the data processing diagram linked in your post and our docs, the HLS is bottom-of-atmopshere / surface reflectance. However, it’s important to note that this is processed differently than a Sentinel-2 L2A product.
I think the final two bulletpoints summarise it nicely:
S30: MSI harmonized surface reflectance resampled to 30 m into the Sentinel-2 tiling system and adjusted to Landsat 8 spectral response function.
L30: OLI harmonized surface reflectance and Top-of-Atmosphere (TOA) brightness temperature resampled to 30 m into the Sentinel-2 tiling system.
