Why does radiometric response vary between images?

The PlanetScope constellation includes satellites with different relative spectral responses
(RSRs). Using PS data therefore requires an additional harmonization step to compare between
different generations of satellites. Furthermore, PS surface reflectance (SR) data is often variable for a
variety of other reasons, resulting in the need for a normalization step in addition to harmonization.

Directional reflectance effects are not corrected for in PS SR data, therefore stray light can be present in some satellites, and the radiative transfer models used for SR corrections are based on approximations that may have errors at low sun angles and other adverse conditions.

In addition, SR corrections for PS data rely on MODIS estimates of atmospheric optical thickness collected at a different time of day and at a coarse spatial resolution. This leads to a significant degree of variability in PS SR data even within data with the same RSR.

In addition, there is variability between generations of satellites even after applying a global harmonization correction.
As a result, using Planetscope data often requires additional processing to correct for both factors
before analysis. The scene-level normalization tool provides a method to both approximately
harmonize between the varying RSRs and to normalize the data to correct for variability due to other
parameters. This provides increased consistency, though absolute radiometric accuracy can suffer in
some cases.