All surface reflectance ground measurements were spectrally resampled to the spectral resolution of the satellite sensors by applying a spectral convolution with the sensor’s response function. For the comparison, the four closest pixels to the station location were selected and their reflectance values retrieved. For a small number of sites, the number of pixels was reduced due to heterogeneous terrain. The estimated reflectances for each processor were then averaged over each target area, and the mean and standard deviation were compared to the validation measurement and its uncertainty.
The Spectral Angle Mapper (SAM) metric was used to give an indication of the similarity between retrieval and reference across the whole spectrum. For the general assessment, APU plots were used to compare the processors’ performances, comparing the hyperspectral in-situ measurements at every band.
The Accuracy (A), Precision (P), and Uncertainty (U) metrics assess the processors’ performance in retrieving AOD. They are computed starting with the difference between the retrieved value 𝑦_𝑖 and the reference 𝑦_(𝑅𝐸𝐹,_𝑖 ) for observation 𝑖, i.e.,
Then, the APU metrics are calculated based on Equations (A) to (U). The APU metrics are given in the respective units i.e., % for SR.