For any model _m that simulates a BRF in a given band (λ), viewing (Ω_v) and illumination (Ω₀) geometry over a surface target with structure (ζ) one can compute a corresponding E_n number:

where x is the model simulated BRF, X is the reference BRF under identical conditions, and U is the expanded uncertainty of the model (x) or reference (X). Within RAMI-IV the expanded uncertainty of the model was not known. The expanded uncertainty of the reference was set to twice the standard uncertainty of the reference which was known. If one assumed the standard uncertainty of the model to lie just within the limits of the tolerable uncertainty (that is 3% of the reference BRF here) one can compute Ux assuming a coverage factor of 2 (just as was done for the reference BRF).

Should the absolute value of the resulting En number be larger than unity then one knows that model _m does not satisfy the imposed 3% accuracy criterion.

En numbers were computed for each one of the several thousand BRF simulations of model m, their absolute value was taken and the resulting series was sorted in increasing order. Next, the 95th percentile of this series was chosen. Should the corresponding |E_n|0.95 value be less than unity then at least 95% of all BRF simulations agree with the 3% accuracy criteria for BRFs.