The verisimilitude of the independent pixel approximation used in cloud remote sensing

We assess the validity of the “independent pixel approximation” (IPA) using Monte Carlo simulations of realistic scale-invariant clouds modeled with 2-D horizontal variations in optical depth. The IPA uses a plane-parallel approximation for each pixel, and is used in virtually all cloud remote sensing algorithms. We confirm the validity of the IPA at the largest scales and demonstrate its shortcomings on the smallest scales: a) It overestimates the variability of the radiation field when the optical depth field is known, and b) it underestimates the variability of the optical depth field when the radiation field is known. Both effects are due to smoothing by horizontal fluxes.