Inhomogeneity effects on cloud shortwave absorption measurements: Two-aircraft simulations

Abstract

We set up an Observation System Simulation in order to quantify the idiosyncratic uncertainties caused by cloud inhomogeneity in two-aircraft measurements. An independently validated fractal model is used to simulate the horizontal variability of optical depth for boundary layer stratus, with and without gaps. A spectral band between 0.9 and 1.0 μm, with a combination of strong and weak water vapor absorption and negligible weak liquid water absorption, is selected for a detailed study of column absorption. Measurements of upwelling and downwelling radiative fluxes are simulated by the Monte Carlo method at two altitudes in a realistic cloudy atmosphere. We consider three methods of estimating column absorption: cloud forcing ratio, reflectance (R) versus transmittance (T) regression, and Ackerman-Cox correction. Emphasis is on the second method, R versus T, which is shown to be biased in the direction of enhanced absorption as long as the R and T data points are affected by horizontal fluxes. In other words, to reduce the bias to an acceptable level, (R, T) measurements must be at a scale well within the regime where the independent pixel approximation is accurate. Under completely cloudy skies, radiative smoothing processes dominate the bias in the absorption estimate at small scales; hence spatial averages over at least 30 times the radiative smoothing scale, which is commensurate with cloud thickness, are required. If there are substantial gaps in the cloud cover, geometrical shadowing and cloud side-illumination effects overwhelm the radiative smoothing. As a result, the bias in R versus T methodology is worsened, and aircraft positioning becomes more critical. We also examine the effect of Sun angle and instrumental parameters such as vertical separation and horizontal offset between aircraft. The Ackerman-Cox correction offers an alternative to spatial averaging; a modification of this intrinsically small-scale approach that can virtually eliminate horizontal flux contributions in the data is presented.