Profiling Capability of High-Resolution Oxygen A-band Spectroscopy for Stratus Cloud Cover

Davis, A. B., Jet Propulsion Laboratory

Cloud Distributions/Characterizations

Cloud Properties

Davis AB, IN Polonsky, and A Marshak. 2009. Space-Time Green Functions for Diffusive Radiation Transport, in Application to Active and Passive Cloud Probing. In Light Scattering Reviews, Volume 4, pp. 169-292. Ed. by A.A. Kohkanovsky, Heidelberg, Germany: Springer.

Transmission: (a) Ratio of mean path <ct>Τ to cloud thickness Η times (1−g)τ plotted versus cosine of SZA μ0 and cloud optical depth τ; asymmetry factor g was set to 0.85, then delta-rescaled to 0.46. Given this ratio (>1/2) and Η or τ, one can infer the other cloud parameter.
(b) As in panel (a) but for the non-dimensional moment ratio (RMS/mean) for ct divided by the asymptotic (high τ) value of √7/5 ≈ 1.18. There is only a slight deviation from unity at high sun for optically thin clouds.

Reflection: (a) Same ratio as in Fig. 1b but for reflection (and without dividing by √7/5). Knowing the SZA, one can infer from this pair of oxygen A-band observables the cloud optical thickness τ.
(b) Same as in Fig. 1a but for the log of the ratio <ct>R/Η (with no further division). Knowing this weakly varying proportionality factor from μ0 and τ (i.e., from the moment ratio), one can infer Η from <ct>R.

Oxygen A-band spectroscopy has long been recognized as a potentially powerful and diverse probe of the Earth's atmosphere from ground, aircraft, and space. However, it is still very much underexploited, particularly in the case of cloudy skies. DOE's ARM Program has supported groundbreaking instrument and algorithm development in O2 A-band spectroscopy at ever higher resolutions and continues to invest in this promising technology both through Science Team efforts by Min and coworkers at SUNY-Albany and through the Small Business Innovation Research (SBIR) program. It has thus been demonstrated, in theory and in practice, that the higher the spectral resolving power and, simultaneously, the more aggressive the out-of-band rejection, the more inherent atmospheric information can be extracted from the A-band data.

The source of radiation, the Sun, is steady but, curiously, the primary products of A-band spectroscopy are actually temporal statistics: the mean and, resolution permitting, a few higher-order moments of the light path. We denote this quantity as ct and define it as the total geometric path covered by the sunlight from cloud top to either cloud base (whence detection by ground-based sensors) or else back to cloud top (where it can eventually be detected by airborne and space-based sensors). Because of multiple scatterings and/or surface reflections, ct is a random variable with a distribution dependent on atmospheric (and, secondarily, surface) conditions. So this is a golden opportunity for cloud remote sensing. What kind of derived cloud products can one obtain from the short sequence of available moments <(ct)q> (q=1,2,...)? Also, what kind of spatial average do they represent?

In their recent review paper on "Space-Time Green Functions for Diffusive Radiation Transport, in Application to Active and Passive Cloud Probing" in Light Scattering Reviews (Vol. 4), Davis, Polonsky and Marshak address these questions systematically for cloud layers with significant optical depth, enough for diffusion theory to apply. Assuming the above-mentioned resolution and out-of-band rejection are sufficient, the answers depend foremost on whether the light is transmitted or reflected by the clouds:

  • If the single cloud layer is reasonably plane-parallel, but not necessarily uniform, one can infer reliably from transmitted light only the optical depth of the layer given its physical thickness, or vice-versa (Figure 1a). Second- and higher-order moments add no further information (Figure 1b). Sensitivity to solar zenith angle (SZA) is small and limited to high sun over tenuous clouds (Figures 1ab). Sensitivity to internal cloud stratification (not illustrated) is also second-order.
  • In sharp contrast, one can obtain from reflected light the cloud extinction profile; this is just as with lidar but for layers far too opaque to be probed directly with a pulsed laser beam. However, the profile is averaged horizontally over the spatial extent of the multiple-scattering Green function, which is on the order of a cloud thickness. More precisely, one can infer cloud-top altitude (a standard A-band product), cloud thickness (hence cloud-base altitude), and volume-average extinction (via cloud optical depth). The proposed method for the 2-parameter retrieval is illustrated in the two panels of Fig. 2, and we recall that the standard cloud top altitude product is biased low, unless we account for the multiple scattering discussed here.

Refined signal modeling by Davis et al. (not illustrated) shows that, with sufficient resolving power, one also can obtain the mean vertical gradient of the extinction coefficient across the cloud, and possibly even a measure of the degree of turbulence-driven random variability superposed on the mean vertical gradient. Alternatively, or with lesser resolution, the A-band observations can be fused profitably with data from other cloud probing instrumentation to derive profile characteristics.

In the reflective case, the authors draw a physics-based analogy between high-resolution O2 A-band spectroscopy and wide-field-of-view (FOV)/multiple-scattering cloud lidar observations (see WAIL highlight) from space. At that extreme stand-off distance, there is no access to spatial information, i.e., no estimation of the horizontal transport from laser beam to point of escape from cloud boundary. (So far, only NASA's 1994 LITE mission had a sufficient FOV for multiple scattering studies; all recent space lidar missions have had an aerosol focus, hence ultra-narrow FOVs.)

In summary, Davis et al.'s comprehensive assessment of the cloud-related information content of high-resolution O2 A-band spectroscopy from below and from above is very encouraging for future spatially and/or temporally continuous observations. This applies equally to instruments at ground-based facilities such as ARM's and to their counterparts aboard aircraft and satellites such as the replacement for OCO, which NASA will hopefully procure very soon.