DOE’s Atmospheric Radiation Measurement (ARM) Program has strongly supported efforts to improve the realism of the radiative transfer models used in climate science for cloudy skies, both for radiation budget estimation in global climate models (GCMs) and in cloud remote sensing from ground and from space. This improvement is registered in the spectral domain of course, but also in the spatial domain by going from 1D to 3D representations of clouds. ARM’s Science Team has also taken a leading role in cloud-process modeling with large-eddy simulation (LES) techniques and cloud-resolving models (CRMs). Recently, the CSU group has embedded CRMs in GCM grid-cells as so-called “super-parameterizations.”

This championship of innovative modeling of radiation and dynamics notwithstanding, ARM is primarily an observational program and it has been realized that its extensive federations of instruments and networked observation facilities has an inherent bias towards vertical profiling versus full 3D imaging of the cloudy atmosphere. This is partly due to technological limitations at the time when measurement strategies were designed almost two decades ago, but is also traceable to the conventional wisdom about synoptic-scale atmospheric “dynamics” being quasi-2D (unfolding in the horizontal) while small-scale—hence parameterized—atmospheric “physics” are quasi-1D processes (unfolding in the vertical), driven by the 2D advective tendencies and providing dynamical feedback through mass and energy balances). This conceptualization impacts the modeling of radiative transfer (plane-parallel/slab cloud models arise), cloud microphysics (parcel models arise), radiative-convective balance models, and so on, but also how we collect data. It is undeniable that—gravity oblige—the atmosphere (clouds included) is highly stratified at the largest scales. However, everyday observation tells us that, arguably up to scales commensurate with a few pressure scale-heights, structure of the cloudy atmosphere is truly 3D. Observation and data-processing technologies are now at the point where 3D cloud imaging is feasible on such scales. True to its programmatic commitments, ARM’s management is giving serious consideration to a staged investment of resources into the proposed ARM Volume-imaging Array (AVA) of scanning and profiling cloud-sensitive radars.

AVA is designed to address issues in at least three cutting-edge research areas in climate science that will benefit directly from continuous access to detailed 3D cloud tomography over a limited region:

1. radiative transfer in atmospheric columns populated by 3D clouds, where AVA will provide the required input;
2. satellite-based remote sensing of cloud properties over a wide range of pixel sizes, where AVA will provide critical information about unresolved variability when pixels are large and help understand 3D adjacency effects when pixels are small;
3. in cloud process modeling with LESs and CRMs, where AVA will provide badly needed validation data.

For technical details about AVA and another ARM initiative in 3D cloud tomography using microwave emission, we refer to the Symposium presentations by Wiscombe and Kollias as well as Dong and Wiscombe. In the present paper, we will argue that there is another emerging observational technology that complements cloud tomography: high-resolution differential absorption spectroscopy of molecular oxygen lines in the so-called “A-band” (760–770 nm).

The primary product of O₂-line spectroscopy under cloudy skies is a short sequence of low-order moments of the non-trivial distribution of path lengths for solar radiation resulting from the multiple scattering. Somewhat paradoxically in view of the steady solar source involved, the underpinning radiative transfer of O₂ A-band spectroscopy is time-dependent. This is also the case in atmospheric lidar with, however, one major difference: lidar is predicated on a single scattering event, usually in the backwards direction with respect to the well-collimated laser-beam illumination. Consequently, lidar receivers are designed with very narrow fields-of-view: no more than is necessary to enclose the transmitted beam, within tolerance for misalignment. That simple propagation geometry is key to the “ranging” in lidar. In contrast, A-band spectroscopy uses light that has undergone any number of scatterings. (We exclude from this comparison the few experimental off-beam/multiple scattering cloud lidar systems in existence; otherwise, lidars and A-band spectrometers should in fact be considered as the same class of instruments as defined by the information content of their signals. For instance, airborne lidar yields cloud-top height and so does A-band in the single-/first-scattering limit.)

It is widely recognized that 1D profiling cloud radars and lidars work exceptionally well together, largely because they operate at such widely separated wavelengths that they bring complementary information about clouds; to wit, ARM’s popular ARSCL (Active Remote Sensing of CLouds) product and NASA’s co-manifested CloudSat and CALIPSO missions. Similarly, we see A-band as the natural complement to radar- or microwave-based 3D cloud tomography for three simple reasons:

• Tomography determines 3D cloud structure and path length statistics from A-band are known to be highly sensitive to spatial variability in cloudiness; in our talk, we will survey O₂ A-band response to spatial variability of clouds. However, A-band’s characteristically solar wavelength is directly relevant to climate while the “cloud picture” painted at radar/microwave frequencies may be less germane, hence the possibility of a climate-driven reality check for AVA using A-band data.
• While AVA is an imaging tool, in full 3D, A-band is a powerful integrator in space using the natural smoothing scale of multiple scattering, which is commensurate with the distance between the base of the lowest cloud to the top of the highest one (and may be further amplified by multiple ground reflections). In the presence of multiple scattering, the radiative smoothing scale is the smallest that matters in climate work.
• Finally, oxygen A-band spectroscopy is all about the key climate process of atmospheric absorption. Because it involves a well-mixed non-GHG, A-band can inform us about how well or poorly we compute absorption by any gas in the presence of clouds, e.g., using GCM shortwave parameterizations (assuming perfect knowledge of clouds, from AVA of course).

In conclusion, we believe that robust and continuous A-band spectroscopy will be the keystone of the bridge between upcoming 3D cloud tomography, whether using active (radar) or passive (microwave) modalities, and their above-stated goals (1) and (2) in direct support of atmospheric radiation science.