“The cloud problem” is that uncertainties in representing clouds in numerical models of Earth's climate and weather prevent accurate predictions. The U.S. Department of Energy's ARM (Atmospheric Radiation Measurement) Program stepped up to continuous, multi-decade measurements aimed at “the cloud problem” starting in the early 1990s, including measurements in every part of the electromagnetic spectrum from ultraviolet to radar. Today, ARM operates state-of-the-art cloud radars, lidars, and microwave radiometers, all in vertically-pointing mode, that have substantially improved our ability to detect clouds and obtain range-resolved measurements of their internal microphysical and dynamical structure.

ARM began in the mid-1990s to combine cloud radar and lidar measurements to delineate cloud boundaries, an effort called ARSCL (Active Remote Sensing of Cloud Location). Neither radar nor lidar alone are able to accomplish this seemingly simple task for all cloud conditions. This task proved surprisingly difficult and took years to get right. Once accomplished, this led to a decade-long record that has proved invaluable in testing the validity of, and suggesting improvements to, cloud-overlap parameterizations in climate models. Later, ARM upgraded its cloud radars to include polarimetric capability, Doppler spectral recording, and higher time resolution; it also moved beyond ARSCL to include retrievals of particle size and hydrometeor phase.

To close the problem of radiative fluxes in atmospheric columns populated by 3D clouds, ARM must leave behind its vertically-pointing paradigm and combine scanning radar, lidar and radiometer measurements to map cloud and precipitation over a horizontal domain of roughly 50 km. Plans are currently being laid for this advance, involving networks of scanning radars, some inherited from the CASA program and some custom-built for ARM in collaboration with NASA's GPM Ground Validation Program. Studies are also being done of cloud tomography to supplement the scanning radars. The multifarious challenges in putting together such a system will be outlined, including difference in sensitivity and attenuation, cost and data interpretation.