A multi-sensor approach to the retrieval and model validation of global cloudiness

Abstract

The ephemeral clouds have represented a daunting challenge to the atmospheric modeling community from the very beginning. Namely, their far-reaching importance to the energy balance and hydrological cycle of our planet is eclipsed only by the level of complexity at which they are nested within the non-linear climate feedback system. Too often are clouds oversimplified in atmospheric models simply for lack of a sound physical understanding of their integral role in the system. Historically, our knowledge of atmospheric variables has been steeped in a tradition of observations, and clouds are no exception. Hence, the onus falls squarely upon atmospheric remote sensing to provide this burden of truth through detailed, accurate observations of the physical system. Our inability to resolve by means of traditional passive sensors the detail of information required to characterize clouds has challenged us to explore other resources at our disposal. In this spirit, this research seeks to illustrate and, where applicable, quantify the ways in which active (e.g., radar and lidar) remote sensing devices on existing and proposed platforms can serve to improve our current understanding of cloud and cloud processes. Two perspectives are emphasized: 1) the role of active sensors in the improvement of cloud property retrievals and 2) their application to the validation/development of clouds in numerical weather prediction models. First, a new retrieval technique which employs active-sensors (e.g., lidar and cloud radar) to constrain cloud boundaries in the vertical is shown to decrease significantly the errors incurred by traditional passive satellite retrievals (which approximate cloud heights by various emission or shadow-geometry techniques), and those of night-time cirrus in particular. Based on results from twelve independent single-layer cloud case studies using reflected radiance/brightness-temperature information from GOES, the CSU SSP, and MAS, the a priori data were shown to improve the retrieval uncertainty of night-time cirrus optical properties in excess of 20% for effective particle radius, and 10-20% for optical depth. Uncertainties associated with solar reflection based single-layer retrievals showed improvements typically less than 5% (however, the unexplored benefits of active cloud profile information in the case of day-time multi-layered clouds remain significant). The retrieval results are brought together with detailed cloud profile sampling from the Lidar In-space Technology Experiment (LITE) to conduct the first global-scale active sensor validation of the ECMWF prognostic cloud scheme. Short-range (20-30 hr) forecasts matched to 66 night-time LITE orbits display remarkable agreement in cloud spatial distribution both in the horizontal and vertical. A weighted statistical analysis revealed hit rates between 75-90%, threat scores 45-75%, probabilities of detection ≈ 80%, and false alarm rates 10-45%. Retrieved cloud properties were converted to equivalent cloud liquid/ice water paths and compared against model forecasts. The findings of these comparisons reveal that while there is obvious room for improvement, the level of realism displayed currently by the ECMWF prognostic cloud scheme forecasts suggest that the reanalysis data may be considered as a new resource for global cloud information. A practical application of these findings has been outlined in the context of defining Cloud-Sat instrument requirements based on virtual orbital observations created from ECMWF global cloud distributions of liquid and ice water contents. The outcomes of the research are a near-operational GOES retrieval scheme with performance diagnostics, an quantified assessment of the benefits provided to cloud property retrievals using a multi-sensor observing system that includes active components, and a new perspective on the currently ability of numerical weather prediction models to forecast clouds. This final outcome gives cause for new hope in capturing the complex radiative, convective, and dynamical feedback mechanisms associated with clouds in the climate feedback system. Insodoing, this research appeals to the need for an improved collaborative rapport between the now largely disjoint modeling and measurement communities. Establishment of this link is argued to be the key step toward realizing the true potential of multi-sensor data in addressing the cloud/climate problem.