Item

Abstract

The next frontier in radiation science is to resolve cloud-aerosol-radiation interactions at the so-called process scale, which means at one kilometer or better – a significant advance from monitoring radiation at the 20 km footprint scale as done by the NASA’s Earth Observing System radiometers. Consequently, recent missions pursue a different radiation approach. Both EarthCARE and NASA’s AOS employ radiative transfer calculations that ingest imagery-based and active remote sensing products at their native resolution to calculate radiation fields. The imagery-scale calculations are then evaluated using independent radiance observations. In a process called radiance closure, discrepancies between calculations and observations for select observation angles and wavelength ranges are used to quantify and attribute errors, and perhaps even to nudge the remote sensing products towards higher fidelity. We will explain how this approach may address remote sensing biases of aerosol and cloud parameters for inhomogeneous scenes, which are fairly small at the 20 km scale, but can no longer be ignored at 1 km. Part of the solution will likely be convolutional neural networks, which are outgrowing the stigma to be merely qualitative tools without quantitative use for atmospheric remote sensing – to the point that they are now fueling the necessary transition from single-pixel to context-aware imagery retrieval algorithms. They are starting to outperform so-called physics-based algorithms when assessed through the lens of radiance closure. We will illustrate this with real-world examples, and lay out a vision for future cloud-aerosol-radiation interaction studies from aircraft and satellites.