Predicting atmospheric optical properties for radiative transfer computations using neural networks.

The radiative transfer equations are well known, but radiation parametrizations in atmospheric models are computationally expensive. A promising tool for accelerating parametrizations is the use of machine learning techniques. In this study, we develop a machine learning-based parametrization for the gaseous optical properties by training neural networks to emulate a modern radiation parametrization (RRTMGP). To minimize computa- tional costs, we reduce the range of atmospheric conditions for which the neural networks are applicable and use machine-specific optimized BLAS functions to accelerate matrix computations. To generate training data, we use a set of randomly perturbed atmospheric profiles and calculate optical properties using RRTMGP. Predicted optical properties are highly accurate and the resulting radiative fluxes have average errors within 0.5 W m-2 compared to RRTMGP. Our neural network-based gas optics parametrization is up to four times faster than RRTMGP, depending on the size of the neural networks. We further test the trade-off between speed and accuracy by training neural networks for the narrow range of atmospheric conditions of a single large-eddy simulation, so smaller and therefore faster networks can achieve a desired accuracy. We conclude that our machine learning-based parametrization can speed-up radiative transfer computations while retaining high accuracy. This article is part of the theme issue 'Machine learning for weather and climate modelling'.

Advancing understanding of land-atmosphere interactions by breaking discipline and scale barriers.

Vegetation and atmosphere processes are coupled through a myriad of interactions linking plant transpiration, carbon dioxide assimilation, turbulent transport of moisture, heat and atmospheric constituents, aerosol formation, moist convection, and precipitation. Advances in our understanding are hampered by discipline barriers and challenges in understanding the role of small spatiotemporal scales. In this perspective, we propose to study the atmosphere-ecosystem interaction as a continuum by integrating leaf to regional scales (multiscale) and integrating biochemical and physical processes (multiprocesses). The challenges ahead are (1) How do clouds and canopies affect the transferring and in-canopy penetration of radiation, thereby impacting photosynthesis and biogenic chemical transformations? (2) How is the radiative energy spatially distributed and converted into turbulent fluxes of heat, moisture, carbon, and reactive compounds? (3) How do local (leaf-canopy-clouds, 1 m to kilometers) biochemical and physical processes interact with regional meteorology and atmospheric composition (kilometers to 100 km)? (4) How can we integrate the feedbacks between cloud radiative effects and plant physiology to reduce uncertainties in our climate projections driven by regional warming and enhanced carbon dioxide levels? Our methodology integrates fine-scale explicit simulations with new observational techniques to determine the role of unresolved small-scale spatiotemporal processes in weather and climate models.