Collaborative Research: A Flexible Framework for Radiation Parameterizations Traceable to Benchmarks

点击翻译按钮获取中文摘要

Earth's climate is set by the balance between the flow of radiant energy into the atmosphere, largely as visible sunlight, and the return flow of infrared radiation to space. Simulations of earth's climate thus require an approximate representation of these flows which is fast enough to be practical and accurate enough to capture phenomena of interest, such as the warming effect of carbon dioxide (CO2). The development of such representations, referred to as radiative transfer parameterizations or more simply as radiation schemes, is challenging given that the transparency of the atmosphere to infrared radiation can vary abruptly from one wavelength to another. Practical radiation schemes exist, but they are challenging to develop and update and it is not generally feasible to tailor them to specific applications. For example the version of the Community Earth System Model (CESM) available in 2020 uses a radiation scheme published in 2008, despite the availability of new spectroscopic observations, and the same version is used for full-complexity present-day simulations and for idealized simulations that would benefit from a faster but less accurate version.This award supports the creation of a toolkit that would greatly enhance the ability of climate researchers to create radiative transfer parameterizations suitable for their needs. The user would supply a set of benchmark atmospheres, specified as profiles of temperature and composition (water vapor and CO2, for instance), that cover the range of conditions anticipated in the particular application, and the toolbox would use a sophisticated line-by-line radiative transfer model and a database of up-to-date spectroscopic observations to create parameterization options with varying degrees of accuracy and computational cost. Users would then be able to choose the best trade-off between cost and accuracy for their application. In addition, the project creates parameterizations for specific research goals: one for simulations of paleoclimates with high CO2 concentrations and two for studies of the interaction between radiation and convective clouds (one emphasizing speed, which can be invoked more frequently, and one emphasizing accuracy). A third, developed with collaborators at the Geophysical Fluid Dynamics Laboratory (funded through other sources), is optimized for simulations of present-day climate and prediction of climate fluctuations including El Nino events.The project also includes work on alternative methods for parameterizing radiative transfer, one of which is a discrete frequency approximation, in which an optimally chosen set of strictly monochromatic spectral lines is used instead of a representation in terms of frequency bands. The lines are determined from the spectral database using a fast optimization technique such as simulated annealing. The second is a machine learning approach designed to emulate the underlying exact solutions to the radiative transfer equations.The work has broader impacts through the development of a key piece of infrastructure for weather and climate models. The work will thus have benefit for the worldwide community that relies on these models as tools for basic science research. Workshops and tutorials are supported to facilitate community engagement. The work also enhances the value of model-based predictions and projections as guidance for decision makers concerned with climate variability and change. The project supports a postdoc and a graduate student, thus providing for the future workforce in climate model development.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.