Editors’ Vox is a blog from AGU’s Publications Department.
Computer models are essential for understanding atmospheric phenomena and predicting Earth’s climate and weather. Many complex processes occur on small scales and short time frames that are not accurately captured in large-scale models. A new book published by AGU in the Geophysical Monograph Series, Fast Processes in Large-Scale Atmospheric Models: Progress, Challenges and Opportunities, explores how to represent these processes in models. We asked the book’s editors to explain the concept of fast processes, outline some recent advances in understanding and suggest what research questions remain.
In simple terms, what are “fast processes”?
Computer models are essential tools for understanding atmospheric phenomena and for making accurate predictions of any changes in the Earth’s climate and weather.
In such models, the whole atmosphere (Earth) is divided into many grids with the number of grid points (model resolution) limited by computational resources. Weather and climate models are based on the area of interest being divided into equal grids.
Fast processes are the physical processes that occur at spatial scales smaller than a single grid in a large-scale model and at temporal scales shorter than the timestep in a large-scale model.
Fast processes, or subgrid (scale) processes, are the physical processes that occur at spatial scales smaller than a single grid in a large-scale model and at temporal scales shorter than the timestep in a large-scale model.
A myriad of such fast processes exists and we use the term “fast physics” to generically represent the collection of all fast processes.
Fast processes cannot be directly resolved or predicted in the model grid and, as a result, they are represented using approximate equations called “parameterizations”.
Why is it important to understand fast processes and how they relate to atmospheric models?
Fast processes constitute an integral part of the atmospheric and climate system and their effects need to be represented as accurately as possible in large-scale models.
First, fast processes, such as those related to clouds and aerosols, constitute an integral part of the atmospheric and climate system and their effects need to be represented as accurately as possible in large-scale models. Studies have shown that deficient fast process parameterizations, especially those related to clouds, have been primarily responsible for the stubborn large uncertainty in model climate sensitivity; how climate models respond to climate forcing perturbations.
Second, fast processes cannot be resolved directly at the grid level and have to be approximately represented through various “parameterizations”. It is obviously desirable to make such parameterizations as accurate as possible within the available computational power and resources.
Third, fast processes are challenging to measure and occur in complex, interacting, and non-linear ways which we refer to as “4M-2N complexity” in the book. As a result, we understand much less about them than the resolved processes in large-scale models with significant knowledge gaps in some important processes such as turbulent entrainment-mixing and aerosol-cloud interactions.
What are some recent developments in representing fast processes in models?
We would highlight the following as exciting new developments:
- Fast physics parameterizations have been gradually advancing from being primarily a practical fix for individual fast processes to deep theoretical underpinning of stochastic scale interactions, and process interactions within the complex “4M-2N” system.
- Several different approaches/ideas have been developed to unify the representation of distinct processes and their interactions. The importance and necessity of top-down approaches have been increasingly recognized and ideas in other disciplines, such as statistical physics and information theory, have found valuable applications in atmospheric and climate sciences.
- The essential role of organic integration of numerical models of different types, synthesis of measurements from different observing platforms, and effective evaluation of various parameterization schemes has been increasingly emphasized for further development. The growing interest in a systematic, process-oriented model-measurement synthesis further reinforces the pressing need for interdisciplinary research.
- With the advancement in computer technology and growing computational power, model resolutions have been continually increasing, along with the development of unstructured or resolution-adaptive computing meshes.
- More physical processes that were ignored in early studies have started to be considered, for example, turbulent entrainment-mixing processes.
What do you think are the next big areas for research in this field?
Recent advances pose new challenges that will likely drive future research and lead to new directions in atmospheric modeling. To highlight a few points:
- Fast physics parameterizations will continue to be major sources of uncertainty in weather and climate prediction models. In particular, those occurring at sub-Large Eddy Simulation (LES) scales and fraught with significant knowledge gaps will likely become more acute as the resolutions of weather and climate models increase. Process modeling at ultra-high resolutions such as particle-resolved direct numerical simulation models that can resolve the smallest turbulent eddies and follow individual cloud particles, together with measurement technologies at similarly high resolutions, hold great potentials to fill these critical knowledge gaps.
- Most fast physics parameterizations remain essentially one dimensional in the vertical direction; such 1D representation will likely experience a shift to more realistic 3D representation for processes such as radiative transfer and gravity waves. This shift will pose new challenges not only to representing individual processes in question but more so to considering the process interactions (e.g., gravity-turbulence interactions) and complex coherent and irregular structures.
- It is anticipated that artificial intelligence and machine learning will find increasing use in virtually every area of atmospheric modeling, including parameterization development, model evaluation and parameter calibration, synthesis of data collected at different resolutions and with different platforms, and development of digital twins for model-measurement integration.
How is your book structured?
Addressing the topic of fast physics in large-scale models requires developing adequate parameterizations, conducting measurements for process understanding and model evaluation, and seamless integration between modeling and measurements. A unique feature of the book is bringing these different aspects together.
The book is divided into three parts. Part I discusses major subgrid processes covering different processes more or less in the conventional compartmentalized format that emphasizes individual processes. Part II explores promising approaches that have recently emerged to unify the treatment of individual processes and thus allow for consideration of process interactions. Part III looks at remote-sensing and laboratory measurements, model evaluation, and model-measurement integration.
What type of reader will find this book most useful?
The book is targeted at researchers and graduate students. Readers will find the first chapter a helpful introduction to the historical development of fast physics parameterizations and the involved complexities. Thereafter, each chapter can be read separately as a stand-alone chapter or in conjunction with other chapters. The final chapter summarizes emerging challenges, new opportunities, and future research directions.
Fast Processes in Large-Scale Atmospheric Models: Progress, Challenges, and Opportunities, 2023. ISBN: 978-1-119-52899-9. List price: $230 (hardcover), $184 (ebook).
Chapter 1 is freely available. Visit the book’s page on Wiley.com and click on “Read an Excerpt” below the cover image.
—Yangang Liu ([email protected]; 
0000-0003-0238-0468), Brookhaven National Laboratory, USA; and Pavlos Kollias (0000-0002-5984-7869), Brookhaven National Laboratory and Stony Brook University, USA
Editor’s Note: It is the policy of AGU Publications to invite the authors or editors of newly published books to write a summary for Eos Editors’ Vox.
