Climate models are complicated software systems that approximate atmospheric and oceanic fluid mechanics at a coarse spatial resolution. Typical climate forecasts only explicitly resolve processes larger than 100 km and approximate any process occurring below this scale (e.g. thunderstorms) using so-called parametrizations. Machine learning could improve upon the accuracy of some traditional physical parametrizations by learning from so-called global cloud-resolving models. We compare the performance of two machine learning models, random forests (RF) and neural networks (NNs), at parametrizing the aggregate effect of moist physics in a 3 km resolution global simulation with an atmospheric model. The NN outperforms the RF when evaluated offline on a testing dataset. However, when the ML models are coupled to an atmospheric model run at 200 km resolution, the NN-assisted simulation crashes with 7 days, while the RF-assisted simulations remain stable. Both runs produce more accurate weather forecasts than a baseline configuration, but globally averaged climate variables drift over longer timescales.
The Late Pleistocene Antarctic temperature variation curve is decomposed into two parts: cyclic and stochastic. These two parts represent different but tightly interconnected processes and also represent two different types of self-organization of the Earth climate system. The self-organization in the cyclic component is the non-linear auto-oscillation reaction of the Earth climate system, as a whole, to the input of solar radiation. The self-organization in the stochastic component is a nonlinear critical process, taking energy from, and fluctuating around the cyclic component of the temperat…
The Late Pleistocene Antarctic temperature variation curve is decomposed into two components: cyclic and high frequency, stochastic. For each of these components, a mathematical model is developed which shows that the cyclic and stochastic temperature variations are distinct, but interconnected, processes with their own self-organization. To model the cyclic component, a system of ordinary differential equations is written which represent an auto-oscillating, self-organized process with constant period. It is also shown that these equations can be used to model more realistic variations in tem…
We compare the performance of a recently proposed empirical climate model based on astronomical harmonics against all available general circulation climate models (GCM) used by the IPCC (2007) to interpret the 20th century global surface temperature. The proposed model assumes that the climate is resonating with, or synchronized to a set of natural harmonics that have been associated to the solar system planetary motion, mostly determined by Jupiter and Saturn. We show that the GCMs fail to reproduce the major decadal and multidecadal oscillations found in the global surface temperature record…
Power spectra of global surface temperature (GST) records reveal major periodicities at about 9.1, 10-11, 19-22 and 59-62 years. The Coupled Model Intercomparison Project 5 (CMIP5) general circulation models (GCMs), to be used in the IPCC (2013), are analyzed and found not able to reconstruct this variability. From 2000 to 2013.5 a GST plateau is observed while the GCMs predicted a warming rate of about 2 K/century. In contrast, the hypothesis that the climate is regulated by specific natural oscillations more accurately fits the GST records at multiple time scales. The climate sensitivity to …
We explore the potential of feed-forward deep neural networks (DNNs) for emulating cloud superparameterization in realistic geography, using offline fits to data from the Super Parameterized Community Atmospheric Model. To identify the network architecture of greatest skill, we formally optimize hyperparameters using ~250 trials. Our DNN explains over 70 percent of the temporal variance at the 15-minute sampling scale throughout the mid-to-upper troposphere. Autocorrelation timescale analysis compared against DNN skill suggests the less good fit in the tropical, marine boundary layer is driven…
Cloud microphysical parameterizations in atmospheric models describe the formation and evolution of clouds and precipitation, a central weather and climate process. Cloud-associated latent heating is a primary driver of large and small-scale circulations throughout the global atmosphere, and clouds have important interactions with atmospheric radiation. Clouds are ubiquitous, diverse, and can change rapidly. In this work, we build the first emulator of an entire cloud microphysical parameterization, including fast phase changes. The emulator performs well in offline and online (i.e. when coupl…
University of Lausanne（洛桑大学）；Columbia University（哥伦比亚大学）；Massachusetts Institute of Technology（麻省理工学院）；University of California, Irvine（加利福尼亚大学尔湾分校）；Google Research（谷歌研究院）；University of California, Los Angeles（加利福尼亚大学洛杉矶分校）；University of California, San Diego（加利福尼亚大学圣迭戈分校）；NVIDIA（英伟达）
Reply to "Comment on 'Advanced Testing of Low, Medium, and High ECS CMIP6 GCM Simulations Versus ERA5-T2m' by N. Scafetta (2022)" by Schmidt, Jones, and Kennedy (2023)
The equilibrium climate sensitivity (ECS) of the CMIP6 global circulation models (GCMs) varies from 1.83 °C to 5.67 °C. Herein, 38 GCMs are grouped into three ECS classes (low, 1.80-3.00 °C; medium, 3.01-4.50 °C; high, 4.51-6.00 °C) and compared against the ERA5-T2m records from 1980-1990 to 2011-2021. We found that all models with ECS > 3.0 °C overestimate the observed global surface warming and that spatial t-statistics rejects the data-model agreement over 60% (using low-ECS GCMs) to 81% (using high-ECS GCMs) of the Earth's surface. Thus, the high and medium-ECS GCMs are unfit for predictio…
From stable online coupling to decade-long climate simulations: A machine learning parameterization for cloud microphysics in ICON
Deutsches Zentrum für Luft- und Raumfahrt e.V.（德国航空航天中心）；University of Oxford（牛津大学）；European Centre for Medium-Range Weather Forecasts（欧洲中期天气预报中心）；University of Bremen（不来梅大学）
Evaluating Skill and Stability of ArchesWeather and ArchesWeatherGen under Multi-Decadal Climate Simulations
Inria Paris（巴黎国家信息与自动化研究所）；Google DeepMind（谷歌DeepMind）；Otto-von-Guericke-University Magdeburg（马格德堡奥托·冯·格里克大学）；University of Potsdam（波茨坦大学）；European Centre for Medium-Range Weather Forecasting（欧洲中期天气预报中心）；University of Colorado Boulder（科罗拉多大学博尔德分校）
Forecasting the El Nino-Southern Oscillation (ENSO) has been a subject of vigorous research due to the important role of the phenomenon in climate dynamics and its worldwide socioeconomic impacts. Over the past decades, numerous models for ENSO prediction have been developed, among which statistical models approximating ENSO evolution by linear dynamics have received significant attention owing to their simplicity and comparable forecast skill to first-principles models at short lead times. Yet, due to highly nonlinear and chaotic dynamics (particularly during ENSO initiation), such models hav…
Navigating the Noise: Bringing Clarity to ML Parameterization Design with O(100) Ensembles
Expertise Center for Climate Extremes, University of Lausanne（洛桑大学气候极端事件专业中心）；NVIDIA Research（英伟达研究院）；Department of Biomedical Data Science, Stanford University School of Medicine（斯坦福大学医学院生物医学数据科学系）；Department of Earth System Sciences, University of California at Irvine（加州大学欧文分校地球系统科学系）；Multimodal Cognitive AI, Intel Labs（英特尔实验室多模态认知人工智能部门）；Faculty of Geosciences and Environment, University of Lausanne（洛桑大学地球科学与环境学院）；Department of Statistics, University of California at Irvine（加州大学欧文分校统计系）；Department of Earth and Planetary Sciences, Harvard University（哈佛大学地球与行星科学系）；LEAP Science and Technology Center, School of Engineering and Applied Sciences, Climate School, Columbia University（哥伦比亚大学气候学院工程与应用科学学院LEAP科技中心）；Berkeley AI Research (BAIR), Department of Electrical Engineering and Computer Science, University of California at Berkeley（加州大学伯克利分校电气工程与计算机科学系伯克利人工智能研究中心）
