An official website of the United States government
Subgrid parameterizations, which represent physical processes occurring below the resolu- tion of current climate models, are an important component in producing accurate, long-term predictions for the climate. A variety of approaches have been tested to design these com- ponents, including deep learning methods. In this work, we evaluate a proof of concept illustrating a multiscale approach to this prediction problem. We train neural networks to predict subgrid forcing values on a testbed model and examine improvements in prediction accuracy that can be obtained by using additional information in both fine-to-coarse and coarse-to-fine directions.
more »
« less
The high-frequency and rare events barriers to neural closures of atmospheric dynamics
Recent years have seen a surge in interest for leveraging neural networks to parameterize small-scale or fast processes in climate and turbulence models. In this short paper, we point out two fundamental issues in this endeavor. The first concerns the difficulties neural networks may experience in capturing rare events due to limitations in how data is sampled. The second arises from the inherent multiscale nature of these systems. They combine high-frequency components (like inertia-gravity waves) with slower, evolving processes (geostrophic motion). This multiscale nature creates a significant hurdle for neural network closures. To illustrate these challenges, we focus on the atmospheric 1980 Lorenz model, a simplified version of the Primitive Equations that drive climate models. This model serves as a compelling example because it captures the essence of these difficulties.
more »
« less
- Award ID(s):
- 2108856
- PAR ID:
- 10509778
- Publisher / Repository:
- IOP publishing
- Date Published:
- Journal Name:
- Journal of Physics: Complexity
- Volume:
- 5
- ISSN:
- 2632-072X
- Page Range / eLocation ID:
- 025004
- Subject(s) / Keyword(s):
- Neural networks parameterization slow-fast systems atmospheric dynamics
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract We use neural networks and large climate model ensembles to explore predictability of internal variability in sea surface temperature (SST) anomalies on interannual (1–3 years) and decadal (1–5 and 3–7 years) timescales. We find that neural networks can skillfully predict SST anomalies at these lead times, especially in the North Atlantic, North Pacific, Tropical Pacific, Tropical Atlantic and Southern Ocean. The spatial patterns of SST predictability vary across the nine climate models studied. The neural networks identify “windows of opportunity” where future SST anomalies can be predicted with more certainty. Neural networks trained on climate models also make skillful SST predictions in reconstructed observations, although the skill varies depending on which climate model the network was trained. Our results highlight that neural networks can identify predictable internal variability within existing climate data sets and show important differences in how well patterns of SST predictability in climate models translate to the real world.more » « less
-
Čanađija, Marko (Ed.)Neural mechanisms and underlying directionality of signaling among brain regions depend on neural dynamics spanning multiple spatiotemporal scales of population activity. Despite recent advances in multimodal measurements of brain activity, there is no broadly accepted multiscale dynamical models for the collective activity represented in neural signals. Here we introduce a neurobiological-driven deep learning model, termedmultiscale neuraldynamicsneuralordinarydifferentialequation (msDyNODE), to describe multiscale brain communications governing cognition and behavior. We demonstrate that msDyNODE successfully captures multiscale activity using both simulations and electrophysiological experiments. The msDyNODE-derived causal interactions between recording locations and scales not only aligned well with the abstraction of the hierarchical neuroanatomy of the mammalian central nervous system but also exhibited behavioral dependences. This work offers a new approach for mechanistic multiscale studies of neural processes.more » « less
-
As populations worldwide show increasing levels of stress, understanding emerging links among stress, inflammation, cognition, and behavior is vital to human and planetary health. We hypothesize that inflammation is a multiscale driver connecting stressors that affect individuals to large-scale societal dysfunction and, ultimately, to planetary-scale environmental impacts. We propose a “central inflammation map” hypothesis to explain how the brain regulates inflammation and how inflammation impairs cognition, emotion, and action. According to our hypothesis, these interdependent inflammatory and neural processes, and the inter-individual transmission of environmental, infectious, and behavioral stressors—amplified via high-throughput digital global communications—can culminate in a multiscale, runaway, feed-forward process that could detrimentally affect human decision-making and behavior at scale, ultimately impairing the ability to address these same stressors. This perspective could provide non-intuitive explanations for behaviors and relationships among cells, organisms, and communities of organisms, potentially including population-level responses to stressors as diverse as global climate change, conflicts, and the COVID-19 pandemic. To illustrate our hypothesis and elucidate its mechanistic underpinnings, we present a mathematical model applicable to the individual and societal levels to test the links among stress, inflammation, control, and healing, including the implications of transmission, intervention (e.g., via lifestyle modification or medication), and resilience. Future research is needed to validate the model’s assumptions and conclusions against empirical benchmarks and to expand the factors/variables employed. Our model illustrates the need for multilayered, multiscale stress mitigation interventions, including lifestyle measures, precision therapeutics, and human ecosystem design. Our analysis shows the need for a coordinated, interdisciplinary, international research effort to understand the multiscale nature of stress. Doing so would inform the creation of interventions that improve individuals’ lives; enhance communities’ resilience to stress; and mitigate the adverse effects of stress on the world.more » « less
-
Predicting the thermal properties of nanoporous materials is a major challenge that affects their applications in efficient thermal insulation and energy storage. This narrative review discusses the application of machine learning models in nanoporous materials, including covalent organic frameworks, metal–organic frameworks, aerogels, and zeolites. It discusses model advancements, with a focus on predictive accuracy and computational efficiency. This includes models such as convolutional neural networks, graph neural networks, and physics-informed neural networks. This study also addresses the limitations of these data-driven models, including data availability, challenges in maintaining physical consistency, and difficulties in generalizing across various material families. Additionally, it covers emerging approaches such as multimodal and transfer learning, which are explored for their potential to reduce computational costs. Moreover, the benefits of interpretable machine learning methods for understanding underlying physical mechanisms are introduced and highlighted. This review provides comprehensive and practical guidelines for researchers using machine learning approaches in the study and design of nanoporous materials.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript
- Journal Article:
- https://doi.org/10.1088/2632-072X/ad3e59
