Sonnewald, Maike, Krissy Anne Reeve, and Redouane Lguensat, May 2023: A Southern Ocean supergyre as a unifying dynamical framework identified by physics-informed machine learning. Communications Earth and Environment, 4, 153, DOI:10.1038/s43247-023-00793-7. Abstract
The Southern Ocean closes the global overturning circulation and is key to the regulation of carbon, heat, biological production, and sea level. However, the dynamics of the general circulation and upwelling pathways remain poorly understood. Here, a physics-informed unsupervised machine learning framework using principled constraints is used. A unifying framework is proposed invoking a semi-circumpolar supergyre south of the Antarctic circumpolar current: a massive series of leaking sub-gyres spanning the Weddell and Ross seas that are connected and maintained via rough topography that acts as scaffolding. The supergyre framework challenges the conventional view of having separate circulation structures in the Weddell and Ross seas and suggests that idealized models and zonally-averaged frameworks may be of limited utility for climate applications. Machine learning was used to reveal areas of coherent driving forces within a vorticity-based analysis. Predictions from the supergyre framework are supported by available observations and could aid observational and modelling efforts to study this climatologically key region undergoing rapid change.
Clare, Mariana C., Maike Sonnewald, Redouane Lguensat, Julie Deshayes, and V Balaji, November 2022: Explainable artificial intelligence for Bayesian neural networks: Toward trustworthy predictions of ocean dynamics. Journal of Advances in Modeling Earth Systems, 14(11), DOI:10.1029/2022MS003162. Abstract
The trustworthiness of neural networks is often challenged because they lack the ability to express uncertainty and explain their skill. This can be problematic given the increasing use of neural networks in high stakes decision-making such as in climate change applications. We address both issues by successfully implementing a Bayesian Neural Network (BNN), where parameters are distributions rather than deterministic, and applying novel implementations of explainable AI (XAI) techniques. The uncertainty analysis from the BNN provides a comprehensive overview of the prediction more suited to practitioners' needs than predictions from a classical neural network. Using a BNN means we can calculate the entropy (i.e., uncertainty) of the predictions and determine if the probability of an outcome is statistically significant. To enhance trustworthiness, we also spatially apply the two XAI techniques of Layer-wise Relevance Propagation (LRP) and SHapley Additive exPlanation (SHAP) values. These XAI methods reveal the extent to which the BNN is suitable and/or trustworthy. Using two techniques gives a more holistic view of BNN skill and its uncertainty, as LRP considers neural network parameters, whereas SHAP considers changes to outputs. We verify these techniques using comparison with intuition from physical theory. The differences in explanation identify potential areas where new physical theory guided studies are needed.
Ocean acidification is a consequence of the absorption of anthropogenic carbon emissions and it profoundly impacts marine life. Arctic regions are particularly vulnerable to rapid pH changes due to low ocean buffering capacities and high stratification. Here, an unsupervised machine learning methodology is applied to simulations of surface Arctic acidification from two state-of-the-art coupled climate models. We identify four sub-regions whose boundaries are influenced by present-day and projected sea ice patterns. The regional boundaries are consistent between the models and across lower (SSP2-4.5) and higher (SSP5-8.5) carbon emissions scenarios. Stronger trends toward corrosive surface waters in the central Arctic Ocean are driven by early summer warming in regions of annual ice cover and late summer freshening in regions of perennial ice cover. Sea surface salinity and total alkalinity reductions dominate the Arctic pH changes, highlighting the importance of objective sub-regional identification and subsequent analysis of surface water mass properties.
Irrgang, Christopher, Niklas Boers, and Maike Sonnewald, et al., August 2021: Towards neural Earth system modelling by integrating artificial intelligence in Earth system science. Nature Machine Intelligence, 3, DOI:10.1038/s42256-021-00374-3667-674. Abstract
Earth system models (ESMs) are our main tools for quantifying the physical state of the Earth and predicting how it might change in the future under ongoing anthropogenic forcing. In recent years, however, artificial intelligence (AI) methods have been increasingly used to augment or even replace classical ESM tasks, raising hopes that AI could solve some of the grand challenges of climate science. In this Perspective we survey the recent achievements and limitations of both process-based models and AI in Earth system and climate research, and propose a methodological transformation in which deep neural networks and ESMs are dismantled as individual approaches and reassembled as learning, self-validating and interpretable ESM–network hybrids. Following this path, we coin the term neural Earth system modelling. We examine the concurrent potential and pitfalls of neural Earth system modelling and discuss the open question of whether AI can bolster ESMs or even ultimately render them obsolete.
Sonnewald, Maike, and Redouane Lguensat, August 2021: Revealing the impact of global heating on North Atlantic circulation using transparent machine learning. Journal of Advances in Modeling Earth Systems, 13(8), DOI:10.1029/2021MS002496. Abstract
The North Atlantic ocean is key to climate through its role in heat transport and storage. Climate models suggest that the circulation is weakening but the physical drivers of this change are poorly constrained. Here, the root mechanisms are revealed with the explicitly transparent machine learning (ML) method Tracking global Heating with Ocean Regimes (THOR). Addressing the fundamental question of the existence of dynamical coherent regions, THOR identifies these and their link to distinct currents and mechanisms such as the formation regions of deep water masses, and the location of the Gulf Stream and North Atlantic Current. Beyond a black box approach, THOR is engineered to elucidate its source of predictive skill rooted in physical understanding. A labeled data set is engineered using an explicitly interpretable equation transform and k-means application to model data, allowing theoretical inference. A multilayer perceptron is then trained, explaining its skill using a combination of layerwise relevance propagation and theory. With abrupt CO2 quadrupling, the circulation weakens due to a shift in deep water formation regions, a northward shift of the Gulf Stream and an eastward shift in the North Atlantic Current. If CO2 is increased 1% yearly, similar but weaker patterns emerge influenced by natural variability. THOR is scalable and applicable to a range of models using only the ocean depth, dynamic sea level and wind stress, and could accelerate the analysis and dissemination of climate model data. THOR constitutes a step toward trustworthy ML called for within oceanography and beyond, as its predictions are physically tractable.
Sonnewald, Maike, Redouane Lguensat, Daniel C Jones, Peter D Dueben, Julien Brajard, and V Balaji, July 2021: Bridging observations, theory and numerical simulation of the ocean using machine learning. Environmental Research Letters, 16(7), DOI:10.1088/1748-9326/ac0eb0. Abstract
Progress within physical oceanography has been concurrent with the increasing sophistication of tools available for its study. The incorporation of machine learning (ML) techniques offers exciting possibilities for advancing the capacity and speed of established methods and for making substantial and serendipitous discoveries. Beyond vast amounts of complex data ubiquitous in many modern scientific fields, the study of the ocean poses a combination of unique challenges that ML can help address. The observational data available is largely spatially sparse, limited to the surface, and with few time series spanning more than a handful of decades. Important timescales span seconds to millennia, with strong scale interactions and numerical modelling efforts complicated by details such as coastlines. This review covers the current scientific insight offered by applying ML and points to where there is imminent potential. We cover the main three branches of the field: observations, theory, and numerical modelling. Highlighting both challenges and opportunities, we discuss both the historical context and salient ML tools. We focus on the use of ML in situ sampling and satellite observations, and the extent to which ML applications can advance theoretical oceanographic exploration, as well as aid numerical simulations. Applications that are also covered include model error and bias correction and current and potential use within data assimilation. While not without risk, there is great interest in the potential benefits of oceanographic ML applications; this review caters to this interest within the research community.
Sonnewald, Maike, Redouane Lguensat, Aparna Radhakrishnan, Zoubero Sayibou, V Balaji, and Andrew T Wittenberg, 2021: Revealing the impact of global warming on climate modes using transparent machine learning and a suite of climate models In ICML 2021 Workshop on Tackling Climate Change with Machine Learning, . Abstract
