The current configuration of the ocean overturning involves upwelling predominantly in the Southern Ocean and sinking predominantly in the Atlantic basin. The reasons for this remain unclear, as both models and paleoclimatic observations suggest that sinking can sometimes occur in the Pacific. We present a six-box model of the overturning in which temperature, salinity and low-latitude pycnocline depths are allowed to vary prognostically in both the Atlantic and Pacific. The overturning is driven by temperature, winds, and mixing and modulated by the hydrological cycle. In each basin there are three possible flow regimes, depending on whether low-latitude water flowing into northern surface boxes is transformed into dense deep water, somewhat lighter intermediate water, or light water that is returned at the surface. The resulting model combines insights from a number of previous studies and allows for nine possible global flow regimes. For the modern ocean, we find that although the interbasin atmospheric freshwater flux suppresses Pacific sinking, the equator-to-pole flux enhances it. When atmospheric temperatures are held fixed, seven possible flow regimes can be accessed by changing the amplitude and configuration of the modern hydrological cycle . North Pacific overturning can strengthen with either increases or decreases in the hydrological cycle, as well as under reversal of the interbasin freshwater flux. Tipping-point behavior of both transient and equilibrium states is modulated by parameters such as the poorly constrained lateral diffusive mixing. If hydrological cycle amplitude is varied consistently with global temperature, northern polar amplification is necessary for the Atlantic overturning to collapse.@en

Natural variability at subinertial frequencies (time scale of several days) plays an important role in the interaction between Greenland’s fjords, the continental shelf, and shelf-break exchange with the deep basins. In this study we identified the nature and driving mechanisms of this variability in four fjords in Southeast Greenland, in three high-resolution numerical simulations. We find two dominant frequency ranges in along-fjord velocity, volume transport of Atlantic Water, and along-fjord heat transport: one around 2–4 days and one around 10 days. The higher frequency is most prominent in the two smaller fjords (Sermilik Fjord and Kangerdlugssuaq Fjord), while the lower frequency peak dominates in the larger fjords (Scoresby Sund and King Oscar Fjord). The cross-fjord structure of variability patterns is determined by the fjord's dynamic width, while the vertical structure is determined by the stratification in the fjord. The dominant frequency range is a function of stratification and fjord length, through the travel time of resonant internal Kelvin waves. We find that the subinertial variability is the imprint of Coastal Trapped Waves, which manifest as Rossby-type waves on the continental shelf and as internal Kelvin-type waves inside the fjords. Between 50% and 80% of the variability in the fjord is directly forced by Coastal Trapped Waves propagating in from the shelf, with an additional role played by alongshore wind forcing on the shelf.@en

In the modern ocean, the transformation of light surface waters to dense deep waters primarily occurs in the Atlantic basin rather than in the North Pacific or Southern Oceans. The reasons for this remain unclear, as both models and paleoclimatic observations suggest that sinking can sometimes occur in the Pacific. We present a six-box model of overturning that combines insights from a number of previous studies. A key determinant of the overturning configuration in our model is whether the Antarctic Intermediate Waters are denser than the northern subpolar waters, something that de-pends on the magnitude and configuration of atmospheric freshwater transport. For the modern ocean, we find that al-though the interbasin atmospheric freshwater flux suppresses Pacific sinking, the poleward atmospheric freshwater flux out of the subtropics enhances it. When atmospheric temperatures are held fixed, North Pacific overturning can strengthen with either increases or decreases in the hydrological cycle, as well as under reversal of the interbasin freshwater flux. Tipping-point behavior, where small changes in the hydrological cycle may cause the dominant location of densification of light waters to switch between basins and the magnitude of overturning within a basin to exhibit large jumps, is seen in both transient and equilibrium states. This behavior is modulated by parameters such as the poorly constrained lateral dif-fusive mixing coefficient. If hydrological cycle amplitude is varied consistently with global temperature, northern polar amplification is necessary for the Atlantic overturning to collapse. Certain qualitative insights incorporated in the model can be validated using a fully coupled climate model.

@en

The east Greenland shelf contains southward currents that carry relatively cold and fresh water of Arctic origin that are important to the Arctic’s freshwater budget and the Atlantic Meridional Overturning Circulation (AMOC). Near the shore lies a current named the East Greenland Coastal Current (EGCC), which is a surface intensified southward jet estimated to have a 0.5-2.0 Sv volume transport. The EGCC carries fresher water than the East Greenland Current (EGC), which is found at the shelf break. Due to sparse observations, little understanding exists of the EGCC mean flow, variability, or drivers north of Denmark Strait. Moreover, literature definitions of the EGCC are inconsistent. Therefore, a high-resolution numerical simulation of the region is being used to study the EGCC, with a focus north of Denmark Strait. The performance of five proposed EGCC definitions is assessed by computing volume transport time-series and analyzing velocity cross sections at five model sections along the east Greenland shelf. In addition, the proposed EGCC definitions are applied to an observational time-series at 60°N from the OSNAP East extension. Results show that the mean EGCC transport values differ by up to ~3 Sv at the southern sections (south of 66°N) and ~0.6 Sv at the northern sections (north of 66°N). EGCC definitions based on a specific isohaline performed well in the southern sections but include part of the EGC in the northern sections and, on some occasions, exclude the bottom part of the EGCC. EGCC definitions based on velocity thresholds work better to capture the EGCC but can extend too far towards the EGC. Therefore, we recommend using an EGCC definition based on a velocity threshold and including a criterion on distance from the shore.@en

We describe a new multidisciplinary effort to understand how Artificial Intelligence (AI) could be used to improve climate tipping point discovery using the collapse of the Atlantic Meridional Overturning Circulation (AMOC) as a case study. Our methodology includes an AI simulated environment where a Generative Adversarial Network (GAN) is used to play an adversarial game between two deep networks. One network, the generator, tries to invoke an AMOC collapse tipping point. The other network, the discriminator, tries to predict which model configurations lead to an AMOC collapse, and which do not. The discriminator uses a climate surrogate model to run model configurations suggested by the generator network. In this study we have explored using a well-researched reduced oceanography four-box model (Gnanadesikan et al. 2018) as the surrogate. In addition, we introduce a new type of surrogate model that is used for expanding to larger models, which is based on modeling stochastic differential equations to estimate bifurcation escape times, enabling a deeper understanding of the unstable areas and how to escape them. We describe how we apply this method to larger Global Circulation Models (GCMs) by means of a CESM2-calibrated dataset. Included in this methodology is a neuro-symbolic representation of the model configurations that translates into natural language questions enabling climate modelers to ask questions of the learned adversarial space and to provide explainability. We describe the results of applying this method to study the AMOC, and evaluate our method by comparing how well it learns to predict AMOC collapse, how well it learns the area of uncertainty in state space, and how effective the neuro-symbolic language is in providing a natural language interface into the adversarial game.@en

Fjords are the primary conduit for oceanic heat transport to the Greenland Ice Sheet. This heat transport is highly time-variable, with typical frequencies in the subinertial range (with periods of several days). We studied four fjords along the southeast coast of Greenland to determine what drives oceanic heat into these fjords. The four fjords cover a range of widths and lengths, and three different high-resolution model simulations provided a range in stratifications and surface forcings. We find that the heat transport is associated with resonant modes of internal subinertial waves. The variability in transport of deep warm (Atlantic) water is highly correlated to the principal component that is associated with the gravest vertical normal mode (EOF 1 in narrow fjords, EOF 3 in a wide fjord). The dominant frequency in the heat transport time series is thus determined by the shape of the fjord and the stratification, the latter through the wave speed of the gravest normal mode. The oceanic heat supply to the Greenland Ice Sheet is thus governed by variables that are relatively easy to measure observationally.@en

We propose a hybrid Artificial Intelligence (AI) climate modeling approach that enables climate modelers in scientific discovery using a climate-targeted simulation methodology based on a novel combination of deep neural networks and mathematical methods for modeling dynamical systems. The simulations are grounded by a neuro-symbolic language that both enables question answering of what is learned by the AI methods and provides a means of explainability. We describe how this methodology can be applied to the discovery of climate tipping points and, in particular, the collapse of the Atlantic Meridional Overturning Circulation (AMOC). We show how this methodology is able to predict AMOC collapse with a high degree of accuracy using a surrogate climate model for ocean interaction. We also show preliminary results of neuro-symbolic method performance when translating between natural language questions and symbolically learned representations. Our AI methodology shows promising early results, potentially enabling faster climate tipping point related research that would otherwise be computationally infeasible.@en

In 2018, the IPCC summarized in a special report the potential risks surrounding climate tipping point. In 2019, Lenton et al. highlighted climate tipping points that could contribute to irreversible changes to our world including ice melt, deforestation, and circulation slowing. In this work we describe a study of the Atlantic Meridional Overturning Circulation (AMOC), and the use of an Artificial Intelligence (AI) assisted climate model methodology capable of performing unsupervised tipping point discovery to discover conditions which could lead to AMOC collapse.

We developed a novel AI generative adversarial network (GAN), where a set of deep learning generators attempt to invoke AMOC collapse by perturbing a constrained set of parameters, while another deep learning network, the discriminator, tries to learn how to avoid AMOC collapse. A surrogate model is used to run model configurations to test this adversarial method. We show that our methodology can be used to discover areas in model space that are consistent with fold bifurcations where the system moves from an on state to an off state. We measured the performance of this method by comparing it to an AMOC four box model and experiments described in (Gnanadesikan 2018) which uses the four box model to understand overturning stability.

We have found that the deep learning method can be used to exploit the area of uncertainty that is consistent with the area that separates the two stable states in a fold bifurcation model. When we compared the results of the adversarial network to the Gnanadesikan experiments we observed that when incorporating information regarding the uncertainty in the loss function, increasing the number of AI generators caused the AI agents to become more focused on this area of uncertainty. This area of uncertainty is consistent with what is described as the separatrix. In this study, we show the benefit of using this novel unsupervised approach as part of an AI assisted climate modeling methodology.@en

This study examines the applicability of the Gnanadesikan four-box model of the Atlantic Meridional Overturning Circulation to a global earth system model. The Community Earth System Model 2 (CESM2) is used to quantify the extent to which pycnocline depth and AMOC strength as predicted by the four-box model from regional densities and Southern Ocean fluxes match CESM2 values. The dynamics of a reduced meridional density difference driving a deepening pycnocline and weakening AMOC are qualitatively matched between the 4-box model and CESM2. We will quantify the fraction of variability in the CESM2 Large Ensemble of historical and SSP3-7.0 runs that is explained by the 4-box model; initial results suggest about two-thirds of the variability in AMOC is explained and over 90% of the variability in the pycnocline depth. This analysis may be repeated for any CMIP6 model, and a comparison of SSP5-8.5 across a handful of models is included as a demonstration.@en

Denmark Strait, the channel located between Greenland and Iceland, is a critical gateway between the Nordic Seas and the North Atlantic. Mesoscale features crossing the strait regularly enhance the volume transport of the Denmark Strait overflow. They interact with the dense water masses descending into the subpolar North Atlantic and therefore are important for the Atlantic Meridional Overturning Circulation. Using a realistic numerical model, we find new evidence of the causal relationship between overflow surges (i.e., mesoscale features associated with high-transport events) and overflow cyclones observed downstream. Most of the cyclones form at the Denmark Strait sill during overflow surges and, because of potential vorticity conservation and stretching of the water column, grow as they move equatorward. A fraction of the cyclones form downstream of the sill, when anticyclonic vortices formed during high-transport events start collapsing. Regardless of their formation mechanism, the cyclones weaken starting roughly 150 km downstream of the sill, and potential vorticity is only materially conserved during the growth phase.@en

Multimodel Arctic Ocean “climate response function” experiments are analyzed in order to explore the effects of anomalous wind forcing over the Greenland Sea (GS) on poleward ocean heat transport, Atlantic Water (AW) pathways, and the extent of Arctic sea ice. Particular emphasis is placed on the sensitivity of the AW circulation to anomalously strong or weak GS winds in relation to natural variability, the latter manifested as part of the North Atlantic Oscillation. We find that anomalously strong (weak) GS wind forcing, comparable in strength to a strong positive (negative) North Atlantic Oscillation index, results in an intensification (weakening) of the poleward AW flow, extending from south of the North Atlantic Subpolar Gyre, through the Nordic Seas, and all the way into the Canadian Basin. Reconstructions made utilizing the calculated climate response functions explain ∼50% of the simulated AW flow variance; this is the proportion of variability that can be explained by GS wind forcing. In the Barents and Kara Seas, there is a clear relationship between the wind‐driven anomalous AW inflow and the sea ice extent. Most of the anomalous AW heat is lost to the atmosphere, and loss of sea ice in the Barents Sea results in even more heat loss to the atmosphere, and thus effective ocean cooling. Release of passive tracers in a subset of the suite of models reveals differences in circulation patterns and shows that the flow of AW in the Arctic Ocean is highly dependent on the wind stress in the Nordic Seas.@en

The Southeast Greenland shelf region harbors various exchange flows that play an important role in the climate system, such as the Denmark Strait Overflow and fjord-shelf exchange. Most of the literature on these flows focuses on slow processes such as estuarine circulation and hydraulically-controlled flow over a sill. Nevertheless, there is overwhelming evidence that the volume transport variability of these processes is dominated by subinertial oscillations (timescale of several days). In this study we use a high-resolution ocean model to explore and characterize subinertial variability along the Southeast Greenland coast. We study time series and spatial patterns of sea level anomalies along the shelf break, and time series of dense-water volume transport across several sections along the coast. Using a wavelet analysis of the time series and an evaluation of the time evolution of spatial anomaly patterns, we identify strong evidence for coastal-trapped wave modes at around 4 and 10 day frequencies, propagating along the coast. We discuss the importance of these results in the light of sparse observational data and their interpretation.@en

The Denmark Strait Overflow (DSO) is an important contributor to the lower limb of the Atlantic Meridional Overturning Circulation (AMOC). Determining DSO formation and its pathways is not only important for local oceanography but also critical to estimating the state and variability of the AMOC. Despite prior attempts to understand the DSO sources, the upstream pathways and circulation remain uncertain due to short term (3-5 days) variability, mixing and watermass transformation. This makes it challenging to study the DSO from observations. Given this complexity, this study maps the upstream pathways and along-pathway changes in its water properties, using Lagrangian backtracking of the DSO sources in a realistic numerical ocean simulation. The model is compared with Eulerian and Lagrangian observations. Eulerian fields are compared with historical mooring and CTD sections, and particle pathways are compared with RAFOS float data. The Lagrangian pathways confirm that several branches contribute to the DSO from the North, namely, the East Greenland Current, the separated EGC, and the North Icelandic Jet. Moreover, the results reveal an additional pathway from south of Iceland along the Iceland shelf, which supplies 26% of the DSO during winter of 2008 when the NAO index is positive. The southern contribution is as high as 34% by the end of March. This southern pathway presents a short circuit for the North Atlantic Deep Water and needs to be explored further, especially in observations. Finally, an updated schematic DSO circulation diagram is proposed.@en

About half of the dense overflow across the Greenland-Iceland-Scotland ridge is supplied by the Denmark Strait Overflow (DSO), making Denmark Strait a critical gateway between the Arctic and the subpolar North Atlantic. Boluses and pulses are mesoscale features that cross the Denmark Strait during high-overflow transport events. They occur with the same frequency as the DSO cyclones observed downstream. However, the connection between boluses, pulses, and DSO cyclones is not well established. To investigate this connection, we designed an automated vortex detection algorithm and we applied it to a realistic, high-resolution numerical model solution. Our goal is threefold: (i) To better understand the dynamics associated with stretching of the water column in the DSO; (ii) To determine how the hydrographic properties of DSO mesoscale features evolve in space and time; (iii) To advance the understanding of the high-frequency variability characteristic of the Denmark Strait Overflow. The model shows that both boluses and pulses produce DSO cyclones, but the underlying mechanisms are different. DSO cyclones associated with boluses form in Denmark Strait and, because of potential vorticity conservation and stretching of the water column, they quickly grow as they move south. In contrast, DSO cyclones associated with pulses form downstream of the sill as slow-moving anticyclones near the East Greenland shelfbreak collapse. Regardless of their formation mechanism, DSO cyclones weaken when they encounter a rapid steepening of the bathymetric slope, and the rate of decay is slower than the earlier growth rate. The mean stratification of the cyclones decreases during their life cycle, while their initial increase in relative vorticity due to stretching is followed by a decrease. As a result, potential vorticity is only conserved during the growth phase. Last, composites of the deep cyclones show that they are preceded by downwelling and followed by upwelling during their life cycle.@en

OceanSpy is an open-source and user-friendly Python package that enables scientists and inter- ested amateurs to analyze and visualize oceanographic data sets. OceanSpy builds on software packages developed by the Pangeo community, in particular Xarray (Hoyer & Hamman, 2017), Dask (Dask Development Team, 2016), and Xgcm (“Xgcm,” n.d.). The integration of Dask facilitates scalability, which is important for the petabyte-scale simulations that are becoming available. OceanSpy can be used as a standalone package for analysis of local circulation model output, or it can be run on a remote data-analysis cluster, such as the Johns Hopkins University SciServer system (Medvedev, Lemson, & Rippin, 2016), which hosts several sim- ulations and is publicly available. OceanSpy enables extraction, processing, and visualization of model data to (i) compare with oceanographic observations, and (ii) portray the kinematic and dynamic space-time properties of the circulation.@en

Increasingly warm Atlantic Water inflow has been linked to decreased sea ice cover in the Barents Sea. The heat brought in by the Atlantic Water contains enough heat to melt all the sea ice in the Arctic. However, the strong Arctic halocline successfully traps the heat well below the sea ice. With retreating ice cover and changing Atlantic Water inflow, it is not clear to what extent that will still hold true in the future. We use a moderately high-resolution ocean-sea ice model with realistic bathymetry, but idealized forcing cycles, to study the change in vertical heat flux to the sea ice from the Atlantic layer through the halocline for various Atlantic inflow scenarios. We show how far the influence of the Atlantic Water reaches, how that affects the sea ice, and how that depends on varying forcing conditions.@en

Denmark Strait is dynamically relevant to the global climate system because the dense water that overflows through the strait is a major contributor to the Deep Western Boundary Current. Observations show that the Denmark Strait Overflow (DSO) pulses with high frequencies (2-5 days). The hydrography and circulation of our high-resolution numerical simulations covering the east Greenland shelf and the Iceland and Irminger Seas is consistent with available observations in Denmark Strait. The model shows that boluses and pulses, the two dominant mesoscale features observed in the strait, play a major role in controlling the variability of the DSO thickness and transport into the Irminger Sea. Indeed, the mean southward volume flux of the DSO in the model is about 30% greater in the presence of boluses and pulses. In this study we take advantage of our ~2 km resolution model to interpret the sparse observations available for this area. We describe the evolution of boluses and pulses downstream of Denmark Strait by investigating several vertical sections across the east Greenland shelf. Moreover, we use a Lagrangian particle code to track boluses and pulses and to study their hydrographic properties along the descending pathways. Combining the Eulerian and Lagrangian approaches we isolate boluses and pulses from the background state and we investigate the circulation in the Irminger basin associated with the mesoscale features.@en