A hybrid data assimilation algorithm is developed for complex dynamical systems with partial observations. The method starts with applying a spectral decomposition to the entire spatiotemporal fields, followed by creating a machine learning model that builds a nonlinear map between the coefficients of observed and unobserved state variables for each spectral mode. A cheap low-order nonlinear stochastic parameterized extended Kalman filter (SPEKF) model is employed as the forecast model in the ensemble Kalman filter to deal with each mode associated with the observed variables. The resulting ensemble members are then fed into the machine learning model to create an ensemble of the corresponding unobserved variables. In addition to the ensemble spread, the training residual in the machine learning-induced nonlinear map is further incorporated into the state estimation that advances the quantification of the posterior uncertainty. The hybrid data assimilation algorithm is applied to a precipitation quasi-geostrophic (PQG) model, which includes the effects of water vapor, clouds, and rainfall beyond the classical two-level QG model. The complicated nonlinearities in the PQG equations prevent traditional methods from building simple and accurate reduced-order forecast models. In contrast, the SPEKF model is skillful in recovering the intermittent observed states, and the machine learning model effectively estimates the chaotic unobserved signals. Utilizing the calibrated SPEKF and machine learning models under a moderate cloud fraction, the resulting hybrid data assimilation remains reasonably accurate when applied to other geophysical scenarios with nearly clear skies or relatively heavy rainfall, implying the robustness of the algorithm for extrapolation.

How to cite: Mou, C., Smith, L. M., and Chen, N.: Combining Stochastic Parameterized Reduced-Order Models with Machine Learning for Data Assimilation and Uncertainty Quantification with Partial Observations , EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-1335, https://doi.org/10.5194/egusphere-egu23-1335, 2023.

On average every 4 years, the sea-surface temperature in the Eastern Equatorial Pacific is a few degrees higher than normal. This phenomenon, which reaches its maximum usually around Christmas is known as El Niño. This event has a strong influence on the climate all around the globe through well-known tele-connections. The occurrence of El Niño is related to extreme weather events, that affect people and properties. For these reasons is important to better understand the behavior of this climatic phenomenon. The property of EL Niño we have focused on during our project is related to the following research question: Is El Niño only due to external noise, or it is a self-sustained phenomenon, which amplitude is amplified by the noise?

To answer to this question, we have applied a Machine Learning tool called Reservoir Computer. After the training procedure, through feedback connections, the Reservoir model can be transformed into an autonomous evolving system. Our results show that the autonomous evolving Reservoir can delete the noise from the training data. The signal produced in output by the autonomous evolving Reservoir reflects the patterns of the training data, without noise. This method can therefore be used to understand if the EL Niño oscillations is only due to random noise, that excites a steady state, or it is a periodic phenomenon, which amplitude is randomly increased by external noise. To understand its limitations, our approach has been first applied to data produced by different models, that simulate EL Niño (Jin Timmerman, Zebiak Cane and CESM). After these first experiments, performed in a controlled scenario, our method has been applied to real data, to see what the self-evolving Reservoir model can tell us about the real EL Niño phenomenon.

How to cite: Guardamagna, F., Dijkstra, H., and Weiners, C.: Is El Niño only due to the noise or it is a self-sustained phenomenon?, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-8645, https://doi.org/10.5194/egusphere-egu23-8645, 2023.

In this study, we present a deep learning approach to deriving a reduced-order model of stochastically forced atmospheric zonal jets. The approach provides a four orders of magnitude speed-up in simulating the jets, over numerical integration, together with a lower-degrees-of-freedom latent representation of the system- used to yield insight into the underlying dynamics.

We consider the behaviour of zonal jets on a beta plane as represented by a two-dimensional model driven by stochastic forcing, which parameterises the turbulence due to baroclinic instability. This idealised model gives a useful analogue for week-to-week variations in the large-scale dynamics of the tropospheric midlatitude jet - the driver of European weather. We establish that the time evolution of the jets depends both on the nonlinear two-way interaction between the mean flow and the eddies and, crucially, the time history of the stochastic forcing. As a result, the current state or recent history of the system does not predict the forward evolution but instead determines a distribution of possible time evolutions.

To model the flow, we utilise methods in manifold learning to learn a transformation to a latent representation of the system and then use a probabilistic neural network to model the stochastic latent dynamics. We verify the neural network’s performance by comparing the statistical and spectral properties of an ensemble from the neural network, obtained via sampling in the latent space, with an ensemble of numerical integrations, with different realisations of the stochastic forcing- with identical initial conditions. To study jet variability, we use ensembles of trajectories in both the latent and observation space to quantify to what extent different system states are driven by deterministic or stochastic dynamics.

How to cite: Shokar, I., Haynes, P., and Kerswell, R.: Learning Stochastic Dynamics with Probabilistic Neural Networks to study Zonal Jets, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-9121, https://doi.org/10.5194/egusphere-egu23-9121, 2023.

Due to their fast computational performance, neural networks (NNs) are nowadays commonly used in the context of remote sensing. The issue of performance is especially important in the context of big data and operational processing. Classical retrieval algorithms often use a radiative transfer model (RTM) as forward model with which an optimization algorithm can then solve the inverse problem of inferring the quantities of interest from the measured spectra. However, these RTMs are usually computationally very expensive and therefore replacing them by a NN is desirable to increase performance. But the application of NNs is not straightforward and there are at least two main approaches:

1. NNs used as forward model, where a NN approximates the radiative transfer model and can thus replace it in the inversion algorithm

2. NNs for solving the inverse problem, where a NN is trained to infer the atmospheric parameters from the measurement directly

The first approach is more straightforward to apply. However, the inversion algorithm still faces many challenges, as the spectral fitting problem is generally ill-posed. Therefore, local minima are possible and the results often depend on the selection of the a-priori values for the retrieval parameters.

For the second case, some of these issues can be avoided: no a-priori values are necessary, and as the training of the NN is performed globally, i.e. for many training samples at once, this approach is potentially less affected by local minima. However, due to the black-box nature of a NN, no indication about the quality of the results is available. In order to address this issue, novel methods like Bayesian neural networks (BNNs) or invertible neural networks (INNs) have been presented in recent years. This allows the characterization of the retrieved values by an estimate of uncertainty describing a range of values that are probable to produce the observed measurement. We apply and evaluate these new BNN and INN methods for the retrieval of cloud properties from TROPOMI in order to demonstrate their potential as operational algorithms for current (Sentinel-5P) and future (Sentinel-4 and Sentinel-5) Copernicus atmospheric composition missions.

How to cite: Romahn, F., Loyola, D., Doicu, A., Molina García, V., Lutz, R., and Argyrouli, A.: Uncertainty quantification for the retrieval of cloud properties with deep neural networks for TROPOMI / Sentinel-5 Precursor, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-10972, https://doi.org/10.5194/egusphere-egu23-10972, 2023.

Reduced order quasi-geostrophic ocean-atmosphere coupled models provide a platform that preserve key atmosphere behaviours, while still being simple enough to allow for analysis of the system dynamics. These models produce typical atmospheric dynamical features like atmospheric blocking and other low-frequency variability, while having a low number of degrees of freedom. For this reason, these models are well suited to investigating tipping points or bifurcations in the Earth's climate due to their simplified but insightful dynamics.

In our present work we compare the dynamics of an ocean-atmosphere coupled model, previously implemented with linearised temperature equations (Vannitsem et al., 2015), but here we solve the equations including the non-linear Stefan-Boltzmann law in the radiative temperature term. When compared with the original version of the model with linearised temperature equations, the modified version of the model is found to produce multiple stable flows in the coupled ocean-atmosphere system. We find, for increasing atmospheric emissivity, there is an increase in the number of stable attractors, and these stable attractors present distinct flows in the ocean and atmosphere and distinct Lyapunov stability properties.

Vannitsem, S., Demaeyer, J., De Cruz, L., & Ghil, M. (2015). Low-frequency variability and heat transport in a low-order nonlinear coupled ocean–atmosphere model. Physica D: Nonlinear Phenomena, 309, 71-85.

How to cite: Hamilton, O., Demaeyer, J., Vannitsem, S., and Crucifix, M.: Multistability in a Coupled Ocean-AtmosphereReduced Order Model: Non-linear TemperatureEquations, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-5496, https://doi.org/10.5194/egusphere-egu23-5496, 2023.

The observational research of tipping elements in the climate system relies largely on time series analysis via so-called Early Warning Signals. An upward trend in the estimated variance or lag-1 autocorrelation of the observable may be a sign for Critical Slowing Down (CSD), a phenomenon exhibited during the destabilization a system’s fixed point. This approach has been employed extensively both for assessing contemporary tipping risks [1] and understanding the dynamics in the advent of past abrupt climate change [2]. However, this inference of destabilization from statistical observations is in general only valid under certain model assumptions with regard to both the deterministic dynamics and the stochastic component (noise). While the assumption of additive white noise is the most canonical approach to representing unresolved dynamics, it has long been understood that certain variabilities in the climate system exhibit correlation and persistence [3]. In this case, trends in the above indicators should no longer be attributed solely to CSD, since they may also be rooted in possibly changing correlation characteristics of the driving noise. While there has been progress in the development of indicators for discrete-time models incorporating correlated noise [4], the task of assessing discrete-time data from continuous-time models has not received as much attention. We present a simple linearly restoring stochastic model with red noise as its driving force and discuss possible avenues of estimating system stability from time series data through the autocorrelation structure and power spectral density of the observable. We quantitatively compare these methods to conventional Early Warning Signals, highlighting the potential pitfalls of the latter in this setting.

[1] Boers, N. (2021). Observation-based early-warning signals for a collapse of the Atlantic Meridional Overturning Circulation. Nature Climate Change 11

[2] Rypdal, M. (2016). Early-Warning Signals for the Onsets of Greenland Interstadials and the Younger Dryas–Preboreal Transition, Journal of Climate, 29(11)

[3] Mann, M.E., Lees, J.M. (1996). Robust estimation of background noise and signal detection in climatic time series. Climatic Change 33

[4] Rodal, M., Krumscheid, S., Madan,G. , LaCasce, J.H., and Vercauteren, N. (2022). Dynamical stability indicator based on autoregressive moving-average models: Critical transitions and the Atlantic meridional overturning circulation, Chaos 32

How to cite: Morr, A. and Boers, N.: Detecting Critical Slowing Down under the influence of continuous-time Red Noise, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-5250, https://doi.org/10.5194/egusphere-egu23-5250, 2023.

Over the last two decades, tipping points have become a hot topic due to the devastating consequences that they may have on natural and human systems. Tipping points are typically associated with a system bifurcation when external forcing crosses a critical level, causing an abrupt transition to an alternative, and often less desirable, state. However, the rate of change in forcing is arguably of even greater relevance in the human-dominated anthropocene, but is rarely examined as a potential sole mechanism for tipping points. Thus, I will introduce the related phenomenon of rate-induced tipping: an instability that occurs when external forcing varies across some critical rate, usually without crossing any bifurcations. First, I will explain when to expect rate-induced tipping. Then, using illustrating examples of differing complexity I will highlight universal and generic properties of rate-induced tipping in a range of natural and human systems.

How to cite: Ritchie, P., Alkhayuon, H., Cox, P., and Wieczorek, S.: When to Expect Rate-Induced Tipping in Natural and Human Systems, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-8105, https://doi.org/10.5194/egusphere-egu23-8105, 2023.

Climate Change is forcing Earth System tipping elements rapidly, in some cases this forcing occurs on a similar timescale to the intrinsic timescale of the tipping element itself. This poses challenges for our ability to get good early warning signals for these tipping elements, as typical approaches require a clear timescale separation between the assumed slow forcing and the timescale of the system. We demonstrate that by calculating early warning signals ‘over space’ instead of ‘over time’ better early warning signals can be obtained for faster forcing. We compare the relative merits of these two ways of calculating early warning signals.

How to cite: Clarke, J., Cox, P., Ritchie, P., and Huntingford, C.: Spatial Early Warning Signals for Rapidly Forced Systems, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-12923, https://doi.org/10.5194/egusphere-egu23-12923, 2023.

Multiple equilibria are found in all members of the hierarchy of climate models, ranging from simple planetary energy balance models to fully coupled general circulation models. They arise from the physical and biogeochemical coupling of different climate system components, and hence they are a general feature of planetary dynamics. Transitions from one equilibrium to another can be triggered by a temporary perturbation of the system which crosses a tipping point. Greenland ice cores and many other paleoclimate archives have abundantly demonstrated that the Earth System had limited stability during the last ice ages and that tipping has occurred in the past. A particularly dynamic period was the transition from the last ice age to the present. We present recent model simulations that reconcile different paleoceanographic indicators and so permit the quantitative reconstruction of the transient changes of the Atlantic meridional overturning circulation. This circulation may also tip in the future depending on the level and rate of increases in greenhouse gas concentrations. However, reducing the uncertainties where such tipping points lie and how close the climate system is to them, requires much better resolved climate models.

The tipping of regional systems has come into recent focus because the impacts on humans and ecosystems may be substantial. Among them are the various monsoon systems, parts of the Antarctic ice sheet, shifts in the statistics of extreme climate and weather events, the extent of the Amazon rain forest, or the grassland distribution in Eastern Africa, and hence biodiversity. Such changes would all have regional consequences that are not yet reflected in current climate change projections.

Therefore, regional tipping needs to be assessed systematically by the scientific community using a new generation of climate models at kilometer-scale resolution. A cross-working group IPCC Special Report on “Climate Tipping Points and Consequences for Habitability and Resources” in its forthcoming 7^th assessment cycle would help strengthening a consensus on this topic and trigger the much needed advances in scientific understanding to more comprehensively inform adaptation and mitigation strategies.

How to cite: Stocker, T.: Tipping Points: A challenge for climate change projections, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-12900, https://doi.org/10.5194/egusphere-egu23-12900, 2023.

During past glacial intervals the high northern latitude’s climate was punctuated by abrupt warming events which were accompanied by a sudden loss of sea ice, a reinvigoration of the Atlantic Meridional Overturning Circulation (AMOC), and cooling of the Nordic Seas. Despite being considered the archetype of past abrupt climatic change, to date there is no consensus about the physical mechanism behind these so-called Dansgaard-Oeschger events and the subsequent milder interstadial phase. Here, we propose an excitable model system to explain the DO cycles, in which interstadials are regarded as noise-induced state space excursions. Our model comprises the mutual multi-scale interactions between four dynamical variables representing Arctic atmospheric temperatures, Nordic Seas’ temperatures and sea ice cover, and AMOC. Crucially, the model’s atmosphere-ocean heat flux is moderated by the sea ice variable, which in turn is subject to large perturbations dynamically generated by fast evolving intermittent noise. If supercritical, these perturbations trigger interstadial-like state space excursions seizing all four model variables. As a physical source for such a driving noise process we propose convective events in the ocean or atmospheric blocking events. The key characteristics of DO cycles are reproduced by our model with remarkable resemblance to the proxy record; in particular, their shape, return time, as well as the dependence of the interstadial and stadial durations on the background temperatures are reproduced accurately. In contrast to the prevailing understanding that the DO variability showcases bistability in the underlying dynamics, we conclude that multi-scale, monostable excitable dynamics provides a promising alternative candidate to explain the millennial-scale climate variability associated with the DO events.

How to cite: Riechers, K., Gottwald, G., and Boers, N.: Glacial abrupt climate change as a multi-scalephenomenon resulting from monostable excitabledynamics, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-11899, https://doi.org/10.5194/egusphere-egu23-11899, 2023.

In response to abruptly increasing atmospheric CO₂ concentrations, general circulation model experiments typically evidence a rapid reduction or full collapse of the Atlantic Meridional Overturning Circulation (AMOC) from its current, strongly overturning state, into one characterized by weak overturning and reduced northward oceanic heat transport. This tipping point is frequently discussed in the context of present and past global climate changes. Less understood, however, is the evolution of the circulation towards a new equilibrium state, which occurs over many centuries or millennia following the initial AMOC response. To revisit this problem, we have performed multi-millennial simulations of the Community Earth System Model version 1 (CESM1) in a low-resolution configuration (T31 gx3v7), appropriate for paleoclimate studies. We consider a pre-industrial control (284.7ppm) simulation, as well as abrupt 2x, 4x, 8x, 16x, and 0.5x pre-industrial control atmospheric CO₂ concentrations whereby atmospheric concentrations are increased at the start of integration and held constant for the duration of the experiment. In all global warming scenarios, we observe a rapid collapse to the AMOC within the first 250 years, attributed mechanistically to the complex interplay between surface salinity and temperature which inhibits deep-water formation in the sub-polar North Atlantic. Then, in our abrupt doubling and quadrupling of atmospheric CO₂ experiments we observe a recovery to the circulation after some 1,000 years, and 3,500 years, respectively. After initially collapsing, our 8xCO₂ experiment remains in this weakened state even after 10,000 years of integration have been performed, potentially indicating that a new equilibrium may have been met in this very warm climate.

We have further observed other intriguing bifurcations which arise stochastically in the forced system. First, in our abrupt 4xCO₂ experiment, with the AMOC in a collapsed state we observe a spontaneous activation of the Pacific Meridional Overturning Circulation (PMOC) some 2,500 years following the initial forcing. The circulation persists for 1,000 years and has a notable effect on climate in the North Pacific region, for instance raising surface temperatures through the associated increase in Pacific Ocean northward heat transport. At 3,500 years the circulation collapses concomitantly with an AMOC recovery in the experiment, demonstrating a AMOC/PMOC seesaw. Secondly, in our abrupt global cooling experiment, we observe a spontaneous collapse of the AMOC after 2,000 years, which precedes a recovery over the next 1,500 years, before a secondary, rapid collapse to the circulation at 3,500 years. The behavior resembles a Dansgaard-Oeschger Event. Overall, our results highlight the rich quasi-equilibrium dynamical behavior of the Global Meridional Overturning Circulation in past climates for which atmospheric CO₂ concentrations were markedly different.

How to cite: Curtis, P. E. and Fedorov, A.: The Tipping Points of the Atlantic Meridional Overturning Circulation in Warm and Cold Climates, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-9907, https://doi.org/10.5194/egusphere-egu23-9907, 2023.

Greenland ice core records feature Dansgaard–Oeschger (DO) events; abrupt warming episodes followed by a gradual cooling phase during mid-glacial periods. Here, we analysis spontaneous self-sustained D-O type oscillations reproduced in three climate models: CCSM4, MPI-ESM and HadCM3. The three models show D-O type oscillatory behaviour in a remarkably similar, narrow window of atmospheric CO2 concentrations between approximately 185 to 230 parts per million (ppm). This CO2 range also compares particularly well with Marine Isotopic Stage 3 (MIS 3 - between 27.8 – 59.4 thousand of years BP, hereafter ka) atmospheric CO2 values (∼ 233-187.5 ppm), when D-O events occurred with most regularity. Outside this CO2 window of oscillatory behaviour, two different stable states are shown in the three models; warm high CO2 (strong AMOC) and cold low CO2 (weak AMOC) states. The weak state remains stable below the first critical tipping point near 185-195 ppm and the strong state remains stable above the second tipping point near 217-230 ppm. In all three models, the oscillatory experiments with higher CO2 show an increased built-up of stadial salinity in the upper ocean in the subtropics, especially in the eastern edge of the North Atlantic Current, compared with the ensemble mean: the tendency to re-invigorate the Atlantic Meridional Overturning Circulation (AMOC) is increased and so the system spend less time in the stadial phase. CO2 also affects North Atlantic and Arctic sea ice, determining interstadial and stadial duration. Similar sensitivity CO2 experiments performed with other climate models may help in further constraining the here-identified range of atmospheric CO2 (∼185-230 ppm) bounding this D-O sweet-spot. 

How to cite: Malmierca Vallet, I. and Sime, L. C.: Atmospheric CO2 impact on spontaneous Dansgaard–Oeschger type oscillations: oscillatory sweet-spot for three climate models, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-4149, https://doi.org/10.5194/egusphere-egu23-4149, 2023.

Some simple models of Arctic sea ice show bifurcations associated with the loss of sea ice under increased surface radiative forcing (Eisenman and Wettlaufer 2009). However, experiments using GCMs typically show a smooth loss of sea ice under increasing CO₂. This mismatch adds to uncertainty on the existence of tipping point behaviour in the Arctic and the processes that stabilise or destabilise it from this behaviour.

Simple models exhibiting tipping points typically omit many features of the Arctic climate system. Their bifurcations usually arise from the ice-albedo feedback. The purpose of my work is to use a bottom-up hierarchical approach to investigate how additional features of Arctic climate not included in simple models affect the existence of bifurcations in the system.

I started with a base ice model (Eisenman and Wettlaufer 2009) and investigate the role of local ice-atmosphere feedbacks using a coupled atmospheric column model. This allowed me to analyse the impact of the following on the possible states for the model to exist in:

• Changes to the atmospheric temperature profile – particularly the transition from a stable atmosphere with a strong temperature inversion to a less stable atmosphere as the Arctic warms.
• Explicitly resolved changes in surface heat fluxes and downwelling longwave radiation.
• Changes in low level Arctic clouds – particularly as the atmospheric structure changes.

I also explored the sensitivity of the model to changes and variation in atmospheric heat transport.

I will present results of this work and demonstrate how local atmospheric feedbacks affect the stability of tipping points in Arctic sea ice.

How to cite: Derby, E. and Pierrehumbert, R.: Model complexity and Arctic sea ice tipping points – a single column model approach, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-7989, https://doi.org/10.5194/egusphere-egu23-7989, 2023.

Global sea level rise due to the melting of the Greenland ice sheet (GrIS) in response to anthropogenic global warming poses a severe threat to ecosystems and human society (IPCC, 2021). Modelling and paleoclimatic evidence suggest that rapidly increasing temperatures in the Arctic can trigger positive feedback mechanisms, and the GrIS is hypothesised to exhibit multiple stable states (Gregory et al., 2020). 
Consequently, critical transitions are expected when the global mean surface temperature crosses specific thresholds, and there is substantial hysteresis between the alternative stable states (Robinson et al., 2012). 
Here, we investigate the impact of different climate scenarios that overshoot temperature goals and then return to lower temperatures at different pace. Our results show that both the maximum GMT and the time span of overshooting given GMT targets are critical in determining GrIS stability. We find an abrupt loss of the ice sheet for a threshold temperature, preceded by several intermediate stable states. We show that even temporarily overshooting the temperature threshold may lead to catastrophic consequences in specific scenarios. On the other hand, overshoots might be tolerable if GMTs are subsequently reduced below 1.5°C GMT above pre-industrial levels within a few centuries. Even without a transition to a new ice sheet state the short-term global sea level rise can exceed several metres before returning to moderate GMTs.

Allan, R. P., Hawkins, E., Bellouin, N., & Collins, B. (2021). IPCC, 2021: Summary for Policymakers (V. Masson-Delmotte, P. Zhai, A. Pirani, S. L. Connors, C. Péan, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M. I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J. B. R. Matthews, T. K. Maycock, T. Waterfield, O. Yelekçi, R. Yu, & B. Zhou, Eds.). Cambridge University Press. https://centaur.reading.ac.uk/101317/

Gregory, J. M., George, S. E., & Smith, R. S. (2020). Large and irreversible future decline of the Greenland ice sheet. The Cryosphere, 14(12), 4299–4322. https://doi.org/10.5194/tc-14-4299-2020

Robinson, A., Calov, R., & Ganopolski, A. (2012). Multistability and critical thresholds of the Greenland ice sheet. Nature Climate Change, 2(6), 429–432. https://doi.org/10.1038/nclimate1449

How to cite: Bochow, N., Poltronieri, A., Rypdal, M., Robinson, A., and Boers, N.: Overshooting the critical threshold for the Greenland ice sheet, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-3074, https://doi.org/10.5194/egusphere-egu23-3074, 2023.

The Atlantic Meridional Overturning Circulation (AMOC) transports warm, saline water towards the northern North Atlantic, contributing substantially to the meridional heat transport in the climate system. Measurements of the Atlantic freshwater divergence show that the AMOC may be in a bistable state and hence subject to collapsing under anthropogenic greenhouse gas forcing. We aim at computing the probability of such a transition, focusing on time scales up to the end of this century.  

Simulating trajectories in a climate model is very expensive. To minimize the amount of data required to compute the probability of such rare AMOC transitions, we use a rare-events algorithm called TAMS (Trajectory-Adaptive Multilevel Sampling), that encourages the transition without changing the statistics. In TAMS, N trajectories are simulated and evaluated with a score function; the poorest-performing trajectories are discarded, and the best ones are re-simulated.

The optimal score function is the committor function, defined as the probability that a trajectory reaches a zone A of the phase space before another zone B. To avoid the difficulties raised by its exact computation, we estimate it using a feedforward neural network. Because of the expense of simulating data in a climate model, we also minimize the amount of data needed to train the neural network by reusing data processed through TAMS.

As a first step, using simulated data from an idealized stochastic AMOC model, where forcing and white noise are applied via a surface freshwater flux, we compute the transition probabilities versus noise and forcing amplitudes. Then, we apply the same protocol to compute these transition probabilities in the much more sophisticated climate model FAMOUS. This model is a coarse resolution Atmosphere-Ocean General Circulation Model that has been shown to exhibit a collapse of the AMOC via hosing experiments. In this new setup, we compute once again the transition probabilities of the AMOC versus noise and forcing, where the forcing amplitude is a hosing flux, and the atmosphere dynamics plays the role of the noise.

How to cite: Jacques-Dumas, V., van Westen, R. M., and Dijkstra, H. A.: Computation of the AMOC collapse probability using a rare-event algorithm, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-8648, https://doi.org/10.5194/egusphere-egu23-8648, 2023.

The Amazon forest is at risk of dieback due to climate change, in particular decreasing mean annual precipitation (MAP) and increasing mean annual temperature (MAT). This study assesses the influence on South American vegetation under two possible future climate change scenarios: global warming, and global warming combined with an AMOC collapse. We consider MAT and MAP as control parameters and use their projected changes from climate model simulations with the Earth System Model HadGEM3. We then estimate the most probable states of vegetation based on empirical relationships between these parameters and tree cover. Our results suggest that an AMOC collapse would not contribute to further rainforest dieback over most of the Amazon basin. Instead, in parts of tropical South America, MAP increases and MAT decreases after AMOC collapse, which tends to stabilize the Amazon forest and hence delay the Amazon dieback compared to the default global warming scenario.

How to cite: Nian, D., Bathiany, S., Ben-Yami, M., Blaschke, L., Hirota, M., Rodrigues, R., and Boers, N.: The combined effect of global warming and AMOC collapse on the Amazon Forest, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-3291, https://doi.org/10.5194/egusphere-egu23-3291, 2023.

Decisions are usually taken sequentially in climate change policy: every certain amount of years, new agreements and promises are made about greenhouse gas emission reduction etc. In the intersection of decision theory and climate science, sequential decision problems can be formulated and solved, to find optimal sequences of policies and support policy makers with some advice.

There are, however, many uncertainties affecting the outcome of these optimisations. Since these decision problems tend to be very simple in comparison with the complexity of the real world, knowing how different uncertainties affect optimal policies might be more important than what the optimal policy comes out to be. In this work, we explore how some uncertainties affect optimal policies and the possible trajectories associated with those optimal policies. 
  
For this aim we formulate a sequential decision problem with a single "global" policy maker. The decision problem starts with the world state in 2020 and decisions are taken every 10 years till 2100. The policy maker has options regarding CO2 emissions reduction, geoengineering in the form of solar radiation modification and carbon dioxide removal.

We simulate the effects of the decisions on the world’s state with SURFER. SURFER is a simple and fast model featuring a carbon cycle responsive to positive and negative emissions, it allows for geoengineering and accounts for sea level rise from ice sheets (containing tipping points) and from ocean expansion and glacier melt. SURFER has been shown to reproduce the globally averaged behavior of earth system models and models of intermediate complexity from decades to millennia. As opposed to some optimal decision problems in the context of climate change which use integrated assessment models of the climate and the economy, here, with the aim of transparency and simplicity, we consider only a climate model. 

We define a modular and transparent cost function that contains what the policy maker cares about. This function is a linear sum of costs associated with: green transition, geoengineering use and risks, temperature and ocean acidification damages and long term sea level rise commitments.

Using this decision problem we investigate how different kinds of uncertainties affect the sequence of optimal policies obtained and the optimal trajectories associated with those optimal policies. We consider three different kinds of uncertainties: uncertainties in the priorities of the decision maker (i.e., in the reward, cost or utility function), uncertainties on some physical parameters (in particular, climate sensitivity and ice sheet tipping points) and political uncertainty (policymaker’s decisions may not be implemented). 

How to cite: Martinez Montero, M., Crucifix, M., Brede, N., Botta, N., and Couplet, V.: Effects of different uncertainties on optimal policies, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-13893, https://doi.org/10.5194/egusphere-egu23-13893, 2023.

Drylands occupy 45% emerged lands on Earth, are home of more than 2 billion people and are extremely vulnerable to climate change. Aridity increases is expected to influence the structure and functioning of drylands in a non-linear fashion. Yet, the prevalence and drivers of these abrupt changes in ecosystem structure and function remain poorly studied. We especially lack investigations of the changes of dynamical properties of these systems and how these dynamical properties relate to aridity. Those are key to understand the real menace of experiencing abrupt shifts with aridity increases in the near future.

Here we used remote sensing tools to acquire dynamical trajectories of normalized vegetation indices (NDVI, surrogates of plant fractional cover) for more than 50,000 dryland sites. With this information we conducted analysis using machine learning processes to examine the relationship of aridity with some key dynamical properties of dryland ecosystems, including several aspects of resilience (ability to withstand fluctuations without changing the functioning of ecosystems), dynamical drivers of productivity, complexity of dynamical trajectories and abruptness of productivity changes.

By doing so we provide a comprehensive assessment of aridity thresholds on dynamical properties of dryland productivity that show clear vulnerability of certain zones of the Earth exhibiting critical aridity thresholds previously identified through space. In particular, we show accumulation of abrupt shifts on aridity values characteristic of transition areas from semiarid to arid ecosystems. Furthermore, these values exhibit also nonlinear shifts in resilience of ecosystems and on the identity of key dynamical drivers. Our work paves the way to expand the incidence of aridity threshold from spatial to temporal implications, and highlights the necessity of developing strategies to protect and monitor especially vulnerable areas affecting more than one fifth of emerged lands.

How to cite: Berdugo, M., Delgado-Baquerizo, M., Guirado, E., Gaitan, J. J., Fournier, C., Crowther, T. W., and Dakos, V.: Prevalence and drivers of abrupt shifts in global drylands: gathering dynamical evidences of aridity thresholds, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-11666, https://doi.org/10.5194/egusphere-egu23-11666, 2023.

Global warming is expected to lead to increase in amplitude and autocorrelation in climate variability in most locations around the world. These changes could have a great and imminent impact on ecosystems. In this work, we demonstrate that changes in climate variability can drive cyclic predator-prey ecosystems to extinction via so-called phase tipping (P-tipping), a new type of instability that occurs only from certain phases of the predator-prey cycle. We coupled a simple mathematical model of climate variability to a self-oscillating paradigmatic predator-prey model. Most importantly, we combine realistic parameter values for the Canada lynx and snowshoe hare with actual climate data from the boreal forest to demonstrate that critically important species in the boreal forest have increased likelihood of extinction under predicted changes in climate variability. The cyclic populations of these species are most vulnerable during stages of the cycle when the predator population is near its maximum. We identify stochastic resonance as the underlying mechanism for the increased likelihood extinction.

How to cite: Alkhayuon, H., Tyson, R., and Wieczorek, S.: Stochastic resonance, climate variability, and phase-tipping: The increasing risk of extinction in cyclic ecosystems, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-687, https://doi.org/10.5194/egusphere-egu23-687, 2023.

Whether current-day ecosystems, often heavily modified by humans, can adapt to climate change is one of the most pressing scientific questions. Coastal ecosystems are at the forefront of climate impact, as salt marshes and intertidal flats may drown if these systems cannot follow sea level rise.

We developed a model to investigate how the emergence of complex creek networks during early salt marsh development affects the ability of marsh ecosystems to accumulate sediment, thereby compensating for sea level rise. This model is based on a scale-dependent feedback relation between vegetation growth and sedimentation, as plants locally block water flow, which then diverts to their surroundings. The model revealed that this self-organization process drives the emergence of a complex creek network of ever smaller creeks nested in between larger ones.

We used the model to analyze the importance of creek network complexity for the rate at which marshes accumulate sediment. The model highlights that in salt marshes, plant traits have a defining effect on the development of creek network complexity. Yet, it is the emergent effect of creek network complexity on sedimentation, rather than plant traits directly, that controlled sedimentation rates, determining the adaptive capacity of the marsh to sea level rise. Self-organized creek complexity proved a defining characteristic determining the resilience of this ecosystem to climate change.

We used our model to study whether restored coastal wetlands can be designed in such a way as to improve the adaptive capacity to sea level rise. We explored 14 realigned coastal wetlands and related the established, real-world creek network, being either entirely artificial dug-out channels or naturally formed creeks, to their potential, model-predicted sedimentation rate.

We observed that the developing channel networks in restored wetlands had much lower creek development and channel branching than natural systems, resulting in less efficient channel networks. Model simulations showed that if artificial creek networks deviated more from the creek pattern observed in natural ecosystems, or from the ones predicted from our model, they had lower sediment transport efficiency. Our findings suggest that if a more natural organization is followed when designing climate-proof coastal ecosystems, they are more resilient to climate change.

How to cite: Van de Koppel, J., Cornacchia, L., Van de Vijsel, R., and Van der Wal, D.: Using self-organization to build climate-resilient ecosystems, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-3328, https://doi.org/10.5194/egusphere-egu23-3328, 2023.

The persistence and functionality of forest ecosystems are highly dependent on their resilience to the ongoing rapid changes in climate conditions and in natural and anthropogenic pressures. Experimental evidences of a sudden increase in tree mortality across different biomes are rising concerns about the ongoing changes in forest resilience. However, how forest resilience, which is the capacity to withstand and recover from perturbations, is evolving in response to global changes is not yet explored. Here, we integrate satellite-based vegetation indices with machine learning to show how forest resilience, quantified in terms of critical slowing down indicators, has changed over the period 2000-2020. We show that tropical, arid and temperate forests are experiencing a significant decline in resilience, likely related to the increase in water limitations and climate variability. On the contrary, boreal forests show an increasing trend in resilience, likely for the benefits of climate warming and CO₂ fertilization in cold biomes, which may outweigh the adverse effects of climate change. These patterns emerge consistently in both managed and intact forests corroborating the existence of common large-scale climate drivers. Reductions in resilience are statistically linked to abrupt declines in forest productivity, occurring in response to a slow drifting toward a critical resilience threshold. We estimate that about 22% of intact undisturbed forests, corresponding to 3.32 Pg C of GPP, have already reached such critical threshold and are experiencing a further degradation in resilience. Altogether, these signals reveal a widespread and increasing instability of global forests and should be accounted for in the design of land-based mitigation and adaption plans.

How to cite: Forzieri, G., Dakos, V., G Mc Dowell, N., Alkama, R., and Cescatti, A.: Emerging signals of a global drift in forest resilience under climate change, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-5180, https://doi.org/10.5194/egusphere-egu23-5180, 2023.

The resilience of tropical forests against climate change and deforestation is vital for biodiversity and carbon drawdown. This resilience is hard to measure directly, but is suspected to be decreasing. There is particular concern that the Amazon rainforest may be approaching a “tipping point” where the large-scale loss of species and carbon pools amplifies substantially. Candidate mechanisms for such threshold effects often involve positive feedbacks and span a large range of scales. For example, individual trees can die from hydraulic failure when soil moisture decreaseses, forest fires can mediate a regional transition to a savanna state, and by synchronising remote regions, the moisture recycling feedback could cause a continental-scale forest dieback. Conceptual dynamical systems suggest that the loss of resilience that accompanies such transitions can be measured by statistical indicators like increasing autocorrelation. Satellite observations of vegetation indices related to greenness and biomass seem to support these theoretical expectations.

Here we analyse dynamic global vegetation models (DGVMs) from CMIP6, as well as idealised simulations with LPJ, in order to bridge the complexity gap between conceptual models and the real world. First, we assess how resilience of terrestrial carbon pools in the tropics depends on mean annual rainfall (MAP). We find that this relationship differs between models, and can also differ substantially from the observed positive relationship, depending on how the models capture carbon pool dynamics on the grid-cell level. Second, we show that changes in resilience do not necessarily require any atmosphere-vegetation feedbacks, fire feedback or ecological interactions, suggesting that observed relationships may capture physiological effects in individual trees rather than the stability of the entire forest. Third, we also find that the coexistence of vegetation types affects vegetation resilience in DGVMs. In particular, plant types with faster dynamics can replace slower ones (e.g., grass replacing trees), leading to decreased autocorrelation but not necessarily larger sensitivity to MAP. We conclude that suitable indicators of tropical vegetation resilience should be determined by (i) using DGVMs to understand better what mechanisms are at play, and (ii) using observations to rule out certain model approaches (e.g. area-averaged versus individual-based models).

How to cite: Bathiany, S., Nian, D., and Boers, N.: Indicators of tropical forest resilience in vegetation models, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-9297, https://doi.org/10.5194/egusphere-egu23-9297, 2023.

Spatial and temporal aggregations are common when preparing remote sensing data for analysis. Aggregations often serve to enhance the underlying signal of interest while suppressing noise, and can improve estimations of mean states and long-term trends in data. However, aggregating means that the highest-resolution parts of a signal can no longer be resolved, and rapid or fine-scale fluctuations are removed, potentially biasing analyses that rely on these parts of the signal. Further, data aggregation often goes along with gap-filling, which can further dilute the signals of interest.

In this work, we examine the impact of spatial aggregation on estimates of vegetation resilience by comparing MODIS vegetation data sets at a range of spatial resolutions (native 250 m – 25 km). We first use synthetic data to investigate various de-seasoning and de-trending schemes and their responsiveness to gaps in the underlying data. Based on these insights, we calculate two estimates of vegetation resilience at the global scale and at multiple spatial resolutions to determine the optimal level of spatial aggregation for MODIS data, considering the tradeoffs between fine-scale (gappy, noisy) and aggregated (continuous, smooth) vegetation data in terms of resilience estimation. Our results provide best practices for the aggregation, deseasoning, detrending, and analysis of vegetation resilience at the global scale.

How to cite: Smith, T. and Boers, N.: How Low Can You Go? Implications of Spatial Aggregation for the Estimation of Ecosystem Resilience, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-5449, https://doi.org/10.5194/egusphere-egu23-5449, 2023.

The projected increase of drought occurrence under future climates will affect terrestrial ecosystems, particularly by increasing drought-induced tree mortality. The capacity to simulate drought mortality in vegetation models is therefore essential to understand climate change impacts on ecosystem functions and services, as well as on functional diversity. Using the trait-based vegetation model aDGVM2, we assessed tree mortality under drought conditions in tropical Asia under future climate, and if vegetation is resilient to drought or if tipping point behavior occurs. We further assessed how drought impacts are related to pre-drought community composition and diversity. We conducted model simulations for multiple sites in tropical Asia, representing a biogeographic gradient ranging from savannas to tropical forests. Responses of vegetation attributes and mortality rates were simulated until 2099 under the RCP8.5 scenario. Repeated droughts of different length were modeled to test drought impacts and resilience. Finally, the diversity of pre-drought communities was constrained by removing different trait syndromes to test how community composition and diversity influence drought resistance and resilience. Model simulations showed substantial biomass dieback during drought which was attributed to increased mortality rates, primarily among tall and old trees. Drought response differed between current and elevated CO₂ levels under RCP8.5, with higher biomass recovery under elevated CO₂ due to fertilization effects. Pre-drought community composition influenced biomass dieback and mortality during drought, and the presence or absence of drought-adapted plants had the highest effect on drought impacts. Despite severe drought impacts, recovery of most vegetation attributes was possible after drought periods. We conclude that repeated droughts under future conditions will have vast impacts on vegetation attributes and mortality in tropical ecosystems. Conserving functional diversity in ecosystems buffers drought impacts. However, according to model results, vegetation is resilient, and irreversible transitions to alternative vegetation states do, for the investigated scenarios, not occur. Improved models representing lagged drought impacts, irreversible damage of individual plants, and the interactions between drought regimes, CO₂ fertilization and trait diversity are required.

How to cite: Scheiter, S., Kumar, D., Pfeiffer, M., and Langan, L.: Drought mortality and resilience of savannas and forests in tropical Asia, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-1283, https://doi.org/10.5194/egusphere-egu23-1283, 2023.

It is thought that tropical forests can exist as an alternative stable state to savanna. Therefore, perturbation by climate change or human impact may lead to crossing of a tipping point beyond which there is rapid large-scale forest dieback that is not easily reversed. Modelling studies of alternative stable tree cover states have either relied on mean-field assumptions or not included the spatiotemporal dynamics of fire, making it hard to compare their output to spatial data. In this talk, we analyse a microscopic model of tropical forest and fire and show how dynamics of forest area are linked to its emergent spatial structure. We find that the relation between forest perimeter and area determines the nonlinearity in forest growth while forest perimeter weighted by adjacent grassland area determines the nonlinearity in forest loss. Together with the linear changes, which are independent of spatial structure, these two effects lead to an emergent relation between forest area change and forest area, defining a single-variable ordinary differential equation. Such a relation between pattern and dynamics enables empiricists to assess forest stability and resilience directly from a single spatial observation of a tropical forest-grassland landscape.

How to cite: Wuyts, B. and Sieber, J.: Emergent structure, dynamics and abrupt transitions in a cellular automaton of tropical forest and fire, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-16946, https://doi.org/10.5194/egusphere-egu23-16946, 2023.