Natural short-lived halogens exert an indirect cooling effect on climate.

Observational evidence shows the ubiquitous presence of ocean-emitted short-lived halogens in the global atmosphere1-3. Natural emissions of these chemical compounds have been anthropogenically amplified since pre-industrial times4-6, while, in addition, anthropogenic short-lived halocarbons are currently being emitted to the atmosphere7,8. Despite their widespread distribution in the atmosphere, the combined impact of these species on Earth's radiative balance remains unknown. Here we show that short-lived halogens exert a substantial indirect cooling effect at present (-0.13 ± 0.03 watts per square metre) that arises from halogen-mediated radiative perturbations of ozone (-0.24 ± 0.02 watts per square metre), compensated by those from methane (+0.09 ± 0.01 watts per square metre), aerosols (+0.03 ± 0.01 watts per square metre) and stratospheric water vapour (+0.011 ± 0.001 watts per square metre). Importantly, this substantial cooling effect has increased since 1750 by -0.05 ± 0.03 watts per square metre (61 per cent), driven by the anthropogenic amplification of natural halogen emissions, and is projected to change further (18-31 per cent by 2100) depending on climate warming projections and socioeconomic development. We conclude that the indirect radiative effect due to short-lived halogens should now be incorporated into climate models to provide a more realistic natural baseline of Earth's climate system.

Non-Linear Dimensionality Reduction With a Variational Encoder Decoder to Understand Convective Processes in Climate Models.

Deep learning can accurately represent sub-grid-scale convective processes in climate models, learning from high resolution simulations. However, deep learning methods usually lack interpretability due to large internal dimensionality, resulting in reduced trustworthiness in these methods. Here, we use Variational Encoder Decoder structures (VED), a non-linear dimensionality reduction technique, to learn and understand convective processes in an aquaplanet superparameterized climate model simulation, where deep convective processes are simulated explicitly. We show that similar to previous deep learning studies based on feed-forward neural nets, the VED is capable of learning and accurately reproducing convective processes. In contrast to past work, we show this can be achieved by compressing the original information into only five latent nodes. As a result, the VED can be used to understand convective processes and delineate modes of convection through the exploration of its latent dimensions. A close investigation of the latent space enables the identification of different convective regimes: (a) stable conditions are clearly distinguished from deep convection with low outgoing longwave radiation and strong precipitation; (b) high optically thin cirrus-like clouds are separated from low optically thick cumulus clouds; and (c) shallow convective processes are associated with large-scale moisture content and surface diabatic heating. Our results demonstrate that VEDs can accurately represent convective processes in climate models, while enabling interpretability and better understanding of sub-grid-scale physical processes, paving the way to increasingly interpretable machine learning parameterizations with promising generative properties.

Reactive halogens increase the global methane lifetime and radiative forcing in the 21st century.

CH4 is the most abundant reactive greenhouse gas and a complete understanding of its atmospheric fate is needed to formulate mitigation policies. Current chemistry-climate models tend to underestimate the lifetime of CH4, suggesting uncertainties in its sources and sinks. Reactive halogens substantially perturb the budget of tropospheric OH, the main CH4 loss. However, such an effect of atmospheric halogens is not considered in existing climate projections of CH4 burden and radiative forcing. Here, we demonstrate that reactive halogen chemistry increases the global CH4 lifetime by 6-9% during the 21st century. This effect arises from significant halogen-mediated decrease, mainly by iodine and bromine, in OH-driven CH4 loss that surpasses the direct Cl-induced CH4 sink. This increase in CH4 lifetime helps to reduce the gap between models and observations and results in a greater burden and radiative forcing during this century. The increase in CH4 burden due to halogens (up to 700 Tg or 8% by 2100) is equivalent to the observed atmospheric CH4 growth during the last three to four decades. Notably, the halogen-driven enhancement in CH4 radiative forcing is 0.05 W/m2 at present and is projected to increase in the future (0.06 W/m2 by 2100); such enhancement equals ~10% of present-day CH4 radiative forcing and one-third of N2O radiative forcing, the third-largest well-mixed greenhouse gas. Both direct (Cl-driven) and indirect (via OH) impacts of halogens should be included in future CH4 projections.

Deep Learning Based Cloud Cover Parameterization for ICON.

A promising approach to improve cloud parameterizations within climate models and thus climate projections is to use deep learning in combination with training data from storm-resolving model (SRM) simulations. The ICOsahedral Non-hydrostatic (ICON) modeling framework permits simulations ranging from numerical weather prediction to climate projections, making it an ideal target to develop neural network (NN) based parameterizations for sub-grid scale processes. Within the ICON framework, we train NN based cloud cover parameterizations with coarse-grained data based on realistic regional and global ICON SRM simulations. We set up three different types of NNs that differ in the degree of vertical locality they assume for diagnosing cloud cover from coarse-grained atmospheric state variables. The NNs accurately estimate sub-grid scale cloud cover from coarse-grained data that has similar geographical characteristics as their training data. Additionally, globally trained NNs can reproduce sub-grid scale cloud cover of the regional SRM simulation. Using the game-theory based interpretability library SHapley Additive exPlanations, we identify an overemphasis on specific humidity and cloud ice as the reason why our column-based NN cannot perfectly generalize from the global to the regional coarse-grained SRM data. The interpretability tool also helps visualize similarities and differences in feature importance between regionally and globally trained column-based NNs, and reveals a local relationship between their cloud cover predictions and the thermodynamic environment. Our results show the potential of deep learning to derive accurate yet interpretable cloud cover parameterizations from global SRMs, and suggest that neighborhood-based models may be a good compromise between accuracy and generalizability.

Climate policy implications of nonlinear decline of Arctic land permafrost and other cryosphere elements.

Arctic feedbacks accelerate climate change through carbon releases from thawing permafrost and higher solar absorption from reductions in the surface albedo, following loss of sea ice and land snow. Here, we include dynamic emulators of complex physical models in the integrated assessment model PAGE-ICE to explore nonlinear transitions in the Arctic feedbacks and their subsequent impacts on the global climate and economy under the Paris Agreement scenarios. The permafrost feedback is increasingly positive in warmer climates, while the albedo feedback weakens as the ice and snow melt. Combined, these two factors lead to significant increases in the mean discounted economic effect of climate change: +4.0% ($24.8 trillion) under the 1.5 °C scenario, +5.5% ($33.8 trillion) under the 2 °C scenario, and +4.8% ($66.9 trillion) under mitigation levels consistent with the current national pledges. Considering the nonlinear Arctic feedbacks makes the 1.5 °C target marginally more economically attractive than the 2 °C target, although both are statistically equivalent.