In the present study, production and loss pathways of tropospheric ozone and their rates are identified by an algorithm for the automatic determination of reaction pathways in complex chemical systems. For this purpose, reaction rates were provided by the chemistry-transport model IFS(MOZART) (Integrated Forecasting System - Model for Ozone and Related chemical Tracers). A detailed analysis is carried out for three different scenarios: clean air (Palau), intermediate emissions (Athens), and large emissions (Beijing). At each location the processes at the surface and at an altitude of 500 m are analysed. The ozone production rate is largest in Beijing, intermediate in Athens and smallest on Palau. Nevertheless, there is net ozone loss at the surface in Beijing because of a strong net conversion of ozone to NO _2 by freshly emitted NO. The ozone production is dominated by methane oxidation on Palau and by the oxidation of short-lived, i.e. emitted nearby, VOCs (volatile organic compounds) in Beijing, where the strongest individual contributor at the surface is isoprene. Athens represents an intermediate situation. The pathways determined show in detail all intermediate steps of the degradation of individual VOCs, including the interaction with NO _x and HO _x species, and permit the calculation of the number of ozone molecules formed per VOC molecule consumed. For instance, at the surface in Beijing the average net production of ozone in pathways leading to the full degradation of isoprene (to CO _2 ) is 10.1 ozone molecules per isoprene molecule consumed. At the same location pathways producing up to 18 ozone molecules per isoprene molecules have been found. However, the rates of these extreme pathways are very small.

The intensive catalytic chemical ozone loss cycles involving inorganic chlorine compounds are known to cause the ozone hole, that regularly forms in the Antarctic polar vortex each winter and spring. One key point of the explanation of the ozone hole are heterogeneous reactions on the surface of Polar Stratospheric Cloud (PSC) particles, which are present in the polar stratosphere owing to the very low stratospheric temperatures.Ozone mixing ratios can reach values of a few ppbv which is below detection limit of ozone sonde observations. Under these extreme conditions, fast chlorine deactivation into the reservoir HCl does occur even though polar stratospheric clouds are still present, that are normally causing chlorine activation.In this study we revisit this issue and investigate the occurring chlorine chemistry in more detail. The explanation of the chemical mechanism of the fast net HCl formation is based on the automatic determination of reaction pathways by the Pathway Analysis Program (PAP) (Lehmann, 2004).The simulations of chemical composition are performed by the model CLaMS in box model mode along an ensemble of about 600 trajectories in the Antarctic spring 2018. The simulated rapid complete chlorine deactivation into HCl in the presence of PSCs is in line with satellite observations by the Microwave Limb Sounder (MLS). ReferenceLehmann, R.: An algorithm for the determination of all significant pathways in chemical reaction systems, J. Atmos. Chem. 47, 45-78 (2004).

Simulations of Antarctic chlorine and ozone chemistry in previous work show that in the core of the Antarctic vortex (16-18 km, 85-55 hPa, 390-430 K) HCl null cycles (initiated by reactions of Cl with CH4 and CH2O) are effective. These HCl null cycles cause both HCl molar mixing ratios to remain very low throughout Antarctic winter and spring. They cause ozone-destroying chlorine (ClOx) to remain enhanced so that rapid ozone depletion proceeds. Here we investigate the impact of the observed dehydration in Antarctica, which strongly reduces ice formation and the uptake of HNO3 from the gas phase; however the efficacy of HCl null cycles is not affected. Moreover, also when using the observed very low HCl molar mixing ratios in Antarctic winter as an initial value, HCl null cycles are efficient in maintaining low HCl (and high ClOx) throughout winter and spring. Further, the reaction CH3O2+ClO is important for the efficacy of the HCl null cycle initiated by the reaction CH4+Cl. Using the current kinetic recommendations instead of earlier ones has very little impact on the simulations. All simulations presented here for the core of the Antarctic vortex show extremely low minimum ozone values (below 50 ppb) in late September to early October in agreement with observations.

In climate modeling, the stratospheric ozone layer is typically only considered in a highly simplified form due to computational constraints. For climate projections, it would be of advantage to include the mutual interactions between stratospheric ozone, temperature, and atmospheric dynamics to accurately represent radiative forcing. The overarching goal of our research is to replace the ozone layer in climate models with a machine-learned neural representation of the stratospheric ozone chemistry that allows for a particularly fast, but accurate and stable simulation. We created a benchmark data set from pairs of input and output variables that we stored from simulations of the ATLAS Chemistry and Transport Model. We analyzed several variants of multilayer perceptrons suitable for physical problems to learn a neural representation of a function that predicts 24-h ozone tendencies based on input variables. We performed a comprehensive hyperparameter optimization of the multilayer perceptron using Bayesian search and Hyperband early stopping. We validated our model by replacing the full chemistry module of ATLAS and comparing computation time, accuracy, and stability. We found that our model had a computation time that was a factor of 700 faster than the full chemistry module. The accuracy of our model compares favorably to the full chemistry module within a 2-year simulation run, also outperforms a previous polynomial approach for fast ozone chemistry, and reproduces seasonality well in both hemispheres. In conclusion, the neural representation of stratospheric ozone chemistry in simulation resulted in an ozone layer that showed a high accuracy, significant speed-up, and stability in a long-term simulation.

It is currently uncertain as to whether methane exists on Mars. Data from the Curiosity Rover suggests a background methane concentration of a few tenths parts per billion whereas data from the Trace Gas Orbiter suggest an upper limit of twenty parts per trillion. If methane exists on Mars then we do not understand fully the physical and chemical processes affecting its lifetime. Atmospheric models suggest an over-estimate in the lifetime by a factor of around six hundred compared with earlier observations. In the present work we assume the Curiosity Rover background methane value and estimate the uncertainty in atmospheric chemistry and mixing processes in our atmospheric column model 1D TERRA. Results suggest that these processes can only explain a factor of ~sixteen lowering in the methane lifetime. This implies that if methane is present then additional, currently unknown processes are required to explain the observed lifetime.

The Polar SWIFT model is a fast scheme for calculating the chemistry of stratospheric ozone depletion in the polar vortex in winter. It is intended for use in general circulation models (GCMs) and earth system models (ESMs) to enable the simulation of interactions between the ozone layer and climate when a full stratospheric chemistry scheme is computationally too expensive. In addition to the simulation of chemistry, ozone has to be transported in the GCM. As an alternative to the general schemes for the transport and mixing of tracers in the GCMs, a parameterization of the transport of ozone can be used in order to obtain the total change of ozone as the sum of the change by transport and by chemistry. One of the benefits of this approach is the easy and self-contained coupling to a GCM. Another potential advantage is that a transport parameterization based on reanalysis data and measurements can avoid deficiencies in the representation of transport in the GCMs, such as deficits in the representation of the Brewer–Dobson circulation caused by the gravity wave parameterization. Hence, we present a transport parameterization for the Polar SWIFT model that simulates the change in vortex-averaged ozone by transport in a fast and simple way without the need for a complex transport scheme in the GCM.

The Polar SWIFT model is a fast scheme for calculating the chemistry of stratospheric ozone depletion in the polar vortex in winter.It is intended for use in general circulation models (GCMs) and earth system models (ESMs) to enable the simulation of interactions between the ozone layer and climate when a full stratospheric chemistry scheme is computationally too expensive.In addition to the simulation of chemistry, ozone has to be transported in the GCM.As an alternative to the general schemes for the transport and mixing of tracers in the GCMs, a parameterization of the transport of ozone can be used in order to obtain the total change of ozone as the sum of the change by transport and by chemistry.One of the benefits of this approach is the easy and selfcontained coupling to a GCM.Another potential advantage is that a transport parameterization based on reanalysis data and measurements can avoid deficiencies in the representation of transport in the GCMs, such as deficits in the representation of the Brewer-Dobson circulation caused by the gravity wave parameterization.Hence, we present a transport parameterization for the Polar SWIFT model that simulates the change in vortex-averaged ozone by transport in a fast and simple way without the need for a complex transport scheme in the GCM.

Common representations of the stratospheric ozone layer in climate modeling are widely considered only in a very simplified way. Neglecting the mutual interactions of ozone with atmospheric temperature and dynamics has the effect of making climate projections less accurate. Although, more elaborate and interactive models of the stratospheric ozone layer are available, they require far too much computation time to be coupled with climate models. Our aim with this project was to break new ground and pursue an interdisciplinary strategy that spans the fields of machine learning, atmospheric physics and climate modelling. In this work, we present an implicit neural representation of the extrapolar stratospheric ozone chemistry (SWIFT-AI). An implicitly defined hyperspace of the stratospheric ozone chemistry offers a continuous and even differentiable representation that can be parameterized by artificial neural networks. We analysed different parameter-efficient variants of multilayer perceptrons. This was followed by an intensive, as far as possible energy-efficient search for hyperparameters involving Bayesian optimisation and early stopping techniques. Our data source is the Lagrangian chemistry and transport model ATLAS. Using its full model of stratospheric ozone chemistry, we focused on simulating a wide range of stratospheric variability that will occur in future climate (e.g. temperature and meridional circulation changes). We conducted a simulation for several years and created a data-set with over 200E+6 input and output pairs. Each output is the 24h ozone tendency of a trajectory. We performed a dimensionality reduction of the input parameters by using the concept of chemical families and by performing a sensitivity analysis to choose a set of robust input parameters. We coupled the resulting machine learning models with the Lagrangian chemistry and transport model ATLAS, substituting the full stratospheric chemistry model. We validated a two-year simulation run by comparing to the differences in accuracy and computation time from both the full stratospheric chemistry model and the previous polynomial approach of extrapolar SWIFT. We found that SWIFT-AI consistently outperforms the previous polynomial approach of SWIFT, both in terms of test data and simulation results. We discovered that the computation time of SWIFT-AI is more than twice as fast as the previous polynomial approach SWIFT and 700 times faster than the full stratospheric chemistry scheme of ATLAS, resulting in minutes instead of weeks of computation time per model year – a speed-up of several orders of magnitude. To ensure reproducibility and transparency, we developed a machine learning pipeline, published a benchmark dataset and made our repository open to the public. In summary, we could show that the application of state-of-the-art machine learning methods to the field of atmospheric physics holds great potential. The achieved speed-up of an interactive and very precise ozone layer enables a novel way of representing the ozone layer in climate models. This in turn will increase the quality of climate projections, which are crucial for policy makers and of great importance for our planet.

It is currently uncertain as to whether methane exists on Mars. Data from the Curiosity Rover suggests abackground methane concentration of a few tenths parts per billion whereas data from the Trace Gas Orbitersuggest an upper limit of twenty parts per trillion. If methane exists on Mars then we do not understand fully thephysical and chemical processes affecting its lifetime. Atmospheric models suggest an over-estimate in thelifetime by a factor of around six hundred compared with earlier observations. In the present work we assume theCuriosity Rover background methane value and estimate the uncertainty in atmospheric chemistry and mixingprocesses in our atmospheric column model 1D TERRA. Results suggest that these processes can only explain afactor of ~sixteen lowering in the methane lifetime. This implies that if methane is present then additional,currently unknown processes are required to explain the observed lifetime.

Motivated by previous measurements of very low tropospheric ozone concentrations in the Tropical West Pacific (TWP) and the implied low oxidizing capacity of this key region for transport into the stratosphere in boreal winter (e.g. Rex et al. 2014), we set up an atmospheric research station in Palau (7°N 134°E) as part of the StratoClim campaign. Our analysis of regular balloon-borne tropospheric ozone observations at Palau from 01/2016-12/2019 gives unprecedented insights into transport processes and air mass origin in the TWP. We confirm the year-round dominance of a low ozone background in the mid-troposphere. Layers of enhanced ozone are often anti-correlated with water vapor and occur frequently. Moreover, the occurrence of respective layers shows a strong seasonality. Dry and ozone-rich air masses between 5 and 10 km altitude were observed in 71 % of the profiles from February until April compared to 25 % from August until October. By defining monthly atmospheric background profiles for ozone and relative humidity based on observed statistics, we found that the deviations from this background reveal a bimodal distribution of RH anomalies. A previously proposed universal bimodal structure of free tropospheric ozone in the TWP could not be verified (Pan et al. 2015).Back trajectory calculations (ATLAS) confirm that throughout the year the mid-tropospheric background is controlled by local convective processes and the origin of air masses is thus close to or East of Palau in the Pacific Ocean. Dry and ozone-rich air originates in tropical Asia and reaches Palau in anticyclonic conditions over an area stretching from India to the Philippines. This supports the controversial hypothesis of several studies which attribute ozone enhancement against the ozone-poor background to remote pollution events on the ground such as biomass burning (e.g. Andersen et al. 2016). A potential vorticity analysis revealed no stratospheric influence and we thus propose large-scale descent within the tropical troposphere as responsible for dehydration of air masses on their way to Palau.

Motivated by previous measurements of very low tropospheric ozone concentrations in the Tropical West Pacific (TWP) and the implied low oxidizing capacity of this key region for transport into the stratosphere (e.g. Rex et al. 2014), we set up an atmospheric research station in Palau (7° N 134° E). Our analysis of regular balloon-borne tropospheric ozone observations at Palau from 01/2016-10/2019 confirms the year-round dominance of a low ozone background in the mid-troposphere. Layers of enhanced ozone are often anti-correlated with water vapor and occur frequently. Moreover, the occurrence of respective layers shows a strong seasonality. Dry and ozone-rich air masses between 5 and 10 km altitude were observed in 71 % of the profiles from February until April compared to 25 % from August until October. By defining monthly atmospheric background profiles for ozone and relative humidity based on observed statistics, we found that the deviations from this background reveal a bimodal distribution of RH anomalies. A previously proposed universal bimodal structure of free tropospheric ozone in the TWP could not be verified (Pan et al. 2015). Back trajectory calculations confirm that throughout the year the mid-tropospheric background is controlled by local convective processes and the origin of air masses is thus close to or East of Palau in the Pacific Ocean. Dry and ozone-rich air originates in tropical Asia and reaches Palau in anticyclonic conditions over an area stretching from India to the Philippines. This supports the hypothesis of several studies which attribute ozone enhancement against the low ozone background to remote pollution events on the ground such as biomass burning (e.g. Andersen et al. 2016). A potential vorticity analysis revealed no stratospheric influence and we thus propose large-scale descent within the tropical troposphere as responsible for dehydration of air masses on their way to Palau.

In the Antarctic ozone hole, ozone mixing ratios have been decreasing to extremely low values of 0.01–0.1 ppm in nearly all spring seasons since the late 1980s, corresponding to 95–99% local chemical loss. In contrast, Arctic ozone loss has been much more limited and mixing ratios have never before fallen below 0.5 ppm. In Arctic spring 2020, however, ozonesonde measurements in the most depleted parts of the polar vortex show a highly depleted layer, with ozone loss averaged over sondes peaking at 93% at 18 km. Typical minimum mixing ratios of 0.2 ppm were observed, with individual profiles showing values as low as 0.13 ppm (96% loss). The reason for the unprecedented chemical loss was an unusually strong, long-lasting, and cold polar vortex, showing that for individual winters the effect of the slow decline of ozone-depleting substances on ozone depletion may be counteracted by low temperatures.

The winter 2019/2020 showed the lowest ozone mixing ratios ever observed in the Arctic winter stratosphere. It was the coldest Arctic stratospheric winter on record and was characterized by an unusually strong and long‐lasting polar vortex. We study the chemical evolution and ozone depletion in the winter 2019/2020 using the global Chemistry and Transport Model ATLAS. We examine whether the chemical processes in 2019/2020 are more characteristic of typical conditions in Antarctic winters or in average Arctic winters. Model runs for the winter 2019/2020 are compared to simulations of the Arctic winters 2004/2005, 2009/2010, and 2010/2011 and of the Antarctic winters 2006 and 2011, to assess differences in chemical evolution in winters with different meteorological conditions. In some respects, the winter 2019/2020 (and also the winter 2010/2011) was a hybrid between Arctic and Antarctic conditions, for example, with respect to the fraction of chlorine deactivation into HCl versus ClONO 2 , the amount of denitrification, and the importance of the heterogeneous HOCl + HCl reaction for chlorine activation. The pronounced ozone minimum of less than 0.2 ppm at about 450 K potential temperature that was observed in about 20% of the polar vortex area in 2019/2020 was caused by exceptionally long periods in the history of these air masses with low temperatures in sunlight. Based on a simple extrapolation of observed loss rates, only an additional 21–46 h spent below the upper temperature limit for polar stratospheric cloud formation and in sunlight would have been necessary to reduce ozone to near zero values (0.05 ppm) in these parts of the vortex.

Ozone loss pathways and their rates in the ozone quasi-biennial oscillation (QBO), which is simulated by a chemistry-climate model developed by the Meteorological Research Institute of Japan, are evaluated using an ob- jective pathway analysis program (PAP). The analyzed chemical system contains catalytic cycles caused by NOx , HOx , ClOx , Ox , and BrOx . PAP quantified the rates of all significant catalytic ozone loss cycles, and evaluated the partitioning among these cycles. The QBO amplitude of the sum of all cycles amounts to about 4 and 14 % of the annual mean of the total ozone loss rate at 10 and 20 hPa, respectively. The contribution of catalytic cycles to the QBO of the ozone loss rate is found to be as follows: NOx cycles contribute the largest fraction (50 – 85 %) of the QBO amplitude of the total ozone loss rate; HOx cycles are the second-largest (20 – 30 %) below 30 hPa and the third-largest (about 10 %) above 20 hPa; Ox cycles rank third (5 – 20 %) below 30 hPa and second (about 20 %) above 20 hPa; ClOx cycles rank fourth (5 – 10 %); and BrOx cycles are almost negligible. The relative contribution of the NOx and Ox cycles to the QBO amplitude of ozone loss differs by up to 10 % and 20 %, respectively, from their contribution to the annual mean ozone loss rate. The ozone QBO at 20 hPa is mainly driven by ozone transport, which then alters the ozone loss rate. In contrast, the ozone QBO at 10 hPa is driven chemically by NOx and the temperature dependence of [O]/[O3], which results from the temperature dependence of the reaction O + O2 + M → O3 + M. In addition, the ozone QBO at 10 hPa is influenced by the overhead ozone column, which affects [O]/[O3] (through ozone photolysis) and the ozone production rate (through oxygen photolysis).

In the Antarctic ozone hole, ozone mixing ratios have been decreasing to extremely low values of 0.01–0.1 ppm in nearly all spring seasons since the late 1980s, corresponding to 95–99% local chemical loss. In contrast, Arctic ozone loss has been much more limited and mixing ratios have never before fallen below 0.5 ppm. In Arctic spring 2020, however, ozonesonde measurements in the most depleted parts of the polar vortex show a highly depleted layer, with ozone loss averaged over sondes peaking at 93% at 18 km. Typical minimum mixing ratios of 0.2 ppm were observed, with individual profiles showing values as low as 0.13 ppm (96% loss). The reason for the unprecedented chemical loss was an unusually strong, long‐lasting, and cold polar vortex, showing that for individual winters the effect of the slow decline of ozone‐depleting substances on ozone depletion may be counteracted by low temperatures.

We present a Lagrangian convective transport scheme developed for global chemistry and transport models, which considers the variable residence time that an air parcel spends in convection. This is particularly important for accurately simulating the tropospheric chemistry of short-lived species, e.g., for determining the time available for heterogeneous chemical processes on the surface of cloud droplets. In current Lagrangian convective transport schemes air parcels are stochastically redistributed within a fixed time step according to estimated probabilities for convective entrainment as well as the altitude of detrainment. We introduce a new scheme that extends this approach by modeling the variable time that an air parcel spends in convection by estimating vertical updraft velocities. Vertical updraft velocities are obtained by combining convective mass fluxes from meteorological analysis data with a parameterization of convective area fraction profiles. We implement two different parameterizations: a parameterization using an observed constant convective area fraction profile and a parameterization that uses randomly drawn profiles to allow for variability. Our scheme is driven by convective mass fluxes and detrainment rates that originate from an external convective parameterization, which can be obtained from meteorological analysis data or from general circulation models. We study the effect of allowing for a variable time that an air parcel spends in convection by performing simulations in which our scheme is implemented into the trajectory module of the ATLAS chemistry and transport model and is driven by the ECMWF ERA-Interim reanalysis data. In particular, we show that the redistribution of air parcels in our scheme conserves the vertical mass distribution and that the scheme is able to reproduce the convective mass fluxes and detrainment rates of ERA-Interim. We further show that the estimated vertical updraft velocities of our scheme are able to reproduce wind profiler measurements performed in Darwin, Australia, for velocities larger than 0.6 m s−1. SO2 is used as an example to show that there is a significant effect on species mixing ratios when modeling the time spent in convective updrafts compared to a redistribution of air parcels in a fixed time step. Furthermore, we perform long-time global trajectory simulations of radon-222 and compare with aircraft measurements of radon activity.