ALBERT

Description: SWIFT-AI-DS is a benchmark dataset that consists of samples that have been derived from two simulation runs (each 2.5 years long) of the chemistry and transport model ATLAS (Wohltmann and Rex, 2009; Wohltmann et al., 2010). This data set of nearly 200 million samples meets the requirements of a labelled data set and is ideally suited for training and testing of a machine learning based surrogate model. Two time periods were considered in the simulation runs: first from November 1998 to March 2001 and the second from November 2004 to March 2007. The dataset covers the entire Earth geographically, but is vertically restricted to the altitudes of the lower to middle stratosphere, for which the SWIFT (Rex et al., 2014; Kreyling et. al, 2017; Wohltmann et al., 2017) approach of 24-hour ozone tendencies can be applied. Applicability was determined in terms of the chemical lifetime of stratospheric ozone, which is a function of solar irradiance and altitude. It can be described by a dynamic upper bound [Kreyling et. Al, 2017]. Within the range where the chemical lifetime is longer than 14 days, ozone is not in quasi-chemical equilibrium. Moreover, this data set focuses on the region of the lower to middle stratosphere because it is the region with the largest contribution to the total ozone column. State-of-the-art physical process models for stratospheric chemistry require enormous computational time. Our research is focused on developing much faster, yet accurate, surrogate models for computing the 24-hour tendencies of stratospheric ozone. Much faster models of stratospheric ozone provide a new application area such as for climate models. These surrogate models benefit greatly from the methodological and hardware improvements of the last decade. Each simulation run uses the full stratospheric chemistry model to solve a system of differential equations involving 47 chemical species and 171 chemical reactions at a very high (〈〈 seconds) and variable temporal resolution. The ATLAS model is driven by ECMWF reanalysis data (either ERA-I or ERA5). The air parcel state has been sampled at a 24-hour time step (00:00 UTC model time). During postprocessing some variables are stored as 24-hour averages, as 24-hour tendencies or as the state at the beginning of the 24-hour time step. The dataset is stored in 12 monthly netCDF-files.

Description: The Extrapolar SWIFT model is a fast ozone chemistry scheme for interactive calculation of the extrapolar stratospheric ozone layer in coupled general circulation models (GCMs). In contrast to the widely used prescribed ozone, the SWIFT ozone layer interacts with the model dynamics and can respond to atmospheric variability or climatological trends. The Extrapolar SWIFT model employs a repro-modelling approach, where algebraic functions are used to approximate the numerical output of a full stratospheric chemistry and transport model (ATLAS). The full model solves a coupled chemical differential equations system with 55 initial and boundary conditions (mixing ratio of various chemical species and atmospheric parameters). Hence the rate of change of ozone over 24 h is a function of 55 variables. Using covariances between these variables, we can find linear combinations in order to reduce the parameter space to the following nine basic variables: latitude, pressure altitude, temperature, local ozone column, mixing ratio of ozone and of the ozone depleting families (Cly, Bry, NOy and HOy). We will show that these 9 variables are sufficient to characterize the rate of change of ozone. An automated procedure fits a polynomial function of fourth degree to the rate of change of ozone obtained from several simulations with the ATLAS model. One polynomial function is determined per month which yields the rate of change of ozone over 24 h. A key aspect for the robustness of the Extrapolar SWIFT model is to include a wide range of stratospheric variability in the numerical output of the ATLAS model, also covering atmospheric states that will occur in a future climate (e.g. temperature and meridional circulation changes or reduction of stratospheric chlorine loading). For validation purposes, the Extrapolar SWIFT model has been integrated into the ATLAS model replacing the full stratospheric chemistry scheme. Simulations with SWIFT in ATLAS have proven that the systematic error is small and does not accumulate during the course of a simulation. In the context of a 10 year simulation, the ozone layer, simulated by SWIFT, shows a stable annual cycle, with inter-annual variations comparable to the ATLAS model. The application of Extrapolar SWIFT requires the evaluation of polynomial functions with 30–100 terms. Nowadays, computers can calculate such polynomial functions at thousands of model grid points in seconds. SWIFT provides the desired numerical efficiency and computes the ozone layer 104 times faster than the chemistry scheme in the ATLAS CTM.

Description: The Extrapolar SWIFT model is a fast ozone chemistry scheme for interactive calculation of the extrapolar stratospheric ozone layer in coupled general circulation models (GCMs). In contrast to the widely used prescribed ozone, the SWIFT ozone layer interacts with the model dynamics and can respond to atmospheric variability or climatological trends. The Extrapolar SWIFT model employs a repro-modelling approach, in which algebraic functions are used to approximate the numerical output of a full stratospheric chemistry and transport model (ATLAS). The full model solves a coupled chemical differential equation system with 55 initial and boundary conditions (mixing ratio of various chemical species and atmospheric parameters). Hence the rate of change of ozone over 24 h is a function of 55 variables. Using covariances between these variables, we can find linear combinations in order to reduce the parameter space to the following nine basic variables: latitude, pressure altitude, temperature, overhead ozone column and the mixing ratio of ozone and of the ozone-depleting families (Cly, Bry, NOy and HOy). We will show that these nine variables are sufficient to characterize the rate of change of ozone. An automated procedure fits a polynomial function of fourth degree to the rate of change of ozone obtained from several simulations with the ATLAS model. One polynomial function is determined per month, which yields the rate of change of ozone over 24 h. A key aspect for the robustness of the Extrapolar SWIFT model is to include a wide range of stratospheric variability in the numerical output of the ATLAS model, also covering atmospheric states that will occur in a future climate (e.g. temperature and meridional circulation changes or reduction of stratospheric chlorine loading). For validation purposes, the Extrapolar SWIFT model has been integrated into the ATLAS model, replacing the full stratospheric chemistry scheme. Simulations with SWIFT in ATLAS have proven that the systematic error is small and does not accumulate during the course of a simulation. In the context of a 10-year simulation, the ozone layer simulated by SWIFT shows a stable annual cycle, with inter-annual variations comparable to the ATLAS model. The application of Extrapolar SWIFT requires the evaluation of polynomial functions with 30–100 terms. Computers can currently calculate such polynomial functions at thousands of model grid points in seconds. SWIFT provides the desired numerical efficiency and computes the ozone layer 104 times faster than the chemistry scheme in the ATLAS CTM.

Description: The Extrapolar SWIFT model is a fast ozone chemistry scheme for interactive calculation of the extrapolar stratospheric ozone layer in coupled general circulation models (GCMs). In contrast to the widely used prescribed ozone, the SWIFT ozone layer interacts with the model dynamics and can respond to atmospheric variability or climatological trends. The Extrapolar SWIFT model employs a repro-modelling approach, where algebraic functions are used to approximate the numerical output of a full stratospheric chemistry and transport model (ATLAS). The full model solves a coupled chemical differential equations system with 55 initial and boundary conditions (mixing ratio of various chemical species and atmospheric parameters). Hence the rate of change of ozone over 24  h is a function of 55 variables. Using covariances between these variables, we can find linear combinations in order to reduce the parameter space to the following nine basic variables: latitude, pressure altitude, temperature, local ozone column, mixing ratio of ozone and of the ozone depleting families (Cly, Bry, NOy and HOy). We will show that these 9 variables are sufficient to characterize the rate of change of ozone. An automated procedure fits a polynomial function of fourth degree to the rate of change of ozone obtained from several simulations with the ATLAS model. One polynomial function is determined per month which yields the rate of change of ozone over 24 h. A key aspect for the robustness of the Extrapolar SWIFT model is to include a wide range of stratospheric variability in the numerical output of the ATLAS model, also covering atmospheric states that will occur in a future climate (e.g. temperature and meridional circulation changes or reduction of stratospheric chlorine loading). For validation purposes, the Extrapolar SWIFT model has been integrated into the ATLAS model replacing the full stratospheric chemistry scheme. Simulations with SWIFT in ATLAS have proven that the systematic error is small and does not accumulate during the course of a simulation. In the context of a 10 year simulation, the ozone layer, simulated by SWIFT, shows a stable annual cycle, with inter-annual variations comparable to the ATLAS model. The application of Extrapolar SWIFT requires the evaluation of polynomial functions with 30–100 terms. Nowadays, computers can calculate such polynomial functions at thousands of model grid points in seconds. SWIFT provides the desired numerical efficiency and computes the ozone layer 104 times faster than the chemistry scheme in the ATLAS CTM.

Description: The SWIFT model is a fast yet accurate chemistry scheme for calculating the chemistry of stratospheric ozone. It is mainly intended for use in Global Climate Models (GCMs), Chemistry Climate Models (CCMs) and Earth System Models (ESMs). For computing time reasons these models often do not employ full stratospheric chem- istry modules, but use prescribed ozone instead. This can lead to insufficient representation between stratosphere and troposphere. The SWIFT stratospheric ozone chemistry model, focuses on the major reaction mechanisms of ozone production and loss in order to reduce the computational costs. SWIFT consists of two sub-models. 1) Inside the polar vortex, the model calculates polar vortex averaged ozone loss by solving a set of coupled differential equations for the key species in polar ozone chemistry. 2) The extrapolar regime, which this poster is going to focus on. Outside the polar vortex, the complex system of differential equations of a full stratospheric chemistry model is replaced by an explicit algebraic polynomial, which can be solved in a fraction of the time needed by the full scale model. The approach, which is used to construct the polynomial, is also referred to as repro-modeling and has been successfully applied to chemical models (Turanyi (1993), Lowe & Tomlin (2000)). The procedure uses data from the Lagrangian stratospheric chemistry and transport model ATLAS and yields one high-order polynomial for global ozone loss and production rates over 24h per month. The stratospheric ozone change rates can be sufficiently described by 9 variables. Latitude, altitude, temperature, the overhead ozone abundance, 4 mixing ratios of ozone depleting chemical families (chlorine, bromine, nitrogen-oxides and hydrogen-oxides) and the ozone concentrations itself. The ozone change rates in the lower stratosphere as a function of these 9 variables yield a sufficiently compact 9-D hyper-surface, which we can approximate with a polynomial. In the upper stratosphere (roughly above 30km) the ozone chemical lifetime becomes shorter than the transport time scales, thus the ozone concentrations are determined by the local atmospheric conditions. We therefore introduce an additional regime in the upper stratosphere, where the ozone concentrations, instead of the 24h change rates, are fitted. The fitted polynomial for upper stratospheric ozone is dependent on the same variables, except the ozone concentration, naturally. This poster shows results of simulations employing the polynomial scheme and discusses constraints on the method.

Description: The goal of this PhD-thesis was the development of a fast yet accurate chemistry scheme for an interactive calculation of the extrapolar stratospheric ozone layer. The SWIFT-model is mainly intended for use in Global Climate Models (GCMs). For computing-time reasons GCMs often do not employ full stratospheric chemistry modules, but use prescribed ozone instead. This method does not consider the interaction between atmospheric dynamics and the ozone layer and can neither resolve the inter-annual variability of the ozone layer nor respond to climatological trends. Various studies [Calvo et al., 2015, Gillett and Thompson, 2003, Thompson and Solomon, 2002] have pointed out these insufficiencies. Existing fast ozone schemes, as in Cariolle und Teyssedre [2007] and McLinden et al. [2000], use a Taylor expansion of the first order to expand the rate of change of ozone about reference conditions of ozone mixing ratio, temperature and the locale ozone column and thus can not sufficiently adept to climate change scenarios, differing from the reference conditions. The SWIFT-model, in contrast, considers the full chemical system of a stratospheric chemistry model, including non-linearities and fluctuations of ozone depleting species, to determine the rate of change of ozone. The SWIFT-model consists of two modules, a polar and an extrapolar module. The polar module calculates vortex-averaged ozone loss by solving a set of coupled differential equations for the key species in polar ozone chemistry. Coefficients of the equation system are determined by simulations with a full chemistry model [Wohltmann et al., 2016]. This dissertation presents the extrapolar SWIFT-module, where we use algebraic functions to approximate the rate of change of ozone of the full model. In the full model, 55 initial and boundary conditions (e.g. various chemical species and atmospheric parameters) determine the function of rate of change of ozone, creating a 55-dimensional hypersurface. The numerical output of several simulations with the full model characterize the shape of the hypersurface. Using linear combinations of these variables, we can reduce the parameter space to the following nine dimensions: latitude, pressure, temperature, local ozone co- lumn, mixing ratio of ozone and of the ozone depleting families (Cly , Bry, NOy and HOy ). These nine variables sufficiently describe the shape of the 55-dimensional hypersurface. An automated procedure fits 9-dimensional polynomials of degree four to the reduced function. One global polynomial per month is determined which calculates the rate of change of ozone over 24 h. The full model used to fit the polynomials is the chemistry- and transport-model ATLAS. Two 2.5-years ATLAS-simulations from separate decades constitute the fitting-dataset. A key aspect for the robustness of the SWIFT-model is the incorporation of a wide range of stratospheric variability in the fitting-datasets. The systematic error between ATLAS and SWIFT causes the ozone mixing ratios to drift by less than 0.5% per day in the central regions of the 9-dimensional parameter space. Higher errors are located in the boundary regions, where the sampling density of the fitting-dataset is low, i.e. for rarely occurring atmospheric conditions. Here, the errors can rise to 4% per day. However, steep ozone gradients and non-linearities in the rate of change function are not the sources of significant errors. The extrapolar SWIFT-module has been integrated into the ATLAS-CTM as an optional chemistry scheme. Simulations with SWIFT in ATLAS have proven that the systematic error does not accumulate in the course of a run. In a 10 year simulation SWIFT has continuously produced a stable annual cycle, with inter-annual variations of the ozone layer well comparable to the full ATLAS-CTM. Horizontal gradients in the ozone distribution due to planetary waves, are well resolved by SWIFT. The average deviations between partial ozone columns in ATLAS and SWIFT are less than ±15 DU. Especially in the mid- and high-latitudes the extrapolar SWIFT-module yields better results than existing fast ozone schemes. The application of SWIFT requires the calculation of polynomials with 30 – 100 terms. Nowadays, computers can solve such polynomials at thousands of grid points in seconds. Therefore SWIFT provides the desired numerical efficiency and computes the ozone layer 10000 times faster than the chemistry model in the ATLAS-CTM.

Description: We have made a climatology of the diurnal variation of short-lived atmospheric compounds, such as ClO, BrO, HO2, and HOCl, as well as longer life time longer-lived species: O3, the hydrogen chloride isotopes H35Cl and H37Cl, and HNO3 from measurements by the Superconducting SubMIllimeter9 wave Limb-Emission Sounder (SMILES) on International Space Station (ISS). We performed the observation with very low noise on the emission spectrum for measuring of vertical profiles of atmospheric compositions with altitude range from the lower stratosphere to the lower thermosphere (20 – 100 km), thus observing at all local times due to a non-sun-synchronous orbit of ISS. The diurnal variation climatologies are based on data periods of two months. Consideration of the SMILES time-space sampling patterns with respect to the averaging coordinates is a key issue for climatology creation. Biases induced by inhomogeneous sampling are minimized by carefully choosing the size of averaging bins. The sampling biases of the diurnal variation climatology of ClO and BrO are investigated in a comparison of homogeneously sampled model data versus SMILES sampled model data from the stratospheric Lagrangian chemistry and transport model ATLAS. Mostly the relative error is in the range of 0 – 20%.

Description: We present a climatology of the diurnal variation of short-lived atmospheric compounds, such as ClO, BrO, HO2, and HOCl, as well as longer-lived species: O3, the hydrogen chloride isotopes H35Cl and H37Cl, and HNO3. Measurements were taken by the Superconducting Submillimeter-wave Limb-Emission Sounder (SMILES). This spectrally resolving radiometer, with very low observation noise and altitude range from the lower stratosphere to the lower thermosphere (20–100km), was measuring vertical profiles of absorption spectra along a non-sun-synchronous orbit, thus observing at all local times. We used the retrieved volume mixing ratio profiles to compile climatologies that are a function of pressure, a horizontal coordinate (latitude or equivalent latitude), and a temporal coordinate (solar zenith angle or local solar time). The main product presented are climatologies with a high resolution of the temporal coordinate (diurnal variation climatologies). In addition, we provide climatologies with a high resolution of the horizontal coordinate (zonal climatologies).The diurnal variation climatologies are based on data periods of 2 months and the zonal climatologies on monthly data periods. Consideration of the SMILES time-space sampling patterns with respect to the averaging coordinates is a key issue for climatology creation, especially in case of diurnal variation climatologies. Biases induced by inhomogeneous sampling are minimized by carefully choosing the size of averaging bins. The sampling biases of the diurnal variation climatology of ClO and BrO are investigated in a comparison of homogeneously sampled model data versus SMILES-sampled model data from the stratospheric Lagrangian chemistry and transport model ATLAS. In most cases, the relative sampling error is in the range of 0–20%. The strongest impact of sampling biases is found where the species' temporal gradients are strongest (mostly at sunrise and sunset), with a relative error of 60–100%. The SMILES climatology data sets are available via the SMILES data distribution home page.

Description: Interactions between climate change and stratospheric ozone modify both, the evolution of surface climate and the recovery of the stratospheric ozone layer. Accounting for the climate feedbacks from changing ozone as well as the impact of climate change on the evolution of the ozone layer requires the interactive representation of stratospheric chemistry in Earth System Models. Our understanding of stratospheric ozone chemistry is now mature at the process scale and state of the art Chemical Transport Models (CTM) result in a realistic representation of the global ozone layer and the chemical processes affecting it. But the huge computational effort of these models makes it difficult to include the ozone layer interactively in Earth System Models (ESMs). We have developed SWIFT, an extremely fast module for interactive ozone chemistry in climate models. SWIFT allows for an interactive treatment of stratospheric ozone in standard ESMs with little numerical overhead. We will present the current status of SWIFT and results from coupling SWIFT to a climate model.