ML-TOMCAT: machine-learning-based satellite-corrected global stratospheric ozone profile data set from a chemical transport model

2023

Preprint

Abstract. Monitoring the atmospheric concentrations of greenhouse gases (GHGs) is crucial in order to improve our understanding of their climate impact. However, there are no long-term profile data sets of important GHGs that can be used to gain a better insight into the processes controlling their variations in the atmosphere. Here, we merge chemical transport model (CTM) output and profile measurements from two solar occultation instruments, the HALogen Occultation Experiment (HALOE) and the Atmospheric Chemistry Experiment – Fourier Transform Spectrometer (ACE-FTS), to construct long-term (1991–2021), gap-free stratospheric profile data sets (hereafter, TCOM) for two important GHGs. The Extreme Gradient Boosting (XGBoost) regression model is used to estimate the corrections needed to apply to the CTM profiles. For methane (TCOM-CH4), we use both HALOE and ACE satellite profile measurements (1992–2018) to train the XGBoost model while profiles from three later years (2019–2021) are used as an independent evaluation data set. As there are no nitrous oxide (N2O) profile measurements for earlier years, XGBoost-derived correction terms to construct TCOM-N2O profiles are derived using only ACEFTS profiles for the 2004–2018 time period, with profiles from 2019–2021 again being used for the independent evaluation. Overall, both TCOM-CH4 and TCOM-N2O profiles show excellent agreement with the available satellite measurement-based data sets. We find that compared to evaluation profiles, biases in TCOM-CH4 and TCOM-N2O are generally less than 10 % and 50 %, respectively, throughout the stratosphere. Daily zonal mean profile data sets on altitude (15–60 km) and pressure (300–0.1 hPa) levels are publicly available via https://doi.org/10.5281/zenodo.7293740 for TCOM-CH4 (Dhomse, 2022a) and https://doi.org/10.5281/zenodo.7386001 for TCOM-N2O (Dhomse, 2022b).

Section: Methodsmentioning

confidence: 99%

Using machine-learning to construct TOMCAT model and occultation measurement-based stratospheric methane (TCOM-CH4) and nitrous oxide (TCOM-N2O) profile data sets

2023

Preprint

“…(2) the incomplete presentation of complex atmospheric processes and their feedbacks, (3) the incorrect parameterisation for photochemical reactions in a CTM, or (4) the uncertainties in observational data sets (e.g. Dhomse et al, 2021).…”

Section: Total Column Ozone Trends and Explanatory Variablesmentioning

confidence: 99%

“…These biases might be associated with deficiencies in the representation of the photochemical reactions and dynamical processes in the model (e.g. Mitchell et al, 2020;Dhomse et al, 2013Dhomse et al, , 2016Dhomse et al, , 2021.…”

Section: Comparison Of Vertical Ozone Profilesmentioning

confidence: 99%

“…However, model-simulated ozone distributions generally show some biases with respect to observationbased data sets due to uncertain photochemical parameters, transport errors, and other simplifications of computationally expensive processes (e.g. WMO, 2014WMO, , 2018Dhomse et al, 2018Dhomse et al, , 2021. The inability of chemical models to simulate the observed lower stratospheric ozone decrease can be largely attributed to the model deficiencies in, for example, transport (Chipperfield et al, 2018;Ball et al, 2018Ball et al, , 2020.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Effects of reanalysis forcing fields on ozone trends and age of air from a chemical transport model

et al. 2022

Atmos. Chem. Phys.

Self Cite

Abstract. We use TOMCAT, a 3-dimensional (3D) offline chemical transport model (CTM) forced by two different meteorological reanalysis data sets (ERA-Interim and ERA5) from the European Centre for Medium-Range weather Forecasts (ECMWF) to analyse seasonal behaviour and long-term trends in stratospheric ozone and mean age of air. The model-simulated ozone variations are evaluated against two observation-based data sets. For total column ozone (TCO) comparisons, we use the Copernicus Climate Change Service (C3S) data (1979–2019), while for ozone profiles we use the Stratospheric Water and OzOne Satellite Homogenized (SWOOSH) data set (1984–2019). We find that the CTM simulations forced by ERA-Interim (A_ERAI) and ERA5 (B_ERA5) can both successfully reproduce the spatial and temporal variations in stratospheric ozone. Also, modelled TCO anomalies from B_ERA5 show better agreement with C3S than A_ERAI, especially in Northern Hemisphere (NH) mid latitudes, except that it gives somewhat larger positive biases (> 15 DU, Dobson units) during winter–spring seasons. Ozone profile comparisons against SWOOSH data show larger differences between the two simulations. In the lower stratosphere, ozone differences can be directly attributed to the representation of dynamical processes, whereas in the upper stratosphere they can be directly linked to the differences in temperatures between ERAI and ERA5 data sets. Although TCO anomalies from B_ERA5 show relatively better agreement with C3S compared to A_ERAI, a comparison with SWOOSH data does not confirm that B_ERA5 performs better at simulating the variations in the stratospheric ozone profiles. We employ a multivariate regression model to quantify the TCO and ozone profile trends before and after peak stratospheric halogen loading in 1997. Our results show that, compared to C3S, TCO recovery trends (since 1998) in simulation B_ERA5 are significantly overestimated in the Southern Hemisphere (SH) mid latitudes, while for A_ERAI in the NH mid latitudes, simulated ozone trends remain negative. Similarly, in the lower stratosphere, B_ERA5 shows positive ozone recovery trends for both NH and SH mid latitudes. In contrast, both SWOOSH and A_ERAI show opposite (negative) trends in the NH mid latitudes. Furthermore, we analyse age of air (AoA) trends to diagnose transport differences between the two reanalysis data sets. Simulation B_ERA5 shows a positive AoA trend after 1998 and somewhat older age in the NH lower stratosphere compared to A_ERAI, indicating that a slower Brewer–Dobson circulation does not translate into reduced wintertime ozone buildup in the NH extratropical lower stratosphere. Overall, our results show that models forced by the most recent ERA5 reanalyses may not yet be capable of reproducing observed changes in stratospheric ozone, particularly in the lower stratosphere.

“…Here, we use the merged Stratospheric Water and OzOne Satellite Homogenized (SWOOSH, version 2.7) data set to 75 assess the stratospheric ozone trends (Davis et al, 2016) for the 1984-2020 time period. In addition, a machine-learning-based satellite-corrected gap-free global stratospheric ozone profile dataset from a chemical transport model (ML-TOMCAT, Dhomse et al, 2021) is also used for comparison.…”

mentioning

confidence: 99%

Stratospheric ozone trends and attribution over 1984–2020 using ordinary and regularised multivariate regression models

et al. 2023

Preprint

Abstract. Accurate quantification of long-term trends in stratospheric ozone can be challenging due to their sensitivity to natural variability, the quality of the observational datasets, non-linear changes in forcing processes as well as the statistical methodologies. Multivariate linear regression (MLR) is the most commonly used tool for ozone trend analysis, however, the complex coupling in most atmospheric processes can make it prone to the over-fitting or multi-collinearity-related issues when using the conventional Ordinary Least Squares (OLS) setting. To overcome this issue, we adopt a regularised (Ridge) regression method to estimate ozone trends and quantify the influence of individual processes. Here, we use the Stratospheric Water and OzOne Satellite Homogenized (SWOOSH) merged data set (v2.7) to derive stratospheric ozone profile trends for the period 1984–2020. Beside SWOOSH, we also analyse a machine-learning-based satellite-corrected gap-free global stratospheric ozone profile dataset from a chemical transport model (ML-TOMCAT), and output from two chemical transport model (TOMCAT) simulations forced with ECMWF reanalyses ERA-Interim and ERA5. With Ridge regression, the stratospheric ozone profile trends from SWOOSH data show smaller declines during 1984–1997 compared to OLS with the largest differences in the lowermost stratosphere (> 4 % per decade at 100 hPa). Upper stratospheric ozone has increased since 1998 with maximum (~2 % per decade near 2 hPa) in local winter for mid-latitudes. Negative trends with large uncertainties are observed in the lower stratosphere with the most pronounced in the tropics. The largest differences in post-1998 trend estimates between OLS and Ridge regression methods appear in the tropical lower stratosphere (with ~7 % per decade difference at 100 hPa). Ozone variations associated with natural processes such as the quasi-biennial oscillation (QBO), the solar variability, the El Niño–Southern Oscillation (ENSO), the Arctic oscillation (AO) and the Antarctic oscillation (AAO) also indicate that Ridge regression coefficients are somewhat smaller and less variable compared to the OLS-based estimates. Additionally, ML-TOMCAT based trend estimates are consistent with SWOOSH data set. Finally, we argue that the large differences between the satellite-based data and model simulations confirm that there are still large uncertainties in ozone trend estimates especially in the lower stratosphere, and caution is needed when discussing results if explanatory variables used are correlated.