scholarly journals IntelliO3-ts v1.0: A neural network approach to predict near-surface ozone concentrations in Germany

2020 ◽  
Author(s):  
Felix Kleinert ◽  
Lukas H. Leufen ◽  
Martin G. Schultz
Keyword(s):  
Time Series ◽  
Surface Ozone ◽  
Training Data ◽  
Horizontal Wind ◽  
Daily Maximum ◽  
High Exposure ◽  
Near Surface ◽  
Reference Models ◽  
Ozone Concentrations ◽  
Single Station

Abstract. The prediction of near-surface ozone concentrations is important to support regulatory procedures for the protection of humans from high exposure to air pollution. In this study, we introduce a data-driven forecasting model named IntelliO3-ts, which consists of multiple convolutional neural layers (CNN), grouped together as inception blocks. The model is trained with measured multi-year ozone and nitrogen oxides concentrations of more than 300 German measurement stations in rural environments, and six meteorological variables from the meteorological COSMO reanalysis. This is by far the most extensive dataset used for time series predictions based on neural networks so far. IntelliO3-ts allows predicting daily maximum 8-hour average (dma8eu) ozone concentrations for a lead time of up to four days, and we show that the model outperforms standard reference models like persistence. Moreover, we demonstrate that IntelliO3-ts outperforms climatological reference models for the first two days, while it does not add any genuine value for longer lead times. We attribute this to the limited deterministic information that is contained in the single station time series training data. We applied a bootstrapping technique to analyse the influence of different input variables and found, that the previous day ozone concentrations are of major importance, followed by 2 m temperature. As we did not use any geographic information to train IntelliO3-ts in its current version and included no relation between stations, the influence of the horizontal wind components on the model performance is minimal. We expect that the inclusion of advection-diffusion terms in the model could improve results in future versions of our model.

scholarly journals IntelliO3-ts v1.0: a neural network approach to predict near-surface ozone concentrations in Germany

2021 ◽  
Vol 14 (1) ◽  
pp. 1-25
Author(s):  
Felix Kleinert ◽  
Lukas H. Leufen ◽  
Martin G. Schultz
Keyword(s):  
Neural Network ◽  
Time Series ◽  
Surface Ozone ◽  
Training Data ◽  
Horizontal Wind ◽  
High Exposure ◽  
Near Surface ◽  
Reference Models ◽  
Ozone Concentrations ◽  
Single Station

Abstract. The prediction of near-surface ozone concentrations is important for supporting regulatory procedures for the protection of humans from high exposure to air pollution. In this study, we introduce a data-driven forecasting model named “IntelliO3-ts”, which consists of multiple convolutional neural network (CNN) layers, grouped together as inception blocks. The model is trained with measured multi-year ozone and nitrogen oxide concentrations of more than 300 German measurement stations in rural environments and six meteorological variables from the meteorological COSMO reanalysis. This is by far the most extensive dataset used for time series predictions based on neural networks so far. IntelliO3-ts allows the prediction of daily maximum 8 h average (dma8eu) ozone concentrations for a lead time of up to 4 d, and we show that the model outperforms standard reference models like persistence models. Moreover, we demonstrate that IntelliO3-ts outperforms climatological reference models for the first 2 d, while it does not add any genuine value for longer lead times. We attribute this to the limited deterministic information that is contained in the single-station time series training data. We applied a bootstrapping technique to analyse the influence of different input variables and found that the previous-day ozone concentrations are of major importance, followed by 2 m temperature. As we did not use any geographic information to train IntelliO3-ts in its current version and included no relation between stations, the influence of the horizontal wind components on the model performance is minimal. We expect that the inclusion of advection–diffusion terms in the model could improve results in future versions of our model.

scholarly journals A numerical process study on the rapid transport of stratospheric air down to the surface over western North America and the Tibetan Plateau

2019 ◽  
Vol 19 (9) ◽  
pp. 6535-6549 ◽  
Author(s):  
Bojan Škerlak ◽  
Stephan Pfahl ◽  
Michael Sprenger ◽  
Heini Wernli
Keyword(s):  
Boundary Layer ◽  
Tibetan Plateau ◽  
Surface Ozone ◽  
Vertical Transport ◽  
The Tibetan Plateau ◽  
Near Surface ◽  
Air Mass ◽  
Ozone Concentrations ◽  
Rapid Transport ◽  
Upper Level

Abstract. Upper-level fronts are often associated with the rapid transport of stratospheric air along tilted isentropes to the middle or lower troposphere, where this air leads to significantly enhanced ozone concentrations. These plumes of originally stratospheric air can only occasionally be observed at the surface because (i) stable boundary layers prevent an efficient vertical transport down to the surface, and (ii) even if boundary layer turbulence were strong enough to enable this transport, the originally stratospheric air mass can be diluted by mixing, such that only a weak stratospheric signal can be recorded at the surface. Most documented examples of stratospheric air reaching the surface occurred in mountainous regions. This study investigates two such events, using a passive stratospheric air mass tracer in a mesoscale model to explore the processes that enable the transport down to the surface. The events occurred in early May 2006 in the Rocky Mountains and in mid-June 2006 on the Tibetan Plateau. In both cases, a tropopause fold associated with an upper-level front enabled stratospheric air to enter the troposphere. In our model simulation of the North American case, the strong frontal zone reaches down to 700 hPa and leads to a fairly direct vertical transport of the stratospheric tracer along the tilted isentropes to the surface. In the Tibetan Plateau case, however, no near-surface front exists and a reservoir of high stratospheric tracer concentrations initially forms at 300–400 hPa, without further isentropic descent. However, entrainment at the top of the very deep boundary layer (reaching to 300 hPa over the Tibetan Plateau) and turbulence within the boundary layer allows for downward transport of stratospheric air to the surface. Despite the strongly differing dynamical processes, stratospheric tracer concentrations at the surface reach peak values of 10 %–20 % of the imposed stratospheric value in both cases, corroborating the potential of deep stratosphere-to-troposphere transport events to significantly influence surface ozone concentrations in these regions.

scholarly journals Sensitivity of Surface Ozone Simulation to Cumulus Parameterization

2008 ◽  
Vol 47 (5) ◽  
pp. 1456-1466 ◽  
Author(s):  
Zhining Tao ◽  
Allen Williams ◽  
Ho-Chun Huang ◽  
Michael Caughey ◽  
Xin-Zhong Liang
Keyword(s):  
Cloud Cover ◽  
Surface Ozone ◽  
Meteorological Conditions ◽  
Cumulus Parameterization ◽  
Daily Maximum ◽  
Boundary Layer Height ◽  
Ozone Concentrations ◽  
Simulated Precipitation ◽  
Cumulus Schemes ◽  
The Impact

Abstract Different cumulus schemes cause significant discrepancies in simulated precipitation, cloud cover, and temperature, which in turn lead to remarkable differences in simulated biogenic volatile organic compound (BVOC) emissions and surface ozone concentrations. As part of an effort to investigate the impact (and its uncertainty) of climate changes on U.S. air quality, this study evaluates the sensitivity of BVOC emissions and surface ozone concentrations to the Grell (GR) and Kain–Fritsch (KF) cumulus parameterizations. Overall, using the KF scheme yields less cloud cover, larger incident solar radiation, warmer surface temperature, and higher boundary layer height and hence generates more BVOC emissions than those using the GR scheme. As a result, the KF (versus GR) scheme produces more than 10 ppb of summer mean daily maximum 8-h ozone concentration over broad regions, resulting in a doubling of the number of high-ozone occurrences. The contributions of meteorological conditions versus BVOC emissions on regional ozone sensitivities to the choice of the cumulus scheme largely offset each other in the California and Texas regions, but the contrast in BVOC emissions dominates over that in the meteorological conditions for ozone differences in the Midwest and Northeast regions. The result demonstrates the necessity of considering the uncertainty of future ozone projections that are identified with alternative model physics configurations.

The differing impact of air stagnation on near-surface ozone across Europe

2020 ◽  
Author(s):  
José M. Garrido-Pérez ◽  
Carlos Ordóñez ◽  
Ricardo García-Herrera ◽  
Jordan L. Schnell
Keyword(s):  
Climate Model ◽  
Surface Ozone ◽  
Weather Conditions ◽  
Ozone Production ◽  
Maximum Temperature ◽  
Daily Maximum Temperature ◽  
Daily Maximum ◽  
Southern Europe ◽  
Near Surface ◽  
Photochemical Production

<p>Daily maximum temperature is known to be the meteorological variable that mostly controls the afternoon near-surface ozone concentrations during summer. Air stagnation situations, characterised by stable weather conditions and poor ventilation, also lead to the accumulation of pollutants and regional ozone production close to the surface. This work evaluates the joint effect of daily maximum temperature and a simplified air stagnation index on surface ozone observations in eight regions of Europe during summer 1998-2015.</p><p>As expected, the correlations of MDA8 O<sub>3</sub> (maximum daily 8-h running average ozone) with temperature are higher than with stagnation for most regions. Nevertheless, stagnation can also be considered as a good predictor of ozone, especially in the regions of central/southern Europe, where the correlation coefficients between MDA8 O<sub>3</sub> and the percentage of stagnant area are within the range 0.50–0.70. MDA8 O<sub>3</sub> consistently increases over central/southern Europe under stagnant conditions, but this is not always the case in the north. Under non-stagnant conditions and daily maximum temperatures within 20-25 ºC (typical temperatures of fair weather conditions that allow photochemical production), northern Europe is affected by southerly advection that often brings aged air masses from more polluted areas, increasing the MDA8 O<sub>3</sub> mixing ratios.</p><p>We have also found that the ozone diurnal cycles in the central/southern regions exhibit large amplitudes, with above-average daytime and below-average night-time concentrations, when stagnation occurs. Stagnant nights are often associated with stable shallow planetary boundary layer and, presumably, enhanced dry deposition and chemical destruction of ozone. After sunrise, mixing with air from air from the residual layer, accumulation of ozone and precursors, and photochemical production seem to be the main mechanisms involved in the build-up of daytime ozone.</p><p>According to previous studies, some of the central/southern European regions where stagnation has a clear impact on ozone have undergone significant upward trends in air stagnation in the past and are also likely to experience increases in the future. However, our study has identified other regions with unclear responses of summer ozone to the occurrence of stagnation. This indicates that climate model projections of increases in stagnation should not directly be translated into enhanced summer ozone pollution if the sensitivity of this pollutant to stagnation has not been proved for a particular region.</p>

scholarly journals Changes of daily surface ozone maxima in Switzerland in all seasons from 1992 to 2002 and discussion of summer 2003

2005 ◽  
Vol 5 (5) ◽  
pp. 1187-1203 ◽  
Author(s):  
C. Ordóñez ◽  
H. Mathis ◽  
M. Furger ◽  
S. Henne ◽  
C. Hüglin ◽  
...  
Keyword(s):  
Statistical Model ◽  
Surface Ozone ◽  
Global Radiation ◽  
Vertical Mixing ◽  
Analysis Of Covariance ◽  
Daily Maximum ◽  
Ozone Concentrations ◽  
Four Seasons ◽  
Lower Effect ◽  
The Impact

Abstract. An Analysis of Covariance (ANCOVA) was used to derive the influence of the meteorological variability on the daily maximum ozone concentrations at 12 low-elevation sites north of the Alps in Switzerland during the four seasons in the 1992–2002 period. The afternoon temperature and the morning global radiation were the variables that accounted for most of the meteorological variability in summer and spring, while other variables that can be related to vertical mixing and dilution of primary pollutants (afternoon global radiation, wind speed, stability or day of the week) were more significant in winter. In addition, the number of days after a frontal passage was important to account for ozone build-up in summer and ozone destruction in winter. The statistical model proved to be a robust tool for reducing the impact of the meteorological variability on the ozone concentrations. The explained variance of the model, averaged over all stations, ranged from 60.2% in winter to 71.9% in autumn. The year-to-year variability of the seasonal medians of daily ozone maxima was reduced by 85% in winter, 60% in summer, and 50% in autumn and spring after the meteorological adjustment. For most stations, no significantly negative trends (at the 95% confidence level) of the summer medians of daily O3 or Ox (O3+NO2) maxima were found despite the significant reduction in the precursor emissions in Central Europe. However, significant downward trends in the summer 90th percentiles of daily Ox maxima were observed at 6 sites in the region around Zürich (on average −0.73 ppb yr-1 for those sites). The lower effect of the titration by NO as a consequence of the reduced emissions could partially explain the significantly positive O3 trends in the cold seasons (on average 0.69 ppb yr-1 in winter and 0.58 ppb yr-1 in autumn). The increase of Ox found for most stations in autumn (on average 0.23 ppb yr-1) and winter (on average 0.39 ppb yr-1) could be due to increasing European background ozone levels, in agreement with other studies. The statistical model was also able to explain the very high ozone concentrations in summer 2003, the warmest summer in Switzerland for at least ~150 years. On average, the measured daily ozone maximum was 15 ppb (nearly 29%) higher than in the reference period summer 1992–2002, corresponding to an excess of 5 standard deviations of the summer means of daily ozone maxima in that period.

scholarly journals Can a global model reproduce observed trends in summertime surface ozone levels?

2012 ◽  
Vol 12 (15) ◽  
pp. 6983-6998 ◽  
Author(s):  
S. Koumoutsaris ◽  
I. Bey
Keyword(s):  
Transport Model ◽  
Control Strategies ◽  
Surface Ozone ◽  
The United States ◽  
Daily Maximum ◽  
Chemical Transport Model ◽  
Ozone Concentrations ◽  
Western Us ◽  
Ozone Trends ◽  
Rural Sites

Abstract. Quantifying trends in surface ozone concentrations is critical for assessing pollution control strategies. Here we use observations and results from a global chemical transport model to examine the trends (1991–2005) in daily maximum 8-h average concentrations in summertime surface ozone at rural sites in Europe and the United States (US). We find a decrease in observed ozone concentrations at the high end of the probability distribution at many of the sites in both regions. The model attributes these trends to a decrease in local anthropogenic ozone precursors, although simulated decreasing trends are overestimated in comparison with observed ones. The low end of observed distribution show small upward trends over Europe and the western US and downward trends in Eastern US. The model cannot reproduce these observed trends, especially over Europe and the western US. In particular, simulated changes between the low and high end of the distributions in these two regions are not significant. Sensitivity simulations indicate that emissions from far away source regions do not affect significantly summer ozone trends at both ends of the distribution in both Europe and US. Possible reasons for discrepancies between observed and simulated trends are discussed.

scholarly journals Two-scale multi-model ensemble: is a hybrid ensemble of opportunity telling us more?

2018 ◽  
Vol 18 (12) ◽  
pp. 8727-8744 ◽  
Author(s):  
Stefano Galmarini ◽  
Ioannis Kioutsioukis ◽  
Efisio Solazzo ◽  
Ummugulsum Alyuz ◽  
Alessandra Balzarini ◽  
...  
Keyword(s):  
Time Series ◽  
Air Quality ◽  
Surface Ozone ◽  
Regional Scale ◽  
Probability Of Detection ◽  
Daily Maximum ◽  
Regional Models ◽  
Global Models ◽  
Ensemble Performance ◽  
Scale Models

Abstract. In this study we introduce a hybrid ensemble consisting of air quality models operating at both the global and regional scale. The work is motivated by the fact that these different types of models treat specific portions of the atmospheric spectrum with different levels of detail, and it is hypothesized that their combination can generate an ensemble that performs better than mono-scale ensembles. A detailed analysis of the hybrid ensemble is carried out in the attempt to investigate this hypothesis and determine the real benefit it produces compared to ensembles constructed from only global-scale or only regional-scale models. The study utilizes 13 regional and 7 global models participating in the Hemispheric Transport of Air Pollutants phase 2 (HTAP2)–Air Quality Model Evaluation International Initiative phase 3 (AQMEII3) activity and focuses on surface ozone concentrations over Europe for the year 2010. Observations from 405 monitoring rural stations are used for the evaluation of the ensemble performance. The analysis first compares the modelled and measured power spectra of all models and then assesses the properties of the mono-scale ensembles, particularly their level of redundancy, in order to inform the process of constructing the hybrid ensemble. This study has been conducted in the attempt to identify that the improvements obtained by the hybrid ensemble relative to the mono-scale ensembles can be attributed to its hybrid nature. The improvements are visible in a slight increase of the diversity (4 % for the hourly time series, 10 % for the daily maximum time series) and a smaller improvement of the accuracy compared to diversity. Root mean square error (RMSE) improved by 13–16 % compared to G and by 2–3 % compared to R. Probability of detection (POD) and false-alarm rate (FAR) show a remarkable improvement, with a steep increase in the largest POD values and smallest values of FAR across the concentration ranges. The results show that the optimal set is constructed from an equal number of global and regional models at only 15 % of the stations. This implies that for the majority of the cases the regional-scale set of models governs the ensemble. However given the high degree of redundancy that characterizes the regional-scale models, no further improvement could be expected in the ensemble performance by adding yet more regional models to it. Therefore the improvement obtained with the hybrid set can confidently be attributed to the different nature of the global models. The study strongly reaffirms the importance of an in-depth inspection of any ensemble of opportunity in order to extract the maximum amount of information and to have full control over the data used in the construction of the ensemble.

Trends in near-surface ozone concentrations in Switzerland: the 1990s

2002 ◽  
Vol 36 (17) ◽  
pp. 2841-2852 ◽  
Author(s):  
Stefan Brönnimann ◽  
Brigitte Buchmann ◽  
Heinz Wanner
Keyword(s):  
Surface Ozone ◽  
Near Surface ◽  
Ozone Concentrations

scholarly journals Supplementary material to "Reanalysis of and attribution to near-surface ozone concentrations in Sweden during 1990–2013"

2017 ◽  
Author(s):  
Camilla Andersson ◽  
Heléne Alpfjord ◽  
Lennart Robertson ◽  
Per Erik Karlsson ◽  
Magnuz Engardt
Keyword(s):  
Surface Ozone ◽  
Near Surface ◽  
Ozone Concentrations ◽  
Supplementary Material

scholarly journals A multi-model comparison of meteorological drivers of surface ozone over Europe

2018 ◽  
Author(s):  
Noelia Otero ◽  
Jana Sillmann ◽  
Kathleen A. Mar ◽  
Henning W. Rust ◽  
Sverre Solberg ◽  
...  
Keyword(s):  
Time Series ◽  
Air Quality ◽  
Relative Humidity ◽  
Surface Ozone ◽  
Meteorological Variables ◽  
Maximum Temperature ◽  
Quality Models ◽  
The Balkans ◽  
Ozone Concentrations ◽  
The Relationship

Abstract. The implementation of European emission abatement strategies has led to significant reduction in the emission of ozone precursors during the last decade. Ground level ozone is also influenced by meteorological factors such as temperature, which exhibit interannual variability, and are expected to change in the future. The impacts of climate change on air quality are usually investigated through air quality models that simulate interactions between emissions, meteorology and chemistry. Within a multi-model assessment, this study aims to better understand how air quality models represent the relationship between meteorological variables and surface ozone concentrations over Europe. A multiple linear regression (MLR) approach is applied to observed and modelled time series across ten European regions in springtime and summertime for the period of 2000–2010 for both models and observations. Overall, the air quality models are in better agreement with observations in summertime than in springtime, and particularly in certain regions, such as France, Mid-Europe or East-Europe, where local meteorological variables show a strong influence on surface ozone concentrations. Larger discrepancies are found for the southern regions, such as the Balkans, the Iberian Peninsula and the Mediterranean basin, especially in springtime. We show that the air quality models do not properly reproduce the sensitivity of surface ozone to some of the main meteorological drivers, such as maximum temperature, relative humidity and surface solar radiation. Specifically, all air quality models show more limitations to capture the strength of the relationship ozone-relative humidity detected in the observed time series in most of the regions, in both seasons. Here, we speculate that dry deposition schemes in the air quality models might play an essential role to capture this relationship. We further quantify the relationship between ozone and maximum temperature (mo3-T, climate penalty) in observations and air quality models. In summertime, most of the air quality models are able to reproduce reasonably well the observed climate penalty in certain regions such as France, Mid-Europe and North Italy. However, larger discrepancies are found in springtime, where air quality models tend to overestimate the magnitude of observed climate penalty.

Sign in / Sign up

Export Citation Format

Share Document