This title appears in the Scientific Report :
2020
Please use the identifier:
http://hdl.handle.net/2128/26422 in citations.
Please use the identifier: http://dx.doi.org/10.5194/gmd-2020-169 in citations.
IntelliO3-ts v1.0: A neural network approach to predict near-surface ozone concentrations in Germany
IntelliO3-ts v1.0: A neural network approach to predict near-surface ozone concentrations in Germany
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 (...
Saved in:
Personal Name(s): | Kleinert, Felix (Corresponding author) |
---|---|
Leufen, Lukas H. / Schultz, Martin G. | |
Contributing Institute: |
Jülich Supercomputing Center; JSC |
Published in: | Geoscientific model development discussions, 2020 (2020) S. 1--69 |
Imprint: |
Katlenburg-Lindau
Copernicus
2020
|
DOI: |
10.5194/gmd-2020-169 |
Document Type: |
Preprint |
Research Program: |
Earth System Data Exploration Doktorand ohne besondere Förderung Artificial Intelligence for Air Quality Data-Intensive Science and Federated Computing |
Link: |
Get full text OpenAccess |
Publikationsportal JuSER |
Please use the identifier: http://dx.doi.org/10.5194/gmd-2020-169 in citations.
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. |