Preprints
https://doi.org/10.5194/egusphere-2024-2676
https://doi.org/10.5194/egusphere-2024-2676
29 Oct 2024
 | 29 Oct 2024
Status: this preprint is open for discussion.

Knowledge-inspired fusion strategies for the inference of PM2.5 values with a Neural Network

Matthieu Dabrowski, José Mennesson, Jérôme Riedi, Chaabane Djeraba, and Pierre Nabat

Abstract. Ground-level concentrations of Particulate Matter (more precisely PM2.5) are a strong indicator of air quality, which is now widely recognized to impact human health. Accurately inferring or predicting PM2.5 concentrations is therefore an important step for health hazard monitoring and the implementation of air quality related policies. Various methods have been used to achieve this objective, and Neural Networks are one of the most recent and popular solutions.

In this study, a limited set of quantities that are known to impact the relation between column AOD and surface PM2.5 concentrations are used as input of several networks architectures to investigate how different fusion strategies can impact and help explain predicted PM2.5 concentrations. Different models are trained on two different sets of simulated data, namely global scale atmospheric composition reanalysis provided by the Copernicus Atmospheric Monitoring Service (CAMS) as well as higher resolution data simulated over Europe with the Centre National de Recherches Météorologiques ALADIN model.

Based on an extensive set of experiments, this work proposes several models of knowledge-inspired Neural Networks, achieving interesting results both from the performance and interpretability points of view. Specifically, novel architectures based on BC-GANs (which are able to leverage information from sparse ground observation networks) and on more traditional UNets, employing various information fusion methods, are designed and evaluated against each other. Our results can serve as baseline benchmark for other studies and be used to develop further optimised models for the inference of PM2.5 concentrations from AOD at either global or regional scale.

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.
Matthieu Dabrowski, José Mennesson, Jérôme Riedi, Chaabane Djeraba, and Pierre Nabat

Status: open (until 24 Dec 2024)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
Matthieu Dabrowski, José Mennesson, Jérôme Riedi, Chaabane Djeraba, and Pierre Nabat

Data sets

Knowledge-inspired fusion strategies for the inference of PM2.5 values with a Neural Network - CAMS data for experiments Matthieu Dabrowski https://doi.org/10.5281/zenodo.13929498

CNRM-ALADIN64 - Regional climate simulation over the Euro-Mediterranean region Marc Mallet and Pierre Nabat http://dx.doi.org/10.25326/703

Model code and software

Knowledge-inspired fusion strategies for the inference of PM2.5 values with a Neural Network - code for experiments Matthieu Dabrowski https://doi.org/10.5281/zenodo.13947256

Matthieu Dabrowski, José Mennesson, Jérôme Riedi, Chaabane Djeraba, and Pierre Nabat

Viewed

Total article views: 104 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
80 19 5 104 0 0
  • HTML: 80
  • PDF: 19
  • XML: 5
  • Total: 104
  • BibTeX: 0
  • EndNote: 0
Views and downloads (calculated since 29 Oct 2024)
Cumulative views and downloads (calculated since 29 Oct 2024)

Viewed (geographical distribution)

Total article views: 99 (including HTML, PDF, and XML) Thereof 99 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 
Latest update: 08 Nov 2024
Download
Short summary
This work focuses on the prediction of aerosol concentration values at ground level, which are a strong indicator of air quality, using Artificial Neural Networks. A study of different variables and their efficiency as inputs for these models is also proposed, and reveals that the best results are obtained when using all of them. Comparison of networks architectures and information fusion methods allows the extraction of knowledge on the most efficient methods in the context of this study.