Harmonizing atmospheric ozone column concentrations over the Tibetan Plateau from 2005 to 2022 using OMI and Sentinel-5P TROPOMI: A deep learning approach

Abstract

Atmospheric ozone plays a pivotal role in Earth's climate system, influencing solar radiation absorption in the stratosphere and regulating ultraviolet light reaching the surface. Accurate monitoring of ozone concentration is crucial for environmental assessments, air quality monitoring, and climate change studies. The Ozone Monitoring Instrument (OMI) and Sentinel-5 Precursor TROPOspheric Monitoring Instrument (TROPOMI) provide valuable data for such monitoring. While OMI offers a long data record since 2004, but its effectiveness is hindered by its limitations in spatial resolution and signal-to-noise ratio, stemming from satellite hardware and retrieval algorithms. Sentinel-5P TROPOMI provides higher spatial resolution and improved signal-to-noise ratio, nevertheless, data record from it is rather short. Harmonizing these two datasets by taking the best use of their specific advantages is essential for creating a comprehensive and accurate atmospheric ozone concentration dataset. To maximize the advantages of these multi-source data products, our method utilizes a neural network to learn the mapping relationship between OMI and Sentinel-5P TROPOMI ozone column concentration products, constructing a harmonized model that optimizes the spatial and temporal sequence of historical OMI ozone column concentrations while considering topographic factors. The reconstructed ozone column concentration product is a long time series with the high spatial resolution and accuracy characteristics of Sentinel-5P TROPOMI. This research leverages powerful nonlinear modeling and spatial feature mapping capabilities based on deep learning networks to create a harmonized dataset of atmospheric ozone column concentrations, offering a comprehensive understanding of ozone distribution across the Tibetan Plateau. This dataset not only improves accuracy and precision in ozone concentration measurements but also facilitates in-depth analysis of local ozone variations, providing reliable dataset for scientific investigations into the atmospheric environment. The complete dataset is openly accessible at https://doi.org/10.5281/zenodo.10430751.

Keywords

• Ozone column concentration
• Data Harmonization
• OMI OMDOAO3E
• Sentinel-5P TROPOMI OFFL
• Deep Learning