Two Decades from Space: Satellite-Constrained Parameterization Delivers Computationally Efficient Global Ozone Production Rates and Sensitivity Records

Over the last two decades there have been diverse variabilities and trends in anthropogenic, pyrogenic, and biogenic emissions, resulting in varied responses of net ozone production rates (PO₃) to their precursors. A quantitative understanding of the underlying ozone precursor sources and the extent to which they influence PO₃ typically requires running computationally demanding chemical transport models, often constrained by satellite observations, under various modeling scenarios.Instead, we provide a much more efficient approach to predicting PO₃ using a deep neural network trained on more than 6 million observationally constrained data points collected from suborbital atmospheric composition missions. The parameterization meaningfully captures the non-linear relationships between O₃-NOX-VOC, as well as photolysis rates and water vapor. The parameterization inputs are constrained by various datasets including reanalysis models, ground remote sensing, TROPOMI and OMI retrievals, enabling us to provide a long-term record of global net ozone production rates along with magnitude-dependent sensitivity maps that advance beyond the conventionally binary maps (i.e., NOX-sensitive or VOC-sensitive) obtained from ozone indicators such HCHO/NO₂.We reveal predominantly positive trends in PO₃ over Asia and the Middle East (>30% relative to 2005) and negative trends across the eastern U.S., Europe, and parts of Africa during 2005-2019, based on stable long-term records from OMI. We demonstrate how rapid evolution of heat waves can substantially increase PO₃ and its sensitivity to NOₓ and VOC. Our high-resolution TROPOMI-based product reveals high locally-produced ozone in less-documented regions such as Johannesburg (South Africa), Rio de Janeiro (Brazil), São Paulo (Brazil), Santiago (Chile), Hanoi (Vietnam), Cairo (Egypt), and Tehran (Iran). This product, along with a comprehensive error budget, is freely available to our community (ozonerates.space)