CO emissions inversion using satellite CO observations and uncertainty in tandem identify and Attribute New and Unknown Sources over Central Asia

This study focuses on TROPOMI based observations and retrieval uncertainties of tropospheric CO and HCHO in Central Asia, an area experiencing rapid development. The data is applied within the Model Free Inversion Estimation Framework (MFIEF), to calculate daily, grid-by-grid CO emissions and uncertainty range from 2019 to 2023.

The framework herein has been expanded by integrating the data with explicit observational uncertainties, and applying a new and unbiased analytical system to perform comprehensive uncertainty analysis. Results show that the observational uncertainties have a significant impact on the calculated emissions, with approximately 55% of the data deemed unreliable. After filtration, the remaining data reveals more distinct high-value areas, while excluding the small number of extremely high values which were as likely to be due to observational noise as the large number of very low emissions pixels. Remaining pixels generally conform to know industrial, power, coal, steel, mining, and urban areas, enhancing the reliability of emission estimates. A few interesting exceptions, as discussed herein include large underground coal fires.

The CO emissions exhibit distinct temporal and spatial patterns. In urban and industrial areas of China from 2019 to 2022, emissions show a downward trend, followed by a slight increase in 2023, while in underground coal fire areas and in non-Chinese areas of Central Asia there are different trends observed. Emissions are highest during the months with the least UV radiation and coldest temperatures, such as December and January. Spatially, high emissions are concentrated in urban and industrial areas, while natural areas have relatively lower emissions, with the notable exception of underground burning coal fields which are found to be roughly as significant as large steel, power, and industrial sites.

Comparisons with EDGAR indicate our results have both different spatial distribution and temporal variation. Our results show a greater likelihood of decreasing over time and more variability (daily to weekly scale). This provides a scientific basis for understanding CO emissions in Central Asia while also contributing to the improvement of emission inventories, air quality models, as a basis dataset for CO and CH₄ retrievals, and even for attribution studies to be performed.