China is one of the largest emitters of sulfur dioxide (SO₂) and nitrogen dioxide (NO₂) in the world. Its growing economy heavily depends on coal fuels as an energy source. China alone is now responsible for about 30% of the total global emissions of SO₂ into the atmosphere while over 90% of these emissions are caused by coal consumption.

To reduce SO₂ emissions in China, the authorities have implemented several environmental regulations: the desulfurization of coal-fired power plants in 2005 & 2006, followed by a 12-year plan (2011-2015) with additional filtering efforts, phasing out obsolete capacity in coal-using industry and gradually using more oil, gas, and renewable energies from 2011 onward.

Since this long-term plan required a substantial update of the emission inventories, the research of MariLiza Koukouli from the Aristotle University of Thessaloniki and colleagues from the Royal Netherlands Meteorological Institute and the Royal Belgian institute for Aeronomy came in very handy.

Koukouli and her team used state-of-the-art OMI/Aura (Ozone Monitoring Instrument) of SO₂ satellite observations alongside a Chemical Transport Model, XTM, to update the existing emission estimates from 2005 to 2015. The team provided not only updates on emissions but also a hindsight calculation that allows an update on emission levels used by multiple different services, such as air quality forecasters.

Koukouli and colleagues used computing resources provided by the HellasGrid infrastructure, the Greek HTC infrastructure, which is part of the EGI Federation.

“The sheer volume of satellite and modelling data required extensive preparation before attempting to extract the updated emissions and made the usage of other types of computational infrastructure unmanageable”, says Koukouli.

Furthermore, a large amount of CPU power was required so that the input datasets were prepared for further analysis. “For example, 1 typical pre- and post-processing analysis of both satellite observations as well as chemical transport model simulations would require around 150 hours of runtime and approximately 1GB of memory”, Koukouli adds.

Hundreds of computing runs were required in order to complete the analysis and produce the updated emission inventory. “The HellasGrid infrastructure offers a complete set of tools, from the initial data ingestion into the system, to the computing language packages required to read the observations and extract the relevant information, making it indispensable for our work”, she concludes.

The satellite-based emission inventory shows a sharp decline in the emitted SO₂ for the entire Eastern China region, after a slight increase between years 2005 and 2010.

Koukouli is confident that satellite observations will be able to capture the state of emissions on a daily basis. The latest European satellite placed in orbit in October 2017, Sentinel-5P, carries a very high spatial resolution instrument, the TROPOspheric Monitoring Instrument, TROPOMI, on board the Sentinel-5P satellite, which will be able to provide a great insight in highly varying atmospheric pollution sources. The next step in Koukouli’s work is to use TROPOMI NO₂ and SO₂ observations to quantify emission levels for Europe.