Evaluation of SO2 emissions from the OMI point source catalog

Pollutant emission measurements from the Ozone Monitoring Instrument (OMI) satellite are a valuable source of data on emissions given their spatial and temporal coverage especially for regions where no inventory data are available. These measurements are updated annually and include sources that may be missed by bottom-up inventories. Satellite measurements are combined with a meteorological re-analysis to derive a catalog of roughly 500 point sources.[1] However, to understand and validate this point source catalog, they need to be compared to definitive bottom-up inventory data. In this project we compared the Sulfur Dioxide (SO2) emissions from the OMI point source catalog for the contiguous US with SO2 emissions inventory data. Specifically, we compared the satellite emissions with 3 different datasets namely, the Energy Information Administration (EIA) power plants dataset (which includes SO2 emissions largely from Continuous Emission Monitoring Systems (CEMS) for power plants), the National Emissions Inventory (NEI) maintained by the Environmental Protection Agency (EPA) (which includes all point sources, power plant and non-power plant) and the Emissions and Grid Resource Integrated Database (EGRID) also maintained by the EPA (which covers point sources for power plants). Through comparisons with such diverse datasets, we intend to validate the emissions from the point source catalog both in terms of the source of emissions (power plants vs other sources) and temporal coverage (including multiple datasets allows us to examine trends over time in more detail). To make the emissions from the point source catalog and the inventories more comparable, we aggregated emissions from the inventories within a radius of 40 kms. This approach allowed us to categorize emissions from individual sources in inventories as either detected by the satellite or not detected thus also allowing us to quantify the error structure (satellite measurement inventory value) from detected sources. We find, as expected, that emission sources not detected by the satellite are the largest aggregate source of error between the satellite estimates and the inventories, especially in more recent years. For sources that are detected, we find that errors in aggregate (total of all detected sources) are relatively low. However, when exploring errors for individual sources over time, we find that the errors are not necessarily random i.e., there are consistent positive biases (over-estimates) or consistent negative biases (under-estimates) for many sources. Moreover, the errors for individual sources in any given year can be significant with large over or under-estimates ranging from -90% to +500% (roughly 10 - 90th percentile) in an asymmetric distribution with a long tail. [1] Fioletov at al 2015, Beirle et al 2011