Inferring Fossil Fuel NOx and CO2 Emissions: Combining Bottom-Up Inventories with Improved OMI NO2 Data

Nitrogen dioxide (NO₂) is a short-lived trace gas that has significant impacts on air quality and the ozone layer. It is emitted predominantly from anthropogenic sources. Our team is responsible for developing and improving the NO₂ standard product (stratospheric and tropospheric Vertical Column Densities, VCDs, current version 3.1) derived from the hyperspectral measurements by Ozone Monitoring Instrument (OMI) on board the NASA EOS Aura satellite. We first introduce the forthcoming OMI NO₂ tropospheric VCDs (version 4.0), processed with an improved tropospheric NO₂ Air-Mass Factor (AMF) algorithm. The algorithm is based on a conceptually new, geometry-dependent Lambertian surface equivalent reflectivity (GLER) operational product. GLER is calculated using the vector radiative transfer model VLIDORT, which assimilates high–resolution bidirectional reflectance distribution (BRDF) information from NASA’s Aqua MODIS instrument over land and the wind-dependent Cox–Munk wave-facet slope distribution over water, the latter with a contribution from the water-leaving radiance. The GLER and the corresponding consistently retrieved effective cloud fraction and optically centroid cloud pressure provide inputs to the new NO₂ AMF algorithm, which increases tropospheric VCDs (in the upcoming version 4.0) by up to 50% in highly polluted areas; the differences include both cloud and surface BRDF effects as well as biases between the MODIS and OMI-based surface reflectance data sets. These operational updates accommodate, on a global scale, the most important improvements proposed for regional OMI NO₂ tropospheric VCDs by expert users of the OMI data.

OMI tropospheric VCDs have been widely used to derive NO_x emissions at the surface owing to the relatively short NO_x lifetime. Independently, we have developed and contrasted both modelling and statistical fitting approaches to derive “top-down” emissions to compare with traditional “bottom-up” emissions inventories. In the modelling approach, we develop daily Global Modeling Initiative (GMI)-derived sensitivity factors to enable updates of “bottom-up” gridded (0.5^o x 0.625^o) NO_x emissions directly with the OMI measured tropospheric NO₂ VCDs. Here we discuss uncertainties in sensitivity factors and emissions by comparing current operational (v3.1) and new (v4.0) OMI tropospheric NO₂ VCDs. With the statistical fitting approach, we combine the tropospheric OMI NO₂ VCDs with global wind re-analysis data to estimate the “top-down” NO_x emissions by fitting modified exponential Gaussian plume model to the measured annual average downwind plume patterns for isolated point sources (e.g., power plants, smelters). We discuss emissions uncertainties by changing algorithm assumptions, wind data, and OMI NO₂ VCDs.

Broadening applicability of the high-quality OMI data, we develop a method to estimate trends in power plant CO₂ emissions from OMI-derived NO_x emissions. We develop bottom-up gridded datasets of NO_x/CO₂ emission ratios for US, China, and India over the OMI era based on detailed unit-level emission inventories of thermal power plants. We classify NO_x/CO₂ emission ratios for isolated power plants by fuel type, boiler types, and control technologies and apply the ratios to the OMI-derived NO_x emissions to estimate the corresponding CO₂ emissions. We also explore the combined use of OMI data and space-borne CO₂ data collected by recent carbon observing satellites such as GOSAT and OCO-2.