Assessment of Spatial Downscaling Used in Global Greenhouse Gas Emission Estimates

Abstract

Accurate fossil fuel CO2 (FFCO2) emission estimates are crucial for tracking mitigation and constraining atmospheric‐based emissions estimation. Global gridded emissions often downscale national total emissions using spatial proxies. Here, we examine such downscaling by comparing a recently released global gridded FFCO2 inventory (GRACED) to a bottom‐up, high‐resolution estimate (Vulcan) in the U.S. which has demonstrated consistency with independent atmospheric‐based emission estimates. While sector‐specific national totals are similar, large differences are found at the gridcell scale (e.g., total FFCO2 median gridcell‐scale relative difference is 128.7%). Examination of the downscaling approaches suggests that gridcell‐scale residential/commercial differences are driven by GRACED use of a population spatial proxy, missing the spatial influence of other driving factors such as surface air temperature or building size. Population is also a poor choice for industrial building emissions as they better correlate with transportation infrastructure such as rail and marine shipping hubs. The use of population and road length as an onroad FFCO2 emissions spatial proxy does not reflect differences in urban versus rural driving behavior. Finally, the use of the CARMA powerplant data set as a spatial proxy is ill‐advised due to its inconsistency with measured powerplant CO2 emissions. Similar implications were found in a previous study comparing GRACED to a European high‐resolution inventory (CAMS‐REG) suggesting some universality to biases in the GRACED emission estimates at fine scales. Aggregation to larger urban area aggregates ameliorates the large gridcell scale differences between GRACED and Vulcan, suggesting a more viable scale at which global downscaling proxies can be used.