Improving estimates of surface carbon fluxes to support emissions monitoring, reporting and verification at local and regional scales: quantifying uncertainty and the effects of spatial scaling.

Abstract:

Current approaches to quantifying surface-atmosphere fluxes of carbon often combine inventories of fossil fuel carbon emissions (ffCO₂) and biosphere flux estimates with atmospheric measurements to drive forward and inverse-atmospheric modeling at high spatial and temporal resolutions (1km grids, hourly time steps have become common). Given that over 70% of total ffCO2 emissions are attributable to urban areas, accurate estimates of ffCO₂ at urban scales are critical to support emissions mitigation policies at state and local levels. A successful regional or national carbon monitoring system requires a careful quantification of the uncertainties associated with estimates of both ffCO₂ and biogenic carbon fluxes.

Errors in the spatial distribution of ffCO₂ priors used to inform atmospheric transport models can bias posterior flux estimates, and potentially provide misleading information to decision makers on the impact of policies. Most current ffCO₂ priors are either too coarsely resolved in time and space, or suffer from poorly quantified errors in spatial distributions at local scales. Accurately downscaling aggregate activity data requires a careful understanding of the potentially non-linear relationships between source processes and spatial proxies. We report on ongoing work to develop an integrated, high-resolution carbon monitoring system for the Northeastern U.S., and discuss insights into the impact of spatial scaling on model uncertainty.

We use a newly developed dataset of hourly surface carbon fluxes for all human and biogenic sources at 1km grid resolution for the years 2013 and 2014. To attain these spatial and temporal resolutions, ffCO₂ flux estimates were subject to varying degrees of aggregation and/or downscaling depending on the native source data for each sector. We will discuss several important examples of how the choice of scaling variables and priors influences the spatial distribution CO₂ and CH₄ retrievals.