Baker, David: CIRA/Colorado State University (Project Lead)
Project Funding:
2014 - 2018
NRA: 2014 NASA: Carbon Monitoring System
Funded by NASA
Abstract:
Measurements of atmospheric CO2 concentration have provided a top-down view of the global carbon cycle, clarifying the impact of the anthropogenic fossil fuel input, and giving a rough latitudinal breakdown of the uptake of the fossil input by the oceans and land biosphere. NASA and other space agencies around the world have invested great effort in designing satellite missions to measure column-integrated CO2 concentrations from space, in hopes of getting enough spatio-temporal coverage to resolve surface CO2 fluxes at regional scales -- it is hoped that the processes driving the uptake and release of CO2 will be easier to identify at these scales, leading to better predictions of CO2 levels and global warming in the future. These CO2 measurements complement direct measurements of plant biomass nicely, since they sense the impact of other land ecosystem processes less easily measured (e.g., carbon stored below the ground in roots and soils, and carbon running off into streams and groundwater), as well as the impact of fossil fuel and biomass burning, and air-sea fluxes. Global flux inversion studies based on atmospheric measurements could thus be used as a check on the more direct measurement of plant biomass. Alternatively, they could be used as a framework for interpreting the biomass measurements in the context of the broader carbon cycle. If the flux estimates from such a system could be made at a spatial resolutions fine enough to parse the results across geopolitical boundaries, with reliable uncertainty estimates, they could be suitable for carbon trading and treaty verification purposes. The density and reliability of current satellite CO2 measurements have limited their usefulness towards this end so far, but the expected explosion of satellite CO2 data in the coming decade or two, including eventually from satellites in geosynchronous and highly-elliptic orbits rapidly scanning the land surface, should make this feasible. If CO2 fluxes must be resolved at scales of 1x1 deg or better to attribute them to individual countries reliably, then there is also a computational challenge to overcome in implementing such an inversion system: atmospheric transport models take roughly an order of magnitude longer to run each time the spatial resolution is doubled; if the resolution of the fluxes is increased from current levels (order 4x4 deg) to 1x1 deg, then the inversions should take roughly a hundred times longer to complete. Running the models at even finer scales is desirable, to come closer to the scales at which the measurements are actually made (e.g. of order 3 km2 for an OCO-2 pixel FOV). Here we propose a new inversion method that will efficiently estimate fluxes at sub-degree resolution, while at the same time producing a high-rank covariance matrix that quantifies flux uncertainty at the same scales. It solves the same Euler-Lagrange equations as the currently-used variational methods do, but does so with a direct matrix inversion rather than with an iterative descent method. The measurements are grouped into blocks, and a basis function is run through the transport model for each block, with the highest spatial resolution being coarsened as mixing spreads out the signal. The variational method is thus effectively parallelized, since the basis functions can be run on separate processors. Once the matrix inversion is done, the resulting covariance matrix may be used as a preconditioner in the standard iterative search to refine the finest scales. The rank of the covariance matrix produced by the method is limited only by the size of the matrix that can be inverted in memory, e.g. O(10,000), as compared with the O(100) matrices currently produced by ensemble Kalman filter and iterative variational methods. We test the accuracy of the uncertainties given by this covariance matrix and use it to compare the ability of different CO2-measuring satellite concepts to constrain country-scale annual mean fluxes.
Publications:
Basu, S., Baker, D. F., Chevallier, F., Patra, P. K., Liu, J., Miller, J. B. 2018. The impact of transport model differences on CO<sub>2</sub> surface flux estimates from OCO-2 retrievals of column average CO<sub>2</sub>. Atmospheric Chemistry and Physics. 18(10), 7189-7215. DOI: 10.5194/acp-18-7189-2018
More details may be found in the following project profile(s):
