Toward improved regional estimates of carbon dioxide sources and sinks through coupled carbon–atmospheric data assimilation

Open Access
Chen, Hans Weiteng
Graduate Program:
Meteorology and Atmospheric Science
Doctor of Philosophy
Document Type:
Date of Defense:
August 06, 2018
Committee Members:
  • Fuqing Zhang, Dissertation Advisor/Co-Advisor
  • Fuqing Zhang, Committee Chair/Co-Chair
  • Thomas Lauvaux, Committee Member
  • Steven B. Feldstein, Committee Member
  • Richard B. Alley, Outside Member
  • David W. Titley, Outside Member
  • carbon cycle
  • data assimilation
  • atmospheric inversion
  • climate
  • climate change
  • ensemble kalman filter
  • Bayesian inference
  • act-america
  • co2
  • carbon dioxide
Accurate estimates of regional carbon dioxide (CO2) sources and sinks are necessary to further our understanding of the carbon cycle and improve predictions of future climate change. CO2 surface fluxes can be constrained using atmospheric CO2 observations combined with atmospheric transport models through so-called top-down or inverse methods. At regional scales, however, inverse estimates of CO2 fluxes have been shown to be sensitive to errors in model representation of atmospheric transport. How to account for such atmospheric transport errors in inversions is currently not well understood. This dissertation examines the impact of atmospheric transport errors on simulated atmospheric CO2 mole fractions and inferred CO2 fluxes at subcontinental scales and hourly to monthly time scales. We first investigate how much space for improvement there is in two contemporary CO2 analysis datasets by comparing CO2 mole fractions from the analyses with airborne in situ measurements of CO2 from the Atmospheric Carbon and Transport - America field campaigns in summer 2016 and winter 2017. The analyses show an overall good agreement with observations except for large biases in near-surface CO2 mole fractions in the Mid-Atlantic region of the United States during summer, which suggests that CO2 fluxes can be further optimized in this region. Next, we quantify how transport errors due to uncertainties in meteorological initial conditions propagate to errors in atmospheric CO2 mole fractions through ensemble sensitivity experiments in a regional mesoscale model. Transport errors in CO2 are found to be of comparable magnitude and share similar spatiotemporal characteristics as errors due to uncertainties in CO2 fluxes on sub-monthly time scales. Finally, we present the development of a coupled carbon--atmospheric data assimilation system for regional CO2 flux inversion. This data assimilation system uses the ensemble Kalman Filter to optimize both meteorological variables and CO2 mole fractions and fluxes. Coupling the atmospheric and carbon states allows us to investigate the role of atmospheric transport errors in the CO2 flux optimization. The data assimilation system is tested in a series of perfect model experiments with synthetic observations to examine how well the CO2 flux inversion performs when different types of errors are introduced.