Design and evaluation of CO<sub>2</sub> observation network to optimize surface CO<sub>2</sub> fluxes in Asia using observation system simulation experiments

Abstract

Continuous efforts have been made to monitor atmospheric CO2 as it is one of the most influential greenhouse gases in Earth's atmosphere. Inverse modeling, which is one of the methods to 10 carry out such monitoring, derives estimated CO2 mole fractions in the air from calculated surface carbon fluxes using model and observed CO2 mole fraction data. Although observation data is crucial for successful modeling, comparatively fewer in-situ observation sites are located in Asia compared to Europe or North America. Based on the importance of the terrestrial ecosystem of Asia for global carbon exchanges, more observation stations and an effective observation network design are required. In this 15 paper, several observation network experiments were conducted to optimize the surface carbon flux of Asia using CarbonTracker and observation system simulation experiments (OSSE). The impacts of the redistribution of and additions to the existing observation network of Asia were evaluated using hypothetical in-situ observation sites. In the case of the addition experiments, 10 observation stations,which is a practical number for real implementation, were added through three strategies: random addition, 20 the influence matrix (i.e., self-sensitivity), and ecoregion information within the model. The simulated surface carbon flux in Asia in summer can be improved by redistributing the existing observation network.The addition experiments revealed that considering both the distribution of normalized self-sensitivity and ecoregion information can yield better simulated surface carbon fluxes compared to random addition, regardless of the season. This study provides a diagnosis of the existing observation network and useful 25 information for future observation network design in Asia to estimate the surface carbon flux, and also suggests the use of an influence matrix for designing carbon observation networks. Unlike other previous