The CarbonTracker Data Assimilation System for CO₂ and δ¹³C (CTDAS-C13 v1.0): retrieving information on land–atmosphere exchange processes

Abstract. To improve our understanding of the global carbon balance and its representation in terrestrial biosphere models, we present here a first dual-species application of the CarbonTracker Data Assimilation System (CTDAS). The system's modular design allows for assimilating multiple atmospheric trace gases simultaneously to infer exchange fluxes at the Earth surface. In the prototype discussed here, we interpret signals recorded in observed carbon dioxide (CO₂) along with observed ratios of its stable isotopologues ¹³CO₂∕¹²CO₂ (δ¹³C). The latter is in particular a valuable tracer to untangle CO₂ exchange from land and oceans. Potentially, it can also be used as a proxy for continent-wide drought stress in plants, largely because the ratio of ¹³CO₂ and ¹²CO₂ molecules removed from the atmosphere by plants is dependent on moisture conditions.

The dual-species CTDAS system varies the net exchange fluxes of both ¹³CO₂ and CO₂ in ocean and terrestrial biosphere models to create an ensemble of ¹³CO₂ and CO₂ fluxes that propagates through an atmospheric transport model. Based on differences between observed and simulated ¹³CO₂ and CO₂ mole fractions (and thus δ¹³C) our Bayesian minimization approach solves for weekly adjustments to both net fluxes and isotopic terrestrial discrimination that minimizes the difference between observed and estimated mole fractions.

With this system, we are able to estimate changes in terrestrial δ¹³C exchange on seasonal and continental scales in the Northern Hemisphere where the observational network is most dense. Our results indicate a decrease in stomatal conductance on a continent-wide scale during a severe drought. These changes could only be detected after applying combined atmospheric CO₂ and δ¹³C constraints as done in this work. The additional constraints on surface CO₂ exchange from δ¹³C observations neither affected the estimated carbon fluxes nor compromised our ability to match observed CO₂ variations. The prototype presented here can be of great benefit not only to study the global carbon balance but also to potentially function as a data-driven diagnostic to assess multiple leaf-level exchange parameterizations in carbon-climate models that influence the CO₂, water, isotope, and energy balance.