Potential Predictability Limits of the Global Carbon Cycle

The growth rate of atmospheric CO₂ on inter-annual time-scales is mainly driven by the response of the land and ocean carbon sinks to climate variability. As of now, it is unknown for how long atmospheric CO₂ variations are predictable and how the ocean and land carbon sinks contribute to this predictability of atmospheric CO₂.

Using perfect-model simulations representing perfect initialization, which yield an upper bound of predictability in an Earth-System-Model, we show that inter-annual atmospheric CO₂ variations are potentially predictable for three years. We find predictability of two years for the annual oceanic CO₂ fluxes with areas of higher predictability regionally. Terrestrial CO₂ flux is also predictable for two years and dominated by the ENSO-affected tropical forests and mid-latitudes. The isolated effect of temporally accumulated global terrestrial carbon sink predictability on atmospheric CO₂ of five years dominates over its oceanic counterpart of 12 years. Therefore the terrestrial carbon cycle limits predictability of atmospheric CO₂ concentration.

Our research shows the potential of ESM-based prediction systems to predict near-term variations in atmospheric CO₂. These prediction systems can inform policy-makers about the expected natural variations of the global carbon cycle to assess the expected mitigation efforts efficiency within the time-scale of the global stocktakes.

Reference:

Spring and Ilyina (2019). “Predictability Horizons in the Global Carbon Cycle Inferred from a Perfect-Model Framework”. Geophysical Research Letters, submitted manuscript