Widespread Shift in the Seasonal Phase of Long-Term and Multi-Site Atmospheric CO2 Observations

Observing atmospheric CO₂ concentrations is akin to taking the pulse of the global carbon cycle. Long-term observations at sampling stations in the Northern Hemisphere reveal a pronounced seasonal cycle, reflecting the annual course of terrestrial photosynthetic carbon uptake and respiratory release. This cycle has changed over the last decades: The amplitude has increased due to increased respiration during winter and greater carbon uptake during the growing season. At the same time, the phase defined as the timing of the zero-downward crossing of the seasonal cycle has shifted to earlier in the season. However, the drivers of the increase in amplitude, and particularly of the phase shift, are still uncertain.

Here, we analyse this phase shift and its drivers over the last six decades using observational data, process-based models and statistical learning. Our analysis reveals a statistically significant phase shift towards earlier dates of 1.5 – 2.0 days per decade at multiple stations in the Northern Hemisphere and even at the South Pole (2 ± 1 days per decade). In contrast to the increase in amplitude and phenological changes, the phase shift does not increase with latitude and is rather consistent across the latitudinal gradient.

To understand what is behind the observed phase changes, we next analyse simulations from different experiments using Earth system models and land surface model outputs from the TRENDY protocol. The carbon fluxes from the TRENDY models are transported using an atmospheric transport model. Additionally, we train statistical learning models to predict phase changes based on various potential drivers, such as observed temperature and pressure fields and evaluate their performance and feature importance.

By combining long-term atmospheric CO₂ observations, process-based model simulations and statistical learning, this study will shed light on the driving forces of seasonal CO₂ phase shifts and provide key insights into the changing land carbon dynamics.