Model-Data Fusion for Constraining the Carbon Cycle Response to Interannual Climate Variations

Predicting future biosphere-atmosphere interactions and feedbacks to climate change requires accuracy across a range of scales, from individual ecosystem processes to the globally integrated atmospheric CO2 growth rate. This challenge requires insights from observations and the coordinated use of these observations for improving the structure and function of predictive models. Here, we focus specifically on observations and simulations of interannual variability in the carbon cycle. The impact of interannual climate variations on terrestrial ecosystems is an emergent response, based on the covariation of climate drivers and the independent responses of several ecosystem processes, including productivity, respiration, and disturbance. While these feedbacks are often diagnosed by considering the coincident patterns of climate and carbon cycle variability, the timing of the forcing in the context of the seasonal cycle or multi-annual responses must also be considered. For example, an ecosystem response to a climate anomaly may be phase-locked to the seasonal cycle. At longer timescales, ecosystems may also have lagged responses to climate forcing. Such memory in ecosystems may provide a mechanism to damp the impacts of interannual climate stress. Alternatively, ecosystem memory may permit larger cumulative impacts from slow variations in climate. Here, we use a combination of Earth system model output, atmospheric CO2, and satellite observations to demonstrate the interaction among these timescales. Our results suggest that careful analysis of autocorrelation within datasets and model output may facilitate development of more predictive models for future climate simulations.