Underestimated Ecosystem Carbon Turnover Time and Sequestration under the Steady State Assumption: a Perspective From Long-term data Assimilation

It is critical to accurately estimate carbon (C) turnover time as it dominates the uncertainty in ecosystem C sinks and their response to future climate change. However, C turnover time is mainly estimated under the C cycle steady state assumption (SSA), an ideal situation that rarely happens in reality. Considerable biases may arise when improperly invoking SSA in estimating C turnover time and its covariance with climate at local or large scales with substantial C sinks, which may deeply affect ecosystem C sequestration evaluation. But this issue has not yet been carefully examined. Here we developed a novel model-data fusion (MDF) framework and systematically analyzed the SSA-induced biases compared with the realistic dynamic disequilibrium state using time-series data collected from 10 permanent forest plots in the eastern China monsoon region. The results showed that: (1) the SSA significantly underestimated the magnitude of mean turnover times (MTTs) by 30%, thereby leading to a 4.33-fold underestimation of the net ecosystem productivity (NEP) in these forest ecosystems, a major C sink globally; (2) the SSA-induced bias in MTT and NEP negatively correlates with forest age, which offers a significant caveat for SSA applied in young-aged ecosystems; (3) the sensitivity of MTT to temperature and precipitation under the SSA was 30% and 67% lower, respectively. Thus, under the expected climate change, the spatiotemporal changes in MTT are likely to be underestimated, thereby potentially resulting in large errors in the variability of predicted global NEP. These explicit quantifications on the substantial uncertainties in MTT and its response to climate arising from SSA would facilitate the future research toward better understanding of regional or global C cycle dynamics and C-climate feedback at the disequilibrium state, and the analytical framework developed here may effectively reduce the uncertainty in ecosystem C sequestration estimations with long-term data assimilation.