Atmospheric water demand constrains net ecosystem production in subtropical mangrove forests

Abstract

Subtropical mangroves have great potential to sequester atmospheric carbon dioxide (CO2) and thus significantly contribute to climate change mitigation. Meanwhile, the carbon cycling processes in subtropical mangroves are vulnerable to the changing climatic conditions, such as the warming-induced vapor pressure deficit (VPD) increases. However, the impacts of VPD on net ecosystem production (NEP) in subtropical mangroves remain poorly understood due to a lack of detailed assessment of NEP responses to VPD among different mangrove forests over a long-term observational period. In this study, we deployed eddy-covariance systems to measure net ecosystem CO2 exchange at three subtropical mangrove forests for 16 site-years, comprising two natural and one artificial forests. We employed a state-of-the-art data-driven modeling approach (i.e., Shapley additive explanations (SHAP) framework based on extreme gradient boosting model), which enabled us to explore the interactive effects of meteorological and tidal factors (e.g. salinity) with VPD on the mangrove NEP. We revealed that air temperature, global solar radiation (GR) and wind speed have significant interactions on the response of NEP to VPD stress in subtropical mangroves. For instance, when GR was high, the SHAP interaction values of VPD and GR on NEP decreased with increasing VPD, but when GR was low, the trends were the opposite. However, instead of identifying interactive effects between tidal salinity and VPD on mangrove NEP, we came across potential independent influence of salinity on the same. SHAP analysis was also able to disentangle the impact of VPD from other abiotic drivers. Thus, we evaluated the threshold effect of VPD stress on NEP loss in subtropical mangroves and observed a range of 2.50–2.95 kPa. Above this range, VPD stress leveled off. The subtropical mangrove responses to VPD should be therefore considered in the dynamic global vegetation models to increase the accuracy in carbon cycle simulations in the future.