Drought Response of a Subtropical Mixed Forest Ecosystem in Dinghushan, China: Insights from Dynamic Process Network and Eddy Covariance Observations

Subtropical forests serve as globally significant carbon sinks, yet how climate–CO2 flux coupling pathways respond to hydroclimatic extremes across multiple timescales remains poorly understood. Conventional flux-based approaches can detect changes in CO2 flux magnitudes but may miss underlying shifts in the directed relationships between meteorological drivers and ecosystem CO2 exchange processes. Dynamic Process Network (DPN) analysis provides an information-theoretic approach for resolving how directional climate–CO2 flux coupling reorganizes across seasons and hydroclimatic conditions. We adapted the DPN approach into a high-resolution, daily-sliding framework to quantify the time-varying, directed information flow from air temperature (Ta) and precipitation (P) to gross primary productivity (GPP), ecosystem respiration (RE), and net ecosystem exchange (NEE) in a subtropical mixed coniferous-broadleaved forest at Dinghushan, southern China (2003–2010). By coupling the DPN with multi-timescale Standardized Precipitation Evapotranspiration Index (SPEI-30, -90, and -360), we diagnosed how climate–CO2 flux coupling pathways reorganize under drought, normal, and wet conditions across meteorological, ecological, and hydrological timescales. Three main patterns emerged. First, although CO2 fluxes remained apparently stable across most hydroclimatic states, the DPN analysis revealed pathway-specific shifts in directed climate–CO2 flux coupling, indicating that climate–CO2 flux coupling pathways can reorganize even when changes in flux magnitude remain limited. Second, temperature and precipitation exhibited strongly asymmetric coupling modes: temperature coupling persisted year-round, whereas precipitation coupling was largely confined to the wet season and remained much weaker. Within this seasonal structure, ecosystem respiration emerged as the principal flux component associated with seasonal variation in net CO2 uptake; the stronger dry-season net CO2 uptake (~64% of annual NEE) was driven mainly by reduced RE rather than enhanced GPP. Third, drought-associated responses varied across SPEI accumulation timescales: wet-season TE under drought was significantly lower than under normal conditions for five of six pathways at SPEI-30 and all six at SPEI-90. In contrast, wet conditions at SPEI-360 were associated with enhanced wet-season net CO2 uptake through concurrent GPP enhancement and RE suppression, with a significantly more negative NEE anomaly than under drought conditions (−0.674 vs. +0.118 gC m−2 d−1; p < 0.001). These findings demonstrate that information-theoretic diagnostics can detect changes in ecosystem climate sensitivity that remain invisible to conventional flux-magnitude analyses, offering a complementary diagnostic framework for assessing changes in ecosystem CO2 exchange sensitivity under intensifying hydroclimatic extremes.