Dissolved Organic Carbon Flux Is Driven by Plant Traits More Than Climate across Global Forest Types

Dissolved organic carbon (DOC) is one of the most important components in the global carbon cycle, which is largely influenced by climate and plant traits. Although previous studies have examined the impacts of climatic factors (e.g., mean annual temperature (MAT) and precipitation (MAP)) or plant traits (e.g., leaf area index, leaf nitrogen) on DOC, the relative importance of climate and plant traits on DOC flux remains unclear on a global scale. In this study, we compiled 153 pairs of DOC observational data from 84 forest sites to explore the relative importance of climate and plant traits on DOC flux with a linear mixed model, variance partitioning, and random forest approaches. Our results showed that DOC fluxes from throughfall and the litter layer were higher in broadleaved forests than those in coniferous forests. Throughfall-DOC flux increased significantly with MAT and MAP in coniferous forests, but that from the litter layer showed no significant correlations with climate factors. In broadleaved forests, throughfall-DOC flux increased with potential evapotranspiration (PET), while that from the litter layer was positively correlated with MAT. Meanwhile, throughfall-DOC flux had negative relationships with specific leaf area (SLA), leaf nitrogen content (LN), and leaf phosphorus content (LP) in broadleaved forests, but it showed a positive correlation with SLA in coniferous forests. Litter-layer-DOC flux increased with LN in broadleaved forests, but this correlation was the opposite in coniferous forests. Using the variance partitioning approach, plant traits contributed to 29.0% and 76.4% of the variation of DOC from throughfall and litter layer, respectively, whereas climate only explained 19.1% and 8.3%, respectively. These results indicate that there is a more important contribution by plant traits than by climate in driving the spatial variability of global forest DOC flux, which may help enhance forest management as a terrestrial carbon sink in the future. Our findings suggest the necessity of incorporating plant traits into land surface models for improving predictions regarding the forest carbon cycle.