Variations of carbon allocation and turnover time across tropical forests

Aim: The carbon sink in tropical forests is a highly uncertain component of the global carbon budget. An understanding of the processes controlling this sink requires better quantification of carbon allocation, stocks and turnover times. Location: Tropical forests. Time period: 2010–2017. Major taxa studied: Tropical forest ecosystem. Methods: We develop a novel data assimilation system using satellite- based annual above- ground biomass derived from L- band vegetation optical depth with 25 km × 25 km grid spacing, together with leaf area, to constrain the 25 km × 25 km carbon allocation patterns of net primary productivity (NPP) into the wood, leaf and root pools, and their turnover times. Our average data- driven estimates of these variables are broadly consistent with independent ground- based estimates of NPP allocation and wood turnover from forest inventory plots. Results: In tropical forest, on average, the NPP allocation into wood (0.30 ± 0.04) is significantly higher than that into leaves (0.24 ± 0.07). From the wet to dry tropics, forest NPP allocation into both wood and leaves declines slightly. The turnover times of forest leaf pools exhibit little spatial variation, whereas the turnover times of wood pools in Africa (median and interquartile range: 50.1+5.5 −4.0 years) are slightly longer than those in South America (48.2+4.0 −3.3 years) and Southeast Asia (48.3+5.4 −3.1 years). Our datasets reveal emergent trade- offs across climatic and vegetation gradients between growth and life span/turnover for both wood and leaves. The spatial gradients of NPP allocation to wood/leaves are associated with canopy height, adjusted by climate condition and nutrient acquisition. The spatial gradients of wood and leaf turnover times are influenced mainly by climate and leaf characteristics. Main conclusions: Our data- driven estimates of carbon allocation and turnover times provide a basis for more detailed exploration of these mechanisms in field studies. This highlights that improved model representation of carbon allocation and turnover is necessary for more accurate prediction of future carbon dynamics.