气候变化领域集成服务门户

Field-measured forest productivity and its time-series are critical for understanding the impact of climate change on forest carbon cycling, and also for validating process-based models. Tree-rings are widely used to reconstruct stand productivity history. However, it remains ambiguous how to ensure a good sample design to estimate forest productivity precisely, and meanwhile minimize the sampling effort. Here we addressed the following questions: (1) how many minimum tree-ring samples are needed to estimate forest productivity history accurately? (2) Can we predict optimum sampling design from climate conditions and forest structure? (3) Are commonly used sampling methods accurate enough? We set up 48 forest plots across four succession stages at four study sites along a latitude gradient in Northeast China. Tree-rings were sampled from all trees in each plot, and stand and individual productivity history over the past 20 years was reconstructed. We simulated different sampling designs by randomly extracting trees in each plot to select an optimal design that could estimate stand productivity history accurately but with the least sampling size, and to evaluate the accuracy of commonly used tree-ring sampling methods. We analyzed the influence of climatic gradient, forest type, distribution of productivity and biomass across DBH classes, and further developed models to predict sampling design. It was found that ca. 100%, 42%, 18% and 10% of individuals should be sampled from the 1/4 largest to the 1/4 smallest trees, respectively, to ensure an accurate estimation of stand productivity history. The optimum sampling design was highly related with the distribution of productivity and biomass across diameter classes of the plots, and changed significantly along climate gradient but showed no clear trend across successional stages. Sampling designs inferred from models based on climate indices and the biomass ratio in each diameter class by each plot could estimate stand productivity history satisfactorily from other validation plots. It was also showed that previous sampling methods to estimate forest productivity may incur large uncertainties. As such, these data should be viewed with caution. We proposed a new tree-ring sampling strategy based on the fact that forest productivity and its temporal variation were largely determined by large trees. We further showed that the optimal sampling design was predictable from forest structure and climate conditions. More studies are needed in other regions to develop models to optimize sampling designs for different forest types under different climate conditions, for a more accurate estimation of forest productivity dynamics under global climate change.