The response of global and biome-level GPP to drought: An analysis of eddy-covariance data and global ecosystem model output

Background/Question/Methods

Extreme weather events are considered principal drivers of terrestrial gross primary productivity (GPP) and are expected to increase in frequency and/or severity in many regions of the world due to climate change. Numerous published studies indicate that GPP in forest ecosystems are heavily influenced by water scarcity. However, most previous analyses focus on single regions, which makes it difficult to draw a generalized and global picture of how GPP responds to drought. In addition, since the response of the terrestrial carbon cycle to drought is driven by multiple biotic processes and abiotic variables, replicating these relationships represents a formidable challenge for terrestrial biosphere models. We implemented a systematic approach to quantify GPP-drought relationships: (1) converting raw precipitation data into standardized precipitation index (SPI), which has a standard-normal distribution, and setting a threshold standard deviation to identify extremely dry conditions in the spatiotemporal data array; (2) applying the flood-in algorithm to join the “dry” grid cells that are adjacent in space and time to form three-dimensional drought voxels; (3) subtracting the GPP seasonality and linear trend in 1-degree grid cell to yield GPP anomalies comparable across space and time; (4) computing GPP anomalies over the spatiotemporal domain of drought voxels.  

Results/Conclusions

We applied this approach to GPP and associated precipitation data (1988-2010) from a global eddy-covariance-derived dataset and simulations from four dynamic global vegetation models (DGVMs) from the Coupled Model Intercomparison Project (CMIP5); and explored the relationships between GPP anomalies and drought events in each of these sources. The relationships extracted from eddy covariance estimates agree qualitatively with those from the DGVMs. For both of eddy-covariance and DGVMs data, the 50-100 largest drought events can explain 80% of the combined negative GPP anomalies across global forest ecosystems, and the loss of GPP (magnitude of the negative anomaly) increases exponentially with increasing water scarcity.  Both eddy-covariance and DGVMs analyses indicated that the GPP anomalies in tropical forests are highly sensitive to drought, GPP anomalies in boreal forests have low sensitivity to drought, and temperate forests show intermediate sensitivity. Compared with the relationships from eddy-covariance data, the relationships between GPP anomalies and drought in DGVMs were more tightly correlated. However, there was considerable variation among different models in the shapes of their response curves of GPP anomalies to drought.