TL;DR
This study reveals a critical precipitation threshold (750-950 mm/year) in the contiguous US where the primary driver of year-to-year carbon exchange shifts from plant production to ecosystem respiration. This finding, robust across multiple datasets, highlights a significant limitation in current vegetation models, which fail to capture this threshold and may underestimate respiration's role in carbon-climate feedbacks.
What Problem Does This Paper Address?
Accurately predicting how terrestrial ecosystems will respond to climate change is crucial for understanding the global carbon cycle and future climate trajectories. A key uncertainty lies in how ecosystem production (GPP) and respiration (TER) – the two main components of net ecosystem exchange (NEE) – are affected by climate variability, particularly water availability. While some studies suggest GPP is the dominant factor in dry regions, others point to respiration in tropical areas. Previous attempts to reconcile these hypotheses have relied on ecosystem-scale data or global vegetation models (DGVMs), which may not accurately represent continental-scale dynamics. This paper aims to empirically determine the primary control on inter-annual NEE variability across the contiguous United States and assess whether current DGVMs can reproduce these observed patterns.
Methods
The researchers employed a multi-pronged approach:
- Temporal Correlation Analysis: They calculated the correlation between inter-annual variability (IAV) of GPP and NEE using various observational datasets and DGVM simulations. This was done at a pixel level across the contiguous US.
- Sensitivity Analysis: They quantified the temporal and spatial sensitivities of GPP and TER to precipitation. This involved regressing carbon fluxes against climate factors over time (temporal sensitivity) and across climate gradients (spatial sensitivity).
- Ecosystem Respiration Modeling: To investigate model-data mismatches, they developed and tested several empirical respiration models with varying complexity (including soil carbon, temperature, moisture, and productivity) and compared their performance against observed respiration data and DGVM outputs.
- Model-Data Comparison: They directly compared the observed sensitivities and correlations with those simulated by an ensemble of ten DGVMs from the TRENDY project.
Datasets
The study utilized a range of datasets:
- Gridded Observation-Based Fluxes:
- GPP estimates derived from MODIS satellite observations (GPPMODIS).
- NEE estimates from an ensemble of atmospheric CO2 inversion models (NEEACI).
- NEE upscaled from eddy covariance flux data for North America (EC-MOD).
- Other GPP proxies: Solar-induced chlorophyll fluorescence (SIF), normalized difference vegetation index (NDVI), and fraction of photosynthetically active radiation (FPAR).
- Climate Data: Monthly gridded temperature and precipitation data from the Climate Research Unit (CRU TS v. 3.25).
- Dynamic Global Vegetation Models (DGVMs): Simulations from ten DGVMs from the TRENDY v2 ensemble.
- Eddy Covariance (EC) Observations: Data from 17 flux sites across the contiguous US (FLUXNET2015 database).
- Global Soil Respiration Database (SRDB v3): For validating respiration models.
- Wildfire Emission Data: From the Global Fire Emission Database (GFED4s).
- Human Influence Index (HII): To assess the impact of human activities.
Key Insights and Ideas
- The Precipitation Threshold is Key: The most striking finding is the identification of a distinct precipitation threshold (750-950 mm/year) where the dominant control on year-to-year carbon exchange shifts. Below this threshold, GPP variability drives NEE, while above it, TER variability becomes more influential. This suggests a fundamental change in ecosystem response to water availability.
- Models Miss the Mark: Current DGVMs consistently fail to reproduce this observed precipitation threshold. They tend to show GPP as the primary driver across all precipitation levels, particularly overestimating GPP's sensitivity in wetter regions.
- Respiration's Underappreciated Role: The study strongly suggests that ecosystem respiration, particularly heterotrophic respiration (Rh), plays a more significant role in regulating NEE variability, especially in mesic (wetter) ecosystems, than is currently captured by models.
- Model-Data Mismatch in Mesic Regions: The discrepancy between model simulations and observational data is most pronounced in mesic regions, pointing to a failure in DGVMs to accurately simulate respiration processes under wetter conditions.
- Heterotrophic Respiration is a Bottleneck: The research highlights heterotrophic respiration as a poorly understood process that is critical for accurate carbon cycle modeling, especially its response to environmental factors like soil moisture.
Key Results
- Spatial Patterns: Mean annual GPP is strongly correlated with precipitation, peaking in the warm and wet southeastern US. In contrast, the largest NEE (strongest carbon sink) is found in intermediate precipitation zones (750-1200 mm/year) and cooler, wetter regions like the North Central US.
- Temporal Correlations: Observational data show a significant positive correlation between GPP and NEE in xeric (drier) ecosystems, but this correlation weakens considerably in mesic ecosystems. TRENDY DGVM simulations, however, show a universally positive temporal correlation between GPP and NEE across all precipitation levels.
- Sensitivity Thresholds: Observational data reveal a precipitation threshold where the sensitivity of NEE to precipitation shifts from GPP control to TER control. This threshold is estimated between 750 mm/year (from eddy covariance data) and 950 mm/year (from gridded inversion data).
- Model Performance: Empirical respiration models that incorporate soil carbon, temperature, and precipitation (TERCTMP) performed better than DGVMs in simulating observed respiration, particularly in mesic regions. The TERCTMP model explained significantly more variation in measured Rh than DGVM simulations in these areas.
Limitations and Future Work
- Data Uncertainties: The study acknowledges potential data uncertainties in large-scale gridded observational fluxes, which might contribute to variations in estimated precipitation thresholds.
- Model Simplifications: While the empirical respiration models showed promise, they are simplifications of complex ecosystem processes.
- Human Land Use: The paper briefly touches on the influence of human land-use activities (e.g., agriculture and harvest) as a potential driver of spatial mismatches, suggesting that DGVMs need to better represent crop yields and harvest.
- Future Work: The authors emphasize the need for improved understanding and modeling of heterotrophic respiration, particularly its response to soil moisture and other environmental factors in mesic ecosystems. This is crucial for reducing uncertainty in future carbon-climate feedback predictions. They also suggest that better representation of human land-use impacts is needed in DGVMs.
Summary updated