TL;DR
This paper addresses the critical need for accurate plant functional type (PFT) maps in Earth system models (ESMs) by developing four new global PFT datasets derived from multiple satellite sensors. These datasets, when used to drive a DGVM, reveal that land-cover uncertainty, particularly in arid regions, can significantly impact estimates of carbon and water fluxes, highlighting the importance of robust PFT mapping for climate change research.
What Problem Does This Paper Address?
Earth system models (ESMs) are crucial tools for understanding how our planet's climate and ecosystems interact, especially in the face of global change. A key component of these models is representing vegetation, which significantly influences carbon and water cycles. However, directly translating the vast diversity of plant species into the simplified categories that ESMs can handle – known as Plant Functional Types (PFTs) – is challenging. Current PFT datasets often rely on single satellite sensors and can be outdated, leading to uncertainties in model simulations. This paper tackles the problem of creating more consistent and up-to-date PFT datasets that are directly usable by ESMs, and quantifies how uncertainties in these datasets propagate to affect climate-sensitive biogeochemical fluxes.
Methods
The authors developed a novel methodology to create four global PFT datasets. This involved:
- Cross-walking Land Cover Legends: They took existing global land-cover datasets from different satellite sensors and reclassified their categories into a consistent set of PFTs based on phenology and physiognomy. This process is inherently subjective, especially for mixed vegetation types.
- Merging with Climate Data: The reclassified land-cover data was then merged with climate zones derived from the Köppen-Geiger classification system to define PFTs.
- Spatial Aggregation: The merged data was aggregated to a 0.5° spatial resolution, a common resolution for ESM input data.
- Uncertainty Analysis: A beta diversity metric was used to quantify the spatial dissimilarity and uncertainty among the four developed PFT datasets.
- DGVM Simulation: The LPJmL dynamic global vegetation model (DGVM) was run in a diagnostic mode, using the four PFT datasets as prescribed inputs, to quantify the impact of land-cover uncertainty on the sensitivity of Gross Primary Productivity (GPP) and transpiration fluxes to climate variables.
Datasets
The study utilized the following land-cover datasets:
- GLC2000: Derived from SPOT-VEGETATION data.
- GlobCover: Derived from ENVISAT-MERIS data.
- MODIS Land Cover (MOD12Q1): Two versions, V004 and V005, derived from MODIS data.
These were merged with the Köppen-Geiger climate classification dataset.
Key Insights and Ideas
- Land Cover Uncertainty Matters: The study strongly demonstrates that uncertainties in land-cover datasets, and consequently in PFT maps, have a significant impact on the simulated sensitivity of key ecosystem processes (GPP and transpiration) to climate. This is a crucial finding for model development and validation.
- Arid Regions are Particularly Sensitive: The research highlights that arid and semi-arid regions are hotspots of uncertainty, with land-cover variations contributing up to 30% uncertainty in GPP sensitivity to precipitation. This is likely due to the complex mix of grasslands, shrublands, and sparse vegetation in these areas and differing definitions of forest cover thresholds.
- PFT Definitions are Key: Disagreements often arise from differences in how land-cover datasets define forest cover thresholds and how they classify mixed vegetation types, natural vs. managed grasslands, and shrublands.
- Beta Diversity as a Metric: The use of beta diversity is a novel and effective way to quantify spatial disagreement and uncertainty across different PFT datasets, providing a clear picture of where and why discrepancies occur.
Key Results
- The study developed four global PFT datasets based on different satellite sensors and classification systems.
- The greatest uncertainty in PFT classifications was found between cropland and grassland categories, and in dryland systems between shrubland, grassland, and forest categories.
- In arid regions, land-cover uncertainty contributed up to 30% uncertainty in the sensitivity of GPP to precipitation and 20% uncertainty in transpiration sensitivity.
- The study found that while global GPP and transpiration estimates were similar across datasets, regional sensitivities to climate varied significantly due to land-cover differences.
Limitations and Future Work
- Subjectivity in Reclassification: The authors acknowledge that manually reclassifying land-cover legends into PFTs is inherently subjective, especially for mixed vegetation categories. While they attempted to mitigate bias by comparing agreement levels, this remains a potential source of error.
- Need for Finer Resolution and Consistency: The paper suggests that future PFT datasets should aim for higher spatial resolution and greater consistency with current satellite products and ESM requirements.
- Improving Dryland and Tropical Classifications: Specific areas like drylands and tropical forests require further refinement in classification to reduce uncertainty.
- Adaptive Traits: The authors point to future ESM developments that will likely incorporate higher diversity of PFTs and entirely new approaches focusing on adaptive, rather than fixed, traits, which will require new methods for detecting life history traits not easily observable from space.
- Crop Functional Types: The paper notes that distinguishing crop types and managed grasslands remains a challenge and is an area for future integration into ESMs.
Summary updated