The three major axes of terrestrial ecosystem function

Mirco Migliavacca, Talie Musavi, Miguel D. Mahecha, Jacob A. Nelson, Jürgen Knauer, Dennis Baldocchi, Óscar Pérez‐Priego, Rune Christiansen, Jonas Peters, Karen Anderson, Michael Bahn, T. Andrew Black, Peter D. Blanken, Damien Bonal, Nina Buchmann, Silvia Caldararu, Arnaud Carrara, Nuno Carvalhais, Alessandro Cescatti, Jiquan Chen, James Cleverly, Edoardo Cremonese, Ankur R. Desai, Tarek S. El‐Madany, Martha M. Farella, Marcos Fernández‐Martínez, Gianluca Filippa, Matthias Forkel, Marta Galvagno, Ulisse Gomarasca, Christopher M. Gough, Mathias Göckede, Andreas Ibrom, Hiroki Ikawa, Ivan A. Janssens, Martin Jung, Jens Kattge, Trevor F. Keenan, Alexander Knohl, Hideki Kobayashi, Guido Kraemer, B. E. Law, Michael J. Liddell, Xuanlong Ma, Ivan Mammarella, David Martini, Craig Macfarlane, Gioṙgio Matteucci, Leonardo Montagnani, Daniel E. Pabon‐Moreno, Cinzia Panigada, Dario Papale, Elise Pendall, Josep Peñuelas, Richard P. Phillips, Peter B. Reich, Micol Rossini, Eyal Rotenberg, Russell L. Scott, Clément Stahl, Ulrich Weber, Georg Wohlfahrt, Sebastian Wolf, Ian J. Wright, Dan Yakir, Sönke Zaehle, Markus ReichsteinNature 2021 · DOI

Abstract

The leaf economics spectrum and the global spectrum of plant forms and functions revealed fundamental axes of variation in plant traits, representing ecological strategies shaped by evolutionary development. Ecosystem functions depend on environmental conditions and the traits of species within ecological communities. However, the axes of variation of ecosystem functions have remained largely unknown, limiting our understanding of how ecosystems respond to anthropogenic drivers and climate variability. Here, we derive a set of ecosystem functions from a global dataset of surface gas exchange measurements across major terrestrial biomes. We find that 71.8% of the variability within ecosystem functions is captured by three key axes: maximum productivity, water-use strategy, and carbon-use efficiency. While state-of-the-art land surface models reproduce the primary productivity axis, they tend to simulate more strongly correlated functions than observed, limiting their predictive accuracy for environmental changes.

Introduction

Ecologists have long sought to identify the fundamental dimensions of biological variation. While frameworks like the leaf economics spectrum have successfully characterized plant-level traits, a comparable global framework for entire ecosystems has remained elusive. Ecosystems exchange carbon, water, and energy with the atmosphere, yet the underlying structure of these functional variations across biomes is not well understood. Identifying these axes is critical for predicting how ecosystems will respond to climate change and land-use transitions.

Methods and Data

The study derived 12 Ecosystem Functional Properties (EFPs)—including maximum photosynthetic capacity ($GPP_{sat}$), water-use efficiency (WUE), and carbon-use efficiency (aCUE)—from the global FLUXNET dataset, comprising 203 sites and 1,484 site-years of high-frequency gas exchange measurements. Satellite-derived GlobBiomass data and site-specific measurements of canopy height and leaf area index were used to characterize vegetation structure.

To analyze this data, the researchers employed:

  1. Principal Component Analysis (PCA): To reduce the dimensionality of the 12 EFPs into major axes of variation.
  2. Random Forest & Partial Dependence: To rank the importance of climate variables versus structural variables.
  3. Invariant Causal Prediction: To identify stable causal links between drivers and ecosystem functions across different environments.
  4. Model Benchmarking: Comparing empirical results against two leading Land Surface Models (LSMs), OCN and JSBACH.

Results: The Three Fundamental Axes

The analysis revealed that 71.8% of the total variability in ecosystem functions is captured by just three principal components (PCs). These axes represent distinct ecological strategies and functional capacities across the globe.

Principal component analysis of ecosystem functional properties (EFPs). (a) Biplot of PC1 and PC2 showing the distribution of plant functional types (PFTs). (b) Explained variance by the first eight PCs. (c) Relative contribution of EFPs to the first three PCs. (d) Loadings of EFPs on the first three PCs.
  • Axis 1: Productivity (39.3% of variance): This axis represents the "capacity" of the system, distinguishing between low-productivity sparse shrublands and high-productivity forests. It is dominated by maximum productivity metrics like $GPP_{sat}$ and $NEP_{max}$.
  • Axis 2: Water-Use Strategy (21.4% of variance): This axis reflects how efficiently an ecosystem uses water relative to carbon gain, characterized by WUE and evaporative fraction.
  • Axis 3: Carbon-Use Efficiency (11.1% of variance): This axis represents the fraction of assimilated carbon stored versus respired, primarily driven by aCUE and respiration ($R_b$).

Discussion: Structural Drivers and Model Performance

A central finding is that vegetation structure often trumps climate as a causal driver. For the primary Productivity axis, structural variables like maximum leaf area index ($LAI_{max}$) and above-ground biomass (AGB) were stronger causal drivers than temperature or radiation. Vegetation height ($H_c$) emerged as a critical "master switch" for water-use strategies, modulating aerodynamic resistance and water loss.

When benchmarking Land Surface Models, the researchers found that while models like OCN and JSBACH successfully reproduced the first axis (Productivity), they struggled with the second and third. The models exhibited "stiffness," simulating ecosystem functions that are more tightly correlated than they are in nature. This lack of flexibility limits their ability to accurately predict the diverse range of responses to environmental changes.

Conclusion

This research establishes a global framework for ecosystem function, highlighting the dominant role of vegetation structure. The identified axes provide a roadmap for understanding how terrestrial ecosystems process carbon, water, and energy. Improving climate forecasts will require moving beyond fixed plant functional types toward more flexible, trait-based representations of vegetation in land surface models to better capture the functional diversity of the real world.

Summary updated

Abstract

The leaf economics spectrum<sup>1,2</sup> and the global spectrum of plant forms and functions<sup>3</sup> revealed fundamental axes of variation in plant traits, which represent different ecological strategies that are shaped by the evolutionary development of plant species<sup>2</sup>. Ecosystem functions depend on environmental conditions and the traits of species that comprise the ecological communities<sup>4</sup>. However, the axes of variation of ecosystem functions are largely unknown, which limits our understanding of how ecosystems respond as a whole to anthropogenic drivers, climate and environmental variability<sup>4,5</sup>. Here we derive a set of ecosystem functions<sup>6</sup> from a dataset of surface gas exchange measurements across major terrestrial biomes. We find that most of the variability within ecosystem functions (71.8%) is captured by three key axes. The first axis reflects maximum ecosystem productivity and is mostly explained by vegetation structure. The second axis reflects ecosystem water-use strategies and is jointly explained by variation in vegetation height and climate. The third axis, which represents ecosystem carbon-use efficiency, features a gradient related to aridity, and is explained primarily by variation in vegetation structure. We show that two state-of-the-art land surface models reproduce the first and most important axis of ecosystem functions. However, the models tend to simulate more strongly correlated functions than those observed, which limits their ability to accurately predict the full range of responses to environmental changes in carbon, water and energy cycling in terrestrial ecosystems<sup>7,8</sup>.
BibTeX
@inproceedings{migliavacca2021three,
  author = {Mirco Migliavacca and Talie Musavi and Miguel D. Mahecha and Jacob A. Nelson and Jürgen Knauer and Dennis Baldocchi and Óscar Pérez‐Priego and Rune Christiansen and Jonas Peters and Karen Anderson and Michael Bahn and T. Andrew Black and Peter D. Blanken and Damien Bonal and Nina Buchmann and Silvia Caldararu and Arnaud Carrara and Nuno Carvalhais and Alessandro Cescatti and Jiquan Chen and James Cleverly and Edoardo Cremonese and Ankur R. Desai and Tarek S. El‐Madany and Martha M. Farella and Marcos Fernández‐Martínez and Gianluca Filippa and Matthias Forkel and Marta Galvagno and Ulisse Gomarasca and Christopher M. Gough and Mathias Göckede and Andreas Ibrom and Hiroki Ikawa and Ivan A. Janssens and Martin Jung and Jens Kattge and Trevor F. Keenan and Alexander Knohl and Hideki Kobayashi and Guido Kraemer and B. E. Law and Michael J. Liddell and Xuanlong Ma and Ivan Mammarella and David Martini and Craig Macfarlane and Gioṙgio Matteucci and Leonardo Montagnani and Daniel E. Pabon‐Moreno and Cinzia Panigada and Dario Papale and Elise Pendall and Josep Peñuelas and Richard P. Phillips and Peter B. Reich and Micol Rossini and Eyal Rotenberg and Russell L. Scott and Clément Stahl and Ulrich Weber and Georg Wohlfahrt and Sebastian Wolf and Ian J. Wright and Dan Yakir and Sönke Zaehle and Markus Reichstein},
  title = {{The three major axes of terrestrial ecosystem function}},
  booktitle = {Nature},
  year = {2021},
  doi = {10.1038/s41586-021-03939-9},
}
Figure