Gross primary production (GPP) by plants is fundamental to life on land, serving as the primary carbon source for terrestrial ecosystems. Remote sensing technologies, particularly satellite measurements of the fractional absorption of light (fAPAR) by vegetation, offer a powerful aid to monitoring and understanding GPP patterns and trends across the globe. But many current remote sensing-based models (RSMs) for GPP are not well-founded, either theoretically or empirically; they disagree with another; and they are tending to become more and more complex, without any discernible improvement in accuracy.

A new Perspective in Nature Reviews Earth and Environment addresses the challenges of making RSMs more robust and reliable, building on recent research by Colin Prentice’s group – and a number of collaborators – towards a new generation of RSMs:

Prentice et al., 2024. Principles for satellite monitoring of vegetation carbon uptake. Nature Reviews Earth and Environment.

This paper has had a long gestation. The story began nearly a decade ago, when the multi-talented Brad Evans (https://www.une.edu.au/staff-profiles/ers/bradley-evans) visited VITO in Flanders with Colin Prentice to discuss with Else Swinnen, Roel van Hoolst and others at VITO the idea of implementing the P model (unpublished at that time) in a remote sensing framework. (Brad had worked with Colin on the very earliest development of the P model in Australia.) VITO were enthusiastic – but there was no money! Serendipitously, however, an opportunity arose not long afterwards to apply to the European Space Agency (ESA) for a project to develop a near real-time monitoring system for GPP and biomass production. The proposal was funded, against stiff competition, and led to what became the TerrA-P project – in which the P model was tested at multiple eddy covariance flux sites, applied globally, and shown to rival existing RSMs in performance.

One of the eventual outcomes of the project was an internal report, led by Colin Prentice, on the application of new approaches to satellite-based monitoring of primary production in global carbon cycle research. It seemed that it would be a missed opportunity if this were not further developed into a paper. So this Perspective was designed and successfully submitted as a proposal for Nature Reviews Earth and Environment (a journal that has acquired the astonishingly high impact factor of 49.7 during its relatively short life).

In addition to the P model, the paper highlights the achievements of two other RSM development efforts: BESS (led by Youngryel Ryu at Seoul National University, a partner in LEMONTREE) and BEPS (led by Jing Chen at the University of Toronto). All three models have a common basis in the standard (FvCB) model of photosynthesis (not used in conventional RSMs) and the increasing deployment of eco-evolutionary optimality approaches, as pioneered in the P model, to constrain model parameters and reduce the uncertainty in modelled GPP and other quantities.

The paper focuses on several principles for the development of new-generation RSMs. It is argued that RSMs should: Make better use of remotely sensed data: Use remote sensing input data as comprehensively as possible, i.e. not only fAPAR. Minimize spatial discontinuities: Avoid artificial boundaries between vegetation types as far as possible, by adopting universal representations of processes. Address uncertainties: Propagate uncertainties in data and models, allowing the uncertainties of outputs to be assessed. Systematically assess performance: Evaluate models against in situ measurements: not only comparing numbers, but also comparing modelled and observed interdependencies among variables.

The paper’s main focus is on GPP, but it is emphasized that biomass production has a relatively conservative ratio to GPP – although there is more research to be done to better quantify this ratio, this general approach is far more promising than the conventional one, in which autotrophic respiration is (incorrectly) modelled as if it were independent of GPP. The paper also discusses the integration of new satellite products that provide (or will soon provide) complementary insights into the various processes involved in photosynthesis, including non-photochemical quenching and solar-induced fluorescence (SIF). This is a fast-moving field, and many groups are scrambling to find the best ways to capitalize on the wealth of data (hyperspectral reflectances and SIF) that are expected to become available after the launch of ESA’s FLEX mission in 2026.

Some take-away points:

  • Integrating the Farquhar-von Caemmerer-Berry (FvCB) model into RSMs has many advantages. One is the relatively conservative nature of parameters such as the CO2 compensation point and the Rubisco enzyme affinities for CO2 and O2­; to first order they can be treated as constant across C3 plants.
  • Conventional RSMs rely on Monteith’s empirical light use efficiency (LUE) principle, but this leaves them detached from decades of research on the environmental controls of photosynthesis. The P model exploits a mathematical property of saturation curves to derive LUE parameters for acclimated photosynthesis explicitly from the FvCB model. On the other hand, it overlooks the differential penetration of diffuse and direct light (better handled by BESS and BEPS) and lacks a diurnal cycle (a deficiency remedied by Giulia Mengoli’s work on the “subdaily” P model).
  • Sharp boundaries between different ecosystems caused by land use—such as forests, fields, and grasslands—are already visible in remote sensing fAPAR data, and so need no special treatment (except for the planting of C4 crops, which is a separate issue). But many RSMs introduce additional, artificial discontinuities when they shift parameter settings between different plant functional types (PFTs). PFT-independent parameterizations avoid such artefacts.
  • C3 and C4 plants respond differently to environmental factors like temperature and CO2 levels, but current global data on their distributions are coarse and inaccurate. Accurate global data on crop distributions and detailed information on C3 and C4 plant distributions are needed to improve models. Efforts like the European Space Agency’s World Cereal project promise higher-resolution data on crop types and extents.
  • Uncertainty in remote-sensing-based primary production models can stem from input data, model structure, and parameter values. Input data: Differences among fAPAR products and the quality of meteorological data contribute to uncertainty. Tools like the Sentinel 3 OLCI Global Vegetation Index and the DATimeS toolbox help quantify these uncertainties. Model structure: Different models may have similar biases due to shared assumptions or structures, making it challenging to quantify uncertainties accurately. Comparing models, ensuring they are well-documented and reproducible, can help identify and mitigate these issues. Parameter values: The impact of parameter values on model performance is less well understood. Methods for combining and propagating uncertainties in parameter estimates are needed.
  • Effective evaluation of remote-sensing models involves several considerations. Footprint matching: Aligning the spatial resolution of remote sensing data with flux tower measurements can be challenging. High-resolution data from sources like CubeSats and Sentinel-2 satellites can improve this alignment. Standardization: The PLUMBER project and its successor, PLUMBER2, provide standardized datasets and metrics for model evaluation, facilitating comparisons across different models. Functional Relationships: Models should be tested for their ability to reproduce ecological relationships that can be inferred from flux-tower data, as shown for example in the Li et al. paper introducing BESS v2.0.
  • Traditional methods estimate biomass production by subtracting plant respiration from GPP. They often fall short because the fraction of GPP allocated to nutrient acquisition varies widely; also many models assume that respiration increases exponentially with temperature, not accounting for longer-term acclimation of the base respiration rate. New approaches are exploring empirical models to estimate biomass production efficiency, which may vary with site fertility and management.
  • fAPAR products can include light absorbed by non-green tissues, and therefore may not accurately reflect photosynthesis. Combining process-based models with SIF and leaf chlorophyll content could provide more precise estimates. SIF measures a small fraction of light energy emitted by chlorophyll, and offers a fairly direct indicator of photosynthesis. Although SIF shows promise, its effectiveness depends on knowing the fraction of SIF escaping the canopy. The photochemical reflectance index (PRI) reflects leaf pigment changes, has potential for estimating LUE. Process-based models for PRI, like those implemented in the SCOPE model, show promise but need further development.
  • Remote sensing of atmospheric composition, including formaldehyde (an oxidation product of isoprene that is emitted by leaves), can indicate vegetation stress and reduced GPP.
  • Transpiration is linked to GPP through stomatal control and can in principle be modelled in parallel with GPP. This is being explored by the P model, as well as being implemented in BEPS and BESS.
  • Looking ahead, advances in technology and data availability, including hyperspectral satellites, high-resolution imagery and improved flux measurements, will enhance our understanding of carbon cycling and vegetation function. But some areas require improvements. Integrating new metrics: Use improved data and models to better represent ecosystem processes and responses to environmental changes. Addressing data limitations: Overcome challenges related to spectral bands, flux-tower matching, and data resolution through upcoming missions and technological advances. Standardizing data collection: Implement standardized protocols for flux measurements to improve the accuracy and comparability of in situ data.

You can read the full paper here:

Prentice, I.C., Balzarolo, M., Bloomfield, K.J., Chen, J.M., Dechant, B., Ghent, D., Janssens, I.A., Luo, X., Morfopoulos, C., Ryu Y., Vicca, S. & van Hoolst, R. 2024. Principles for satellite monitoring of vegetation carbon uptake. Nature Reviews Earth & Environment. https://doi.org/10.1038/s43017-024-00601-6.