| 25 Sep 2025
Representing dynamic grass density in the land surface model ORCHIDEE r9010
Abstract. In semi-arid regions, grasses and shrubs often form spatial heterogeneous patterns interspersed with bare soil, optimizing resource use and productivity. Accurately representing the matrix of vegetation and bare soil in global land surface models is essential for advancing the understanding of the carbon, water, and dust cycles. This study focuses on grasslands using the land surface model ORCHIDEE (ORganizing Carbon and Hydrology In Dynamic EcosystEms), which originally assumes a globally fixed maximum grass density. This assumption, referred to as the fixed maximum density approach, limits the model’s ability to capture grassland responses to environmental changes, resulting in unsustainable productivity and unrealistically frequent mortality events, particularly in resource-limited regions. To address these limitations, we introduced a dynamic density approach that simulates grassland density based on indicators of vegetation growth, such as reserve and labile carbon content in the grass. The emerging positive correlation between precipitation and simulated grass density supported the validity of the approach. Compared to the fixed maximum density approach, the new approach substantially reduced simulated mortality events, raised the aridity threshold for frequent mortality, and maintained realistic grassland productivity in regions where the presence of grassland is indicated by the observed leaf area index (LAI). This study not only demonstrates that simulating grass density as a function of carbon availability improves ORCHIDEE’s capacity to capture grassland dynamics under environmental variability, but also provides a promising foundation for investigating land–atmosphere feedbacks in (semi-)arid regions.
Overall assessment
This manuscript presents a physiology-based dynamic grass density approach for the ORCHIDEE land surface model that addresses key limitations of the fixed-density representation.
Here is this reviewer’s understanding: By linking vegetation density to reserve and labile carbon (C) pools, the model adjusts dynamically to resource availability. This mechanism reduces unrealistic mortality events, produces a more realistic emergent slope between precipitation and density, and generates the bare soil fraction directly rather than prescribing it. Together, these advances increase ecological realism, provide a basis for dust emission modelling and improve IPSL-CM performance across major grassland biomes. The study appears timely, well designed and methodologically sound, but several revisions that could strengthen its impact.
Validation and ecological realism
Validation relies mainly on indirect proxies such as LAI and precipitation correlations. Including regional case studies that use field-based estimates of grass density or bare soil cover would allow for more direct evaluation and strengthen confidence in the model’s realism.
The mortality–recruitment scheme, based on C pool trade-offs, is elegant in its simplicity but assumes asexual recruitment. Explicitly discussing the limitations of this assumption would help readers understand how the approach may underperform in ecosystems dominated by seed banks or sexual reproduction.
Parameter recalibration, particularly for C4 grasslands, improves outcomes, yet the robustness of these changes remains uncertain. Providing an additional sensitivity analysis, for example in supplementary material, to show how density responds to parameter variation would bolster confidence in the results.
External uncertainties and broader implications
The paper also acknowledges uncertainties in prescribed plant functional type (PTF) maps, including the unrealistic placement of grasses in hyper-arid zones. Quantifying the extent to which such mapping errors contribute to remaining mortality artefacts would help distinguish external sources of error from limitations internal to the model.
Although dust flux simulations are planned for future work, the manuscript would benefit from a conceptual schematic linking dynamic density, emergent bare soil fractions and dust emission potential. Such a figure would highlight the broader significance of the study.
Presentation and minor issues
Presentation could be improved through more consistent terminology, particularly in distinguishing “density” from “cover” and in clarifying what constitutes “an individual” in the model. Although the manuscript explains that “density” differs from “plant cover”, it sometimes uses “density” in a way that resembles “cover”, e.g. in the statement “… whereas grassland density reflects grass and bare soil fractions within the grassland PFT” (Line 285), which conflicts with the earlier definition of “the number of individuals per unit area” (line 60).
Likewise, the methods section states that “each individual is assumed to occupy 1 m2” (lines 104-105), yet discussions of biomass allocation and asexual reproduction obscure the line between a biological plant and an abstract unit, potentially confusing readers. For example, in “This approach for increasing grassland density reflects grass recruitment through asexual means, which is a suitable method for representing perennial plants” (lines 153-155), it should be clarified that the “individual” is a conceptual unit, not a physical plant.
The distinction between vegetation type fraction - “a value for its fraction (Vfra), line 91 - and “density” is also sometimes unclear, with “density” referring to surface coverage rather than actual counts of individuals., e.g. “... land cover map represents the fraction of vegetation type (Vfra) for each PFT within one grid cell, whereas grassland density represents grass and bare soil fractions within the grassland PFT” (lines 284–285).
A schematic showing C redistribution during density adjustments would help readers follow the mechanism, and adding explicit mortality thresholds to figure annotations (e.g. Fig. 7) would improve interpretability.
Minor grammatical polishing would further smooth the narrative. For example, awkward phrasing, such as “… the mortality in ORCHIDEE should be infrequent and primarily …” (line 443) would flow better as “… mortality in ORCHIDEE should occur infrequently and mainly …”, or “… grassland dies in the ORCHIDEE model and …” (lines 174-175) would be better if worded as “… the grassland is considered dead in ORCHIDEE, and …”.
Using the simple present tense to model descriptions would also enhance the writing, e.g. changing “Adding to these limitations, a fixed density fails to respond to changes in resource availability, hindering the possibility of studying the response of dust emissions …” (lines 71-72) to “In addition, a fixed density does not respond to resource availability, which hinders the study of dust emission responses …”.
Removing phrases such as "Note that" and "including" from “Note that the carbon of other compartments (including leaf, aboveground stem, root and fruit) in each individual remains …” (lines 124-125) would allow for the following: “The carbon in other compartments (leaf, stem, root, fruit) remains …”. Likewise, “Both of the events …” (line 191) could simply be shortened to “The events …”.
Finally, unit notation should follow SI conventions, with spaces before unit symbols and negative exponents for “per” relationships. For example, “gC m-2 per day” should be written as “g C m-2 d-1”, denoting grams of carbon per square meter per day. Likewise, the unit “m2 gC-1” is ambiguous and could be misread as “square meters times grams per carbon”. To remove this confusion, it should be rewritten as “m2 g-1 C”, which distinctly indicates square meters per gram of carbon.
Summary and recommendation
This study represents a significant methodological advance for ORCHIDEE and makes an important contribution to Earth system modelling. Strengthening validation, clarifying demographic simplifications and refining presentation would further enhance its impact. With these minor revisions, the manuscript will be a valuable and timely addition.