the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 Nov 2025
Seasonal Cycle Biases in DGVM Simulations of Double-Cropping Systems: A Case Study in the Huang-Huai-Hai Plain
Abstract. Global dynamic vegetation models (DGVMs) are essential tools for studying the changes in terrestrial ecosystems and their responses to climate change and human activities. However, these models exhibit substantial uncertainties when applied to croplands, particularly in regions with multiple cropping systems. These uncertainties arise from variations in planting types and phenology, which are influenced by sowing and harvesting schedules. This study focused on the phenological estimation errors of DGVMs in typical double - cropping agricultural regions. The Huang - Huai - Hai Plain in eastern China was chosen, which is one of the most important grain-producing areas with mainly winter wheat-summer crop rotation. A comparative analysis was conducted between the seven models from the TRENDY project and three remote sensing observations over last two decades. The results indicate that remote sensing vegetation indices consistently exhibit a typical bimodal structure in the study area, with peaks in April and August, corresponding to the growth peaks of the two-season crops. However, none of the DGVMs successfully capture this bimodal pattern. Given that multiple cropping systems are widespread in middle- and low-latitude regions with favorable water and temperature conditions, improving the simulation capabilities of DGVMs in such areas is an urgent and critical issue for advancing global vegetation modeling.
- Preprint
(4714 KB) - Metadata XML
-
Supplement
(1319 KB) - BibTeX
- EndNote
Status: open (until 23 Mar 2026)
- RC1: 'Comment on egusphere-2025-4997', Anonymous Referee #1, 09 Feb 2026 reply
-
RC2: 'Comment on egusphere-2025-4997', Anonymous Referee #2, 25 Feb 2026
reply
This paper compares LAI outputs from seven TRENDY DGVMs against three remote sensing products over the Huang-Huai-Hai Plain, a major double-cropping region in eastern China. The authors find that remote sensing consistently reveals a bimodal seasonal LAI cycle (peaks in April and August, corresponding to winter wheat and summer maize), whereas none of the DGVMs reproduce this pattern. They also report that DGVMs tend to overestimate LAI magnitude and underestimate the cropland contribution to regional greening. The paper calls for incorporating explicit double-cropping mechanisms into DGVMs.
I appreciate the authors' effort to draw attention to the limitations of current DGVMs in representing intensively managed croplands. This is an important and, I think, underappreciated problem, particularly as we increasingly rely on these models for carbon budget attribution and greening studies. The Huang-Huai-Hai Plain is a well-chosen study area, and the use of multiple remote sensing products alongside multiple models adds robustness. I do think the core message of this paper is worth communicating. But at the same time, I do have several concerns that prevent the paper from reaching its potential in its current form.
First of all, the introduction needs substantial restructuring. As it stands, the logic chain is fragmented and the narrative wanders through several loosely connected themes without building a coherent argument toward the study's objective. The introduction opens with a broad discussion of vegetation change, greening, and remote sensing (L23–33), which is fine as context-setting. But then it takes a detour into the specific quantitative findings of Zhu et al. (2016) — reporting that 25–50% of vegetated land shows increased LAI, that CO₂ accounts for 70% of greening, nitrogen deposition 9%, and so on (L36–41). This level of numerical detail belongs in the Discussion when interpreting your own results, not in the Introduction. The important point from that work — that DGVMs attribute most greening to CO₂ fertilisation — can be stated in one sentence. The transition at line 42 ("However, managed lands—particularly croplands—have been shown by observations to make a substantial contribution to global greening") is actually the key motivating observation for the study, but it is buried rather than highlighted. And then the transition at line 47 is jarring: the paper suddenly shifts from "croplands challenge DGVM greening attribution" to "it is essential to first validate that DGVMs can accurately capture the fundamental seasonal characteristics of vegetation." The logical connection — that if DGVMs get the seasonal cycle of croplands wrong, their estimates of cropland productivity and greening contribution are unreliable — is never stated. The reader has to infer it. Lines 47–52 then discuss phenology in quite general terms (SOS advancement, EOS delay, peak greenness enhancement, global warming and northern ecosystem productivity), which reads like textbook material that does not connect specifically to the problem of double-cropping. By the time the reader reaches the study objective at line 68, they have passed through global greening drivers, CO₂ fertilisation percentages, generic phenology concepts, and global multiple cropping statistics — none of which flow naturally into each other. I would suggest reorganising the introduction around a cleaner logic chain.
Moving to the core analysis, my main concern is that the paper stays largely at the descriptive level. The paper documents that models fail but does not explore why. Most TRENDY models were never designed to simulate double-cropping, so the mismatch is expected. This does not make the paper pointless as systematic documentation has value, but it needs a diagnostic dimension. I would suggest adding a table summarising each model's cropland representation (crop functional types, phenology scheme, management features, ability to simulate sequential cropping). This information is available from published model descriptions and would immediately contextualise the results. The three-type classification of seasonal patterns (L178–183) is interesting but should be linked to model design choices rather than left hanging. The Discussion similarly needs to move from restating results to interpreting them: which models already have infrastructure for double-cropping? What would need to change in the TRENDY protocol?
The land-use mask inconsistency is a real problem. The analysis uses RESDC (1 km) to define cropland, but TRENDY models are forced with HYDE3.3 (0.5°). Figure 8 shows these datasets disagree substantially. This means part of the model–observation mismatch could reflect differences in what land is classified as cropland, not just the absence of double-cropping mechanisms. A sensitivity test such as applying the HYDE mask to the remote sensing data and checking whether the bimodal pattern persists would resolve this. The HYDE data are already available to the authors from Figure 8, so this requires no new data.
Data processing methods are missing. The three remote sensing products have very different native resolutions and are all aggregated to 0.5°/monthly, but the aggregation procedure is not described. Monthly means vs. maxima, the order of masking vs. upscaling, and MODIS quality flag handling can all affect peak amplitude and timing — exactly what the paper analyses. A brief methods subsection would address this.
The manuscript is redundant in many sections. The same three conclusions appear at L215–222, in the Discussion, and in the Conclusions. The paragraph at L219–222 essentially repeats L215–218 with different wording. I would remove the end-of-Results summary entirely and add a Limitations subsection instead.
The greening trend analysis deserves more attention. This is arguably the most novel part of the paper. The finding that ISAM and LPJmL show browning where all observations show greening is striking, what might explain it? Even a tentative hypothesis based on published model descriptions would add value. The seasonal pattern of greening (strongest in spring, consistent with enhanced wheat productivity) could be connected more explicitly to agricultural drivers like cultivar change or fertilisation.
Specific Comments:
The section heading at L22 has a typo ("Introdution"). Missing spaces before parenthetical references occur throughout. The Figure 1 caption has a placeholder ("at xx resolution"). "LPX-Bern"/"LPXBern" is inconsistent. The second results subsection (L200) is mislabelled "3.2" instead of "3.3." "MOD15A12H" at L355 should be "MOD15A2H."
Figures 5 and 6 would benefit from a summary panel overlaying remote sensing curves with a model ensemble envelope.
Analysis code should be archived for reproducibility.
Citation: https://doi.org/10.5194/egusphere-2025-4997-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 300 | 78 | 29 | 407 | 60 | 20 | 16 |
- HTML: 300
- PDF: 78
- XML: 29
- Total: 407
- Supplement: 60
- BibTeX: 20
- EndNote: 16
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
View my comments in the attached file.