Plant phenology evaluation of CRESCENDO land surface models. Part II: Trough, peak, and amplitude of growing season

Peano, Daniele; Hemming, Deborah; Delire, Christine; Fan, Yuanchao; Lee, Hanna; Materia, Stefano; Nabel, Julia E. M .S.; Park, Taejin; Wårlind, David; Wiltshire, Andy; Zaehle, Sönke

doi:10.5194/egusphere-2024-4114

Preprints

https://doi.org/10.5194/egusphere-2024-4114

Preprints

30 Jan 2025

| 30 Jan 2025

Plant phenology evaluation of CRESCENDO land surface models. Part II: Trough, peak, and amplitude of growing season

Daniele Peano, Deborah Hemming, Christine Delire, Yuanchao Fan, Hanna Lee, Stefano Materia, Julia E. M .S. Nabel, Taejin Park, David Wårlind, Andy Wiltshire, and Sönke Zaehle

Abstract. Leaf area index is an important metric for characterising the structure of vegetation canopies and scaling up leaf and plant processes to assess their influence on regional and global climate. Earth observation estimates of leaf area index have increased in recent decades, providing a valuable resource for monitoring vegetation changes and evaluating their representation in land surface and earth system models. The study presented here uses satellite leaf area index products to quantify regional to global variations in the seasonal timing and value of the leaf area index trough, peak, and amplitude, and evaluate how well these variations are simulated by seven land surface models, which are the land components of state-of-the-art earth system models. Results show that the models simulate widespread delays, of up to three months, in the timing of leaf area index troughs and peaks compared to satellite products. These delays are most prominent across the Northern Hemisphere and support the findings of previous studies that have shown similar delays in the timing of spring leaf out simulated by some of these land surface models. The modelled seasonal amplitude differs by less than 1 m²/m² compared to the satellite-derived amplitude across more than half of the vegetated land area. This study highlights the relevance of vegetation phenology as an indicator of climate, hydrology, soil, and plant interactions, and the need for further improvements in the modelling of phenology in land surface models in order to capture the correct seasonal cycles, and potentially also the long-term trends, of carbon, water and energy within global earth system models.

Received: 23 Dec 2024 – Discussion started: 30 Jan 2025

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 10609 KB)

Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
Preprint (10609 KB)

Supplement (10325 KB)

Download & links

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Journal article(s) based on this preprint

24 Nov 2025

Plant phenology evaluation of CRESCENDO land surface models using satellite-derived Leaf Area Index – Part 2: Seasonal trough, peak, and amplitude

Daniele Peano, Deborah Hemming, Christine Delire, Yuanchao Fan, Hanna Lee, Stefano Materia, Julia E. M. S. Nabel, Taejin Park, David Wårlind, Andy Wiltshire, and Sönke Zaehle

Biogeosciences, 22, 7117–7135, https://doi.org/10.5194/bg-22-7117-2025,https://doi.org/10.5194/bg-22-7117-2025, 2025

Short summary

Daniele Peano, Deborah Hemming, Christine Delire, Yuanchao Fan, Hanna Lee, Stefano Materia, Julia E. M .S. Nabel, Taejin Park, David Wårlind, Andy Wiltshire, and Sönke Zaehle

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2024-4114', Anonymous Referee #1, 02 Mar 2025
The manuscript Peano et al. compared results from seven land surface models (LSMs) to remote sensing estimated leaf area index (LAI) at global scale regarding the seasonal timing of LAI trough (lowest value) and peak (highest value) and amplitude (max. – min.). The results indicate that all studied LSMs tended to show delayed timing of LAI trough and peak, in particular for the Northern Hemisphere while the modeled amplitude is smaller than satellite estimates. While overall the work is pretty straightforward, I do have a few major comments.

Model description of phenology scheme (L85-125) and related discussions do not provide enough details for readers who are not familiar with these models or with the different phenology schemes or both to understand what might be contributing to the current variations in model simulated LAI and its seasonal dynamics. While most models tend to show delayed timing of LAI trough and peak, there’s one model, LPG-GUESS did a better job capturing this (L170, L180). Why so? Similarly, when model differences, model-data discrepancy across different biomes/regions all present, it would be interesting to discuss the potential causes why we see all these differences. While the discussion tried to talk about the different sources of variability (L280) and potential observation biases (L295), it’s quite thin and not well organized to address the observed spatiotemporal data-model discrepancy. I think expanding discussion and organize it in a similar way as how the results are organized, and maybe also add a little more detail on different phenology schemes in each LSM can be very helpful. It would also be nice to see some suggestions for future model development or priority areas for data collections for better model parameterization or better observations for model benchmarking.

How much of the data-model discrepancy in LAI seasonality can be contributed to model structure and parameterization respectively? I know this is not the main focus of this study, but I think it’s very important to know this before we conclude that we need better phenology models (specifically refer to model structure improvement). Related, it is not clear to me how simulated distribution of the different PFTs (evergreen vs deciduous: they vary in phenology in addition to many other features that can influence plant growth and mortality thus phenology) can influence the modeled LAI seasonality both spatially and temporally. Is it possible that the delayed timing of LAI trough and LAI peak is a mismatch between the observed and the simulated vegetation type that dominate a particular grid cell?

Specific comments

L20: add relevant citations.
L40: maybe can expand to add a little more detail here.
L140: how the different domain resolution might influence model results? Also how coarse is the model resolution?
L168: multi-model ensemble mean
Citation: https://doi.org/10.5194/egusphere-2024-4114-RC1
- AC1: 'Reply on RC1', Daniele Peano, 14 Apr 2025
  
  Please find the authors' reply in the attachment.
  
  Citation: https://doi.org/10.5194/egusphere-2024-4114-AC1
RC2:
'Comment on egusphere-2024-4114', Anonymous Referee #2, 05 Mar 2025

Review of
Plant phenology evaluation of CRESCENDO land surface models. Part II: Trough, peak, and amplitude of growing season.
by Peano et al.
General comments:
This is an interesting paper comparing global leaf area index (LAI) simulations with satellite-derived LAI observations. Seven land surface models are considered along with 3 satellite LAI data sets. All products and observations are first projected onto the same 0.5 degree x 0.5 degree grid and monthly averages are considered. Maps of simulated and observed LAI peak, trough, and annual amplitude are presented. Mean monthly LAI time series show that the models tend to simulate LAI peaks later than the observations. The paper is reasonably well written, but the discussion section could be improved.
Recommendation: major revisions.
Particular comment:
- Phenology simulations may be affected by errors in the atmospheric forcing. What is the quality of the CRUNCEP atmospheric forcing database? Has it been evaluated in previous studies?
- I am concerned with the definition of "LAI trough". Unlike the LAI peak, low LAI values (e.g. in winter or during a drought) can persist for several months. Commonly used phenology indicators, in addition to peak time, are leaf onset and leaf offset. Why not use these more common indicators? Is LAI trough equivalent to leaf offset?
- The observed model LAI peak lags at mid and high latitudes may indicate a problem in the representation of temperature. Temperature is a key driver of leaf emergence in these regions. The temperature relevant to phenology is likely to be close to the land surface temperature resulting from the energy budget calculations. This is particularly true for the ISBA model, where phenology is driven by photosynthesis and leaf temperature. Is daytime leaf temperature underestimated in this model? In a number of ISBA papers (e.g. https://doi.org/10.5194/hess-21-4861-2017) the LAI peak time is generally consistent with observations. What has changed in the ISBA settings? Is the surface temperature calculated in the same way as before?
- For all models, it should be stated how the temperature used in the phenology model is calculated and how reliable it is.

Citation: https://doi.org/10.5194/egusphere-2024-4114-RC2
- AC2: 'Reply on RC2', Daniele Peano, 14 Apr 2025
  
  Please find the authors' reply in the attachment.
  
  Citation: https://doi.org/10.5194/egusphere-2024-4114-AC2

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2024-4114', Anonymous Referee #1, 02 Mar 2025
The manuscript Peano et al. compared results from seven land surface models (LSMs) to remote sensing estimated leaf area index (LAI) at global scale regarding the seasonal timing of LAI trough (lowest value) and peak (highest value) and amplitude (max. – min.). The results indicate that all studied LSMs tended to show delayed timing of LAI trough and peak, in particular for the Northern Hemisphere while the modeled amplitude is smaller than satellite estimates. While overall the work is pretty straightforward, I do have a few major comments.

Model description of phenology scheme (L85-125) and related discussions do not provide enough details for readers who are not familiar with these models or with the different phenology schemes or both to understand what might be contributing to the current variations in model simulated LAI and its seasonal dynamics. While most models tend to show delayed timing of LAI trough and peak, there’s one model, LPG-GUESS did a better job capturing this (L170, L180). Why so? Similarly, when model differences, model-data discrepancy across different biomes/regions all present, it would be interesting to discuss the potential causes why we see all these differences. While the discussion tried to talk about the different sources of variability (L280) and potential observation biases (L295), it’s quite thin and not well organized to address the observed spatiotemporal data-model discrepancy. I think expanding discussion and organize it in a similar way as how the results are organized, and maybe also add a little more detail on different phenology schemes in each LSM can be very helpful. It would also be nice to see some suggestions for future model development or priority areas for data collections for better model parameterization or better observations for model benchmarking.

How much of the data-model discrepancy in LAI seasonality can be contributed to model structure and parameterization respectively? I know this is not the main focus of this study, but I think it’s very important to know this before we conclude that we need better phenology models (specifically refer to model structure improvement). Related, it is not clear to me how simulated distribution of the different PFTs (evergreen vs deciduous: they vary in phenology in addition to many other features that can influence plant growth and mortality thus phenology) can influence the modeled LAI seasonality both spatially and temporally. Is it possible that the delayed timing of LAI trough and LAI peak is a mismatch between the observed and the simulated vegetation type that dominate a particular grid cell?

Specific comments

L20: add relevant citations.
L40: maybe can expand to add a little more detail here.
L140: how the different domain resolution might influence model results? Also how coarse is the model resolution?
L168: multi-model ensemble mean
Citation: https://doi.org/10.5194/egusphere-2024-4114-RC1
- AC1: 'Reply on RC1', Daniele Peano, 14 Apr 2025
  
  Please find the authors' reply in the attachment.
  
  Citation: https://doi.org/10.5194/egusphere-2024-4114-AC1
RC2:
'Comment on egusphere-2024-4114', Anonymous Referee #2, 05 Mar 2025

Review of
Plant phenology evaluation of CRESCENDO land surface models. Part II: Trough, peak, and amplitude of growing season.
by Peano et al.
General comments:
This is an interesting paper comparing global leaf area index (LAI) simulations with satellite-derived LAI observations. Seven land surface models are considered along with 3 satellite LAI data sets. All products and observations are first projected onto the same 0.5 degree x 0.5 degree grid and monthly averages are considered. Maps of simulated and observed LAI peak, trough, and annual amplitude are presented. Mean monthly LAI time series show that the models tend to simulate LAI peaks later than the observations. The paper is reasonably well written, but the discussion section could be improved.
Recommendation: major revisions.
Particular comment:
- Phenology simulations may be affected by errors in the atmospheric forcing. What is the quality of the CRUNCEP atmospheric forcing database? Has it been evaluated in previous studies?
- I am concerned with the definition of "LAI trough". Unlike the LAI peak, low LAI values (e.g. in winter or during a drought) can persist for several months. Commonly used phenology indicators, in addition to peak time, are leaf onset and leaf offset. Why not use these more common indicators? Is LAI trough equivalent to leaf offset?
- The observed model LAI peak lags at mid and high latitudes may indicate a problem in the representation of temperature. Temperature is a key driver of leaf emergence in these regions. The temperature relevant to phenology is likely to be close to the land surface temperature resulting from the energy budget calculations. This is particularly true for the ISBA model, where phenology is driven by photosynthesis and leaf temperature. Is daytime leaf temperature underestimated in this model? In a number of ISBA papers (e.g. https://doi.org/10.5194/hess-21-4861-2017) the LAI peak time is generally consistent with observations. What has changed in the ISBA settings? Is the surface temperature calculated in the same way as before?
- For all models, it should be stated how the temperature used in the phenology model is calculated and how reliable it is.

Citation: https://doi.org/10.5194/egusphere-2024-4114-RC2
- AC2: 'Reply on RC2', Daniele Peano, 14 Apr 2025
  
  Please find the authors' reply in the attachment.
  
  Citation: https://doi.org/10.5194/egusphere-2024-4114-AC2

Peer review completion

AR – Author's response | RR – Referee report | ED – Editor decision | EF – Editorial file upload

ED: Reconsider after major revisions (05 Jun 2025) by Mirco Migliavacca

AR by Daniele Peano on behalf of the Authors (05 Aug 2025) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (11 Aug 2025) by Mirco Migliavacca

RR by Anonymous Referee #3 (22 Aug 2025)

Suggestions for revision or reasons for rejection

Comment on "Plant phenology evaluation of CRESCENDO land surface models. Part II: Trough, peak, and amplitude of growing season" by Daniele Peano et al., Biogeosciences

This article described using global remote sensing based LAI products, including GIMMS-LAI3g, Copernicus Global Land Service LAI version 2 and MODIS MOD15A2H collection 6, to evaluate the simulated peak, trough and duration of LAI (or maybe growing season?) from the historical transient simulation of 7 different LSMs, following TRENDY S3 protocol, during 2000 ~ 2011. The outcome of this work is solid and has important values for improving parameterizations related to phenology and leaf dynamics for terrestrial ecosystems and land surface models (LSMs). But I have several major concerns on the way authors present their work (comment #1 and #2) and the lack of discussion about advancements/limitations of phenology parameterizations in different LSMs (comment #3). Thus, I cannot recommend the paper for publishing before authors can make a major revision to address my following comments.

Major:
1. One of my major concerns is that the purpose of this work is not clear.

It seems like authors want to compare modeled phenology against observation through LAI as an indicator, as suggested by the title "Part II: Trough, peak, and amplitude of growing season". But another paper from authors (Peano et al., 2021) already studied the start and end of the growing season and this "Part II" paper has relatively few interesting conclusions. Moreover, the use of NDVI instead of LAI as indicator for trough, peak, amplitude of growing season has at least 2 advantages: 1) RS-based LAI products are derived fully/partially from NDVI through complicated algorithms, which has more uncertainties and 2) metrics describing growing season, e.g., the onset/offset dates are (or can be) model outputs and can be directly compared to the metrics diagnosed from NDVI annual trace.

I was also thinking another possible purpose of this manuscript is trying to compare the value of modelled LAI amplitude vs. observation directly, which not only contains phenology but also productivity information, but I found no discussion on productivity side.

The third possible purpose is that LAI is directly linked to model parameterizations, thus authors want to utilize LAI as an indicator for diagnosing issues of LSM parameterization. If this is (one of) the focus(es) of this draft, I shall recommend authors to explore how these direct comparisons against LAI can be used to diagnose issues in model parameterizations. We see that all 7 different models show at least 3 different types of phenology and LAI parameterizations from Table 1. You can discuss how your findings can help us understand the spatial and temporal advantage/disadvantage of different phenology parameterization?

In summary, I suggest authors clarify my doubt on the purpose of this paper and provide persuasive reasons on why LAI is a good choice once the purpose is clarified.

2. The analysis is performed at a spatial resolution of half by half degree, so I assume most of the grids are covered by mixed plant functional types. It makes sense to use LAI peak, trough and amplitude for the analysis over mixed forest dominated by deciduous or seasonal forest/grassland/cropland since they have relative clear seasonality both from model parameterization and vegetation spectral signal, but may exaggerate the bias of model parameterization when compared to grids dominated by evergreen forest at high latitudes, where both some of the model parameterizations do ignore the LAI variability, and the vegetation spectral signal is contaminated by snow cover. In fact this is one example of my major concern #1, that using LAI trough, peak and amplitude might not be appropriate for describing growing season and phenology. Good examples to examine land surface phenology on a global scale are the work of Buitenwerf et al., 2015 and the 4GTS method from authors' Part I paper, which combined quantitative metrics with mode detection. I would strongly recommend authors to either justify findings over these grids, or at least explain this point as limitation in their work.

3. Direct comparison of models against observation can provide large amounts but only fragmented pieces of information, which is hard for deriving robust conclusions (e.g., Line 192 - 196). Since authors know the different types of parameterization in models, I'm thinking if authors can link model phenology parameterizations with your model-data comparison. One suggestion from me is that authors can group models with similar parameterization and check if models with different parameterizations can have statistically significant difference in three metrics you used through Student's t-test? You can also think about per-biome or per-zonal analysis if necessary? This can improve the value of the work and suggest the direction for improving the current phenology scheme in these LSMs?

Minor:
Did you perform new simulations or analyzed the existing TRENDY model outputs? If you used TRENDY model products, which version did you use? Please clarify.

Authors need to clarify whether crop phenology is activated or not in the model simulations, since the TRENDY S3 simulation includes the dynamic land use change and cropland. If so, please add crop phenology schemes used by different models in Table 1.

Line 59: "the seasonal timing of trough and peak, and amplitude (trough to peak) of LAI." The authors used "growing season" as title but here used "LAI" instead. It is not a good idea to use both in an exchangeable way since growing season and LAI are different. Suggest to revise and keep consistency.

Line 88: "Seven LSM" shall be "Seven LSMs"

Line 87: since you only have one paragraph describing each LSM, you can remove the subsection titles and merge them into one paragraph.

Line 141 and Line 305: "the same spatial land coverage evolution forces them leaving only differences in plant growth and seasonality among them" is not precise. Since TRENDY models have different PFT definitions, the use of the same land use history as forcing only guarantees the consistent total forest/crop/pasture/urban areas across different models. So the land coverage evolutions for each PFT is not the same across LSMs, and only the relative fraction of total forest land on each grid cell is the same in different models.

Line 218: How did you calculate the agreement of LAI seasonal amplitude? Since amplitude is a numerical, not a categorical variable? I see that you defined +-0.25 as the threshold (correct me if I'm wrong) to define "agreement" of LAI amplitude. This shall be explained in methodology.

Figure 1b: The pattern of LAI trough from CGLS in boreal regions near the arctic circle, such as central Siberia and part of northern Europe, is very similar to the error maps (1d, 1f). This makes me doubt the credibility of CGLS products over these regions, and I wonder what reason causes this similarity? Authors listed several caveats for using RS products in section 4.4, so I feel this can be a good example and suggest authors to check the reason behind and add into this section.

The biome specific results (section 3.3) authors presented are derived from which biome mask? Authors shall display the difference between PFT coverage from models and the biome mask used for analysis as supplemental material to consolidate their conclusions.

Hide

RR by Anonymous Referee #2 (03 Sep 2025)

ED: Reconsider after major revisions (12 Sep 2025) by Mirco Migliavacca

AR by Daniele Peano on behalf of the Authors (21 Oct 2025) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (31 Oct 2025) by Mirco Migliavacca

RR by Anonymous Referee #3 (31 Oct 2025)

ED: Publish subject to technical corrections (04 Nov 2025) by Mirco Migliavacca

AR by Daniele Peano on behalf of the Authors (11 Nov 2025) Author's response Manuscript

Journal article(s) based on this preprint

24 Nov 2025

Plant phenology evaluation of CRESCENDO land surface models using satellite-derived Leaf Area Index – Part 2: Seasonal trough, peak, and amplitude

Daniele Peano, Deborah Hemming, Christine Delire, Yuanchao Fan, Hanna Lee, Stefano Materia, Julia E. M. S. Nabel, Taejin Park, David Wårlind, Andy Wiltshire, and Sönke Zaehle

Biogeosciences, 22, 7117–7135, https://doi.org/10.5194/bg-22-7117-2025,https://doi.org/10.5194/bg-22-7117-2025, 2025

Short summary

Daniele Peano, Deborah Hemming, Christine Delire, Yuanchao Fan, Hanna Lee, Stefano Materia, Julia E. M .S. Nabel, Taejin Park, David Wårlind, Andy Wiltshire, and Sönke Zaehle

Supplement

https://doi.org/10.5194/egusphere-2024-4114-supplement

Daniele Peano, Deborah Hemming, Christine Delire, Yuanchao Fan, Hanna Lee, Stefano Materia, Julia E. M .S. Nabel, Taejin Park, David Wårlind, Andy Wiltshire, and Sönke Zaehle

Viewed

Total article views: 950 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
817	106	27	950	66	26	38

HTML: 817
PDF: 106
XML: 27
Total: 950
Supplement: 66
BibTeX: 26
EndNote: 38

Views and downloads (calculated since 30 Jan 2025)

Month	HTML	PDF	XML	Total
Jan 2025	43	7	4	54
Feb 2025	36	17	1	54
Mar 2025	44	7	3	54
Apr 2025	36	15	5	56
May 2025	17	5	1	23
Jun 2025	35	4	4	43
Jul 2025	26	11	0	37
Aug 2025	113	6	3	122
Sep 2025	413	13	1	427
Oct 2025	32	11	2	45
Nov 2025	22	10	3	35

Cumulative views and downloads (calculated since 30 Jan 2025)

Month	HTML	PDF	XML	Total
Jan 2025	43	7	4	54
Feb 2025	36	17	1	54
Mar 2025	44	7	3	54
Apr 2025	36	15	5	56
May 2025	17	5	1	23
Jun 2025	35	4	4	43
Jul 2025	26	11	0	37
Aug 2025	113	6	3	122
Sep 2025	413	13	1	427
Oct 2025	32	11	2	45
Nov 2025	22	10	3	35

Viewed (geographical distribution)

Total article views: 949 (including HTML, PDF, and XML) Thereof 949 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 24 Nov 2025

Download

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Preprint (10609 KB)
Metadata XML

Short summary

Earth System Models are the principal tools for scientists to study past, present, and future climate changes. This work investigates the ability of a set of them to represent the observed changes in vegetation, which are vital to estimating the impact of future climate mitigation and adaptation strategies. This study highlights the main limitations in correctly representing vegetation variability. These tools still need further development to improve our understanding of future changes.


Total:	0
HTML:	0
PDF:	0
XML:	0