Gross primary productivity responses to meteorological drivers: insights from observations and multi-model ensembles

Zheng, Yuxin; Yue, Xu; Lu, Xiaofei; Zhu, Jun

doi:https://doi.org/10.5194/egusphere-2025-1515

Preprints

https://doi.org/10.5194/egusphere-2025-1515

Preprints

24 Apr 2025

| 24 Apr 2025

Gross primary productivity responses to meteorological drivers: insights from observations and multi-model ensembles

Yuxin Zheng, Xu Yue, Xiaofei Lu, and Jun Zhu

Abstract. Climate change has a substantial impact on ecosystem gross primary productivity (GPP), but the specific roles of different meteorological factors across various vegetation types remain unclear. This study investigates GPP responses to variations in temperature, precipitation, and drought, using data from three observational products and 17 dynamic vegetation models. Observed GPP showed a positive response to temperature in boreal regions, with sensitivities ranging from 0.01 to 0.05 g C m² day^-1 K^-1. In contrast, GPP responded negatively to temperature in the tropics, with sensitivities of -0.07±0.15 g C m² day^-1 K^-1 for evergreen broadleaf forests and -0.25±0.11 g C m²day^-1 K^-1 for C₄ grasslands. Precipitation had a relatively low impact on GPP in deciduous and evergreen forests, while non-tree species, such as grasslands and croplands, showed a positive response. GPP sensitivity to drought index (scPDSI) was similar to that of precipitation, except that observed GPP in evergreen forests negatively responded to scPDSI. The models generally reproduced these observed patterns but tended to overestimate the effect of precipitation on GPP. As a result, they predicted higher sensitivity in tropical grasslands to drought stress but lower resilience in trees. Both observations and simulations exhibited negative GPP responses to extreme warming and drought on a global scale, though models tended to overestimate the magnitude of these negative effects. This study distinguished GPP responses to key meteorological factors across vegetation types and numerical models, providing critical insights for improving the prediction of terrestrial carbon sinks and promoting the climatic resilience of ecosystems.

Received: 30 Mar 2025 – Discussion started: 24 Apr 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, 2909 KB)

Supplement (2725 KB)

Download & links

Yuxin Zheng, Xu Yue, Xiaofei Lu, and Jun Zhu

Status: closed

RC1:
'Comment on egusphere-2025-1515', Anonymous Referee #1, 21 May 2025
General comments
This manuscript investigates how gross primary productivity (GPP) responds to temperature, precipitation, and drought across global ecosystems, using two satellite-derived GPP products (GLASS and GOSIF), a data-driven GPP product (JUNG) and outputs from 17 DGVMs within the TRENDY v11 multi-model simulations. This study refers to the mean of three GPP products as “observations” and uses them to evaluate the sensitivity of DGVM-simulated GPP to environmental variables in terms of the global spatial pattern and different plant functional types (PFTs). The topic is important for understanding the terrestrial carbon cycle under a changing climate. However, there are several major issues that still need to be addressed regarding terminology, data selection and benchmarking, and incomplete information, all of which undermine the study’s reliability. Below, I outline major concerns, missing information, and specific comments for revision.

Major concerns：
The authors refer to GLASS (based on MODIS/AVHRR and LUE models), GOSIF (based on SIF, proxy for GPP), and JUNG (data-driven, upscaled via machine learning) products as “Observations of GPP.” This is conceptually inaccurate and misleading. The GLASS and GOSIF GPP products should instead be referred to as “satellite-derived GPP”/ “satellite-based GPP,” as none of them are direct observations. I believe this mislabeling may also confuse other readers.

The authors use the mean of three GPP products as benchmarks to evaluate the sensitivity of TRENDY model simulations to environmental drivers. Although these products have been validated at flux tower sites, their performance is not consistently reliable, especially for long term trends (Zheng et al., 2020, Bai et al., 2023). In addition, Bai et al. (2023) reported large discrepancies in trends among different satellite-derived GPP products. As shown in Figure 4 of this manuscript, for some plant functional types (PFTs), the sensitivity of GPP to climatic variables varies greatly among the three satellite-derived datasets—sometimes with differences as large as the sensitivities themselves (e.g., for EBF in Fig. 4a and Fig. 4c). Therefore, it is questionable to use the average of these three products as a robust benchmark.

the classification of grid cells by plant functional types (PFTs) is based on the 2001–2012 MODIS land cover mean, but the analysis period spans as far back as 1982. Since both real-world vegetation and modeled PFTs can shift over time, this temporal mismatch introduces additional uncertainty. A more robust approach would be to focus on flux tower sites with stable vegetation types and use site-level GPP–climate relationships to evaluate both model and satellite-derived GPP responses.

Incomplete information
The caption of Figure 2 is the only place where the time periods of the three GPP products are mentioned: “For observed GPP, correlation coefficients were calculated at each grid cell over the period 1982–2017 for GLASS, 2001–2018 for GOSIF, and 1982–2011 for JUNG.” The manuscript does not explain how the temporal mismatch was handled, nor how the three datasets were merged. Additionally, it is unclear what time period was used when calculating the correlation between the merged GPP dataset and climatic variables. Please clarify.

Please provide a supplementary figure showing the spatial distribution and number of grid cells classified under each land cover type.

The definition of “sensitivity” is not clearly stated in the manuscript. A formal definition needs to be provided.

In Section 2.3, please include the mathematical formulation of the scPDSI index.

In Section 2.3, please specify the final temporal resolution of the data used in this study. Are the analyses conducted at monthly or annual resolution?

The discussion in Section 4.1 is too general—it only addresses GPP responses to temperature, precipitation, and drought on a global scale. Readers would benefit from a more detailed discussion of how these responses differ across vegetation types (especially where GPP responses differ significantly between PFTs).

The main text lacks a conclusion section.

Specific comments
fig 1a. JUNG and GOSIF GPP time series are missing.

fig 1. Please Specify the exact time period over which GPP trends were calculated.

fig 3. Please adding the model ensemble mean result to the figure and analyzing it in the corresponding text.

fig 4. Suggest adding a global mean bar in this bar plot.

fig 5. The “latitudinal variations” plots lack units.

line 22. In sentence: "Precipitation had a relatively low impact on GPP" Please be clear to state relative low positive or negative impact? — is this referring to the model or to the GPP products?

line 61-62. “While this response is protective in the short term, it ultimately leads to a decline in GPP.” — needs citation.

line 64-68. Only one example is provided. Please add at least one more reference to support the statement: “there has been a notable increase in the sensitivity of global ..."

line 81-83. Also needs at least one more reference to support the statement: “the need of careful calibration and validation using observed data to improve model reliability.”

line 101. LAI is not an “environmental factor,” but rather a vegetation structural parameter. Please revise.

line 101-103. Since the study analyzes GLASS GPP responses and uses GLASS as a benchmark, please also cite literature showing the consistency between long-term GLASS GPP and tower-based observations.

line 136. It is unclear whether the paper uses S2 or S3 TRENDY simulations—please clarify.

line 204-205. The interpolation method is not described—please add.

line 213. “Most models predicted...” — please specify the exact number of models (e.g., X out of 17).

line 229. The phrase “The ensemble of three observational datasets revealed large spatial heterogeneity in GPP trends” is ambiguous—it could be interpreted as inconsistency among datasets. If the intended meaning is that GPP trends themselves are spatially variable, please reword.

line 236-203. The statement that “Overall, the MME captured the latitudinal variations in GPP trends but tended to overestimate positive trends in tropical regions.” is inaccurate. According to Figure 1, the trends in tropical regions differ in sign between MME and satellite products. Please revise.

line 255. The phrase “likely due to an inadequate representation of light dependency” requires a citation.

line 261. “(Figs. 3 and S1–S3)” should be “Figures S1–S3.”

line 296. “Tree species” is first defined in Line 308. Please move or adjust for clarity.

line 289-301. Figure S4 is cited four times in the main text. If it is so central to the analysis, consider moving it into the main text.

line 305. “in C4 grasslands” should be “for C4 grasslands.”

line 303-305. GPP from shrublands also decreases with rising temperature —this information is missing.

line 343. "various dataset" is undefined - please clarify.

line 342-353. Same issue as above with Line 289–301 — avoid repeatedly citing supplementary figures in the main text. Either integrate them or move the relevant discussion to the supplement.

line 359. In addition to coniferous forests, high-latitude regions also include tundra, deciduous broadleaf forests, and wetlands, etc. Please revise.

line 390. The phrase “improper parameterization” is too vague. It sounds like the models are fundamentally flawed. Please revise or provide a specific reference.

line 420. The manuscript does not analyze interannual variability of GPP response to climatic variables. Please revise the statement accordingly.

References:
Bai, Y., Zhang, S., Zhang, J., Zhao, Y., & Yuan, W. (2023). Different satellite products revealing variable trends in global gross primary production. Journal of Geophysical Research: Biogeosciences, 128(2), e2022JG006918. https://doi.org/10.1029/2022JG006918ResearchGate
Zheng, Y., Yuan, W., Zhang, J., & Liu, S. (2020). Improved estimate of global gross primary production for reproducing its long-term variation, 1982–2017. Earth System Science Data, 12(3), 2725–2746. https://doi.org/10.5194/essd-12-2725-2020
Citation: https://doi.org/10.5194/egusphere-2025-1515-RC1
- AC1: 'Reply on RC1', Xu Yue, 16 Sep 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-1515/egusphere-2025-1515-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2025-1515-AC1
RC2:
'Comment on egusphere-2025-1515', Anonymous Referee #2, 14 Aug 2025

General comments:
The authors utilized existing GPP observations and multi-model simulations to examine latitudinal differences in temperature responses (positive in boreal, negative in tropics) and vegetation-type specific sensitivities. While the topic is within the scope of Biogeosciences, I have several major concerns regarding the novelty and methodology of this work.
Numerous previous studies have already investigated GPP responses to temperature, precipitation, and drought across different regions and vegetation types, as well as global GPP responses to specific meteorological factors, i.e., tropical region (Piao et al., 2013, Lomax et al. 2024), boreal forest to drought (Lindroth et al. 2020, Martínez-García et al., 2024). The general relationships between GPP and meteorological variables described in the abstract could essentially be obtained through literature review alone. Therefore, I am not fully convinced by the motivation, novelty, and critical insights in re-examining these well-documented responses.
The authors acknowledged the potential nonlinear relationship between GPP and meteorological variables in the introduction, yet the analysis relied entirely on linear regression/correlation methods. This approach is inadequate for capturing non-linear GPP responses to extremes (e.g., drought thresholds, temperature optima).
Additionally, the observational datasets (GLASS, GOSIF, JUNG) cover different periods (1982–2017 vs. 2001–2018 vs. 1982–2011). The non-overlapping timeframes may introduce biases in trend and sensitivity analyses, particularly given accelerated climate change after the 2000s.
The 17 models included in this study vary substantially in resolution, carbon-nitrogen coupling, and radiation schemes as illustrated in Table 1. The authors should address how structural differences contribute to inter-model variability through sensitivity analyses or other methods.
Specific comments:
Please clarify all the units, Pg C or Pg CO_2, throughout the manuscript.
Only the GLASS GPP product is shown in Figure 1a, while the other two observation datasets are not included. Please explain the rationale for this selective presentation.
Line 380-383: The divergence between observations and models is already evident in the GPP trends shown in Figure 1. This discrepancy should be explained before analyzing the GPP responses to meteorological factors.
Line 385-388: references are needed for this. Is this from your analysis or previous studies? Please clarify and provide appropriate citations.

Citation: https://doi.org/10.5194/egusphere-2025-1515-RC2
- AC2: 'Reply on RC2', Xu Yue, 16 Sep 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-1515/egusphere-2025-1515-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2025-1515-AC2

Status: closed

RC1:
'Comment on egusphere-2025-1515', Anonymous Referee #1, 21 May 2025
General comments
This manuscript investigates how gross primary productivity (GPP) responds to temperature, precipitation, and drought across global ecosystems, using two satellite-derived GPP products (GLASS and GOSIF), a data-driven GPP product (JUNG) and outputs from 17 DGVMs within the TRENDY v11 multi-model simulations. This study refers to the mean of three GPP products as “observations” and uses them to evaluate the sensitivity of DGVM-simulated GPP to environmental variables in terms of the global spatial pattern and different plant functional types (PFTs). The topic is important for understanding the terrestrial carbon cycle under a changing climate. However, there are several major issues that still need to be addressed regarding terminology, data selection and benchmarking, and incomplete information, all of which undermine the study’s reliability. Below, I outline major concerns, missing information, and specific comments for revision.

Major concerns：
The authors refer to GLASS (based on MODIS/AVHRR and LUE models), GOSIF (based on SIF, proxy for GPP), and JUNG (data-driven, upscaled via machine learning) products as “Observations of GPP.” This is conceptually inaccurate and misleading. The GLASS and GOSIF GPP products should instead be referred to as “satellite-derived GPP”/ “satellite-based GPP,” as none of them are direct observations. I believe this mislabeling may also confuse other readers.

The authors use the mean of three GPP products as benchmarks to evaluate the sensitivity of TRENDY model simulations to environmental drivers. Although these products have been validated at flux tower sites, their performance is not consistently reliable, especially for long term trends (Zheng et al., 2020, Bai et al., 2023). In addition, Bai et al. (2023) reported large discrepancies in trends among different satellite-derived GPP products. As shown in Figure 4 of this manuscript, for some plant functional types (PFTs), the sensitivity of GPP to climatic variables varies greatly among the three satellite-derived datasets—sometimes with differences as large as the sensitivities themselves (e.g., for EBF in Fig. 4a and Fig. 4c). Therefore, it is questionable to use the average of these three products as a robust benchmark.

the classification of grid cells by plant functional types (PFTs) is based on the 2001–2012 MODIS land cover mean, but the analysis period spans as far back as 1982. Since both real-world vegetation and modeled PFTs can shift over time, this temporal mismatch introduces additional uncertainty. A more robust approach would be to focus on flux tower sites with stable vegetation types and use site-level GPP–climate relationships to evaluate both model and satellite-derived GPP responses.

Incomplete information
The caption of Figure 2 is the only place where the time periods of the three GPP products are mentioned: “For observed GPP, correlation coefficients were calculated at each grid cell over the period 1982–2017 for GLASS, 2001–2018 for GOSIF, and 1982–2011 for JUNG.” The manuscript does not explain how the temporal mismatch was handled, nor how the three datasets were merged. Additionally, it is unclear what time period was used when calculating the correlation between the merged GPP dataset and climatic variables. Please clarify.

Please provide a supplementary figure showing the spatial distribution and number of grid cells classified under each land cover type.

The definition of “sensitivity” is not clearly stated in the manuscript. A formal definition needs to be provided.

In Section 2.3, please include the mathematical formulation of the scPDSI index.

In Section 2.3, please specify the final temporal resolution of the data used in this study. Are the analyses conducted at monthly or annual resolution?

The discussion in Section 4.1 is too general—it only addresses GPP responses to temperature, precipitation, and drought on a global scale. Readers would benefit from a more detailed discussion of how these responses differ across vegetation types (especially where GPP responses differ significantly between PFTs).

The main text lacks a conclusion section.

Specific comments
fig 1a. JUNG and GOSIF GPP time series are missing.

fig 1. Please Specify the exact time period over which GPP trends were calculated.

fig 3. Please adding the model ensemble mean result to the figure and analyzing it in the corresponding text.

fig 4. Suggest adding a global mean bar in this bar plot.

fig 5. The “latitudinal variations” plots lack units.

line 22. In sentence: "Precipitation had a relatively low impact on GPP" Please be clear to state relative low positive or negative impact? — is this referring to the model or to the GPP products?

line 61-62. “While this response is protective in the short term, it ultimately leads to a decline in GPP.” — needs citation.

line 64-68. Only one example is provided. Please add at least one more reference to support the statement: “there has been a notable increase in the sensitivity of global ..."

line 81-83. Also needs at least one more reference to support the statement: “the need of careful calibration and validation using observed data to improve model reliability.”

line 101. LAI is not an “environmental factor,” but rather a vegetation structural parameter. Please revise.

line 101-103. Since the study analyzes GLASS GPP responses and uses GLASS as a benchmark, please also cite literature showing the consistency between long-term GLASS GPP and tower-based observations.

line 136. It is unclear whether the paper uses S2 or S3 TRENDY simulations—please clarify.

line 204-205. The interpolation method is not described—please add.

line 213. “Most models predicted...” — please specify the exact number of models (e.g., X out of 17).

line 229. The phrase “The ensemble of three observational datasets revealed large spatial heterogeneity in GPP trends” is ambiguous—it could be interpreted as inconsistency among datasets. If the intended meaning is that GPP trends themselves are spatially variable, please reword.

line 236-203. The statement that “Overall, the MME captured the latitudinal variations in GPP trends but tended to overestimate positive trends in tropical regions.” is inaccurate. According to Figure 1, the trends in tropical regions differ in sign between MME and satellite products. Please revise.

line 255. The phrase “likely due to an inadequate representation of light dependency” requires a citation.

line 261. “(Figs. 3 and S1–S3)” should be “Figures S1–S3.”

line 296. “Tree species” is first defined in Line 308. Please move or adjust for clarity.

line 289-301. Figure S4 is cited four times in the main text. If it is so central to the analysis, consider moving it into the main text.

line 305. “in C4 grasslands” should be “for C4 grasslands.”

line 303-305. GPP from shrublands also decreases with rising temperature —this information is missing.

line 343. "various dataset" is undefined - please clarify.

line 342-353. Same issue as above with Line 289–301 — avoid repeatedly citing supplementary figures in the main text. Either integrate them or move the relevant discussion to the supplement.

line 359. In addition to coniferous forests, high-latitude regions also include tundra, deciduous broadleaf forests, and wetlands, etc. Please revise.

line 390. The phrase “improper parameterization” is too vague. It sounds like the models are fundamentally flawed. Please revise or provide a specific reference.

line 420. The manuscript does not analyze interannual variability of GPP response to climatic variables. Please revise the statement accordingly.

References:
Bai, Y., Zhang, S., Zhang, J., Zhao, Y., & Yuan, W. (2023). Different satellite products revealing variable trends in global gross primary production. Journal of Geophysical Research: Biogeosciences, 128(2), e2022JG006918. https://doi.org/10.1029/2022JG006918ResearchGate
Zheng, Y., Yuan, W., Zhang, J., & Liu, S. (2020). Improved estimate of global gross primary production for reproducing its long-term variation, 1982–2017. Earth System Science Data, 12(3), 2725–2746. https://doi.org/10.5194/essd-12-2725-2020
Citation: https://doi.org/10.5194/egusphere-2025-1515-RC1
- AC1: 'Reply on RC1', Xu Yue, 16 Sep 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-1515/egusphere-2025-1515-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2025-1515-AC1
RC2:
'Comment on egusphere-2025-1515', Anonymous Referee #2, 14 Aug 2025

General comments:
The authors utilized existing GPP observations and multi-model simulations to examine latitudinal differences in temperature responses (positive in boreal, negative in tropics) and vegetation-type specific sensitivities. While the topic is within the scope of Biogeosciences, I have several major concerns regarding the novelty and methodology of this work.
Numerous previous studies have already investigated GPP responses to temperature, precipitation, and drought across different regions and vegetation types, as well as global GPP responses to specific meteorological factors, i.e., tropical region (Piao et al., 2013, Lomax et al. 2024), boreal forest to drought (Lindroth et al. 2020, Martínez-García et al., 2024). The general relationships between GPP and meteorological variables described in the abstract could essentially be obtained through literature review alone. Therefore, I am not fully convinced by the motivation, novelty, and critical insights in re-examining these well-documented responses.
The authors acknowledged the potential nonlinear relationship between GPP and meteorological variables in the introduction, yet the analysis relied entirely on linear regression/correlation methods. This approach is inadequate for capturing non-linear GPP responses to extremes (e.g., drought thresholds, temperature optima).
Additionally, the observational datasets (GLASS, GOSIF, JUNG) cover different periods (1982–2017 vs. 2001–2018 vs. 1982–2011). The non-overlapping timeframes may introduce biases in trend and sensitivity analyses, particularly given accelerated climate change after the 2000s.
The 17 models included in this study vary substantially in resolution, carbon-nitrogen coupling, and radiation schemes as illustrated in Table 1. The authors should address how structural differences contribute to inter-model variability through sensitivity analyses or other methods.
Specific comments:
Please clarify all the units, Pg C or Pg CO_2, throughout the manuscript.
Only the GLASS GPP product is shown in Figure 1a, while the other two observation datasets are not included. Please explain the rationale for this selective presentation.
Line 380-383: The divergence between observations and models is already evident in the GPP trends shown in Figure 1. This discrepancy should be explained before analyzing the GPP responses to meteorological factors.
Line 385-388: references are needed for this. Is this from your analysis or previous studies? Please clarify and provide appropriate citations.

Citation: https://doi.org/10.5194/egusphere-2025-1515-RC2
- AC2: 'Reply on RC2', Xu Yue, 16 Sep 2025
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-1515/egusphere-2025-1515-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2025-1515-AC2

Yuxin Zheng, Xu Yue, Xiaofei Lu, and Jun Zhu

Supplement

https://doi.org/10.5194/egusphere-2025-1515-supplement

Yuxin Zheng, Xu Yue, Xiaofei Lu, and Jun Zhu

Viewed

Total article views: 642 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
558	67	17	642	25	17	36

HTML: 558
PDF: 67
XML: 17
Total: 642
Supplement: 25
BibTeX: 17
EndNote: 36

Views and downloads (calculated since 24 Apr 2025)

Month	HTML	PDF	XML	Total
Apr 2025	82	12	4	98
May 2025	87	22	5	114
Jun 2025	68	8	1	77
Jul 2025	67	9	1	77
Aug 2025	127	8	1	136
Sep 2025	127	8	5	140

Cumulative views and downloads (calculated since 24 Apr 2025)

Month	HTML	PDF	XML	Total
Apr 2025	82	12	4	98
May 2025	87	22	5	114
Jun 2025	68	8	1	77
Jul 2025	67	9	1	77
Aug 2025	127	8	1	136
Sep 2025	127	8	5	140

Viewed (geographical distribution)

Total article views: 617 (including HTML, PDF, and XML) Thereof 617 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 19 Sep 2025

Short summary

The sensitivity of gross primary productivity (GPP) to climate varies across regions and ecosystem types. Our study shows that warming enhances GPP in boreal regions but suppresses it in the tropics. Increased precipitation generally promotes GPP but has limited effects on tree species. However, current models tend to overestimate GPP responses to precipitation, resulting in greater GPP variability during drought events.


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