the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 03 Jul 2024
Toward more robust NPP projections in the North Atlantic Ocean
Abstract. Phytoplankton plays a crucial role in both climate regulation and marine biodiversity, yet it faces escalating threats due to climate change. Projecting the future changes in phytoplankton biomass and productivity under climate change requires the utilization of Earth System Models capable of resolving marine biogeochemistry, and exploits the averaged responses across an ensemble of models (within the Coupled Model Intercomparison Project Phase 6, CMIP6) as the most probable projection. However, in the North Atlantic, this straightforward method falls short in providing robust projections of phytoplankton net primary production (NPP) over the 21st century. This is because the processes controlling NPP strongly differ from one model to another, thus causing model divergence. A new inter-comparison approach was therefore developed and applied to 8 CMIP6 models exhibiting substantial divergence in their NPP projections in the North Atlantic. This approach is based on the identification of the mechanisms causing model divergence and the assessment of their reliability, in order to conduct an informed selection of the most reliable models. The basin was first divided into 3 bioregions tailored to the characteristics of each model using a novel method based on a clustering procedure. Two key mechanisms causing model divergence were then identified in the subtropical and subpolar regions (namely, diazotrophy and the presence of an ammonium pool, respectively). This allowed for an informed selection of models in each region, resulting in reduced uncertainty and a more pronounced decrease in total NPP in the subtropical North Atlantic and a stronger increase of small phytoplanton NPP in the subpolar North Atlantic. Our model selection strengthened carbon export and phytoplankton biomass decreases under climate change, but had no impact on zooplankton biomass. By leveraging the diversity of CMIP6 models, this innovative approach identifies the key mechanisms influencing NPP projections and provides valuable insights into the future trajectory of the Earth’s climate system.
- Preprint
(3631 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-1820', Anonymous Referee #1, 10 Sep 2024
General comments
NPP projections in CMIP6 biogeochemical models strongly diverge during the 21st century in the North Atlantic despite a strong consistency between the physical and biogeochemical variables controlling NPP. This manuscript studies an ensemble of 8 ESMs which embed diverse biogeochemical components with significant differences in their parametrizations of primary production. Two mechanisms causing model divergence are identified: diazotrophy (for the subtropical region) and the presence of an ammonium pool (for the subpolar region).
This manuscript is worth publishing. It is well written and I was quite convinced by the explanations. But some further work is needed to make it easier to read, more convincing and impactful. The variety of models (with different PFTs/limiting nutrients/species naming) makes it difficult to disentangle and to interpret results. Throughout the manuscript, there are some results that are not shown in figures. It is a bit confusing. I suggest that you should either show them, or explicitly state that they are “not shown” in the text. Reference to Figures (and which subplot) and Tables could be improved to better guide the reader. In particular, section 3.3.2 on the Subpolar region needs to be better justified (with at least an additional figure or with a sensitivity experiment demonstrating the influence of the ammonium pool).
Detailed comments
Introduction:
While reading the manuscript, I thought about emergent constraints as another method that could be relevant given the objective of this study. You finally mention and discuss this method and relevance for this specific topic only in the conclusion. The same could apply to a selection of CMIP6 models based on their ability to represent past observed changes in NPP. We also somehow lack validation of the ability of the selected models to represent past NPP changes. I would recommend better framing the different methods that could be used (including emergent constraints) and why you chose this one much earlier in the manuscript, as early as in the introduction.
P3-4 Section 2.1:
- Table 1. It would be of interest to list the equilibrium climate sensitivity of the chosen models as well, to inform on their sensitivity to climate change. I would also suggest mentioning the presence (or not) of an ammonium pool in the BGC models.
- Line 88-89: Please specify the models you are referring to (the simplest and most complex BGC models) and refer to Table 1
- Line 89: typo, "the most complex model” : “"complex” is currently missing
- Line 101-102: For the IPSL model, you might be explicit on the version of diazotrophy that is used (given the Bopp et al. 2022 study).
- L91: Why don’t you combine diatoms and microphytoplankton as one large phytoplankton class ? In the manuscript, you justify this because diatoms rely on silicate limitation, but in the end, it seems to me that you never use this silicate limitation to disentangle some processes between the different models. This would allow you to show an additional curve (CanESM5-CanOE) in Fig. 3, 5 and 6, panel B.
- Could you add 1-2 sentences about the differences between the models regarding PO4 and NH4 pools to introduce your explanations about their influences in the following sections? This would immediately explain why some models are missing in the PO4 panel of Fig. 3.
Section 2.2.2. Please provide the explicit identifier (number) of the product to avoid confusion as Copernicus Marine Service products evolve over time
P5 L152: how were coastal areas excluded?
P6 L175: Please refer to Table 2.
P6 L180. Models disagree on sign, but are projected changes large in these regions in the different models? 2D maps of model spread would be useful to complement Fig 1.
P7 Section 3.2. The reference to Fig 2 comes too late. The same applies to the merging of the 2 subpolar subregions. Their separation might be shown in Fig 2 with a dashed line for instance.
P7 L214: yes, surface NO3 concentrations are decreasing in all models but for the 3 ones that project an increase in NPP, it decreases significantly less (2-3 times less) than the others. Same for PO4. Do you think this is a consequence of the increase in N2 fixation and hence of the increase in NPP ?
P7 L213-214: Why do you show only NO3, PO4 and temperature at sea surface while other variables are vertically integrated? This needs to be better justified…
P8, L219: Fig4 -> Fig4a
P8, paragraph on models without any diazotrophy:
- You refer to an observational dataset “the observed value is 0.4 mmol.m-3” but which obs dataset? Over what time period? Perhaps you could add it to Fig.3
- For ACCESS-ESM1-5, you do not mention the increase of MLD in the projections. Is this not the reason for the increase in small phytoplankton NPP (more than the SST increase mentioned in the text)?
- P8 L228. Replace “nanophytoplankton” by “small phytoplankton” for consistency throughout the manuscript.
P8, paragraph on models with explicit diazotrophy: please precise the subplot of Fig 3 you refer to. L233: primary production = total primary production?
P8 L245: For consistency of this section and previous two paragraphs, you might start this paragraph in a more general sense, addressing models with parameterized diazotrophy.
P8, paragraph on models with implicit diazotrophy: please, precise “not shown” at the end of the sentence “However, unlike in IPSL-CM6A-LR, …”. It is not clear from your explanations why the two CanESM models (CanESM5 and CanESM5-CanOE) give such different projections of NPP. L250, does “large phytoplankton” refer to the microphytoplankton of CanESM5-CanOE?
P9 L268: What about diatoms? All models show a diatom NPP decrease but with groups of models showing substantially different decreases.
P9, paragraph L272-283: here an additional figure showing the absolute value of NO3 and NH4 as a function of time is needed so that the reader can see what you are referring to in the paragraph. I am also wondering how you explain the different behaviour between IPSL-CM6A-LR and CESM2/CESM2-WACCM models regarding the projections of diatom NPP and small PHY NPP.
P10, L291: “were still covered” -> will still be covered
P10, L292-293: sentence “However… NO3 and PO4 surface concentrations decrease in all models (fig. 6)”: But it is not the same process as in the subpolar region that causes NO3 and PO4 to decrease. They are probably more consumed due to the increase in NPP.
P10, L289-293: What about changes in upper ocean stratification? Changes in NPP can also occur in areas that were not and will not be covered by sea-ice.
P10, L296: You refer to the half-saturation constant. Is it that of diatoms?
P10. Section 3.4 lacks a reference to Table 2.
P11, L332: I would delete: “and the extreme would be kept in the subset”
P11, L335: “influence diatoms in the former but not the latter”: is it not the opposite?
P13, L390: Typo: “Emergent” --> “emergent”
Fig. 1: Use the same arrangement as in Fig. 3 for the box at the bottom of the figure. The caption could be made more explicit on using the multi-member mean for each ESM.
Fig. 2: Could you add the boundary between the 2 subpolar subregions (with a dotted line for example) so that the reader is aware of their extent in each model and observations?
Fig. 3f: diazotrophy subplot: Please give more details about this “diazotrophy” term. Or give the equations. Is it equal to diazotroph NPP? Why is the unit different between explicit and implicit diazotrophs? It is confusing as it is.
Fig. 3g and h: Please specify that surface NO3 and PO4 concentrations are shown.
Fig. 1, 3, 5, 6: Please make these important figures larger with larger labels. Difficult to distinguish the blue colour of IPSL-CM6A-LR from that of CESM2-WACCM. Please add a title indicating the variable on each subplot to make it easier to read. Could you find a graphical way to remind the reader which model belongs to which group (no diazotrophy, implicit or explicit one) to make it easier to interpret and follow your explanations?
Fig. 4: Instead of “correlations”, I would call it “scatterplot” and then add the correlation numbers to the subplot. Could you explain in more details how you constructed this graph? What does each cross represent? This figure needs to be clarified.
Fig. 7: Please recall the model selection for each region in the figure caption
Table 1: Replace “Diatoms (param.) with Diazotrophs (param.). Could you add to this table the nutrients limiting the growth of each phytoplankton class, so that we know if there is a NH4/PO4/Si… pool? The words “with a parametrization and without an explicit diazotroph group” should be removed in the Table 1 description.
Citation: https://doi.org/10.5194/egusphere-2024-1820-RC1 -
AC1: 'Reply on RC1', Stéphane Doléac, 08 Oct 2024
Response to review 1. October 2024
General comments
NPP projections in CMIP6 biogeochemical models strongly diverge during the 21st century in the North Atlantic despite a strong consistency between the physical and biogeochemical variables controlling NPP. This manuscript studies an ensemble of 8 ESMs which embed diverse biogeochemical components with significant differences in their parametrizations of primary production. Two mechanisms causing model divergence are identified: diazotrophy (for the subtropical region) and the presence of an ammonium pool (for the subpolar region).
This manuscript is worth publishing. It is well written and I was quite convinced by the explanations. But some further work is needed to make it easier to read, more convincing and impactful. The variety of models (with different PFTs/limiting nutrients/species naming) makes it difficult to disentangle and to interpret results. Throughout the manuscript, there are some results that are not shown in figures. It is a bit confusing. I suggest that you should either show them, or explicitly state that they are “not shown” in the text. Reference to Figures (and which subplot) and Tables could be improved to better guide the reader. In particular, section 3.3.2 on the Subpolar region needs to be better justified (with at least an additional figure or with a sensitivity experiment demonstrating the influence of the ammonium pool).
We thank the reviewer for his thorough review and insightful comments, which have significantly contributed to improving our manuscript. Most of the revisions suggested in this review have been incorporated. Notably:
- The figures have been enhanced to improve clarity and make the results section easier to understand.
- A new figure has been added, illustrating the evolution of the proportion of NPP based on NH4 rather than NO3, highlighting the impact of NH4 on small phytoplankton.
- Explanations of the mechanisms controlling NPP projections in each model have been refined and are now more clearly articulated.
Detailed comments
Introduction:
While reading the manuscript, I thought about emergent constraints as another method that could be relevant given the objective of this study. You finally mention and discuss this method and relevance for this specific topic only in the conclusion. The same could apply to a selection of CMIP6 models based on their ability to represent past observed changes in NPP. We also somehow lack validation of the ability of the selected models to represent past NPP changes. I would recommend better framing the different methods that could be used (including emergent constraints) and why you chose this one much earlier in the manuscript, as early as in the introduction.
The last two paragraphs of the introduction have been replaced by the following three paragraphs to take into account the comments of the other reviewer and to better frame the different existing methods to reduce projection uncertainties:
“Several approaches have already been proposed and shown to effectively reduce projection uncertainty. Some studies advocate for model weighting based on their ability to replicate past observations (Sanderson et al., 2017; Knutti et al., 2017), while others highlight the use of emergent constraints as a promising method for reducing projection uncertainty (Eyring et al., 2019; Kwiatkowski et al., 2017). Additionally, in-depth analyses have been conducted to explore the physical drivers of NPP, attributing model discrepancies to variations in ocean physics and their interactions with biogeochemical variables (Whitt, 2019; Xiu et al., 2018; Mousing et al., 2023). However, many of these studies tend to overlook the significant differences in the biogeochemical components of CMIP6 models (Kearney et al., 2021; Séférian et al., 2020). Yet, it is well-established that certain processes, depending on their inclusion and parametrization, can substantially influence NPP projections (Bopp et al., 2022; Rohr et al., 2023).
Building on these insights, the present study seeks to go a step further by linking part of the divergence in NPP projections to specific differences in the biogeochemical processes and pools represented across CMIP6 models. To achieve this, we will identify the processes driving long-term NPP evolution under the SSP5-8.5 scenario in each model and use these insights to assess the reliability of their projections. This process-based evaluation will then inform a model selection aimed at reducing projection uncertainties.
The North Atlantic Ocean, a highly productive region with significant future divergence between model projections (Tagliabue et al., 2021), will serve as a case study. Given that the processes controlling NPP vary across different regions, we will divide the North Atlantic basin into three bioregions (polar, subpolar, and subtropical) using a clustering method and evaluate each subregion individually. Finally, we will explore how our revised estimates of NPP projections influence carbon export and plankton biomass—key indicators of the biological carbon pump and marine ecosystem health—and how our approach integrates with the other methods mentioned above.”
P3-4 Section 2.1:
- Table 1. It would be of interest to list the equilibrium climate sensitivity of the chosen models as well, to inform on their sensitivity to climate change. I would also suggest mentioning the presence (or not) of an ammonium pool in the BGC models.
The nutrients included in each model are now detailed in Table 1. However, we chose not to include ECS in Table 1, as our focus is on explaining the differences in the sign of NPP projections. There is no observed correlation between ECS and the direction of long-term NPP changes (Zelinka et al. 2020). Including ECS would, therefore, add unnecessary complexity to the paper without contributing relevant insights to our approach.
- Line 88-89: Please specify the models you are referring to (the simplest and most complex BGC models) and refer to Table 1
The reference to Table 1 and the names of the marine biogeochemistry components mentioned were added.
- Line 89: typo, "the most complex model” : “"complex” is currently missing
The typo was corrected, thank you for spotting it.
- Line 101-102: For the IPSL model, you might be explicit on the version of diazotrophy that is used (given the Bopp et al. 2022 study).
The following sentence was added in line 101-102 : “In IPSL-CM6A-LR, the initial parametrization from PISCES-v2 was used (Aumont et al., 2015; Bopp et al., 2022).”
- L91: Why don’t you combine diatoms and microphytoplankton as one large phytoplankton class ? In the manuscript, you justify this because diatoms rely on silicate limitation, but in the end, it seems to me that you never use this silicate limitation to disentangle some processes between the different models. This would allow you to show an additional curve (CanESM5-CanOE) in Fig. 3, 5 and 6, panel B.
The reviewer is right, this initial distinction was not used in the manuscript and it was decided to combine the microphytoplankton of CanESM5-CanOE with diatoms. However, the label “diatoms” was kept as it corresponds to the majority of models in the group. The difference of CanESM5-CanOE is now highlighted in figures 3, 5 and 6.
- Could you add 1-2 sentences about the differences between the models regarding PO4 and NH4 pools to introduce your explanations about their influences in the following sections? This would immediately explain why some models are missing in the PO4 panel of Fig. 3.
The following paragraph has been added at the end of section 2.1 : “CMIP6 models can also be distinguished based on how they represent nitrogen and phosphorus (Table 1). Among the eight models studied, three (ACCESS-ESM1-5, CanESM5, and UKESM1-0-LL) use a unique bulk nutrient meant to represent both nitrogen and phosphorus. In contrast, three others (IPSL-CM6A-LR, CESM2, and CESM2-WACCM) include a PO4 pool along with two nitrogen pools, NO3 and NH4. The CanESM5-CanOE model includes two nitrogen pools (NO3 and NH4), but lacks a PO4 pool. Lastly, MPI-ESM1-2-LR represents both a PO4 and a NO3 pool, but does not include NH4.”
Section 2.2.2. Please provide the explicit identifier (number) of the product to avoid confusion as Copernicus Marine Service products evolve over time
The IDs of the Copernicus Marine Service products were added to increase clarity.
P5 L152: how were coastal areas excluded?
The coastal areas were excluded by removing values from grid cells adjacent to the continents. This clarification has been added.
P6 L175: Please refer to Table 2.
The reference was added.
P6 L180. Models disagree on sign, but are projected changes large in these regions in the different models? 2D maps of model spread would be useful to complement Fig 1.
Our study focuses solely on the disagreement in the sign of NPP projections across different models, and not in the emergence of the detected trends. We therefore decided to use only one of the two definitions for projection robustness used in the AR6 WGI IPCC report (figure 4.13) :the criterion that 80 % of models agree on the sign of change. The other definition (“66% of the models show a change greater than the internal-variability threshold”) would provide valuable insights but falls outside the scope of this study.
P7 Section 3.2. The reference to Fig 2 comes too late. The same applies to the merging of the 2 subpolar subregions. Their separation might be shown in Fig 2 with a dashed line for instance.
A new reference to figure 2 was added earlier in the paragraph and the initial separation between the two subpolar regions is now visible.
P7 L214: yes, surface NO3 concentrations are decreasing in all models but for the 3 ones that project an increase in NPP, it decreases significantly less (2-3 times less) than the others. Same for PO4. Do you think this is a consequence of the increase in N2 fixation and hence of the increase in NPP ?
In ACCESS-ESM1-5, the smaller decline in NO3 concentrations cannot be attributed to N2 fixation, as this process is absent in the model. We believe it is likely due to the recovery of the MLD.
For IPSL-CM6A-LR, N2 fixation could indeed serve as an indirect source of NO3, potentially explaining the smaller decrease in NO3 compared to other models. However, other factors, such as gyre dynamics and global ocean circulation, may also contribute. We did not attempt to quantify the individual contributions of these mechanisms to the reduction in surface NO3 concentrations.
Regarding CanESM5 and CanESM5-CanOE, there appears to be some confusion between panels a and g. NO3 concentrations decrease more in CanESM5-CanOE than in CanESM5, while total NPP increases in CanESM5-CanOE and decreases in CanESM5.
P7 L213-214: Why do you show only NO3, PO4 and temperature at sea surface while other variables are vertically integrated? This needs to be better justified…
The reviewer is correct, a justification for this point was missing. The following sentences were added to the methodology section : “In these CMIP6 models, over our study region, there is an almost perfect correlation between projected surface NO3 concentrations and those integrated over the first 100m (not shown). This makes surface nutrient concentrations an excellent proxy for changes occurring throughout the euphotic layer. Therefore, we chose to use surface concentrations for simplicity in the following analysis.”
The following scatter plot represents the long term evolution of NO3 concentrations integrated over 0-100m relative to 1950-1970 versus the evolution of surface NO3 concentrations. These two metrics are nearly perfectly correlated in all models (r > 0.97 ; p << 0.05), thus justifying the use of surface concentrations as a proxy for what happens throughout the euphotic layer.
Evolutions of NO3 concentrations integrated over 0-100m relative to 1950-1970 vs. evolutions of surface NO3 concentrations relative to 1950-1970. The dashed lines represent the linear regressions ; the correlation coefficients and p-values can be found next to model names in the legend.
P8, L219: Fig4 -> Fig4a
This modification was included.
P8, paragraph on models without any diazotrophy:
- You refer to an observational dataset “the observed value is 0.4 mmol.m-3” but which obs dataset? Over what time period? Perhaps you could add it to Fig.3
This value is derived from the World Ocean Atlas 2018 (WOA18). In the case of nitrates, WOA18 only provides climatologies calculated from all available data across all years (refer to WOA18 documentation here). As a result, the exact time period for this value is not specified. The reference to WOA18 has been included in the text.
However, we believe adding this information to Fig. 3 would overly complicate the figure and reduce the clarity of our approach. For consistency, we would need to include observational values for other variables, which might suggest that we select models based on their ability to reproduce observations.
- For ACCESS-ESM1-5, you do not mention the increase of MLD in the projections. Is this not the reason for the increase in small phytoplankton NPP (more than the SST increase mentioned in the text)?
The reviewer is correct; the explanation provided in the text was too brief. The small phytoplankton NPP increase in ACCESS-ESM1-5 is caused by the combined effects of warming and MLD changes. The following detailed description of the mechanisms happening has been added :
“In ACCESS-ESM1-5, nitrates are not a strong limitation on average across the entire subtropical region, and the mechanisms causing NPP increase are more complex. Between 1950 and 2020, the opposing effects of rising SST and decreasing surface nutrient concentrations — partly due to mixed layer shallowing — largely balance each other, resulting in a stable NPP. However, from 2020 to 2100, a recovery in MLD slows the decline in nutrient concentrations, allowing the SST effect to dominate, which leads to an increase in phytoplankton NPP.”
- P8 L228. Replace “nanophytoplankton” by “small phytoplankton” for consistency throughout the manuscript.
Thank you for spotting this. The replacement was done.
P8, paragraph on models with explicit diazotrophy: please precise the subplot of Fig 3 you refer to. L233: primary production = total primary production?
Yes, we are referring to total primary production. Thank you for pointing this out; the clarification has been added. We have also specified the subplots in Fig. 3 in the reference.
P8 L245: For consistency of this section and previous two paragraphs, you might start this paragraph in a more general sense, addressing models with parameterized diazotrophy.
The following two sentences were added at the beginning of the paragraph : “The three models with parameterized diazotrophy exhibit contrasting behaviors in their total NPP projections. While IPSL-CM6A-LR and CanESM5-CanOE project an increase in total NPP, CanESM5 projects a decrease (fig. 3-a)”
P8, paragraph on models with implicit diazotrophy: please, precise “not shown” at the end of the sentence “However, unlike in IPSL-CM6A-LR, …”. It is not clear from your explanations why the two CanESM models (CanESM5 and CanESM5-CanOE) give such different projections of NPP. L250, does “large phytoplankton” refer to the microphytoplankton of CanESM5-CanOE?
The paragraph on models with implicit diazotrophy has been completely revised to clarify the mechanisms controlling NPP across the three models. A new panel showing the evolution of surface NH4 concentrations was added to figure 3 and we added a new figure depicting the evolution of the proportion of total NPP relying on NH4 rather than NO3 as a nitrogen source.
Yes, the 'large phytoplankton' mentioned in this paragraph refers to the microphytoplankton group in CanESM5-CanOE.
The paragraph on models with implicit diazotrophy is now the following :
“The three models with parameterized diazotrophy exhibit contrasting behaviors in their total NPP projections. While IPSL-CM6A-LR and CanESM5-CanOE project an increase in total NPP, CanESM5 projects a decrease (fig. 3-a). In IPSL-CM6A-LR, the absence of an effective control of PO4 concentrations over N2 fixation was identified as responsible for the substantial increase in diazotrophy and primary production in subtropical oceans (Bopp et al., 2022). The increase in N2 fixation replenishes the NH4 pool (Aumont et al., 2015), leading to a greater proportion of total NPP being supported by NH4 rather than NO3 as a nitrogen source (Fig. 5-a). This mechanism primarily benefits small phytoplankton, which are more efficient at consuming nutrients at low concentrations, such as NH4 (Aumont et al., 2015). As a result, small phytoplankton NPP increases in IPSL-CM6A-LR despite the decline in NO3 concentrations. The slight decrease in surface NH4 concentrations can be explained by the increased consumption by small phytoplankton. However, this process does not benefit diatoms, whose long-term growth is regulated by PO4 (Fig. 4-a), leading to a decline in their NPP. In CanESM5-CanOE, N2 fixation similarly replenishes the ammonium pool, and the same trends as in IPSL-CM6A-LR are observed for small phytoplankton NPP and surface NH4 concentrations. Although we cannot compute the proportion of total NPP based on NH4 versus NO3 for CanESM5-CanOE, due to the lack of data (Fig. 5), the similar behaviors between the models suggest that the mechanism is the same: the additional NH4 from N2 fixation supports the increase in small phytoplankton NPP. However, unlike in IPSL-CM6A-LR, CanESM5-CanOE lacks PO4, allowing large phytoplankton to also benefit from N2 fixation, leading to a rise in their NPP as well. Finally, in CanESM5, a single nitrogen pool combines both ammonium and nitrate. Despite the strong increase in N2 fixation, nitrogen concentrations in the subtropical region decrease overall (Fig. 3-g), resulting in a decline in total NPP (Fig. 3-a)”.
The explanations now draw on a new panel in Figure 3, which shows the evolution of surface NH4 concentrations, as well as a new figure depicting the shift in the proportion of total NPP relying on NH4 rather than NO3 as a nitrogen source.
P9 L268: What about diatoms? All models show a diatom NPP decrease but with groups of models showing substantially different decreases.
The spread in diatom NPP projections is substantial, even exceeding that of small phytoplankton. However, all models project either a decline in diatom NPP or a stagnation (large phytoplankton in CanESM5-CanOE). Therefore, the disagreement among models regarding the sign of the future evolution of total NPP can be attributed primarily to small phytoplankton.
P9, paragraph L272-283: here an additional figure showing the absolute value of NO3 and NH4 as a function of time is needed so that the reader can see what you are referring to in the paragraph. I am also wondering how you explain the different behaviour between IPSL-CM6A-LR and CESM2/CESM2-WACCM models regarding the projections of diatom NPP and small PHY NPP.
A new panel has been added to Figure 5, illustrating the evolution of surface NH4 concentrations, along with a new figure showing the evolution of the proportion of total NPP supported by NH4 versus NO3 as a nitrogen source. The paragraph explaining the evolution of small phytoplankton NPP in models with an NH4 pool is now the following :
“Five out of seven models project an increase in small phytoplankton NPP, despite a consistent decrease in NO3 and PO4 concentrations across models (Fig. 6). Of these five, four (IPSL-CM6A-LR, CanESM5-CanOE, CESM2, and CESM2-WACCM) include an ammonium pool. In IPSL-CM6A-LR, CESM2, and CESM2-WACCM, the proportion of total NPP derived from NH4 rather than NO3 significantly increases (Fig. 5-b). As observed in the subtropical region, this rise offsets the impact of decreasing NO3 concentrations, leading to an increase in small phytoplankton NPP, along with a decrease in surface NH4 concentrations in CESM2 and CESM2-WACCM. However, in these models, diatoms are strongly dependent on NO3 (Fig. 4), and their NPP declines accordingly. For CanESM5-CanOE, it is not possible to compute the proportion of total NPP supported by NH4 versus NO3, but it can be hypothesized that NH4 similarly sustains NPP, which would explain the slight increase in small phytoplankton NPP and the stagnation in diatom NPP. Notably, only in IPSL-CM6A-LR does the rise in small phytoplankton NPP fully compensate for the decline in diatom NPP, explaining the overall positive trend in total NPP.”
The differences between IPSL-CM6A-LR and CESM2/CESM2-WACCM can be attributed to several factors:
- A much larger decrease in NO3 concentrations in CESM2/CESM2-WACCM, leading to a more pronounced decline in diatom NPP.
- A greater shift from NO3 to NH4 as a nitrogen source in CESM2/CESM2-WACCM (as shown in the new figure). The higher consumption of NH4 in the CESM models compared to IPSL-CM6A-LR explains the larger increase in small phytoplankton NPP.
P10, L291: “were still covered” -> will still be covered
The past tense was replaced by present tense to keep a descriptive tone and ensure coherence throughout the paragraph.
P10, L292-293: sentence “However… NO3 and PO4 surface concentrations decrease in all models (fig. 6)”: But it is not the same process as in the subpolar region that causes NO3 and PO4 to decrease. They are probably more consumed due to the increase in NPP.
The reviewer is correct; the sentence is now the following : “Surface concentrations of NO3 and PO4 decrease across all models (Fig. 7-g,h), driven both by changes in ocean physics and increased consumption by phytoplankton”
P10, L289-293: What about changes in upper ocean stratification? Changes in NPP can also occur in areas that were not and will not be covered by sea-ice.
The reviewer is correct, but we don't believe this would be the dominant mechanism. All models indicate relatively small changes in MLD, and the polar region is largely characterized by sea ice coverage at the start of our study period.
P10, L296: You refer to the half-saturation constant. Is it that of diatoms?
Yes it is. The precision was added to the text to make it clearer.
P10. Section 3.4 lacks a reference to Table 2.
A new reference to Table 2 was added.
P11, L332: I would delete: “and the extreme would be kept in the subset”
This part of the sentence was removed.
P11, L335: “influence diatoms in the former but not the latter”: is it not the opposite?
The sentence was correct but the whole paragraph was vague, imprecise and did not bring useful information. We therefore decided to remove it.
P13, L390: Typo: “Emergent” --> “emergent”
Thank you for spotting it.
Fig. 1: Use the same arrangement as in Fig. 3 for the box at the bottom of the figure. The caption could be made more explicit on using the multi-member mean for each ESM.
Thank you for pointing out this oversight. The caption of figure 1 has been corrected and the models now appear in the same order in figures 1, 3, 5 and 6.
Fig. 2: Could you add the boundary between the 2 subpolar subregions (with a dotted line for example) so that the reader is aware of their extent in each model and observations?
The boundaries between the 2 subpolar bioregions are now visible in Figure 2. Thank you for your remark.
Fig. 3f: diazotrophy subplot: Please give more details about this “diazotrophy” term. Or give the equations. Is it equal to diazotroph NPP? Why is the unit different between explicit and implicit diazotrophs? It is confusing as it is.
The reviewer is correct; Fig. 3-F was unclear. The following description has been added in the caption : “In panel f, diazotroph NPP is shown on the left axis for models with an explicit diazotroph group, while N2 fixation is plotted on the right axis for models with parameterized N2 fixation”
Fig. 3g and h: Please specify that surface NO3 and PO4 concentrations are shown.
The precision has been added to figures 3, 5 and 6.
Fig. 1, 3, 5, 6: Please make these important figures larger with larger labels. Difficult to distinguish the blue colour of IPSL-CM6A-LR from that of CESM2-WACCM. Please add a title indicating the variable on each subplot to make it easier to read. Could you find a graphical way to remind the reader which model belongs to which group (no diazotrophy, implicit or explicit one) to make it easier to interpret and follow your explanations?
The three model groups are now differentiated in figure 3 : no diazotrophy : dotted lines ; implicit diazotrophs : dashed lines ; explicit diazotrophs : plain lines.
All the suggested modifications have been implemented.
Fig. 4: Instead of “correlations”, I would call it “scatterplot” and then add the correlation numbers to the subplot. Could you explain in more details how you constructed this graph? What does each cross represent? This figure needs to be clarified.
The reviewer is correct; the original version of Figure 4 was quite unclear. The figure has been refined (see figure below) and a new caption was added : “Scatter plots representing the change in diatom NPP relative to 1950–1970 versus the change in surface PO4 concentrations in the subtropical region (a), and NO3 concentrations in the subpolar region (b). CanESM5-CanOE appears only in the subpolar panel due to the absence of PO4 in the model. The dashed lines represent linear regressions, all of which are significant at the p < 0.05 level and show correlations greater than 0.96, except for CanESM5-CanOE, where the correlation is not significant (p = 0.11).”
New version of figure 4
Fig. 7: Please recall the model selection for each region in the figure caption
The caption of fig. 7 now indicates what are the selected models in each region.
Table 1: Replace “Diatoms (param.) with Diazotrophs (param.). Could you add to this table the nutrients limiting the growth of each phytoplankton class, so that we know if there is a NH4/PO4/Si… pool? The words “with a parametrization and without an explicit diazotroph group” should be removed in the Table 1 description.
The caption has been corrected, and a new column added to table 1 to indicate the nutrients limiting phytoplankton growth in each model.
References
IPCC, 2021: Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change [Masson-Delmotte, V., P. Zhai, A. Pirani, S.L. Connors, C. Péan, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J.B.R. Matthews, T.K. Maycock, T. Waterfield, O. Yelekçi, R. Yu, and B. Zhou (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 2391 pp. doi:10.1017/9781009157896
Zelinka, M. D., Myers, T. A., McCoy, D. T., Po-Chedley, S., Caldwell, P. M., Ceppi, P., et al. (2020). Causes of higher climate sensitivity in CMIP6 models. Geophysical Research Letters, 47, e2019GL085782. https:// doi.org/10.1029/2019GL085782
Citation: https://doi.org/10.5194/egusphere-2024-1820-AC1
-
RC2: 'Comment on egusphere-2024-1820', Anonymous Referee #2, 15 Sep 2024
The paper describes a method for making more robust projections of NPP than the ensemble mean normally used for CMIP6 models. The authors use a sub-ensemble of models that shows a strong coherence in the variables that controls NPP. By dividing the North Atlantic into bioregions, the best models in each region is being identified through comparisons of the differences in sensitivity to some mechanisms (diatoms, small phytoplankton, N2 fixation,...). This selection also results in more certain estimates of future carbon export.
The paper is well written and to the point. My main concern is on the statement on the strong coherence between the models based on large scale averaging of processes controlling NPP. This is an oversimplification. Yes, these are important processes, but NPP is controlled by physics on a scale (both spatial and temporal) that is not shown agree by the authors. When that is said, the authors are perfectly right on their statement on model democracy, thus studies like this to make more robust projections are of great significance. I also agree with the authors that such selection should be based both through a comparison with observations (most common) and the present process-based selection. I recommend the paper for publication after a moderate revision. This revision should include a more throughout discussion on the physical drivers, their impact on NPP, and possibly a more detailed comparison of this physics between the selected models (see e.g. https://doi.org/10.1016/j.seares.2023.102366). As it is now, the assumption on coherence is not convincing and thereby also the hypothesis on differences in model parameterization as the only cause for model discrepancy.
I have added a few minor points below.
• line 45. Perhaps one of the most important effects of climate change will be shifts in community structure. I doubt the CMIP6 models include much of this. To allow for full dynamics will require more focused trait based models (e.g. https://doi.org/10.1093/icesjms/fsy090)
• line 58. I am not convinced by this statement about coherence. Neither Figure 1 nor Kwiatkowski et al., 2020, is to a level of detail that this can be concluded on a regional and temporal level.
• line 60. With this in mind I do not agree that the discrepancies observed necessarily originates from the differences in the representation and parameterization of the biological processes. My postulation would be that the main reason for the observed discrepancies are the physics, but on a more detailed level than investigated by authors.
• line 62. An example on how to identify different physical drivers and investigate their impact on modeled primary production can be found in https://doi.org/10.1016/j.seares.2023.102366
• line 67. The hypothesis comes as a consequence of the authors conclusion about coherence. There is a need to elaborate more on the importance of physics both local and regionally.
• line 181. As said above I am not convinced about this strong coherence. Although I recognize that robustness defined as sign agreement is a reasonable metric, regional and decadal to multi-decadal variability is lost in this averaging.
• Section 4.2. It should be mentioned although NPP is important for ecosystem projections, timing and species composition is also of vital importanceCitation: https://doi.org/10.5194/egusphere-2024-1820-RC2 -
AC2: 'Reply on RC2', Stéphane Doléac, 08 Oct 2024
Response to review 2. October 2024
The paper describes a method for making more robust projections of NPP than the ensemble mean normally used for CMIP6 models. The authors use a sub-ensemble of models that shows a strong coherence in the variables that controls NPP. By dividing the North Atlantic into bioregions, the best models in each region is being identified through comparisons of the differences in sensitivity to some mechanisms (diatoms, small phytoplankton, N2 fixation,...). This selection also results in more certain estimates of future carbon export.
The paper is well written and to the point. My main concern is on the statement on the strong coherence between the models based on large scale averaging of processes controlling NPP. This is an oversimplification. Yes, these are important processes, but NPP is controlled by physics on a scale (both spatial and temporal) that is not shown agree by the authors. When that is said, the authors are perfectly right on their statement on model democracy, thus studies like this to make more robust projections are of great significance. I also agree with the authors that such selection should be based both through a comparison with observations (most common) and the present process-based selection. I recommend the paper for publication after a moderate revision. This revision should include a more throughout discussion on the physical drivers, their impact on NPP, and possibly a more detailed comparison of this physics between the selected models (see e.g. https://doi.org/10.1016/j.seares.2023.102366). As it is now, the assumption on coherence is not convincing and thereby also the hypothesis on differences in model parameterization as the only cause for model discrepancy.
We thank the reviewer for his review that significantly contributed to improving our manuscript.
The reviewer is correct in pointing out that the elements presented in this study are insufficient to claim that CMIP6 models are highly consistent in terms of their physical representations, and that the role of physics in NPP projection divergence cannot be fully dismissed. Several assertions in the original manuscript were too strongly stated, creating the impression that we viewed differences in model parameterizations as the sole source of discrepancies, which is not true. In response, the end of the introduction has been completely rephrased to temper these claims and better situate our approach within the broader context of existing techniques, such as emergent constraints and physics-based methods.
Additionally, we have introduced a new subsection in the discussion that explores the complementarity between the biogeochemistry-focused approach developed in this study and a more physics-centered approach, as demonstrated by Mousing et al. (2023). While it is impossible to entirely discount the influence of physics on the divergence of NPP projections, it is equally impossible to overlook the role of varying biogeochemical parameterizations. Therefore, we suggest that the physics-based approach proposed by the author would be a valuable step toward further reducing the uncertainty in NPP projections and would serve as a complementary extension to our study.
However, we have decided not to include a detailed comparison of the physics among the selected models, as this would constitute an entirely separate study. This could be the focus of a future article.
New version of the end of the introduction :
“Several approaches have already been proposed and shown to effectively reduce projection uncertainty. Some studies advocate for model weighting based on their ability to replicate past observations (Sanderson et al., 2017; Knutti et al., 2017), while others highlight the use of emergent constraints as a promising method for reducing projection uncertainty (Eyring et al., 2019; Kwiatkowski et al., 2017). Additionally, in-depth analyses have been conducted to explore the physical drivers of NPP, attributing model discrepancies to variations in ocean physics and their interactions with biogeochemical variables (Whitt, 2019; Xiu et al., 2018; Mousing et al., 2023). However, many of these studies tend to overlook the significant differences in the biogeochemical components of CMIP6 models (Kearney et al., 2021; Séférian et al., 2020). Yet, it is well-established that certain processes, depending on their inclusion and parametrization, can substantially influence NPP projections (Bopp et al., 2022; Rohr et al., 2023).
Building on these insights, the present study seeks to go a step further by linking part of the divergence in NPP projections to specific differences in the biogeochemical processes and pools represented across CMIP6 models. To achieve this, we will identify the processes driving long-term NPP evolution under the SSP5-8.5 scenario in each model and use these insights to assess the reliability of their projections. This process-based evaluation will then inform a model selection aimed at reducing projection uncertainties.
The North Atlantic Ocean, a highly productive region with significant future divergence between model projections (Tagliabue et al., 2021), will serve as a case study. Given that the processes controlling NPP vary across different regions, we will divide the North Atlantic basin into three bioregions (polar, subpolar, and subtropical) using a clustering method and evaluate each subregion individually. Finally, we will explore how our revised estimates of NPP projections influence carbon export and plankton biomass—key indicators of the biological carbon pump and marine ecosystem health—and how our approach integrates with the other methods mentioned above.”
New subsection of the discussion :
“Divergences in NPP projections might in principle result from differences in both biogeochemical model components and the representation of physical processes, which may either amplify of mitigate the effects of biogeochemistry. In this study, we focused on the former, largely setting aside the latter. Our focus was motivated by the observation that, unlike NPP, the evolution with climate change of key oceanic parameters known to influence NPP, such as temperature, MLD, nutrient concentrations and sea-ice cover, had the same sign across all models. Specifically, all models indicated an increase in temperature, a decrease in nutrient concentrations, a reduction in MLD, and a decline in sea-ice cover, while some projected an increase in NPP and others a decrease. However, we should point out that agreement on the sign of the evolution of these variables does not necessarily rule out their contribution to diverging NPP projections. The response of NPP to changes in MLD, for instance, is not always monotonic (Llort et al., 2019), and non-monotononicity may also result from complex interactions between physical and biogeochemical processes on seasonal timescales (Mousing et al., 2023). Future efforts are therefore needed to fully assess how differing physical mechanisms control NPP, further refining projection reliability and reducing uncertainty.”
I have added a few minor points below.
• line 45. Perhaps one of the most important effects of climate change will be shifts in community structure. I doubt the CMIP6 models include much of this. To allow for full dynamics will require more focused trait based models (e.g. https://doi.org/10.1093/icesjms/fsy090)The reviewer is correct; the affirmation line 45 was too strong. It has been modified from “Earth-system models (ESMs) from the Coupled Models Intercomparison Project (CMIP) include all of the above processes “ to “Earth-system models (ESMs) from the Coupled Models Intercomparison Project (CMIP) include most of the above processes”.
• line 58. I am not convinced by this statement about coherence. Neither Figure 1 nor Kwiatkowski et al., 2020, is to a level of detail that this can be concluded on a regional and temporal level.The reviewer is correct; Kwiatkowski et al. (2020) does not provide sufficient evidence to conclude the physical coherence of the models, and our previous assertion was too strongly stated. We therefore have removed this assertion from the revised introduction and the potential influence of physical processes on the divergence of NPP projections is now addressed in a new subsection of the discussion.
• line 60. With this in mind I do not agree that the discrepancies observed necessarily originates from the differences in the representation and parameterization of the biological processes. My postulation would be that the main reason for the observed discrepancies are the physics, but on a more detailed level than investigated by authors.The initial version of the introduction was indeed ambiguous, suggesting that we attributed all discrepancies in NPP projections solely to differences in the biogeochemical components used in CMIP6, which is not accurate. While we maintain that these components account for a significant portion of the divergence in NPP projections, the role of physical processes cannot be entirely disregarded based on the current evidence in the literature. To address this, we have completely rephrased the end of the introduction to remove any ambiguity.
• line 62. An example on how to identify different physical drivers and investigate their impact on modeled primary production can be found in https://doi.org/10.1016/j.seares.2023.102366Thank you for the suggested reference.
• line 67. The hypothesis comes as a consequence of the authors conclusion about coherence. There is a need to elaborate more on the importance of physics both local and regionally.As mentioned in our general response to this review, the final two paragraphs of the introduction were completely rephrased, as the original statements regarding model coherence in terms of physics were too strong.
• line 181. As said above I am not convinced about this strong coherence. Although I recognize that robustness defined as sign agreement is a reasonable metric, regional and decadal to multi-decadal variability is lost in this averaging.The reviewer is correct; the affirmation of model coherence was again too strong and general, making it inaccurate. The sentence is now the following: “This lack of confidence happens despite a coherence between models on the sign of the long-term evolution of MLD and surface nitrate concentrations, particularly in the subpolar North Atlantic Ocean (fig. 1-E,F).”
• Section 4.2. It should be mentioned although NPP is important for ecosystem projections, timing and species composition is also of vital importanceThe purpose of this section is not to discuss the respective impacts of total NPP, bloom timing or species composition on high trophic levels, but how biases in CMIP6 models can affect ISIMIP studies. However, the reviewer is right that species composition is key to understand how high trophic levels will be impacted by climate change. We therefore added the following precision on the connecting variables used between CMIP6 and marine ecosystem models (MEM) to specify that not all MEM rely on species composition information : “Additionally, some models rely on total NPP or biomass (Carozza et al. 2016), while others require inputs for each phytoplankton and zooplankton groups, providing information on species composition (Cheung et al. 2016)”
REFERENCES
Mousing, E. A., Ellingen, I., Hjøllo, S. S., Husson, B., Skogen, M. D., and Wallhead, P.: Why do regional biogeochemical models produce contrasting future projections of primary production in the Barents Sea?, Journal of Sea Research, 192, 102 366, https://doi.org/10.1016/j.seares.2023.102366, 2023.
Citation: https://doi.org/10.5194/egusphere-2024-1820-AC2
-
AC2: 'Reply on RC2', Stéphane Doléac, 08 Oct 2024
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 301 | 97 | 40 | 438 | 10 | 10 |
- HTML: 301
- PDF: 97
- XML: 40
- Total: 438
- BibTeX: 10
- EndNote: 10
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1