the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 21 Feb 2024
Compensatory effects conceal large uncertainties in the modelled processes behind the ENSO-CO2 relationship
Abstract. A large fraction of the interannual variations in the global carbon cycle can be explained and predicted by the impact of El Niño Southern Oscillation (ENSO) on net biome production (NBP). It is therefore crucial that the relationship between ENSO and NBP is correctly represented in Earth system model (ESMs). With this work, we look beyond the top-down ENSO-CO2 relationship in 22 CMIP6 ESMs by describing their characteristic ENSO-NBP pathways. These pathways result from the configuration of three interacting processes which contribute to the overall ENSO-CO2 relationship: ENSO-strength, ENSO-induced climate anomalies, and the sensitivity of NBP to climate. The analysed ESMs agree on the direction of the sensitivity of global NBP to ENSO, but have very large uncertainty in its magnitude, with a global NBP anomaly of -0.15 PgC yr-1 to -2.13 PgC yr-1 per standardised El Niño event. The largest source of uncertainty is the differences in the sensitivity of NBP to climate. The uncertainty among the ESMs grows even further when only the differences in NBP sensitivity to climate are considered. This is because differences in the climate sensitivity of NBP are partially compensated by ENSO strength. There is a similar phenomenon regarding the distribution of ENSO-induced climate anomalies. We show that even model that agree on global NBP anomalies have strong disagreements in the contribution of different regions to the global anomaly. This analysis shows, that while ESMs can have a comparable ENSO-induced CO2 anomaly, the carbon fluxes contributing to this anomaly originate from different regions and are caused by different drivers. The consequence of these alternative ENSO-NBP pathways can be a false confidence in the reproduction of CO2 by assimilating the ocean, and the dismissal of predictive performance offered through ENSO. We suggest to improve the underlying processes by using large-scale carbon flux data for model tuning in order to capture the ENSO-induced NBP anomaly patterns. The increasing availability of carbon flux data from atmospheric inversions and remote sensing products makes this a tangible goal and would lead to a better representation of the processes driving the interannual variability of the global carbon cycle.
- Preprint
(6007 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on esd-2024-7', Chris Jones, 06 Mar 2024
Review of “compensatory effects conceal large uncertainties in the modelled processes behind the ENSO-CO2 relationship”, by Dunkl et al.
This is a well written and explained paper exploring in some depth how the land carbon cycle responds to El Nino events and getting beneath the skin of multiple ESMs. The study breaks apart the response into the magnitude of ENSO, the spatial patterns of climate teleconnections and then the land-carbon sensitivity to these. The latter controls the majority of the spread in modelled responses.
The paper is useful both as a process study on drivers of carbon cycle variability but also in terms of guiding model groups/development plans and evaluation techniques. I recommend publication with minor revisions.
I have one major concern though, and that is the use of TRENDY model results as “observations” against which to evaluate ESMs. This is problematic in a couple of ways
- Firstly these are clearly not “observations” – they have some link to observed meteorology as they are driven by it – but the response is very much a model response
- Secondly, and maybe more important – they are not at all independent of the models you are evaluating. There is a very big overlap between the land schemes in CMIP6 ESMs and the land models used for TRENDY.
So I am afraid you simply cannot use these in the way you do now as observations.
I think this issue has a couple of solutions – depending on your appetite for further study. The simple solution is to drop TRENDY models. You have two other “observation” data sets (which are also not pure obs – as per first objection above – but they are closer to this and they are independent of CMIP6 land schemes). The paper could stand equally well using these two datasets and I don’t think the conclusions would be affected.
A more thorough, and satisfying, outcome could be to make use of the overlap and to see TRENDY results as part-way between the CMIP6 ESMs and the observations. You could even explore a pair-wise comparison for many of the TRENDY/CMIP shared land models (e.g. compare UKESM with JULES, or MPIESM with JSBACH). Where would individual TRENDY models sit on figure 4 for example? I assume they would all be at the same x-axis location (as Nino3.4 is imposed on them), but they would span the same vertical extent as CMIP models?? Can we learn anything by comparing the offline and coupled land schemes. I think in general treating TRENDY as models rather than obs is a better way forward.
Other, minor comments
- I miss any mention of other studies which have tried to constrain CMIP outcomes based on interannual variability. Jones et al (2001) did an early exploration of how a single ESM responded to ENSO and since then several studies have used this as a constrain on future behaviour – notably Cox et al (2013). Do your results have any implications for this approach?
- Re CMIP6 model selection – I would recommend caution when using multiple models which are very close variants – e.g. NoESM2-LM and -MM are essentially the same model except for spatial resolution. The land surface is identical. Likewise the various CMCC variants. Do they really add extra info to this particular study of land response? (maybe they do if the ENSO characteristics differ for example). It might simplify things to reduce the sampling to only one variant from each model family. This might feel like you are taking a smaller sample, but actually by double-sampling the same model you may skew the results.
- It is often quoted that a multi-model mean performs better than individual members (see Jones et al 2023 for a discussion on this for CMIP6 carbon cycle at regional scale). It would be interesting to see the CMIP6 multi-model mean in your evaluation as well as single models.
- Figure 3 – can you zoom in on the panels? It is very hard to read much into the results for regions other than SEA and NSA. I realise this would break the nice feature of having the same x-axis for all panels, but I think the other panels are just too small to see much clearly.
- The inverse relationship between Nino magnitude and NBP sensitivity is interesting – can you comment why you think this might come about? I cannot think of a process-reason for it – why would models with bigger ENSO have lower sensitivity? Is this an artefact of trying to cancel out errors in a model calibration stage? It would be interesting if all model groups had done that!
- I like that you split into NPP and respiration – that’s nice (also seen in Jones et al 2001). Did you think about any obs for this step? I know MODIS NPP is not perfect, but could be useful to identify spatial patterns of NPP for example even if the absolute magnitude is not reliable.
- A final comment – you discuss a lot, and very well, the differences between models and how the two ends of the responses differ. But actually I am also struck that generally most models do OK. For example my first reaction on seeing figure 5 is that generally ESM vs OBS picks up very good extent of the signal between regions. I think it would be useful to say this – that actually CMIP models are not bad. OK they differ in details, and some can be far away from the obs for some metrics. But overall the agreement is encouraging.
- Jones 2001: https://journals.ametsoc.org/view/journals/clim/14/21/1520-0442_2001_014_4113_tccrte_2.0.co_2.xml
- Cox 2013: https://www.nature.com/articles/nature11882
- Jones 2023: https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2023AV001024
Citation: https://doi.org/10.5194/esd-2024-7-RC1 -
AC1: 'Reply on RC1', István Dunkl, 11 Jun 2024
The comment was uploaded in the form of a supplement: https://esd.copernicus.org/preprints/esd-2024-7/esd-2024-7-AC1-supplement.pdf
-
RC2: 'Comment on esd-2024-7', Anonymous Referee #2, 28 May 2024
Comments on “Compensatory effects conceal large uncertainties in the modelled processes behind the ENSO-CO2 relationship”
This manuscript investigates ENSO-CO2 relationship in 22 CMIP6 ESMs by describing their characteristics ENSO-NBP pathways, and explain processes which contribute most to the overall uncertainties in ENSO-CO2 relationships among ESM. And authors find that the largest source of uncertainty is the differences in the sensitivity of NBP to climate. Overall, the manuscript is concise and clear. Here are some minor suggestions.
(1) In abstract: “look beyond the top-down ENSO-CO2 relationship in 22 CMIP6 ESMs”, how to understand the “top-down”
(2) Page2Line30: “Tropical carbon flux anomalies lag behind ENSO by three to six months (Zhu et al., 2017)”, this another paper may be a good reference here which calculate the lead-lag between ENSO and CGR/NBP. “Wang, J., Zeng, N., & Wang, M. (2016). Interannual variability of the atmospheric CO2 growth rate: roles of precipitation and temperature. Biogeosciences, 13(8), 2339-2352.”
(3) Page 4, which periods do you use for reanalysis products?
(4) Page 9, Line 170-171, you may calculate and show the spreads in ENSO-induced temperature and precipitation for each region in the plot.
(5) Page11Line 197-198, “the high NBP anomalies in GFDL-ESM4 are resulting from increased Rh”, increased Rh => reduced Rh? In Figure 6, NBP anomalies in MIROC-ESMs are nearly totally caused by Rh. Maybe need to mention it in the text.
Citation: https://doi.org/10.5194/esd-2024-7-RC2 -
AC2: 'Reply on RC2', István Dunkl, 11 Jun 2024
The comment was uploaded in the form of a supplement: https://esd.copernicus.org/preprints/esd-2024-7/esd-2024-7-AC2-supplement.pdf
-
AC2: 'Reply on RC2', István Dunkl, 11 Jun 2024
Status: closed
-
RC1: 'Comment on esd-2024-7', Chris Jones, 06 Mar 2024
Review of “compensatory effects conceal large uncertainties in the modelled processes behind the ENSO-CO2 relationship”, by Dunkl et al.
This is a well written and explained paper exploring in some depth how the land carbon cycle responds to El Nino events and getting beneath the skin of multiple ESMs. The study breaks apart the response into the magnitude of ENSO, the spatial patterns of climate teleconnections and then the land-carbon sensitivity to these. The latter controls the majority of the spread in modelled responses.
The paper is useful both as a process study on drivers of carbon cycle variability but also in terms of guiding model groups/development plans and evaluation techniques. I recommend publication with minor revisions.
I have one major concern though, and that is the use of TRENDY model results as “observations” against which to evaluate ESMs. This is problematic in a couple of ways
- Firstly these are clearly not “observations” – they have some link to observed meteorology as they are driven by it – but the response is very much a model response
- Secondly, and maybe more important – they are not at all independent of the models you are evaluating. There is a very big overlap between the land schemes in CMIP6 ESMs and the land models used for TRENDY.
So I am afraid you simply cannot use these in the way you do now as observations.
I think this issue has a couple of solutions – depending on your appetite for further study. The simple solution is to drop TRENDY models. You have two other “observation” data sets (which are also not pure obs – as per first objection above – but they are closer to this and they are independent of CMIP6 land schemes). The paper could stand equally well using these two datasets and I don’t think the conclusions would be affected.
A more thorough, and satisfying, outcome could be to make use of the overlap and to see TRENDY results as part-way between the CMIP6 ESMs and the observations. You could even explore a pair-wise comparison for many of the TRENDY/CMIP shared land models (e.g. compare UKESM with JULES, or MPIESM with JSBACH). Where would individual TRENDY models sit on figure 4 for example? I assume they would all be at the same x-axis location (as Nino3.4 is imposed on them), but they would span the same vertical extent as CMIP models?? Can we learn anything by comparing the offline and coupled land schemes. I think in general treating TRENDY as models rather than obs is a better way forward.
Other, minor comments
- I miss any mention of other studies which have tried to constrain CMIP outcomes based on interannual variability. Jones et al (2001) did an early exploration of how a single ESM responded to ENSO and since then several studies have used this as a constrain on future behaviour – notably Cox et al (2013). Do your results have any implications for this approach?
- Re CMIP6 model selection – I would recommend caution when using multiple models which are very close variants – e.g. NoESM2-LM and -MM are essentially the same model except for spatial resolution. The land surface is identical. Likewise the various CMCC variants. Do they really add extra info to this particular study of land response? (maybe they do if the ENSO characteristics differ for example). It might simplify things to reduce the sampling to only one variant from each model family. This might feel like you are taking a smaller sample, but actually by double-sampling the same model you may skew the results.
- It is often quoted that a multi-model mean performs better than individual members (see Jones et al 2023 for a discussion on this for CMIP6 carbon cycle at regional scale). It would be interesting to see the CMIP6 multi-model mean in your evaluation as well as single models.
- Figure 3 – can you zoom in on the panels? It is very hard to read much into the results for regions other than SEA and NSA. I realise this would break the nice feature of having the same x-axis for all panels, but I think the other panels are just too small to see much clearly.
- The inverse relationship between Nino magnitude and NBP sensitivity is interesting – can you comment why you think this might come about? I cannot think of a process-reason for it – why would models with bigger ENSO have lower sensitivity? Is this an artefact of trying to cancel out errors in a model calibration stage? It would be interesting if all model groups had done that!
- I like that you split into NPP and respiration – that’s nice (also seen in Jones et al 2001). Did you think about any obs for this step? I know MODIS NPP is not perfect, but could be useful to identify spatial patterns of NPP for example even if the absolute magnitude is not reliable.
- A final comment – you discuss a lot, and very well, the differences between models and how the two ends of the responses differ. But actually I am also struck that generally most models do OK. For example my first reaction on seeing figure 5 is that generally ESM vs OBS picks up very good extent of the signal between regions. I think it would be useful to say this – that actually CMIP models are not bad. OK they differ in details, and some can be far away from the obs for some metrics. But overall the agreement is encouraging.
- Jones 2001: https://journals.ametsoc.org/view/journals/clim/14/21/1520-0442_2001_014_4113_tccrte_2.0.co_2.xml
- Cox 2013: https://www.nature.com/articles/nature11882
- Jones 2023: https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2023AV001024
Citation: https://doi.org/10.5194/esd-2024-7-RC1 -
AC1: 'Reply on RC1', István Dunkl, 11 Jun 2024
The comment was uploaded in the form of a supplement: https://esd.copernicus.org/preprints/esd-2024-7/esd-2024-7-AC1-supplement.pdf
-
RC2: 'Comment on esd-2024-7', Anonymous Referee #2, 28 May 2024
Comments on “Compensatory effects conceal large uncertainties in the modelled processes behind the ENSO-CO2 relationship”
This manuscript investigates ENSO-CO2 relationship in 22 CMIP6 ESMs by describing their characteristics ENSO-NBP pathways, and explain processes which contribute most to the overall uncertainties in ENSO-CO2 relationships among ESM. And authors find that the largest source of uncertainty is the differences in the sensitivity of NBP to climate. Overall, the manuscript is concise and clear. Here are some minor suggestions.
(1) In abstract: “look beyond the top-down ENSO-CO2 relationship in 22 CMIP6 ESMs”, how to understand the “top-down”
(2) Page2Line30: “Tropical carbon flux anomalies lag behind ENSO by three to six months (Zhu et al., 2017)”, this another paper may be a good reference here which calculate the lead-lag between ENSO and CGR/NBP. “Wang, J., Zeng, N., & Wang, M. (2016). Interannual variability of the atmospheric CO2 growth rate: roles of precipitation and temperature. Biogeosciences, 13(8), 2339-2352.”
(3) Page 4, which periods do you use for reanalysis products?
(4) Page 9, Line 170-171, you may calculate and show the spreads in ENSO-induced temperature and precipitation for each region in the plot.
(5) Page11Line 197-198, “the high NBP anomalies in GFDL-ESM4 are resulting from increased Rh”, increased Rh => reduced Rh? In Figure 6, NBP anomalies in MIROC-ESMs are nearly totally caused by Rh. Maybe need to mention it in the text.
Citation: https://doi.org/10.5194/esd-2024-7-RC2 -
AC2: 'Reply on RC2', István Dunkl, 11 Jun 2024
The comment was uploaded in the form of a supplement: https://esd.copernicus.org/preprints/esd-2024-7/esd-2024-7-AC2-supplement.pdf
-
AC2: 'Reply on RC2', István Dunkl, 11 Jun 2024
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 484 | 67 | 43 | 594 | 41 | 28 |
- HTML: 484
- PDF: 67
- XML: 43
- Total: 594
- BibTeX: 41
- EndNote: 28
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1