NorESMhub
Question on general CMIP inquiry of mass balance
PhD Gang Tang from the University of Melbourne, Australia sent this e-mail to a number of CMIP6 model groups early December. Neither myself or anyone else replied at the time. I do not know if it is still relevant, but I will reply to the author with a link to this issues The topic seems to be on carbon cycle so I have suggested Mats and Joerg as assignees.
Dear Earth System Model Teams,
I hope this email finds you well. My name is Gang. I am a PhD researcher at The University of Melbourne and the Max Planck Institute for Biogeochemistry. I work on the coupled carbon-nitrogen cycle in the MAGICC reduced-complexity model. I am reaching out to inquire about mass conservation checks in the CMIP6 data submission process and to discuss potential solutions for identified issues.
The email is a bit long. In short, I lay out the issue I identified in our paper (link here). Then, I ask a few questions. The point of considering these questions now is to make life for CMIP7 easier through my role on the data request team. If you have any time, I would greatly appreciate your thoughts. The better we do this now, the fewer headaches there will be at the crunch time of submitting model simulation results.
The Mass Conservation Problem: Over the past two years, I have been analyzing CMIP6 data, focusing on C4MIP variables, and have found carbon and nitrogen mass conservation issues in the reported data for many experiments and models. These issues arise for both the total land pool and the subpools such as vegetation, litter, and soil (details in our preprint).
Our analysis suggests that the problem primarily stems from data reporting inconsistencies—either incomplete reporting or flux definitions that differ from C4MIP guidelines. Land-use fluxes and fire emissions are likely contributors to the carbon conservation issue, but even accounting for these, some models remain unconserved. For nitrogen, the situation is more complex due to the fluxes interacting between organic and mineral nitrogen pools. These issues hinder data usability and could compromise analyses like carbon cycle feedback quantifications.
Challenges from MIPs and the CMIP Data Request: As part of the CMIP7 Data Request team (land and land ice), I recognize that some problems may arise from the Data Request itself. For example, for the carbon fluxes, LUMIP and FireMIP request some variables similar to C4MIP (i.e., different variables share the similar/identical definition), which could cause confusion for modeling groups.
Currently, there is no mandatory requirement for modeling groups to submit all variables when participating in MIPs, which is understandable given model differences. However, this means a "black box" for data users - It is impossible to know if the variable is not modeled, not reported, or reported under other variables. Proper documentation of missing fluxes and their causes is the only way I can see to address this, but perhaps there are other ways.
Our Questions: To address these issues, we kindly ask for your input on the following:
Question 1: Does your modeling group currently perform mass conservation checks for CMIP data submissions? If not, would you prefer to implement them internally, or would a general purpose validation tool for quality checks be preferable?
Question 2: Regarding the CMIP7 Data Request (e.g., v1.0beta), have you encountered challenges in reporting variables for different MIPs (e.g., duplicates, inconsistencies)? Would your team be willing to provide comments on challenging variables, such as whether they are unsupported by your model or need redefinition for clarification?
Question 3: For the current CMIP6 data conservation issues, do you have any insights into potential causes or solutions?
The Plan for Moving Forward: As the first step, this email aims to gauge interest from modeling groups in addressing these issues. If your group is interested, a response to question 1 (and, if possible, question 3) would be greatly appreciated.
Later, I will compile a list of CMIP7 Data Request land carbon-nitrogen variables categorized by MIPs. We will need your input on whether these variables are reportable or require adjustments for your specific model. The key point is to identify the incompatibility of model outputs with MIPs.
Finally, I plan to summarize these challenges and present them to the MIPs and CMIP7 Data Request team, with the aim of developing a harmonized data request from MIPs that is convenient for model teams and ultimately resolves the mass conservation issue.
We sincerely appreciate your time and efforts and look forward to your responses!
Best regards, Gang On behalf of the author team Tang, G., Nicholls, Z., Jones, C., Gasser, T., Norton, A., Ziehn, T., Romero-Prieto, A., and Meinshausen, M.: Investigating Carbon and Nitrogen Conservation in Reported CMIP6 Earth System Model Data, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2024-3522, 2024.
This is mostly about the land carbon cycle, so I have added Rosie and Kjetil
This is mostly about the land carbon cycle, so I have added Rosie and Kjetil
Yes, I saw the same. I don't have the answers to the three questions now, but we should be able to respond to this early in the new year.
Hi all. @kjetilaas and i just talked about this 9sorry for the delay) and we thought that our response to the questions would be:
- we don't really have the time and resources for digging back into the NorESM2 C&N balance check issues. This is also partly because none of us who are here now was responsible for this reporting, and so it would be a lot of work and we are oversubscribed.
- We don't have direct experience of the data request process (again, this is due to personnel change) so can't comment on this.
- We are indeed interested in thinking about how to avoid these mass balance problems in CMIP7. Experience with other MIPs that have imposed mass balance constraints is that there are substantial definitional issues to overcome. There is a danger of delaying all of CMIP until we have all understood each other's models! The sooner the parameters of the balance check definition were known the less likely this would be to happen. (hence apologies for the delay in responding!)
@oyvindseland do you want me to reply to the email or do you want to do it? I am fine either way.
@rosiealice Please feel free to respond although I sent the link to this discussion to the original author. I have not seen any replies in this thread, so unless the original author has more comments I think your reply is fine.
@oyvindseland @matsbn @rosiealice @kjetilaas @JorgSchwinger
Dear NorESM team,
Thank you all for your kind responses. I apologize for not replying sooner; I did not receive a notification for the issue comments, otherwise I would have responded earlier.
It is completely understandable that, due to personnel changes, there is no direct answer regarding the CMIP6 data. Our focus is more on how to make the data more reliable and readily usable, particularly with respect to ensuring basic mass conservation, for the upcoming CMIP7 (and future CMIP phases). From your replies, I believe I have already found the key takeaway: establishing a well-defined mass conservation constraint is essential. I fully agree with this approach.
In the near future, I will follow up on the email I previously sent with a summary of responses from the various ESM teams and outline the proposed path forward. Please stay tuned for that.
Thank you again for your valuable responses.
Best regards, Gang
Gang Tang (he/him) PhD Researcher (2022.09-), School of Geography, Earth and Atmospheric Sciences The University of Melbourne, Melbourne, Australia Guest PhD Researcher (2024.09-), Department Biogeochemical Signals Max Planck Institute for Biogeochemistry, Jena, Germany Mobile: (+61) 0466352141 or (+43) 06602738543 E-mail: [email protected] [email protected] Google Scholar: Gang Tang LinkedIn: Gang Tang
@JorgSchwinger @tjiputra @DirkOlivie @kjetilaas @maritsandstad
Hi
I received a reminder about the question added in this thread, e-mail below. No-one was able to answer earlier. Any takers?
Øyvind
Dear ESM groups,
This is a follow-up regarding the mass conservation issues identified in the CMIP6 published data.
First, thank you to all groups who responded to our first email. We greatly appreciate your feedback and are pleased to learn that many models already include internal mass conservation checks.
In this email, we first define the top-level mass conservation equations for the land carbon-nitrogen cycle. Then, we provide further explanations for the C4MIP-requested variables, aiming to minimize confusion around variable definitions—specifically, by clarifying what should be included in each variable. Finally, we again seek feedback from the ESM groups on whether anything remains unclear in the mass conservation validation.
The technical details are included at the end of this message. After reviewing them, we kindly ask your input on the following key questions:
Regarding fDeforest (emission from land cover change) and fHarvest (emission from land management): Are the distinctions between these two fluxes clear now for model outputs?
Regarding fire emissions: Which is preferable for the models: (a) include anthropogenic fires in fAnthDisturb and natural fires in fFireNat, or (b) define total fire emissions as fFire = fFireAnth + fFireNat, separating the components explicitly?
Are there any remaining questions/confusions regarding the five top-level conservation equations?
We hope that with these clarifications and the refined variable definitions, we can collectively resolve the mass conservation issues in the data for the upcoming CMIP7. Thank you again for your continued engagement and valuable contributions.
Best regards, Gang Co-lead, Land & Land-Ice Theme, CMIP7 Data Request Author Team With support from C4MIP
The following is the technical part.
Eq. 1: cLand = cVeg + cLitter + cSoil + cProduct
This is the basic check to make sure the pool size data are good. No need for further explanation.
Eq. 2: dcLand / dt = npp - rh - fAnthDisturb - fFireNat - fProductDecomp - fLandToOcean OR dcLand / dt = npp - rh - fDeforestToAtmos - fHarvestToAtmos - fFireNat - fProductDecomp - fLandToOcean
These two equations are essentially the same. We found that many ESMs do not report fAnthDisturb (Tier-1 variable), instead, the Tier-2 variables fDeforestToAtmos and fHarvestToAtmos are reported. The fAnthDisturb is the carbon flux from all the anthropogenic activities that goes into the atmosphere. It is designed for those models which cannot separate deforestation, harvest, (anthropogenic) fire, etc.
The current definition of fDeforestToAtmos: “Deforested Biomass That Goes into Atmosphere as a Result of Anthropogenic Land-Use Change.” What should actually be included is “Carbon mass flux from all the land cover change, excluding those caused by fire (see later justification), that goes straight into atmosphere, for example, emission from deforestation.”
The current definition of fHarvestToAtmos is “Harvested Biomass That Goes Straight into Atmosphere as Carbon Mass Flux.” What should actually be included is “Carbon mass flux from all the land management, excluding those caused by fire, that goes straight into atmosphere, for example, emission from harvest and grazing.”
The key distinction is: fDeforest focuses on the land cover change and fHarvest focuses on the land management, which align with the C4MIP paper description but here we make it clear in the variable definition.
There is a conflict regarding fire emissions in the previous data request. We noticed the inclusion of a variable fFire from the scenarioMipBaseline, defined as: “Carbon Mass Flux into Atmosphere Due to CO2 Emission from Fire Excluding Land-Use Change.” This variable is not part of the original C4MIP request, yet many ESMs have reported it. In C4MIP, the initial intention was to include anthropogenic fire emissions within fAnthDisturb (or within fHarvestToAtmos for managed agricultural fires), with fFireNat reserved for natural fires.
The presence of fFire introduces potential confusion and risk of double-counting if anthropogenic fire emissions are also reported under fAnthDisturb. To address this, we are considering three options and would appreciate input from ESM groups on which is more practical:
Maintain the current C4MIP definitions: All anthropogenic fire emissions are reported under fAnthDisturb; all natural fire emissions under fFireNat.
Add fFire to the C4MIP request: In this case, fAnthDisturb excludes all fire emissions (like the clarification we made above for the fHarvest and fDeforest); fFire refers to anthropogenic fire emissions; and fFireNat to natural fires.
If needed, we may introduce a new variable fFireAnth to clearly separate anthropogenic emissions, with fFire (revised to represent total fire emissions) = fFireAnth + fFireNat, recognizing that the variable name fFire alone may already be ambiguous.
Eq. 3: nOrganic = nVeg+nLitter+nSoil+nProduct
Similarly, to make sure the pool size data is good.
Eq. 4: dnOrganic / dt = fBNF + fNup - fNnetmin - fNAnthDisturb
There should be no ambiguity regarding this equation. We would like to reiterate that fNAnthDisturb is a composite flux encompassing all anthropogenic nitrogen emissions that exit the organic pools and are directly released into the atmosphere—for example, through deforestation, harvest, product decay, fire, and other processes. Note that we currently do not have a variable representing natural fire nitrogen emissions. If the above mentioned separation of fire carbon emissions is considered favorable, a corresponding separation for fire nitrogen emissions will also be requested.
Eq. 5: dnMineral / dt = fNdep + fNfert + fNnetmin - fNloss - fNup OR dnMineral / dt = fNdep + fNfert + fNnetmin - fNgas - fNleach - fNup
Similar to the carbon flux variables, many ESMs report fNgas and fNleach instead of the Tier-1 variable fNloss. Both approaches are acceptable. We would like to emphasize, however, that regardless of which variables are reported, the corresponding mass conservation equation should be satisfied.