the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
The Representation of Climate Impacts in the FRIDAv2.1 Integrated Assessment Model
Abstract. Feedbacks from the climate to other components of the coupled human-Earth system are expected to strongly influence the co-evolution of human society and its environment. Representing these feedback loops between climate and society, via the Earth system’s response to human activities and the subsequent effect back onto social systems, is essential in order to fully explore the dynamics of the coupled system. However, focus on these feedbacks has traditionally been limited, or excluded, in prior Integrated Assessment Models (IAMs) and IAM-based modelling protocols. This limits the understanding of the effects of climate change, and the response of the overall system to future emissions scenarios and policies.
The new IAM Feedback-based knowledge Repository for IntegrateD Assessments version 2.1 (FRIDAv2.1), documented and explored within this collection, seeks to address this by internalising the feedbacks between subcomponents of the human-Earth system. Within this new IAM, these connections are therefore a key part of the structure, and are documented and discussed here.
These feedbacks from the climate to human societies, conceptualised as climate impacts, are represented as global impact functions within FRIDA. Where possible, they are based on estimates from existing literature, reframed as functions of global climate variables to facilitate their representation within FRIDA. Other impact channels, with insufficient background literature to inform their structure and parameter values, are incorporated via the internal calibration of the IAM.
Since the systematic representation of climate damages within an IAM is a relatively novel endeavour, the approach is constrained by literature limitations and necessary simplifications. In addition, the high level of abstraction of the FRIDA model imposes limits on the set of impacts which can reasonably be implemented, and the level of process detail amongst those included. Nevertheless, FRIDA’s endogenous representation of climate feedbacks to human and natural systems enables valuable insights and intuition building on an underexplored topic. The general nature of the climate damage functions aggregated and documented here allows for their incorporation within other models and frameworks.
- Preprint
(1673 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-2756', Anonymous Referee #1, 22 Jul 2025
- AC1: 'Reply on RC1', Chris Wells, 24 Jul 2025
-
RC2: 'Comment on egusphere-2025-2756', Anonymous Referee #2, 03 Sep 2025
The authors have taken on a difficult task, to present a set of climate impacts linked to an existing integrated assessment model (IAM) called Feedback-based knowledge Repository of IntegrateD Assessments version 2.1 (FRIDA v2.1). The purpose aligns with contemporary scientific needs and research as there is currently substantial effort, by many, to quantify climate impacts and feedbacks between human-earth, human-environment, and human-ecological systems at a variety of scales (some of which coincide with the different h-e terms). In many cases, entire manuscripts focus on only one impact and here the authors try to weave 16 climate impacts through 6 IAM modules for which the reader may have little to know knowledge of FRIDA. The authors have set themselves a nearly impossible task, for a single manuscript, which is to demonstrate scientific contribution and novelty/significance, rigor/quality, and ensuring comprehension by the reader and reproducibility by those highly interested in the content.
Based on the criteria set forth by the journal and their use of the geoscientific model development review criteria, I do not believe the manuscript in its present form is ready for publication and major revisions or a revise and resubmit is required. Below I outline my major concern followed by some minor issues.
Major concerns
- Flow and scope. The manuscript should stand alone without having to read other documents and I did not feel like that was the case. It was also a very slow read with a weak narrative, which led to further loss of enthusiasm. It might be more interesting to reduce the presented set of impacts, highlight their contributions, and provide results that demonstrate their credibility. Consider the potential impact beyond FRIDA.
- Impact Contributions. The way the text was written it was very difficult to distill any contributions from the representation of the 16 impacts. Each impact section (e.g., Section 2.2.1) it felt like I was sifting through content (in this example four paragraphs) to find out what was done in the second last paragraph and how the impact was related to sea level rise in the last paragraph. While the front matter could have been useful, it didn’t motivate the approach used by the authors AND there was insufficient information and details about each impact approach to provide the reader with confidence or to enable replicability of the work. It would be more interesting and impactful to showcase the contribution of the impact approach first, provide information about how the impact was quantified so that others could do so similarly, and then provide additional supporting information if required.
- Paper Contributions. The authors almost immediately reference FRIDA in the introduction and the way the document is written limits the readers interest because it is written with a FRIDA focus. There are no research questions and little discussion of other IAM impact or non-IAM impact representations. It would be more informative to the reader and provide more credibility to the impact approaches to present the impacts in a way that might inspire other modellers to represent your presented impacts in their research. Sure, because you’ve integrated your impacts with FRIDA you have some constraints, but I’m sure it would be possible to suggest different types of data for inputs. As it stands, it seems like the paper is written for FRIDA users only.
- While a full blown comparison of how different approaches to representing impacts would be a huge endeavour, it would be useful to provide some comparisons against how others have represented the same or similar impacts and how those representations may lead to different results.
- Knowing it would be very difficult (let’s say nearly impossible) to retrieve validation data at a global scale, are there local or regional situations where data exist that could provide some form of validation of the impacts presented? Or could you provide at least verification of how the impacts perform under customized scenarios to demonstrate that they behave as expected and have boundary conditions? It would certainly add to the credibility of the presented impacts and therefore the likelihood that they would be used or replicated by others.
Minor Comments
Line 52 – ‘This collection’ – what collection? There are examples like this throughout the manuscript where it seems like it is assumed the reader has knowledge that the authors have. Minimize that assumption and increase the inclusivity of the text. Another easy example here is the text on page 5 where it is written “All six non-climate modules in FRIDA …”. It would be better to introduce this to the reader as “There are six non-climate modules in FRIDA…”, for example, or some other way – hopefully my point is clear.
Line 52 – there is some switching between components and module when describing FRIDA, which was confusing – especially out of the starting gate. Please revise and look to increase consistency throughout.
Line 54 – ‘These climate impact interactions…’ Again, an unclear reference to impact interactions not explicitly mentioned in the previous two sentences. Overall, watch the backward referencing with ‘these’, ‘this’, and ‘those’ without explicit labels used before and alongside the backward references. It forces the reader back to previous text and slows the reader down.
Line 60 – “Climate impact channels” – generally fine, but not a widely used term that I’ve seen. Would be useful to describe what a channel is for the reader.
There is a lot of integration of table and figure references in the text as nouns or focal points/subjects. Using table and figure references in this way degrades the quality of the writing and makes it boring like a textbook. Instead the focus should be on making meaningful and impactful statements and a strong narrative that is supported by tables and figures.
Table 1 – It would seem the table would be more useful, and the research more replicable and transparent, if the actual equations used were included (where possible) in the table.
Line 175 – “… uncertainty was re—sampled” How was this done? How did your resampling deal with covariance and “capture the impacts of covariance” (line 177). Please provide more details.
Section 2.1.1 – It seems like a heavy emphasis on Yalew et al. (2020). Are there other sources to provide additional support?
Several sentences in the document start with “As discussed/detailed in Section #.#.#...” If it was already discussed then it probably doesn’t need to be discussed again. Regardless, every time you start a sentence off like this then the reader starts to think or look back in the document, which slows them down. Make the non-redundant statement and then provide support via citation/reference, which may include another section. Suggest reducing redundancy and limiting those types of references throughout.
Line 321 – HFC not spelled out in full in the manuscript.
Line 599 – “… with FRIDA implementing a process-based, bottom-up approach.” There is little if any evidence that FRIDA works this way in the manuscript. There is no discussion of heterogeneity or variation, aggregation, interaction or other similar themes typically found in bottom-up based approaches (whether equation-based, agent-based, or some other form). Without some actual description of FRIDA in the paper these cursory comments about FRIDA feel unsupported and distracting. If more description of FRIDA was provided and then in this case the Land Use and Agricultural components were provided in more detail then this section would be substantially more interesting. Again, especially if some corresponding results were provided later.
Line 663 – would love to see some more details about the management activities.
Line 689 – Please don’t start a section off with “as discussed in some other section.” Substantial writing improvement to engage the reader is needed.
Line 766 – “… for several important channels this was not possible.” – which ones? Give the reader the specifics.
Line 806 – here you talk about the short runtime of FRIDA but you haven’t given any quantitative data to support it or any comparisons with other software. Therefore, it’s meaningless since your thoughts about length of runtime will most likely differ from the readers.
Citation: https://doi.org/10.5194/egusphere-2025-2756-RC2 -
RC3: 'Comment on egusphere-2025-2756', Anonymous Referee #3, 07 Oct 2025
The study by Wells et al focusses on the representation of climate impact channels in the FRIDA model. It introduces the sectoral damage representations, damage functions, and discusses limitations and opportunities for future development. Fully appreciating the implications of those decisions requires a comprehensive understanding of the full FRIDA model. I studied also Schoenberg et al (in discussion) to get a better feel for how these impacts may represent model dynamics, but some parts remain rather elusive to me.
I want to start by applauding the authors for a massive effort. I enjoyed reading the manuscript and find the approaches well explained and state of the art.
However, I have a few remarks relating to the overall approach as well as to some sector specific outcomes.
- The choice of the impact channels. To some extent the impact channels you model are the ‘easy’ ones, many of which are represented at least partly in complex IAMs (i.e. energy system impacts, cooling needs, crop yields, etc). Not all, of course, but many. Which raises a bit the question of where the major innovation lies in the FRIDA coupling. And then there are some surprising omissions, such as forest fires, that I would have expected to feature more prominently. Also, the representation of damage functions is straight forward and mostly linear. Which is how it is in most models, but when building a model from scratch today, I would have given some thought on how to detect potential non-linear thresholds etc.
- I would have also liked to see a sensitive testing to some of the assumptions to understand which impact channels are the most impactful ones in terms of full century trajectories (I’d suspect the least well constrained ones on government spending and indirect effects, but can’t proof this
- This point relates to the FRIDA development more generally, but is also relevant here. There are models of this class out there, such as the FELIX model developed for more than a decade now. They attempt to do very similar things, and have very similar limitations. Yet, there are not even cited nor is some comparison performed. This is a bit unfortunate, because quite some learnings could have been taken from this class of models also in relation to the specific FRIDA innovations (i.e. in relation to the probabilistic modelling).
- Some of the impact channels remain a bit opaque to me. The first one that comes to mind is the demographic one. From what I understand it’s only mortality that’s modelled, but for a model of that class, other channels, in particular on education may proof much more relevant. I understand from Schoenberg that the female education – pop-growth channel is explicitly modelled. Which means it could be really quite relevant to explore climate impacts on this for 21st century population growth. It seems FRIDA pop growth is close to SSP3 – but with limited variance compared to other variables.
- Similarly, the economic damages modelling is not very clear to me. The authors want to endogenize economic damages, but then assume so fairly simplistic scaling functions of government spending etc. From Schoenberg et al. Fig 10 I take it that economic growth in FRIDA is at the low end of the SSP scenarios. Is this because of climate impacts, or just the economic module of FRIDA? And how do the endogenized impacts compare to other damage functions? I think it would be very useful to understand how FRIDA’s economic performance would look like with and without climate damages and how it compares to macro-economic damage functions in the literature.
- Treatment of adaptation and adaptation costing: I find it reasonable that if the authors had to pick one sector for adaptation it should be sea level rise. Yet, for near-term economic and societal damages, arguably heat related mortality and morbidity all the way to labour productivity are probably more impactful. And somewhat of a miss that adaptation is not coupled to macro-indicators resembling adaptive capacity. The authors cite the Andrijevic et al work, and from what I take their model would fully allow them to endogenize the adaptive capacity modelling also. I understand it’s an avenue for future work, but could be useful – coupled with the government spending module, for example.
Minor comments:
Table 1: Entry for HDD. C&P mistake in the description. I assume it’s “heating energy” that’s required here.
L540: I’m not sure this comprehensive enough in terms of ‘indirect economic effects’. Expectations is an important part, but other dimensions link to dampened economic activity as ripple on effects of destroyed assets or infrastructure e.g. linked to extreme weather events. This may be subsumed here, but it also wouldn’t hurt to call it out a bit more concretely.
L700: I am a bit confused that impacts on forest fires, are not more prominently explored, and that the carbon cycle response is probably largely the representation in FaIR. In such a coupled whole-system model, they strike me as some of the key feedbacks one would like to capture.
PS:
Apologies to the authors for taking so long to complete my review. It’s a very busy time of the year and this a very comprehensive piece of work. I commend their efforts and want to express my regrets of slowing down the review process.
Citation: https://doi.org/10.5194/egusphere-2025-2756-RC3 - AC2: 'Reply on RC2 and RC3', Chris Wells, 23 Oct 2025
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,885 | 68 | 25 | 1,978 | 30 | 32 |
- HTML: 1,885
- PDF: 68
- XML: 25
- Total: 1,978
- BibTeX: 30
- EndNote: 32
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
"Estimates of the total impact of climate change often attempt to measure or model the overall impact on global Gross Domestic Product." The typical impact is a Hicksian Equivalent Variation, rather than an income or output effect.
You can't cite Burke without also citing the by now many papers that make mince meat of that paper.
The set of included impacts is haphazard. Most impacts are based on a single study, ignoring most of the literature. Some impacts are based on primary research, of hair-raising quality. The sections on indirect economic effects and government spending would make an undergraduate blush.
All impacts ignore Schelling (1984). Why on Earth would impacts be a function of climate change and climate change only? What happened to human agency, technological progress, institutional change?