the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 15 Sep 2026
Direct human forcing for earth system and climate impact modelling under CMIP7 scenarios
Abstract. Direct human forcings are important input variables for earth system and climate impact models, describing societal influence on changes to climate and related risks. Here, we present a new multisource dataset consistently describing the spatial and temporal dynamics of the two most important direct human forcings: population and economic production. The dataset was created by harmonizing multiple datasets to achieve smooth transition both the historical (1850–2025) and future (2026–2100) time periods, and to cover many resolutions down to 1 km. Projections were produced for the latest Shared Socioeconomic Pathways (v3.2, 2025), applicable to the seven scenarios designed for Coupled Model Intercomparison Project Phase 7 (CMIP7) earth system models, which require population density as one of the inputs. In a further stage of CMIP7, the dataset will be applied to modelling (societal) climate impacts, particularly in the Inter-Sectoral Impact Model Intercomparison Project 4 (ISIMIP4). This requires additional variables such as population split by degree of urbanization, gross domestic product and fixed asset value, and high resolutions of the data. We show that the new data vastly improves upon quality and availability of direct human forcings compared to previous CMIP phases. It enables, for instance, process-based impact simulations across sectors, impact attribution research, risk assessment and socio-hydrology.
- Preprint
(4245 KB) - Metadata XML
-
Supplement
(131 KB) - BibTeX
- EndNote
Status: open (until 11 Nov 2026)
- RC1: 'Comment on egusphere-2026-5045', Anonymous Referee #1, 12 Oct 2026 reply
-
RC2: 'Comment on egusphere-2026-5045', Anonymous Referee #2, 23 Oct 2026
reply
This study presents a timely and valuable dataset of population and economic variables for Earth system and climate impact modelling under CMIP7 scenarios. The effort to harmonize multiple data sources across historical and future periods, while providing high-resolution spatial information, is a useful contribution to the modelling community. I appreciate the extensive work involved in developing the dataset and comparing it with available reference datasets. My main comments focus on clarifying some methodological choices and discussing the scope and limitations of the validation. Please see the detailed comments below:
Major:
1. Section 2.3.1. It would be helpful to provide some context for choosing OpenStreetMap (OSM) administrative boundaries over alternatives such as GADM, GAUL, or geoBoundaries. Could the authors also describe how consistency between the OSM boundaries and the regions in the subnational GDP data was checked, and summarize the nature and extent of any manual adjustments? Finally, a brief discussion of any differences between the OSM boundaries used here and the GADM level 2 boundaries used in FuturePop, their potential implications for the projected population distribution, and how they were handled would help clarify the approach.2. Lines 283–288 (extrapolation of built-up area). It would be helpful to clarify and discuss the assumption that built-up area scales with population change. GHS-BUILT-S includes both residential and non-residential surfaces, whose development may follow different trajectories. For example, industrial and commercial expansion may be driven by economic activity or policy decisions rather than local population growth. The restriction preventing built-up area from declining addresses the persistence of existing infrastructure, but does not account for these differences in expansion. Given the use of built-up area in GDP and fixed asset allocation and urbanization classification, acknowledging this assumption and its potential implications would help readers interpret the dataset, even if the magnitude and direction of its influence remain uncertain.
Please also clarify whether the population-change ratio used to extrapolate built-up area is calculated for individual grid cells or larger spatial units. If calculated for individual cells, how are built-up cells with zero or very small baseline populations handled, given that the ratio may be undefined or unstable?
3. Sections 2.5 and 3.4 (Validation). The comparisons with several reference datasets are valuable. However, differences in their temporal coverage, spatial units and construction methods would benefit from clearer discussion to help readers interpret the validation results. In particular:
Please clarify the extent to which the reference datasets are independent of the inputs used in this study. The European grids for 1960–2010 and HANZE both draw on Paprotny and Mengel (2023), while the census-based grids may share underlying data with GHS-POP or WorldPop. Common authorship alone does not imply dependence, but shared inputs or processing methods could contribute to agreement. A brief discussion of any such overlap would help clarify what these comparisons demonstrate.
The validation is limited to Europe, where relatively extensive data availability may support better performance. It would be helpful to discuss how differences in data availability and quality may affect the applicability of the European validation results to other regions, and how future validation beyond Europe could strengthen the assessment.Minor:
4. In Fig. 5a, the SSP1 trajectory appears to overlap with SSP5, making it difficult to distinguish. Could the authors make both trajectories more visible, for example by using different line styles or widths, and clarify in the caption if they nearly coincide?5. Suggest providing the full name when the acronyms first appear, like GADM
6. Line 246: Could the authors clarify the rationale for constraining relative subnational GDP per capita to 50–200% in 1850? It would be helpful to explain whether these bounds are informed by historical evidence or previous studies and to report how many regions are affected.
Citation: https://doi.org/10.5194/egusphere-2026-5045-RC2
Data sets
Exposure datasets at multiple scales Dominik Paprotny https://doi.org/10.5281/zenodo.17572467
Model code and software
Exposure datasets at multiple scales Dominik Paprotny https://doi.org/10.5281/zenodo.17572467
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 468 | 135 | 50 | 653 | 49 | 52 | 43 |
- HTML: 468
- PDF: 135
- XML: 50
- Total: 653
- Supplement: 49
- BibTeX: 52
- EndNote: 43
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This study presents a dataset of spatially resolved population and GDP over the 1850-2300 period that stitches together a number of existing datasets at various resolutions (while also newly developing some components), harmonizing them and filling gaps, and is in particular consistent with the most recent, updated versions of the national-level population and GDP projections for the SSPs. This is an important dataset that is sure to have wide use (and indeed an early version has already been put to use), and the paper is largely clear, well-organized, and informative in terms of methods and results. It represents a substantial research effort.
I have three larger comments, followed by a number of more specific issues.
The first regards the framing of the paper. The title and abstract state that the paper is about (1) direct human forcing for (2) earth system models, as well as other uses. But the paper is really only about population and GDP (as well as the related quantity of fixed assets), and their use in ESMs is quite limited. Direct human forcing is a large category that includes not only population and GDP but also land use and a number of types of agriculture and water management (irrigation, fertilizer use, etc.). Readers expecting a paper on the direct human forcers consistent with the CMIP7 scenarios will be surprised to find only a subset of that category here.
As for use in ESMs, GDP or fixed assets is not used at all, and the use of population is quite limited -- its only use is in models that have interactive fire models in which population density is an input.
Further, what is done here is in most cases not to develop new scenarios, but to take existing scenarios and historical data and harmonize them. The spatial population projections for the SSPs, for example, already exist. GDP downscaling from the national level is indeed carried out here and not adopted from elsewhere, and the fixed assets projections are new, but these distinctions are lost in the current framing.
A more accurate and modest title and abstract would include a more straightforward statement of what was done. It can be stated in the abstract and introduction that an important use of these products is as part of a larger package of direct human forcings used in climate research that includes ESMs.
Further comments on this topic are included below.
The second larger issue relates to clarifying the contribution of this paper compared to existing publications and datasets. The abstract and introduction lack clarity on this point. For example, there are essentially no citations in the introduction of the underlying projections and datasets that are put together to create this harmonized dataset. The bullet points on lines 50-57 stating the principles of the dataset do not contain this information either. The reader is left to slowly work out over the course of the paper what is new here and what is pre-existing.
Third, the treatment of urban land area (built up area) projections (section 2.3.5) seems unnecessarily simplistic and will be inconsistent with urbanization projections for the SSPs. There are quite a number of gridded projections of built up area based on the SSPs, as recently evaluated in van Vliet et al (van Vliet, J., Yang, H., Benz, N., Cheng, C., Doelman, J., Gao, J., Huang, Q., Koomen, E., Li, X., Niu, L. and Schrammeijer, E.A., 2025. Hotspots of disagreement across global urban land projections until 2100. Geography and Sustainability, p.100403). While these projections are not based on the most recent SSP population projections (at least I assume not), they would still give a much better basis for future projections, differentiated by scenario, than simply assuming urban land scales with population growth, as is done here. Further, updated projections of the percentage of population that is urban have been produced (based on the population projections used here, and including outcomes based on Degree of Urbanization) by Leiwen Jiang; at a minimum these would provide a better basis than total population for any type of extrapolation of urban area. For some reason IIASA has dropped these projections from its database of updated basic elements of the SSPs (https://ssp.apps.ece.iiasa.ac.at/basic-elements), even though the projections were previously posted and publicly reviewed. It would be worth identifying the status of these projections.
More specific comments:
abstract
25-26: the assertion that these new population and GDP data "vastly" improve on quality and availability compared to previous CMIP phases seems a bit overstated, and there is no discussion in the paper that provides the reasoning that supports the statement.
introduction
29-38: This paragraph blurs distinctions between the concepts of forcing as used in various communities in a way that is likely to further confuse this topic. In CMIP7 forcings are "climate forcings". They are not inputs to modeling the "consequences" of climate change, as is stated in the first sentence, but rather to modeling climate change itself. They are also not drivers of historical and future variability of climate "and its impacts," as is stated in the second sentence, but rather of climate alone. The CMIP7 paper cited in support of these claims (Durack et al., 2025a) is clear about this: "Earth system changes result from natural and human modulation of atmospheric constituents (solar irradiance changes, emissions of reactive gases and aerosols, volcanic eruptions, greenhouse gases), land use, and other Earth system component changes. These natural or anthropogenic drivers are termed “forcing” agents because they drive (i.e., force) Earth system change. " The paper does not describe climate forcings as drivers of impacts on society or ecosystems, only of climate.
Starting on line 34, the paragraph changes to talking about drivers of impacts (or risks), distinguishing climate-related forcers of these risks as well as direct human forcings. This is fine, but the the distinction between impact forcers and climate forcers needs to be made very clear. The term "direct human forcing" needs to be treated carefully as well; it is essentially unique to the ISIMIP project and shouldn't be treated as widely used and understood outside of it.
Further, on line 34 in this paragraph, it is stated that population density is a climate forcing. As noted above, technically this is true, and it is listed in a table of forcers in Durack et al (2025a). But it is misleading to say that it is used "for instance" in fire modeling. It is used *only* in fire modeling, and only in those models with interactive fire that use population density as an input (if this is wrong, then additional instances should be listed in the paper). This relates to the larger point made above about the framing of the paper.
line 57: what "economic model" is referred to here? please clarify
Starting on line 47 and continuing through the end of the introduction, the text describes what is done in this paper but leaves out the fact that updated population and GDP projections for ScenarioMIP-CMIP7 already exist at the national level and, for population, have already been downscaled to the grid level. These existing products should be cited and the additional contribution made here should be clarified.
While it is true, as stated on line 58, that this is the first high resolution global dataset of population and GDP using the updated SSPs, the introduction should mention that there is already a spatial population projection for the UK/European region for the updated SSPs:
Bonatz, H., Nikoletopoulos, T. and Vafeidis, A.T., 2026. Updated spatial population projections for the EU and the UK under the shared socioeconomic pathways provide new insights for climate impact assessments. Earth's Future, 14(3), p.e2025EF006504.
Methods
In the beginning of section 2.1, and in figure 1, it is unclear whether the external gridded datasets (line 75-76 and "gridded level" in the figure) are for historical years only or also include projections. In general figure 1 would be clearer if years were given for all rows in the right-most column.
In section 2.2, the text on GDP is quite confusing. It cites Dellink et al (2017) as the source of the GDP projections, meaning these have not been updated since the original SSPs were produced, yet earlier in the paper it is stated that this is the first gridded projection based on the latest SSP elements. Has an updated reference been forgotten? Even if there is no paper currently available to cite, some statement about the status of the source of these projections is required.
More broadly, this section on the GDP and population trends is confusing and hard to follow (especially for GDP). I suggest a summary table that gives key features of the population and gdp outcomes for the different scenarios, for different parts of the world, and then base the text around that. It probably woud also be useful to discuss population and education outcomes first, since they are then referred to in the GDP description.
130-131: what are the "updated WIC2023" and "WIC2013" projections? What has been referred to so far is only the 2014 SSPs and an update by KC et al (2024).
Also, what is the relationship between the national population and GDP data downscaled here and the official SSP data being used in the IAM scenarios (IAM scenario database here: https://scenariomip.apps.ece.iiasa.ac.at/)? Users will want to know if these are consistent, since other types of direct human forcing will be coming from the IAM simulations.
Related to the larger issue raised above about the contribution of this work, in section 2.3.2 it comes as a surprise to find that the spatial population projections already exist (Bondarenko et al., 2025a and b), as do the historical population grids, and this work provides the service of stitching them together. This is a perfectly acceptable and valuable service, but it should be made clearer, earlier in the paper, what is done here versus what already exists.
206-223: This description of the future gridded population projections is somewhat confusing partly because it seems redundant in places with lines 180-186, which describe the 2015-2030 WorldPop product. More importantly, it does not cite Bondarenko et al 2025b, the source of the projections for 2030-2100, so it is unclear if what is being described is what was done for that publication or for this one. Finally, since the WorldPop product extends to 2030, it is unclear when the text talks about extending beyond 2025 whether it is referring to WorldPop's results for 2025-2030 or the FuturePop product.
line 226: temporarily should be temporally
In section 2.3.4, fixed asset value is introduced as a new national-level time series developed here (as well as a downscaling to gridded representation). This should be distinguished from those series that are not developed here, in the introduction as noted above.
In figure 4, it is worth noting the reason for the sharp decline in available data in recent years. Presumably this is just the lag in data becoming available, but it is worth pointing out for clarity.
Results
line 343: "relentless" introduces a value judgement, a more neutral word should be used
344: figure 4 should be figure 5
In figure 5, and the text referring to it, sources for the projections in panels a-c should be provided.
In figure 7, the confounding influence of population growth makes it hard to draw conclusions about the economic component of GDP change in these maps. I recommend switching this figure to per capita terms, and then moving the total GDP figures to the supplementary material (the reverse would also be a possibility). Both could then be discussed in the main text.
line 416-417: what values of urbanization in the SSP database are being referred to? A particular year? region? time series?