| 04 Aug 2026
Limited reducibility of CMIP6 ensemble size to bracket crop climate impacts globally
Abstract. Global crop–climate impact studies using gridded, process-based models are essential for understanding how climate change may affect agricultural productivity. Due to high computational demand, these models are typically driven by a limited subset of projections from up to five global climate models (GCMs) from CMIP experiments. Yet, there is limited understanding of whether these subsets adequately capture variability in climate drivers within the whole ensemble and how the captured variability in drivers translates into crop yield impacts. Here, we analyze climate and crop yield projections from 29 GCMs contributing to CMIP6 ScenarioMIP that provide all required forcing variables and have been bias-adjusted and downscaled. To address computational constraints, we apply a high-performant crop model emulator to estimate yields of four staple crops across 150 k global locations under the four core emission pathways and for the whole climate ensemble from 1984–2099. We then aggregate results to major regions and use linear programming with the objective to minimize ensemble sizes while bracketing projected changes in temperature, precipitation, or crop yields. I.e., we identify the smallest GCM subsets that capture each variables’ full range. We find that ensemble sizes required to bracket ranges for individual variables increase from temperature (around 5–20 GCMs) to precipitation and further to crop yields (about 10–29 GCMs each), depending on the selection of crops and scenarios, and the strictness of the boundaries. Capturing yield responses across contrasting crops and emission pathways commonly requires the full ensemble. Notably, bracketing climate variables alone fails to capture the range of yield outcomes, highlighting the limited predictive value of individual climatic drivers. Our results emphasize the need for thorough contextualization of GCM selection in impact studies and suggest that emulators can fill a gap in complementing mechanistic model simulations with comprehensive climate uncertainty quantification.
This manuscript demonstrates that commonly used small subsets of climate models (GCMs) may substantially underestimate the range of projected crop-yield responses to climate change. Using an emulator to assess 29 CMIP6 GCMs globally, the authors show that capturing the full range of yield impacts generally requires far more models than capturing the range of individual climate variables such as temperature or precipitation. The results highlight the importance of carefully justifying GCM selection in crop-impact studies and of considering comprehensive climate uncertainty when assessing future agricultural impacts.
I find the manuscript important, cutting-edge, and convincingly presented. The results are new and relevant, with important implications for the uncertainty that is inherent to crop yield projections. As such, this work provides an important corner stone of the interpretation of future crop yield projections. It is also well-written. I am therefore happy to suggest publication after some minor comments I specify below have been addressed.
Specific comments:
l. 1 I am not sure I entirely understand the title, even upon reading it for the 5th time. I encourage the authors to find something more intuitive.
l. 19 Here and elsewhere, specify that you (I think) use the first ensemble member of each of these models. This is important because it may overlook important internal variability uncertainty that can only be distilled with single model large ensembles.
l.20 For the climate modelling community, the use of “forcing” here could be confusing. I believe forcing to the crop models is meant, but forcing in the climate model context could also be external forcing like volcanic eruptions or greenhouse gas concentration changes. Consider specifying.
l. 49 See my comment on line 19: the use of “member” may be misconstrued because it means “model” here, but may also mean “model realisation” in the context of single model ensembles. Please specify.
l. 49ff Since the authors go so heavily into the climate model independence weighting literature here: I find this remains largely inconsequential, which is a bit misleading. Is this depth really needed here?
l. 72 Reference mistakenly in brackets.
l. 79ff Another potential method to address uncertainty in crop model projections is the use of global warming levels in the analysis, and was recently introduced by Mithal et al (https://doi.org/10.1088/1748-9326/ae886b). I think the discussion of that paper would provide more context here and also give a reference point in Figure 3 or lines 521ff, for example.
l. 90 I still do not fully understand the meaning of the term “bracketing” in this context. Re-consider? Would “constrain” work?
l. 281 Typo: enforces.
l. 308 I am not convinced by the use of model numbers instead of actual names. For the people that know the models, full names would be really helpful! Consider getting rid of the coding of model names.
Figure 2 I don’t think I understand this plot. Where am I supposed to look? Which visualisations should I compare? Where is the focus? I don’t get the main message here. Can that be summarised and displayed more efficiently?
Figures 4 & 5 I am unclear why some of the legends are lines while others are crosses. Is there a meaning here? (Or is it my file that is corrupted?)
Figure 6 It is counter-intuitive that more saturated colours are lower values (I think), because higher percentage of low percentage crop yields. I have a feeling this could be also displayed without the double negative. Please consider flipping the color bars.