This study appears to be a very valuable addition to the literature. The topic is particularly important as the risk of nuclear war may be increasing with ongoing deterioration in international relations between nuclear weapon states. Also recent evidence suggests that events such as major solar storms might be more likely than previous thought [1]. The methods of this study seem appropriate and the results make good sense. Some specific issues the authors could consider:

1. In the Introduction – references such as (Barrett et al., 2013) are perhaps somewhat outdated. Consider including findings from a recent major forecasting exercise: [2].
2. The modelling is done on spring wheat which is reasonable as it is more common globally than winter wheat. But a limitation that could be noted is that in ASRS conditions many producer countries could switch to winter wheat which would be more resistant to adverse ASRS impacts (such as cooler temperatures and frosts). It is also usually more productive than spring wheat, eg, for 2023 USDA data: Average winter wheat yields were around 50-55 bushels/acre, compared to spring wheat’s 40-45 bushels/acre.
3. For awhile I was wondering why in the ASRS scenario India and Pakistan still appeared to be involved in the food trade system (given the ASRS scenario involved them in a nuclear war). Then I saw the subsequent results where the modelling removed them. So perhaps explain earlier in the text that the initial results did not include direct nuclear war-related impacts on the nuclear waring nations.
4. The Discussion could consider stating that building country-level food system resiliency and reducing food trade dependence could be partly achieved by greater adoption of plant-based diets and with consumption of locally-produced fruits/vegetables/legumes (rather than imported grains). This is partly because much of the traded grain imported into some countries (especially soya beans, maize) is currently inefficiently used for animal feed (ie, inefficient from the perspective of food energy supplied to humans per energy inputs).
5. The Discussion could consider stating that building country-level food system resiliency and reducing food trade dependence could also be improved by reducing wasteful use of agricultural land that occurs at present eg, growing crops for biofuel (where this is heavily subsidized and where adopting electric vehicles is far more efficient), and growing crops that are a hazard to health (eg, tobacco). The need to reduce the relatively high levels of food waste in many countries is also a potentially very cost-effective way to build food system resiliency.
6. The Discussion could note that although the Xia et al 2022 modelling was very sophisticated – it still had various limitations eg, it did not consider nuclear war impacts on: supply of irrigated water, surface ozone levels, on ultraviolet light damage to agriculture, and “the availability of pollinators, killing frost….”

Trivial points

1. Add an open bracket “(” before:  Bernard de Raymond et al., 2021)
2. Improve wording of the sentence: “Likely because of its less…”
3. Where “teragrams” is first used – could say this is “equivalent to megatonnes” as the latter is probably a bit more understandable to most readers.
4. Fix typo: “Southern Hemisphere.although”

 References

1. Vasilyev V, Reinhold T, Shapiro AI, Usoskin I, Krivova NA, Maehara H, Notsu Y, Brun AS, Solanki SK, Gizon L. Sun-like stars produce superflares roughly once per century. Science. 2024;386(6727):1301-1305.
2. Karger E, Rosenberg J, Jacobs Z, Hickman M, Hadshar R, Gamin K, Smith T, Williams B, McCaslin T, Tetlock P. Forecasting Existential Risks: Evidence from a Long-Run Forecasting Tournament (FRI Working Paper #1): Forecasting Research Institute. https://static1.squarespace.com/static/635693acf15a3e2a14a56a4a/t/64abffe3f024747dd0e38d71/1688993798938/XPT.pdf; 2023.

Regards, Nick Wilson, University of Otago, New Zealand

Summary The authors compute adapted grain trade networks using national yield changes under two scenarios (nuclear fall and major infrastructure loss) and analyze these resulting static networks for wheat, rice, maize, and soybean. They put substantial effort into comparing these networks using a range of metrics.

General comment I thank the authors for their work in this important research field. While reading this comprehensive study, several points arose that, in my view, require further clarification or revision.

Main points I. The central element driving the results is the network model introduced by Hedlund et al. (2022). However, the main text offers no insights into how or why trade connections shift, which hinders interpretation and undermines confidence in the findings. Even after consulting the supplement and the Hedlund et al. paper, it remains unclear how trade changes are modeled—it appears to rely on a gravity model of trade. The authors should explain the main methodology and the points below:

1. Have the authors verified that a gravity model can reproduce observed trade networks (e.g., from FAO data)?

2. How strongly do historical trade patterns influence the gravity-modeled network?

3. These limitations should be explicitly addressed in the discussion.

II. Similarly, further detail on the Louvain algorithm is needed. Since this method underpins the community analysis that drives many key findings, a brief explanation would aid interpretation.

III. Focusing solely on yield changes (e. g., Fig. 1) to explain trade shifts overlooks the role of total (national) production. A 100% yield loss in a major producer has greater impact than in a minor one. Consider including crop production figures or a map showing regional yield or production changes (even in the supplement).

IV. Yield reductions under GCIL: It is unclear why the authors did not use productivity-weighted yield changes, as in Moersdorf et al. They say yield changes from GCIL should be comparable those from ASRS, but the scenarios are not directly comparable due to the much wider yield change range under ASRS—this should be clarified.

Comments on figures

a) Figures 5 and 7: “Import relative difference” vs. “Imports relative difference” should be consistent.

b) Figure 5: Since positive changes also occur (see Fig. 7), the color scale should reflect values above 0%.

c) Figure 6 appears to be missing.

Minor points

1. Graphical abstract: Avoid non-standard abbreviations (ASRS, GCIL).

2. Line 125: “We excluded bilateral trade flows falling below the 75th percentile in trade volume to concentrate on the main trade movements”—how sensitive are results to this threshold?

3. Figure 5 and Figure 7: Also consider showing absolute import calorie losses per capita or total supply losses to see how severe the (imported and domestic) crop losses are.

4. Update Levermann et al. (2024) from preprint to published version: https://doi.org/10.1038/s41893-024-01430-7

5. Consider restructuring the Discussion. Section 4.3 may fit better under Results; Sections 4.4.2 and 4.4.3 could be streamlined or moved, depending on their relevance.

6. Line 150: Please confirm whether the reference to Tukey (1977) is appropriate here.

Further comments • The authors provide access to the code, and its usage is clearly documented in the repository.