Review of “Increased vulnerability of European ecosystems to two compound dry and hot summers in 2018 and 2019” by Bastos et al (esd-2021-19)

The manuscript by Bastos et al explores the existence and extent of compound effects of subsequent dry and hot summers on vegetation, using the case of 2018 and 2019 in Europe.

The manuscript is of general interest to a large audience, across several disciplines. Overall, the manuscript is generally well-written and clear (but see below).

I have however three main suggestions:

1) The manuscript is dense, both in terms of methods and results. The results and even more so discussion are thus inevitably complex, but, unfortunately, at times not very clear, with apparently contrasting statements. A general suggestion would be to give descriptive names to the four clusters, to facilitate the reader throughout the text. Further, I list here some examples of apparently contrasting statements or unclear sentences, but I suggest that the entire discussion is revised for better clarity.

L313: Isn’t the high correlation in contrast with the statement in L 305?

L322-324: This is an important point, but it is not at all clear in the text.

L349: I think this is a potentially controversial point. Deeper roots are an advantage, if the off season has provided water recharge.

L354-355: On which basis can it be stated that this is ‘consistent’?

L356: Why is the vulnerability increased? Is this because of heat? Isn’t this in contrast to L350? Does this apply just to 2019? And, even if so, how could an opposite response in the two years be justified?

2) I find of particular relevance the question of the responses of different land uses. This point is relatively prominent in the abstract, but then only briefly touched upon in the results and discussion. I think the manuscript would gain in terms of impact (and interest within a broader audience) should this question be explored in more detail. To this aim, I would suggest redoing the analyses at the basis of Figure 6, but separating land uses. This could give some insights in how the different vegetation types respond to temporal compound events. Of particular relevance would be the correlation with previous year’s EVI anomalies, temperature, soil moisture and soil available water capacity. This additional analysis would for example reduce the speculation behind statements like that in L349, by allowing exploring the effects of available soil water.

3) One could expect that crops, by being mostly annual plants, would have a radically different response to temporal compound events from all other perennial vegetation. Specifically, one could expect no substantial legacy effects, or even a positive EVI anomaly in 2019, as the result of reduced nutrient use and losses via leaching in the previous year, characterized by low production and low soil moistures. While the discussion is definitely more focused on vegetation dominated by perennial plants, there are few hints at crops also having some legacy effects. I think it would be helpful to discuss in more detail why this is the case (if this is the case), thus better grounding the results in our ecophysiological understanding of plant and ecosystem response to (repeated) heat and drought. The additional analysis suggested above could shed some clarity on this matter, making the discussion less speculative. Authors could also consider how the LSM performance in 2019 is affected by land use: it could be expected that, if LSMs fail to represent the carry over effects, then they would be performing better in ecosystems where carry over effects are intrinsically more limited.

Minor comments:

There are several imprecisions in the text (missing or misplaced blank spaces, inconsistencies in symbols, etc.), in particular in the results and discussion sections.

The readability of Fig 3 would be greatly improved if it was bigger (in particular the right panels). Also, could a different set of colors be used, to highlight differences in the map? C2 and C3 are difficult to distinguish now.

Also Fig. 4 could be a bit bigger, possibly with larger and differently shaped symbols for 20018 and 2019 (which are anyhow outside the regression).

L131: Why not also June-August 2019?

L172: In which sense there is an acclimation to drought, at the scale of one-to-two years?

L244: I think the term ‘recovery’ is not necessarily correct, at least not in all ecosystems. So, I suggest using a more neutral ‘less negative EVI anomalies’.

L 285: ‘stronger’ with respect to what? DH2018?

L357: I find the term ‘natural ecosystems’ potentially confusing here. I suppose it is used to contrast forests/grasslands to croplands, but, in Europe, very few forests can be considered natural (in the sense of unmanaged) ecosystems. Their response to heat and drought is certainly mediated by species choice and other aspects of management.

Review of “Increased vulnerability of European ecosystems to two compound dry and hot summers in 2018 and 2019” by Bastos et al (esd-2021-19)

General comments

The manuscript talks about a relevant topic showing the importance of consecutive dry and hot events for ecosystems. The paper investigates why temporally-compound compound extremes can amplify the damage to the ecosystem focusing on 2018 and 2019 in Europe.

The paper addresses a relevant scientific question and it is within the scope of ESD with the potential to have an audience from a broader community of climate impact studies, agricultural systems, hydrology, and multi-sector dynamics.

The paper concludes that process-based models miss the legacy effects of consecutive compound hot and dry summers. The final results are based on a) regression and correlation analysis, and b) LSM outputs. However, the data preparation has heavily relied on satellite scans from MODIS and simulation outputs of Land Surface Models. 

The paper describes the data sources in great detail which is good. I understand that this is a critical part of the research, but it distracts the reader from the main message. I suggest moving the data construction information to the appendix.

Finally, I suspect the results are reproducible. It would be beneficial if the authors could share the major constructed datasets for regression analysis and related outputs of LSM models.

Below are my technical concerns.

• The title of the paper talks about compound dry and hot summers. The model and data do not include metrics of compound extremes. The T and SM are often separated in the paper and are considered individually. I expected to see some compound indicators. I am not sure the term “compound” in the title is well-represented within the paper.
• The paper is written well. Still, the overall presentation requires revisions. Please describe the main variables in more detail in a table in the appendix. The reader deserves a clear description of the main variables of this study and the main model evaluating the relationship between them (e.g. variables explaining EVI in regression models). There is no single equation neither a table showing the underlying model of the study. While the main work is based on regressions and correlation analysis, the reader should wait until page 7 to learn about them. The problem is that the regression strategy has a vital role here. Are you estimating the marginal impacts for each pixel or a set of clusters? These should be clarified with a written equation with clear indexing of all the variables. In addition, this can be a useful reduced-form model for future studies. In its current form, the regression section looks pretty weak.
• There is no discussion on the goodness of fit for regression and the causal analysis. The lagged EVI used in the model, while can be used to prove the existence of the legacy effect, does not tell us the exact sources of legacy effects. There is a high chance of omitted variables here (e.g. soil moisture in lower layers or disease as discussed in the paper). This could be briefly addressed in the appendix.
• Unfortunately, the interpretations and conclusions are more than what the model results show. For example, the claim in L316-319 is too strong. The correct conclusion is that "the proposed model" and current LSMs did not capture this legacy effect. This effect can be captured in future studies in models with different variables, metrics, and methods. This can be a shortage of the methods of this study. The results can be biased by omitting some variables (e.g. disease, nutrient, radiation, etc). This is the nature of science. Further investigation is required to have such a strong conclusion.
• The paper employs the soil moisture data based on volumetric soil moisture of the top 28 cm. I expected to see other metrics of root-zone soil moisture depending on the vegetation dominance. The soil moisture in lower layers can be “a” major source of legacy effect which is ignored apparently. As mentioned in the paper: “total water storage was lower in 2019 due to the water storage deficit resulting from the 2018 event”. Probably, a more precise hydrological product should be used to capture this.
• Other issues.

1. Define vulnerability
2. L3: “though” does not seem the right word here.
3. L10: Please revise the sentence.
4. L29: “hot and dry” is better to be replaced by “dry and hot” to better represent the DH abbreviation.
5. L65: extra parentheses
6. L124: This is the data section. Maybe change the title to show this.
7. L268: is this your finding? If not move it to the discussion.