Reply on RC1

Comment: This is a highly relevant study, as it addresses spatial and temporal forest disturbances in continental Europe in response to the extreme drought in 2018. The authors draw some important conclusions of the 2018 drought being a lasting trigger of changes in forest disturbances across Europe. While I find this finding very fascinating and relevant, I would wish that the authors could underpin it further. Particularly, the link of low soil moisture/high VPD in 2018 being a main driver of forest disturbances in 2019/2020 should be made clearer. I am convinced that this will increase the relevance of this letter and make it attractive for a large scientific community.

of our analysis was on the total area disturbed, irrespectively of disturbance severity.
Comment: I think it would be good to understand how an already disturbed area/pixel is treated in the following years post disturbance. Are disturbed areas/pixels considered in the next year of your analysis, e.g. from 2019 to 2020 and do they add to the disturbance rate (e.g. because they are more disturbed the next year) or are they omitted because the forest was already disturbed?
Response: We thank the reviewer for this question. Throughout the analysis, a pixel can only be disturbed once, that is a pixel disturbed in 2019 cannot be disturbed again in 2020. This is a limitation of the underlying remote sensing approach, but we argue that this only affects a relatively small proportion of the disturbance areas in Europe. In the initial publication (Senf and Seidl 2021), we estimated that approximately 10 % of the plots identified as disturbed in the manual interpretation of spectral trajectories were affected by > 1 disturbances (that is less than 2 % of Europe's forested area). From those 'double-disturbances', a large proportion relates to management interventions before a natural disturbance (i.e., thinning, removal of bark beetle-infested trees before salvage logging the whole stand) and might thus appear as a longer-duration disturbance event that ultimately results in a single disturbance patch. We here count those longer-duration disturbances event only once in calculating the disturbance rate to not inflate disturbance rate calculations.
Comment: I think it would be worth to assess potential impacts of low soil moisture and high VPD in 2019 on forest disturbance rates. Eventually it was not the one event in 2018, but a repeated drought/heat that increased disturbance rates... This should be at least check and the results presented.

Response:
We thank the reviewer for this comment. We agree that drought conditions continued into 2019 and also 2020. To test this, we reran the model using annual soil moisture data -that is soil moisture from 2018, 2019 and 2020, respectively -instead of only soil moisture from 2018. This model showed very similar results as our initial model, though it had slightly lower support by the data (see Table 1 in revised manuscript). We thus consider the initial model as more appropriate for explaining the drought impact on disturbances. That said, we found that for areas of high disturbance anomalies (> 100 %), there was a strong correlation (r = 0.65) between soil moisture in 2018 and 2019. The spatial patterns of soil moisture anomaly in 2018 are thus likely already representative for the persistent drought impacts in 2019. To consider this important point also in our manuscript, we substantially revised the text (see changes below). We also revised Figure  1 to include maps of the soil moisture and vapor pressure deficit anomalies, giving a better idea about the spatial patterns of the drought and how it developed over the years 2019 and 2020. Finally, we included the model comparison in the manuscript ( Table 1 in the revised manuscript), present the full results of the model as Supplementary Table 1, and added an additional, more detailed version of Figure 2 as Supplementary Figure S1.
Changes in text (L. 43): "We found a substantial increase (up to +500 % compared to the average of 1986 -2015; Fig. 1 a) in forest disturbances in large parts of Europe in 2018, which spatially aligned with observed soil moisture and vapor pressure deficit anomalies in the summer of 2018 ( Fig. 1 b/c). The positive disturbance anomaly was persistent beyond 2018, with disturbance rates remaining considerably above average at least until 2020 (Fig. 1). The elevated levels of disturbance observed in 2019 and 2020 were significantly correlated with negative soil moisture anomalies in 2018 (Fig. 2), suggesting that the 2018 drought had persistent impacts on forest disturbances for at least three years. Soil moisture anomalies in 2019 and 2020 were also significantly correlated to disturbances anomalies in those years, but effects were weaker than those of the soil moisture anomalies in 2018 (Table 1). This suggests that drought conditions in 2018 were already indicative of impacts on disturbances observed in the following years. We further found a significant interaction effect between soil moisture anomalies in 2018 and vapor pressure deficit anomalies in 2019 and 2020, but not in 2018 (Fig. 2 and Supplementary Table S1). Specifically, we found higher positive disturbance anomalies in areas that were affected by both low soil moisture in 2018 and high vapor pressure deficit in 2019 and 2020 (Fig. 2). This result highlights the combined effect of extreme soil moisture deficits and cooccurring atmospheric dryness because of heat, which was characteristic for the drought of 2018 and the following years (Fig. 2 b/c). Overall, summer soil moisture and vapor pressure deficit anomalies alone explained 11.5 % of the total continental-scale variance in disturbance anomalies for 2018 -2020. Yet, we note that there is remaining variability in disturbance not explained by drought and likely related to forest management (Sebald et al., 2021;Senf and Seidl, 2021b), structural drivers (Seidl et al., 2011), and local processes not considered in this analysis (i.e., topography; Senf and Seidl, 2018; Albrich et al., 2020)." Comment: L32 Why is your approach rapid? Did you left something relevant out?
Response: A good comment by the reviewer. 'Rapid' was indeed a wrong description of our approach, as we -of course -did not leave out anything relevant. We thus dropped the word from the manuscript.
Comment: L40 Did you check the soil moisture/VPD anomalie in 2019, in many regions this was a dry and hot year too. Particular two subsequent dry years might have been the trigger for disturbances to last. Could you add some additional analysis/ information on this, please?
Response: Please see answer above.
Comment: L42 To my feeling explaining 11.5% of forest disturbance is not that high, or? Please add some explanation.
Response: While we agree with the reviewer that 11.5 % might not sound a lot of explained variance, we note that this is a continental-scale model including only two predictors. Given the high variability in disturbance drivers across all of Europe -11.5 % of explained variance solely by drought (i.e., soil moisture and vapor pressure deficit) is indeed a substantial proportion. Please also note that this is not only pertaining to the areas affected by drought, but that this is 11.5% of all of Europe's disturbances. We, however, agree that some more context might be needed and we thus added more detail to the text (L. 60): "Yet, we note that there is remaining variability in disturbance not explained by drought and likely related to forest management (Sebald et al., 2021;Senf andSeidl, 2021b), structural drivers (Seidl et al., 2011), and local processes not considered in this analysis (i.e., topography;Senf and Seidl, 2018;Albrich et al., 2020).".
Comment: L57 Can you please add some information on how strong the disturbances were (e.g. stand replacing, 50% of forest canopy lost,…)?

Response:
We thank the reviewer for this question, which addresses and important issue. Based on the maps it is very challenging to estimate the true severity of disturbances, because the severity measure used in Senf and Seidl 2021 is based on the spectral change during disturbance, which only gives an indication of the relative severity. We are currently working on additional analyses converting spectral changes into actual changes in canopy cover (work still under review), but this additional work showed that approximately 75 % of all disturbances in Europe detected in our satellite-based approach were high severity events with > 50 % canopy loss and approximately 10 % of all disturbances in Europe had very high severity (> 90 % canopy loss). This, however, only includes disturbances up to 2016. For the period 2018 to 2020 we estimate the disturbance severity to be higher in many regions, as salvage logging was common in Germany and Czechia, both regions which were historically characterized by relatively low disturbance severities. Hence, as there is no reliable information yet about the proportion of stand-replacing disturbances, we refrained from including those estimates in the manuscript. We however note that our map includes disturbances of variable severity in the revised methods description (L. 162): "The map depicts any abrupt declines in the dominant forest canopy -regardless of its cause -that are detectable at a spatial grain of 30 m, including disturbances that only remove a part of the canopy within a pixel. It does, however, not detect any changes in sub-canopy tree layers.".
Comment: L61-62 But you did not directly assess canopy mortality rates, or?
Response: No, we refrained from assessing absolute canopy mortality rates based on the maps, as those can be biased due to the higher omission error of the maps. We thus prefer relative statements wherever possible. Future research should use our first assessment as basis for a thorough sample-based assessment of the true, absolute disturbance rates (see, e.g., Senf et al. 2018 [Nature Communications] and 2021 [One Earth]); but such a sample-based assessment was beyond the scope of this letter, which aims at informing research and management in a timely manner.