This study presents the modelling of forest responses to extreme climate conditions (i.e., droughts), using the dynamic vegetation model LPJmL-FIT. The research focuses on an area in eastern Germany, applying custom climate scenarios to simulate increased probabilities of the 2018 drought event. Two model configurations were used: one representing a monoculture temperate needleleaf evergreen forest and another representing a mixed forest ecosystem with temperate and boreal needleleaf and broadleaf plant functional types.

This manuscript presents a valuable test case for the vegetation modeling community by showcasing the capability of vegetation demography models within a scenario-testing framework. Such kind of studies are important because illustrate the critical role in defining optimal forest management strategies under climate change conditions. However, I believe there are aspects where the manuscript could be improved to better convey its message to the community. Below, I provide a list of suggestions:

• Authors focused their attention on the response of mixed forest ecosystems based on the PFT classification. However, reliance on PFTs may inadequately capture the diversity of the different tree species defining the resilience of a forest ecosystem. I think authors should address this issue when introducing/justifying the need of their work and when discussing some of the limitations of their study.
• Some recent works (e.g., Forzieri et al., 2022) reported on an emerging signal of declining forest resilience based on trend analysis of remotely sensed information. It would be interesting to apply a similar analysis framework using numerical simulation output presented in this study. In so doing authors would be able to summarize the whole set of model results into more “practical” indicators of ecosystem status.
• The study is based on the pure analysis of numerical simulation results without any level of benchmarking with direct observations and/or forest inventory information. This limitation should be somehow addressed or at least better discussed.

Minor comments:

• I am fine with the creation of ad-hoc drought years in a scenario-testing framework. However, I am not convinced by the arguments used by authors between lines 105-109.
• The color scheme used in Figure 2 makes difficult to easily identify the model response for the different scenarios.
• Please check the statement between lines 277-278. It should be the opposite.
• The first sentence of the Conclusions section does not really apply for mixed forest ecosystems.

This manuscript presents a forest modelling exercise, employing the flexible-individual trait

Dynamic Global Vegetation Model LPJmL-FIT, to test the short- and long-term effects of severe drought frequency on biomass, structure and trait distribution of forests in Berlin and Brandenburg states. The manuscript is overall well written and easy to follow for forest ecosystem modelers. The conclusions, though not very novel, are interesting and in general well sustained by model results.

In my opinion, more model design/formulation details are needed in the Methods section to appropriately interpret the simulation results. In particular, more details should be given with respect to the impacts of drought on demographic processes in the model (growth, mortality, establishment). Does drought impact biomass accumulation via reduced NPP and growth, or is there an increase in mortality? Do increases in mortality rates result from slow growth or are they directly linked to low soil moisture levels? Does the model include explicit inclusion of plant hydraulics? If these details are not included, the authors expect the reader to be acquainted with LPJmL-FIT design. More details on how trait variation, particularly variation in wood density, affects demographic (mortality) rates, should also be given in the Methods section. It would also be important to state whether SLA and WD values are sampled independently, or they are correlated.

Secondly, the consequences of describing plant diversity using PFTs with respect to known species (except P. sylvestris) is not sufficiently acknowledged in the interpretation of results. To which extent the results can be explained by the fact that needle-leaved PFTs have a narrower trait distribution than broadleaves? Moreover, PFT trait variability is at least partly a reflect of higher/lower species diversity within the PFT definition. The authors mention adaptive capacity of broadleaved trees, which may be indeed higher than needleleaves, but this may be explained instead by a broader range of trait values in PFT trait distribution arising from a larger number of species included in the PFT definition. Thus, the “adaptation” of mixed forests would rather mean a shift in species identity within the PFT. The low taxonomic resolution of PFTs affects the comparison of the relative importance of composition changes vs trait variation. Therefore, the authors should careful in use terms like “PFT composition” and not “species composition” when discussing results, specially in section 4.3, and in the implications (e.g. L510).

Thirdly, if I correctly understood the model design, local filtering applies to adults but it does not imply a change in the trait distribution of future offspring (besides the local extinction of PFTs). In other words, there is no trait inheritability between generations. The authors could discuss whether and how they expect their results to change if this model limitation was removed.

Finally, spatial patterns arise from a fine resolution of climate forcing but in my opinion they appear as less interesting for the general reader than other results focusing on the effect of diversity and trait variation, and are a bit disconnected to them. Therefore, they could be omitted or moved to appendix.

Minor comments:

L262: The authors report the following simulation result: The lower drought frequency in Scenarios C and D resulted in a stronger decrease in biomass compared to the higher drought frequency in Scenarios E, F, G in the pine monoculture. I could not find an interpretation of this result, which I found rather surprising.

L277-278: Scenarios A and B had “higher”, not “lower”, biomass and tree density.

L439-440: I wonder to which degree this result is an artifact of model design. The result may be influenced by the random sampling of traits from species-level distribution, instead of having recruits with trait values taken from distributions influenced by adult trait composition. Moreover, wood density may be regarded as a trait varying in time for the same individual, so that all trees in the patch should shift to higher wood density with drought impacts, which would slow vertical growth for all of them.