This papers offers a detailed analysis of drought responses for vegetation in the Midwestern US using an ecohydrologic model and assimilation of MODIS FPAR and LAI data along with flux tower data.  The paper addresses some important ecohydrologic questions about how rapidly transpiration and carbon assimilation decline during drought and how changing phenology, specifically above-ground photosynthetic capacity, accelerates GPP losses.  The paper utilizes advanced modeling techniques and builds on previous work that has established useful ways to assimilated remote sensing data into the DCHM model.

While that paper has significant potential, it did not really make strategic use of the model and observations to address some of the questions posed in the introduction.

For example, the hypothesis posed do not really address the issues of ‘flash’ drought.

In H2 The idea that drought causes both carbon uptake and transpiration to decline is something that is quite well understood - and there is ample evidence that this occurs- we know plant shut down when they run out of water.  There are elements of the timing of this that are perhaps less well understood - and questions about how water use efficiency changes during a drought- and indeed the authors get at this to some extent in the paper - The hypothesis should reflect this. There would also be ways to frame the study (and hypothesis) to look at the relative impact of “flash” drought versus other types of drought that could be interesting. Some additional thinking about how to use the model to test more nuanced (and informative) hypothesis would be strengthen this paper

Throughout the paper, there are statements made that are not well supported by graphs or analysis - for example - that evaporation exceeds infiltration  (this is indirectly shown but it would be much convincing to show this directly - and model results could do this).  In another perhaps more salient example for the paper, the authors state  phenology declines reduce carbon and water exchanges  (H3) - one could argue that because plants have already shut down stomates at that point in the season,  phenological declines do not further reduce transpiration  - I’m not suggesting this is true but graphs presented do not clearly rule this out in the testing of H3.

The author also make statements about flash droughts but do not really distinguish flash drought from other types of drought in their analysis - They have multiple years that they do not really make use of.

Finally there are also important differences between modeled and flux tower data that may be critical in the understanding flash drought responses. These differences need to be more rigorously explored (see detailed comments below)

Some detailed comments.

H1 is actually two hypothesis - it would be useful to separate them

The data sets and modeling proposed here tend to focus on relatively shallow surface soil - including citations of expected rooting depths for the PPTs would be helpful support for the implementation (especially given that flux tower observations of ET tend to be higher than the model in dry years)

line 225 - Some additional (just one or two sentences) information about of how ensembles of phenology parameters are established is needed here (e.g what is done for each of the 3 periods to select the 2000 parameter distributions shown in Figure 4) - There needs to be a bit more context so that reader understands Figure 4 and what controls the variation in parameter sets. 

line 235 - The simulation period is relatively short - and isohydric-anisohydric differences may or not be distinguishable within the 3 years - thus you cannot really state that the vegetation model parameters trained on dry conditions will represent isohydroic vegetation?.  Especially given that parameter values seem to change depending on period (Figure 4) but vegetation PFT does not. 

line 246 - That transpiration is calculated from root water uptake makes sense but it doesn’t follow that this allows you to  “to partition ET”…you would  need to have a separate calculation of total ET to do that. Clarify

Line 259  - For clarity it would be helpful to be consistent in the naming conventions- e.g gamma or growth parameter not both

Line 258- in what way is this in agreement with Lowman and Barros (e.g the choice of longer period for reducing uncertainty) - in a way that’s not so surprising - more information usually reduces uncertainty?

Line 259 That gamma values vary by site could be do to differences in climate (note that game values vary across wet and dry years) - so it is not a given that it varies by plant functional type - rather this is an assumption (e.g I think that you are assigning plant functional type parameters based on this analysis)- The wording of this paragraph could make that point more clearly

Line 265 - “slower” relative to what?

The rationale for the continued focus (beyond Figure 4) of differences due to parameters sets based on wet, dry or both years is unclear - Given that using both wet and dry years clearly reduces uncertainty, I’m not sure why there is a need to compare estimates of FPAR, LAI, The authors may have a reason for this but if so it needs to be emphasized in the text.  Removing this would allow the focus to be on DCHM-PV performance and the “actual” phenological mediated vegetation responses.

Section 3.12 and Section 3.1.3 - If I understand the methods correctly - Figure 6 and 7 show results using parameters conditions on prior information (e.g MODIS assimilation during the calibration period) and MODIS results for 2012 and 2019 - Given that, additional discussion about fit with MODIS would be helpful - How well does DCHM-PV do.  Overall it captures patterns fairly well but there are some notable exceptions (e.g loss of FPAR in July and August at US-KFS in MOIDS that is not tracked by model) - It would be useful to have some presentation of model performance here

Line 315 - which are water stress years (e.g 2012). Also can you note which method (or averaged across all methods) does the 1kgCm2 reduction come from

lLine 380 - The arguments in the first 3 sentences of this paragraph need a bit more detail.  Just because there are fluctuations in evaporation this doesn’t not necessarily mean that “all” water evaporated before it had a chance to infiltrate.

Note the substantial underestimation of ET by the models relative to Ameriflux in 2012 should be noted here as well along with some discussion of why

line 360 - Its worth noting that the drought response can be more complicated than simply shutting stomata - importantly grasses can shift their allocation of carbon - and this will be reflected in above ground biomass (the GPP measured by MODIS) but also in below ground stores and fluxes - For example see  Ingrisch, Johannes, Stefan Karlowsky, Roland Hasibeder, Gerd Gleixner, and Michael Bahn. "Drought and recovery effects on belowground respiration dynamics and the partitioning of recent carbon in managed and abandoned grassland." Global Change Biology 26, no. 8 (2020): 4366-4378.

Similarly on line 380 - there is ample evidence of changing root allocation (and root respiration) for grasslands that would be worth citing here.

line 369 - Note that authors don’t really show that evaporation after rain effects uses all available water - so it doesn’t infiltrate (as stated in the hypothesis) - but they could do this at least with the model since both daily evaporation and precipitation is available.

line 375 - “since GPP is decreasing” as the authors themselves note - declines in GPP do not always reflect what’s happening with transpiration so this statement needs some caveats

In the  discussion, one of the challenges here is that observations and models suggest differences in plant ability to pull water from deeper layers - The first paragraph blurs these distinctions  - so for example the line  “we did find that even during the peak..plants were still” - is this based on flux towers, or model?.  I also note that model and observed in Figure 11 suggest different stories about water use efficiency during the drought.   The observed data suggests plants maintain higher water use efficiency, longer,  during the drought than the model - this is very interesting - its suggests plants are doing something that the model misses which is informative -  but needs to be much more of a focus in the paper.

Figure 12  - its not so easy to see from this figure that when plant are transpiring more they are more efficient in their water use - Simply graphing transpiration vs WUE would show this much more clearly.

Also in figure 12 - some strategic use of color to differentiate wet versus dry years would be helpful here

Line 403 - The idea of flash drought responses is intriguing but I think the paper could do much more to support these ideas (e.g that the rapidness of the change is indicative of a flash drought) - Some more strategic comparison of the declines across different type of drought - flash versus “non-flash” (Of course this would require clearly distinguish what a flash drought is from other types of drought - but that seems to be part of the paper’s motivation)

This study investigates vegetation phenological processes during drought and non-drought periods using a set of DCHM-V and DCHM-PV model simulations. The focus is on three sites in Kansas USA, as they experienced extreme drought and pluvial conditions in recent decades, and also have ground-based and satellite-based observations available to compare with model simulations and study observed processes. The modeling experiments are neatly designed, and the investigation is systematic to study sources of uncertainty in vegetation phenology. I however have a number of comments that need to be addressed – please see below.

Main comments

1)    This study uses 2012 and 2019 to exemplify the contrast of vegetation phenology between flash drought vs non-drought years. It would help to add a non-flash/conventional drought year to the study, as the vegetation phenology could differ considerably between flash drought and conventional drought. It would be interesting to see how the evolution of plant phenology, water use, and productivity may differ between the two drought cases.

2)    Much of the findings are based on the DCHM-V and DCHM-PV simulations, and are thus subject to the performance of the DCHM and its predictive phenology in simulating observed land surface and vegetation processes. The comparison between the model results and independent observations (e.g., MODIS, AmeriFlux) however shows considerable differences: some of the models vs. observations differences are so substantial that they are much larger than the differences between different model experiments (e.g., Figs.5-9). While these differences could be in part due to the data comparison across different spatial and temporal scales (Section 4.5), they also make one wonder about the performance of the DCHM and its predictive phenology. I suggest the study provides more information/results on the fidelity of the DCHM and its predictive component is simulating basic land surface variables (e.g., soil moisture, evapotranspiration) and observed vegetation phonology related fields (e.g., LAI, FPAR), e.g., in terms of climatology, seasonal cycle and year-to-year variations, to make the model-based findings more convincing. Also see some of my detailed comments below.

Detailed comments

1)    It would help to briefly discuss the implications of the findings (e.g., based on WET vs DRY vs 3YR) to subseasonal prediction of vegetation.

2)    Noah-LSM: Noah LSM has multiple versions. If the Noah-LSM used in this study refers to the Noah in NLDAS-2, please specify.

3)    line 259: of gamma => of the growth rate parameter

4)    Figure 12: May want to increase the thickness of curves for 2012 and 2019 to highlight the results for these two years

5)    Line 390: (Figure 10 => (Figure 10)

6)    Figures A3, A5. Middle and deep layer soil moisture for the flash drought year 2012. How to explain the substantial differences between DCHM-V/DCHM-PV and Noah-LSM? Noah-LSM seems to make more sense as it shows a notable decline after June 2012. In contrast, the soil moisture in DCHM-V/DCHM-PV remains relatively steady throughout 2012 and does not seem to be responsive to the strong precipitation deficits during 2012, which looks odd; this is concerning as any issues in simulating soil moisture would adversely impact the simulation of vegetation and evapotranspiration processes etc. Please also see my second main comment.

7)    Figure A6 is identical to Figure A5 and appears to be incorrect. Please check if it plots the results for 2019.

8)    Figure A10: “during 2012”=>”during 2019’?

9)    Figure A11a: The difference between Ameriflux and model simulation is striking. The inclusion of Ameriflux appears to cause confusion rather than providing a truthful evaluation of the model results.