This manuscript investigates how ecosystem primary productivity recovers after experiencing flash droughts using the random forest model and an explainable model. These results reveal the response time of GPP over China and its influencing factors. The topic is of significance to assess flash droughts’ ecological impacts, which is probably a concern to the community of hydrologists, ecologists, and policy-makers. However, the study should clarify some comments before it is accepted by Hydrology and Earth System Sciences.

Major comments:

1. The abstract pointed the most novel finding is proposing a new method of a machine learning method to study the recovery of GPP to flash droughts. In my opinion, the method is not quite new and has been widely used in analyzing the interactions between soil moisture and vegetation. Whereas the recovery of GPP is less involved in previous studies, this study contributes a lot to provide a perspective on this topic.
2. The metric of GPP recovery from flash droughts used in this study may be influenced by data noises, for example, during a flash drought event that persists for 2 months, negative GPP anomalies only occur for 5 days. Such cases should be excluded in the analysis. The precondition of GPP recovery from flash droughts is that GPP has been negatively influenced by flash droughts. Besides, the terminating point of the recovery process is difficult to detect and should not be recognized at the point where the GPP anomaly is above 0, as there are many noises including whether it has experienced another drought or other extreme, the stable condition may be higher or lower than the normal conditions, etc.
3. Does the declining magnitude of GPP caused by flash droughts influence GPP’s recovery time?
4. What are the hydro-meteorological conditions during the recovery stage of GPP? Is there a connection between hydro-meteorological conditions and GPP recovery?
5. The study period is a little short and the available datasets have been updated to 2023 even longer.

Minor comments:

1. L23: “response function functions”
2. L41: “productivity” should be more clear. Maybe “terrestrial ecosystem productivity” is better.
3. L48-54: The phase reviews the previous research about how vegetation recovers from droughts. It seems that they are inconsistent with the recovery of GPP in terms of GPP’s response across different PFTs. Is there any explanation for it?
4. L54:55 & L303:304: Vegetation over humid regions needs more time to recover to its normal condition. As there is more water available over humid regions, why vegetation is more difficult to recover?
5. L57: What is the “background value”?
6. 1 is difficult for readers to understand the metric used in this study. It is better to clarify flash drought and recovery time in Fig.1 more clearly. Perhaps authors can select a case from the observed events.
7. L185: In Fig2.b, should the red line be removed?
8. L199: There is no GPP recovery over northwestern China. Is there no response of GPP to flash droughts? As usual, vegetation is more sensitive to water availability in arid or semi-arid regions than in humid regions. Besides, is the response rate the reverse value of response time? If so, they are presenting the same results.

General Comments

This study investigates the duration required for ecosystems in China to revert to their pre-flash drought state, emphasizing the spatiotemporal patterns of recovery and the factors influencing them. Particularly notable are the findings regarding the impact of post-drought radiation, aridity index, and temperature on recovery time, especially in semi-arid and sub-humid regions. These findings hold significant implications for eco-hydrological research. However, substantial revisions are necessary before the manuscript can be considered for publication.

Major Comments

(1) While the writing is satisfactory and effectively conveys the scientific ideas, the paper would benefit from further polishing to enhance clarity and coherence. Specifically, the introduction should be expanded to provide more detailed information rather than merely listing literature.

(2) A more in-depth discussion is needed, particularly regarding the detailed process analysis and discussion of GPP recovery from flash droughts. The manuscript currently lacks this depth. Incorporating an analysis of the unprecedented 2022 mega-drought in the Yangtze River Basin could serve as a valuable case study to enhance the discussion.

(3) The definition of ecosystem recovery focuses on changes in Gross Primary Productivity (GPP) anomalies. However, in the current results, there are flash droughts lasting more than 100 days where GPP negative anomalies occur for only 5 days. Such cases should be excluded from the analysis.

Specific Comments

Line 23: It should be “response function”.

Line 48-49: It is interesting to study to what extent these ecosystems compensate. Even in the annual carbon balance, will flash drought have a lasting impact.

Line195-197: How were these vegetation classifications determined? Briefly discussing the phenological characteristics of these classifications would be helpful.

Line 231: Please standardize the manuscript by changing all instances of "figure" to "fig".

Line 247-248: Please add the discussion about ecological drought.

Line 251-253: The manuscript should emphasize the mechanisms underlying the study's findings. Adding a discussion on the differences between grasslands and forests, particularly focusing on root depth levels, would be beneficial.