This study is a well-designed experiment that aims to understand how different types of compensatory responses, varying in direction and strength, alter the form of biodiversity-productivity relationship. The study finds that the biodiversity-productivity relationship changes in form over time, both within and between scenarios of extinction.
The sheer size and the duration of the experiment is commendable, and the combination of the two provides important knowledge on the temporal variation of compensatory responses and their role in modifying the biodiversity-productivity relationship.
The study builds a strong case on recent primary literature in the field and provides new insights into the temporal aspects of compensatory responses in plant functional groups. These findings further advances the understanding of post-extinction compensatory responses and the biodiversity-ecosystem functioning research. The conclusion neatly captures how the current study fills the knowledge gap of compensation direction and strength for the form of the biodiversity curve and highlights the importance of incorporating compensation when forecasting post-extinction ecosystem properties.
The introduction lays out key concepts and mechanisms associated with compensatory responses, as well as the caveats of the generalised biodiversity-ecosystem function saturation curve based on synthesised experiments. It builds a strong argument, backed by recent primary literature with a clear defined focus of the study based on the knowledge gap - the importance of compensation direction and strength for the form of the biodiversity-productivity relationship. While extinction scenarios in the experiment are stated to reflect realistic extinctions, it is not exactly clear what it is that makes them realistic.
Results are well illustrated with figures and the interpretation is succinct and consistent with data. In the discussion, the findings are neatly captured in the conceptual framework presenting how alternative types of compensatory responses affect the direction and form of the biodiversity-productivity curve.
For the sake of full transparency, it is encouraged that the authors make the raw data freely available via an online database.
Although, there are some minor language mistakes in the text, which needs correcting - some of which are mentioned below - the manuscript is overall, well-structured and written piece of work. This is a contemporary experiment which furthers the mechanistic understanding of compensatory responses and their implications for post-extinction ecosystem consequences and thus provides a valuable contribution to the field.

More specific comments:
1. Title is good because it sums up the study in a few words. However, it is generic, and not specific. Therefore, it might be worth changing the title to draw out the key message of the paper.
2. General comment. For future reference, please use continual line numbering throughout the manuscript.

INTRODUCTION
1. In Section 1. Page 7, line 5: “a specific PFG was finally remained” – This needs rephrasing.
2. In Section 1. Page 7, line 2-15 – The authors mention that the patterns of PFG extinction of their experimental design better reflects the impacts of realistic extinctions. However, they do not mention what makes the experimental extinction scenarios realistic?

METHODS
3. In Section 2.2. Page 8, line 17-18: Why are pre-treatment composition measurements and removal not undertaken in the same year? And how does this affect the results?
4. In Section 2.2. Page 9, line 5-6: “The measured quadrats in each plot differed among years.” – is this accounted for in the statistical analysis?
5. In Section 2.2. Page 9, line 6-7: “Plants were sorted to species and PFG, dried and weighed.” – Rephrase to: “Plants were sorted by species…”
6. In Section 2.3. Page 9, line 16: Why is the term Cproductivity × A1i included in the calculation of expected productivity? And why isn’t Cproductivity × A2i sufficient for the purpose of calculating the expected productivity?
7. In Section 2.4. Page 11, line 8-9: “The linear and quadratic functions were fit with the lm function, and the power function was fit by the nls function” – please specify that these are R functions.
8. In Section 2.4. Page 11, line 10-11: Clarify how the t-test was used to determine brackets of the compensation index.

RESULTS
9. In Section 3.1. Page 12, line 1: “linear fit best in year 3” change to: “linear fit was best in year 3”
10. In Section 3.4. Page 14, line 6-7: “Abundance played a dominant, though not significantly so, role only in year 2.” – Rephrase. Suggest changing to: “In year 2, abundance played a dominant role, albeit not significant”.
11. In Section 3.4. Page 14, line 7-9: “After this, compensation was the dominant mechanism in year 3 and the co-dominant mechanism with abundance in the following years.” – Rephrase this sentence.

DISCUSSION
12. In Section 4.1. Page 15, line 6-11 - Apart from Annuals, which other PFGs were responsible for differences in the relationship?

FIGURES
13. General comment. Figure legends would benefit from indicating the number of samples that each line and bar represents (n=x).
14. Figure 1 legend. Page 28, line: 6: “were the final PFG remained” – Rephrase this sentence.

SUPPLEMENTARY INFORMATION
15. Include controls in Table S1.