1 General Comments:

I recommend a revise and re-submit. This paper contributes significantly to our understanding of the impact of precipitation events on carbon cycles, particularly from the perspective of individual events using Earth System Models (ESMs). However, there are some concerns that the authors need to address before the paper can be published.

This paper explored how Earth System Models (ESMs) perform in response of vegetation productivity to individual precipitation events. Papers in remote sensing and model simulations usually focus on the interannual change of precipitation and its impact on vegetation productivity, and they have paid little attention to the properties and impact of individual precipitation events. However, manipulated experiments are carried out by monitoring vegetation growth after every rainfall, providing plenty of data about the individual precipitation events. We have limited knowledge of how ESMs simulate vegetation responses after rainfalls. Investigating this question can improve  ESMs by combining findings in manipulated experiments. On the other hand, the research on ESMs could also provide an analytical framework for experiments.

So, this paper is a great attempt to research individual precipitation events. They found distinct simulations in different models by composite precipitation events together over regions. They focused on three propensities: Peak amplitude, post amplitude, and time-lag effect. They didn’t discuss the amount of GPP impacted but used the sensitivity of GPP to precipitation instead.

During their discussion, they explored various reasons for the discrepancies between different models. They narrowed their focus to the processes that related to soil moisture and VPD, which link vegetation productivity and water availability. By comparing model simulations with observations, they found models perform badly in simulating GPP lag and its relationship with soil moisture. Their work told us the timescale of vegetation response should be the main aspect of model improvement.

I highly appreciate that they have thoroughly tested their results by using data from multiple sources and reporting the uncertainty caused by the data sources. This also revealed the inconsistency between observations datasets of vegetation productivities.

2 Specific Comments:

There are some concerns that I think must be handled before re-submit. Firstly, The spatial scale of the precipitation events must be considered. Second, the sample size of the composite should be provided. Addressing these two concerns is essential before resubmitting the manuscript.

The critical assumption they made is that GPP dynamics in a  1 x 1 degree grid box can be a representation of vegetation response to individual precipitation events. However, the scale of precipitation events is highly variable, ranging from 10km to 400 km (Matte et al., 2022; Post et al., 2021). Moreover, spatial variation of daily precipitation is high even on a 2-4 km scale (Augustine, 2010), let alone on a 110 km scale (1 x 1 degree  scale). So, if the spatial resolution of GPP data does not match the spatial scale of precipitation events, it is hard to contribute the GPP anomaly to the individual events. Additionally, using the observation data that aggregated to a coarse resolution, the precipitation event can occur on any part of the grid box, impacting different land covers and causing varied responses of vegetation productivity. So, a supplementary analysis is necessary to show the robustness of the results in Figure 1. The sensitivity test of spatial resolution for aggregation results in Figure 1d. The results in Figure 1d would be highly variable using different spatial resolutions aggregating observation data. If the effect of spatial scale is insignificant, the variation caused by spatial resolution will be less than that from different GPP observation products.

If the supplementary analysis is not possible, some references could be the substitute to demonstrate that  1 x 1 degree grid box has the ability to reflect the vegetation responses to individual precipitation events. For example, Post et al. (2021) found the vegetation response to extreme precipitation can be scaled up to 10 km. However, Zhang et al. (2023) found the coarse-resolution remote sensing images overlooked many greening areas driven by interannual precipitation change.

Secondly, I am wondering about the sample size used to create the composite over regions. The data was from 2000-2014 and there might be many precipitation events in one year. However, due to the coarse spatial resolution, the number of grid boxes in one region might not be that big. Without the sample size provided, it is difficult to have confidence in the results. It would be helpful if you could provide some sample sizes in the results or method section, where the composite is introduced.

Finally, I have a suggestion regarding the presentation of the impact on vegetation productivity. Given that the models used the same precipitation scenario, the sensitivity and the total amount of changes in GPP are essentially equivalent. Therefore, I propose that the authors provide the amount of vegetation productivity change caused by individual events. By comparing the total GPP anomaly of composites from different models, we can directly assess which model overestimates or underestimates the carbon assimilation after rainfalls. It gives a clear indicator of whether the model accurately simulates vegetation productivity and allows for comparison with other remote sensing studies in the future. If authors consider sensitivity to be their primary focus, this suggestion can be dismissed.

Below are some specific questions I want to address, LXXX refer to line XXX in the preprint.

L154: Why does this method have the capability to find the date? Maybe it would be better to have a brief introduction to Lanczos filtering.

L206: The post-event amplitude seems to be too subjective when compared with the other two metrics. I’m curious as to why the researchers chose the time range of 40-60 days for this metric. Was this range based on any references? If they base this range on the paradigm in  Figure 2(a), it might not be suitable for regions or models with lag over 30 days (Figure 2d, see the modeled GPP lag of NorESM2-LM).   In such cases, the GPP only begins to decline from its peak within the 40-60 day range. Consequently, the time range for the post-event period should be adjusted to account for GPP lags.

L213: Why did the GPP be scaled by the ratio of soil moisture? Was it done to eliminate the impact of simulated soil moisture on GPP? By using the scaled GPP, the objective was to measure the sensitivity of GPP to soil moisture instead of precipitation. Readers might be confused. It would be helpful if you could provide an explanation of why the scaling is necessary.

L216: Readers may face difficulties locating relevant analysis later in this section. Maybe it is better to write: “see Figure 4 and analysis later in the section”. Or a clearer way of locating the relevant analysis.

Figure 3: Perhaps there could be an introduction that using the ratio as the radius is a method of normalization.

3 Technical Corrections

Figure 3: “gpp” should be “GPP”.

Figure 4: “vpd” should be “VPD”.

Figure 1A: “gpp” and “vpd” should be capitalized.

Reference

Augustine, D. J. (2010). Spatial versus temporal variation in precipitation in a semiarid ecosystem. Landscape Ecology, 25(6), 913–925. https://doi.org/10.1007/s10980-010-9469-y

Matte, D., Christensen, J. H., & Ozturk, T. (2022). Spatial extent of precipitation events: When big is getting bigger. Climate Dynamics, 58(5), 1861–1875. https://doi.org/10.1007/s00382-021-05998-0

Post, A. K., Davis, K. P., LaRoe, J., Hoover, D. L., & Knapp, A. K. (2021). Semiarid grasslands and extreme precipitation events: Do experimental results scale to the landscape? Ecology, 102(9), e03437. 2023/04/24. https://doi.org/10.1002/ecy.3437

Zhang, J., Zhang, Y., Cong, N., Tian, L., Zhao, G., Zheng, Z., Gao, J., Zhu, Y., & Zhang, Y. (2023). Coarse spatial resolution remote sensing data with AVHRR and MODIS miss the greening area compared with the Landsat data in Chinese drylands. Frontiers in Plant Science, 14, 1129665. 2023/12/22. https://doi.org/10.3389/fpls.2023.1129665

The manuscript titled “Contrasting responses of vegetation productivity to intraseasonal rainfall in Earth System Models” by Bethan L. Harris and other co-authors mainly evaluate vegetation productivity’s response to water availability changes at the intraseasonal scale in in five CMIP6 Earth System Models (ESMs) and also compare the responses with a variety of observation-based products. The results mainly suggests that the models correctly capture that larger increases in GPP at the regional scale are associated with larger increases in surface soil moisture and larger decreases in atmospheric vapour pressure deficit (VPD). However, the sensitivity of the GPP response to these drivers varies between models.

Overall, the results are not new to the reader, and there are also many irregular expressions in terms of the figures and contents. The authors only used five ESMs to study the relationship between productivity and precipitation, and results suggest positive relationship between the two factors. I recommend the authors to dive into the models (or read more relevant model papers) to demonstrate the equations that used in the coupled land surface models to tell the reasons of the different responses between models.

Some other comments:

1. Please check that the abbreviations used in all figures are uniform and standardized. The “GPP” should be capitalized in the figures. And it is suggested to use “VPD” instead of “vpd” throughout the figures to represent vapor pressure deficit. You should use “Observation” rather than “obs”.
2. The legend in Figure 2 (b, c, d) should include an explanation regarding the color of the 1:1 line. And it is suggested to include the fitting equation for each model in the figure.
3. In most of the figures, there is r rather than r2, why?
4. Line 265 to 272 is poor logic flow, please give the explanations.
5. How to see the negative correlation between SSM and VPD in Lines 278 to 279? Please give the explanations.
6. How to calculate VPD from tas and hurs in Line 111 to 112?
7. We have not seen any figure annotations starting from Line 265, and the findings does not correspond to the figures. For example, we could not identify any evidence in Figure 4 that supports your findings in Line 289 to 290. In Line 291 to 299, the article mentions a comparison of slopes. However, we couldn't find any indication of the slope value of the fitted curve in Figure 4, which should be marked in the Figure 4.
8. Line 105, from 1850-2014 should be from ** to ** or during 1850-2014
9. Line 359, the vertical profile of root water uptake may also play a role. We all know that, but how? Give some evidence in terms of model equations or parameters?
10. Line 355, “Further work is therefore needed to understand why different methods for deriving global GPP products result in different relationships with water availability, quantify the uncertainty in these products, and ultimately to obtain observations that will reduce our uncertainty in the response of GPP to intraseasonal rainfall events” Based on my understanding, this is what this work really need to answer, or I can not find new things in this work.