The manuscript aims to verify the variance exhibited by common GPP models and products in repsonse to the driving factors of GPP. The manuscript is well written in general. In my personal opinion, the abstract and discussion sections are excellent as they nicely summarize the key results in a crisp manner. I have few queries/suggestions for the consideration of the authors.

1) What is the reason for not including the accuracy assessment of the models? Is it because of the lack of independent data for validation? Idea about absolute accuracy would further help the readers to know about the GPP models.

2) Why the downscaled SIF product was considered in this study? It will have the effect of LUE model used in the downscaling and any artefacts of downscaling will be present.  Spatial resolution is an issue especially when the SIF based GPP has to be compared with flux tower observations. But  the downscaling procedure might have affected the spatiotemporal characeteristics of the SIF product affecting the variance metrics studied in this work. If the effect of SIF to be studied, it is better to use the original datasets without much modification.

3) What is the temporal frequency of the reflectance data (SPV) that were used to derive vegetation indices? I understand it to be 16-days. If yes, then it will strongly impact the variances at daily and monthly scales. Why can't daily/weekly data be used?

4) As you rightly pointed out, empirical models (Vegetation Indices models) must be developed for each landcover/climate zone combination. There can be significant improvement in the results had it been the case rather than having one generalized regression model. Landcover specific models can be developed to see if there are any changes in the performance of the VI-based models.

5) In Section 2.2, it will be beneficial to add a brief summary of the tests conducted to test the validity of ERA-5 data in place of in-situ observations. If not in the main article, include it as appendix.

6) In page 6, Equation (4), was PAR considered as a constant fraction of incoming solar radiation? A brief mention about PAR estimation will be beneficial.

7) Equation 6 can be explained with an example.

8) When referring to the appendix in the main text, always refer the figure or table to be looked upon. It became little difficult to identify which figure/table to refer. Further, it was difficult to read the figures given in the appendix. Please improve the readability of all the figures.

De Pue and others investigate multiple methods for estimating GPP against observations. The results are insightful and well written and I recommend the manuscript be published after considering the following rather minor comments.

How is 'homogeneous' defined in section 2.1?

Explaining the Papagiannopoulou et al., 2018 delineations would be helpful because these are not in common usage. I see now that they are defined in the text; pointing to this text would help.

Please use the multiplication sign instead of the star for formal equations

table 2: 'shortwave' can be used

the y axis in figure 3 is a bit confusing to me regarding the '-' symbols (and fig. 4, and 5)

Fig. 7 is a bit much to look at and I wonder if this analysis would be better off in an appendix.