This study reported a global product of hourly GPP and ET from 2001 to 2020. In general, this study is important and valuable because none of previous GPP and ET products focused on the hourly scale. However, the authors need introduce the methods more clear especially for the parameterization, which is quite important to judge this dataset is reliable. Besides, I also hope the authors can show the more details of parameters’ distribution and validations. Please see my detailed comments below. Vegetation transpiration is simulated by the supplementary Equ. 8. Did the authors set the same Rn-G to shaded and sunlit leaf? Obviously, there are large differences of latent heat for shaded and sunlit leaves. Line 115: why did you use GLOBMAP LAI data? The spatial resolution of GLOBMAP is quite coarse. Line 150: it is very important to know the details of optimization algorithm. It seems the parameterization method has not been published. Although the authors mentioned the supplementary materials, but I still did not get how the authors optimize the model parameters. Especially, you mentioned the parameters were optimized for each month at each site-year. From 2.3.1 and 2.3.2, I assumed the authors first inversed two model parameters m and Vcmax at eddy covariance sites, and then used machine-learning method to generate global gridded dataset of m and Vcmax, and finally, BEPS model was run based on the gridded m and Vcmax to estimate global GPP? if it is so, why the authors did not just upscale GPP from towers to global scale, just like Jung et al. 2009. Besides, the authors did not show the global patterns of m and Vcmax at all, and we cannot judge if their distributions are reliable. And it is also necessary to show the performance of machine learning method to simulate m and Vcmax at eddy covariance towers, which is quite important than GPP. In addition, I am curious that if the authors did not use gridded parameters, and just used site-based inversed parameters to simulate global GPP, how is the performance? Fig. 4 showed the better performance of hourly simulations than daily simulations. Is it possible? I am curious how the authors examine the performance of hourly simulations, and if the authors included all simulations of night and daytime together, which will result in a false high correlation. As this study aimed to produce an hourly GPP and ET, if there are large difference of parameters diurnal scale. Figure 9: the comparison does not make sense as the products used different LAI datasets, and the different trends basically depends on the trend of LAI datasets.

This study presented hourly estimation of global GPP and ET dataset at a 0.25° resolution from 2001 to 2020. This product could be of great importance in understanding the diurnal ecosystem functionality, and promote monitoring of ecosystem responses to extreme climate events. However, there are several concerns regarding the methods explanation and result presentation that could dampen the science of this study.

Methods: 1. need particular explanation on the newly revised BEPS regarding how exactly the hourly GPP and ET were simulated but not from the previous version. If BEPS was designed to simulate the hourly products, were if just because the hourly inputs were not available before?

1. Need a better presentation and validation on the newly optimized key photosynthesis and stomatal conductance model parameters (i.e., 𝑉cmax and and 𝑚). Was it only revised for the flux sites (as presented in Figure 2)? How were they interpolated into the global scales and what are their uncertainties? How about the spatial and temporal variations of these parameters globally? Also, what will be the differences in the simulated GPP and ET between the new dynamic parameters and original fixed parameters in terms of accuracy and spatial pattern?
2. Figure 1. The bottom-up order is quite counter-intuitive to readers. Suggest using top-down order.

Results: 1. Figure 4, suggest adding the label of 1 in the slope subplots as a reference of good fitting.   

1. Need better presentations on the diurnal patterns of GPP and ET for different vegetation function types. For example, providing hourly curves for different vegetation function types and validated against flux site observations. It still unclear whether these products can capture the diurnal variations of GPP and ET.