The diurnal cycles of terrestrial ecosystem gas excanges determine the land-atmosphere interaction and land ecosystem function feedback to climate change. This study first incorporated 170 flux in-situ observations covering 11 ecosystem types to train and test the canopy irradiative tansfer model (SCOPE), Then latent heat flux, sensible heat flux, soil heat flux, GPP and SIF for each flux site generated by the SCOPE model.  Finally，radom forest model machine learning method integrated with  global gridded meteorology and remote sensing was applied to interpolate the site-level variables  to global gridded hourly fluxes. This is an interesting study, and fill the scope of ESSD. And are attractive to the community.

However, I have some major concerns for its current version. 

For the introduction,  the author did not refer to the STEMMUS-SCOPE. I guess the authors try to use the SCOPE model to retrive SIF timeseries for each flux tower? Then I suggest to illustrate the significance of SIF in generating the global hourly gridded fluxes, since the authors have refered SCOPE model in the Abstract. 

For the result, site-level training and test.

1.The author used site-level GPP and SIF to drive the RF interpplation(RF_OI)? Please show the accuracy comparison between GPP_scope and GPP_EC .

2. Figure 3 and Figure 4 are good ways to show the technical issue of RF_OI. But I suggest to add analyze the diurnal variatons of GPP/LE/H for each IGBP class directly between the SCOPE output and EC tower. For example the the comparison (SCOPE v.s. EC tower) of mean diurnal cycle within one year for each IGBP class? within one season? This could tell the readers of message from IGBP classes. 

For the global gridded fluxes, 

1. If I do not miss the global patterns of hourly products, I did not find the global mean (magnitude) and trends map for each three wate-carbon-energy flux.

2. And then inter-comparison of  the mapping pattern between RF_OI and existing hourly products such as FLUXCOM. Currently, the author only show the intercomparison map of LE (Figure 6)?

2. Please explain the rational for the slected 8 regions to decompose the global product. Why not select the global plant function type classes or K-G climate classes.

3.Also, the author showed the diurnal cycle of global LE and GPP hourly fluxes for the 8 regions, respectively, but we did not find the 8 regions analyze for global hourly fluxes.

Land surface water-energy-carbon fluxes are key for understanding Earth’s climate system. However, high-resolution data on water-energy-carbon fluxes at finer temporal scales remain limited. This study produced a new data that estimates these exchanges hourly during 2000-2020 by using STEMMUS-SCOPE model, field measurements, and machine learning with satellite and meteorological data. I believe this dataset could provide valuable insights into diurnal variability and finer-scale land-air processes. I recommend that this paper be accepted for publication after addressing the following comments. 1. Note that estimating the water-energy-carbon fluxes at regional to global scales depend on interpolation processes. The authors applied the optimal interpolation to merge Rn, LE, and H from STEMMUS-SCOPE simulations with eddy covariance observations. Can the authors provide a reason or explanation to why this interpolation is reasonable or why this method can reduce the interpolation errors in the best possible way? 2. Section 3.5: The authors said that they used three commonly used statistical evaluation metrics. What's the third one, except for RMSE and r? 3. The authors should acknowledge the limitations and biases in the STEMMUS-SCOPE simulations in the Discussion Section. Specific comments: 1. line 12: “First the integrated STEMMUS-SCOPE model” ---> “First, …” Suggested to separate by a comma. Similarly, line 15 ---> “Next, …” 2. line 124: What does “Method ML” mean?