This research provides a large ensemble of water balance ET estimations using existing global datasets. Sound comparisons with mainstream ET products have been reported. Uncertainty assessment is also performed on different scales. The manuscript is within the scope of ESSD journal. Minor revisions need to be considered before publication. Please find details below:

(1) ET-WB incorporates both the natural ET from various sources and human-induced part, e.g., irrigation. What about the products used for validation, particularly for GHM such as WaterGap? Can human activity be a significant reason causing difference between ET-WB and them?

(2) The authors evaluate ET-WB on several time scales longer than a month. What about the performance on the sub-monthly scale? This point should either be supplied or discussed.

(3) The authors explain the spatial pattern of uncertainty in ET-WB by individually presenting the uncertainty in different variables. However, the temporal changes lack necessary justifications, e.g., the decrease of auxiliary ET products in 2016 according to Fig. 9f.

(4) It would be good if the authors could add the uncertainty range of the selected auxiliary ET products in the comparisons with ET-WB in terms of the long-term mean and annual trends.

(5) Fig. 1. The pie chart representing the irrigation percentage rate should be enlarged.

(6) You may indicate the explicit number of datasets used in your workflow in Fig.2.

(7) Table 1 may go to supplementary file as (1) it occupies too much space, (2) it is somehow duplicate with Fig.2.

(8) For colored shading plots like Figs. 3 and 5, the authors should indicate the meaning of the central red line in the legend.

(9) Line 702-704: ‘The human-induced inordinate fluctuations of water balance (e.g., through reservoir management, groundwater extraction) can influence the quality of ET-WB by impacting the accuracy of the specific forcing variable (e.g., R).’ The statement needs to be re-organized for better comprehension.

(10) The authors should either direct the authors to Table S2 or explain the meaning of the rank of basin ID in Fig. 10.

The paper addresses a critical aspect of water cycle analysis by developing a water-balance based global ET dataset. It assesses uncertainties of multiple open access datasets on precipitation, runoff and storage change and performs water balance analysis on their combinations to derive ET by major river basin. This ET data is then compared with four auxiliary datasets developed through different methods. 

The input data, methods and results are described in details and presented in clear and well structured way. The dataset is only provided in MATLAB proprietary format, which may limit its accessibility.

Specific comments:

• Data redundancy and circularity implications could be discussed. For example, runoff models are likely calibrated on GRDC archive, which is also directly used as input data;
• While results are also presented geographically in the paper, the output dataset is only distributed as matrix in mat file. Providing access to the dataset in different formats, including spatially distributed formats, would probably increase its usability;
• the temporal extent of datasets used in the study (Table 1) could be explained. For example the auxilliary ET datasets are used over a shorter temporal span without a clear justification;
• Line 252, CHIRPS data is a combination of remote sensing and stations data, should it be considered as a combined product?
•  Line 35, typo, "for" should read "four";
• Line 96-97, consider rewording for better clarity.