Accurate quantification of the climatic contributions for global land evapotranspiration change is necessary for understanding variability in the global water cycle. This study assembled four ET datasets based on various methodological sources, further adopted the Budyko framework and sensitivity experiment analysis to quantifying the contribution of climatic variables (P, Rn, T, VPD and u) to ET trend. The analysis identified the main climatic factor controls ET trend on a global scale. This research is systemic and detailed, helps reveal the controlling factors of global ET change. The main comments can be found as follows:

1.    The expression should be improved. 

2.    Budyko method was used to conduct a control experiment to compare with ET product results, and the ω parameters were obtained by least squares fitting, did the authors use annual data for the entire period for the fitting? If this were the case, it would not be possible to consider the effect of land use changes on the ω parameters and thus bias the estimated ET simulations, especially considering that such a long study period (1980-2010) with significant land use changes must have an important impact on ω.

3.    Figure 7: As the percentage of grids in each dominant factor controlling annual ET linear trends has been distinguished in Table 2, I suggest to focus on the regions where VPD plays a dominant factor in Figure 7. 

4.    3.3 Validations of attribution method belongs to the 4.2 Uncertainties, as this section discusses the reliability of Budyko method in ET estimation and attribution analysis.

5.    Abstract Line 22: “land-atmosphere interactions” & Page 10 Line 24: “The positive feedbacks”: The main conclusion of this article is demonstrating the main factor affecting ET trend. However, it appears that this study did not address the interaction or feedback between ET and VPD.

6.    As the authors mentioned choice of ET data may add significant uncertainties into the ET attribution. The authors need to show how the impact of the results due to ET datasets uncertainty is reduced and summarize the combined results from multiple data sets, rather than one data set with one result without giving a combined conclusion. And this should also be summarized in Conclusion.

7.    Table 2 gives the percentage of grids in each dominant factor controlling annual ET linear trends with positive and negative. Meanwhile, Figure 2 shows the spatial distribution of annual ET linear trends for 4 datasets, opposite trends between different products in the same pixel can be found. My concern is whether the areas with positive ET trend in one dataset are changing negatively in the other dataset.

Some specific comments:

1    Page 1, Line 25: As you mentioned “terrestrial water flux component”, “accounting for more than 60% of global precipitation” should be “land precipitation”.

2    Page 3, 2.1 Data: Forcing data in Budyko framework and Köppen climate classification should also be summarized.

3    Page 5, Line 35: What’s the meaning of Ci?

4    Figure 4: The image color scheme can be more distinguishable.

5    Page 5, Line 10: How do you define the “dominant factor of ET trends”? Please give an explanation or algorithm.

6    Figure 5 & 8: Please use density scatter plot to improve image quality.

7    Please avoid citing a large number of references in one place.

The manuscript “Attributing of global evapotranspiration trends based on the Budyko framework” by Li et al. investigated the trend of evapotranspiration (ET) at global scale and its contributing factors, including precipitation (P), net radiation (Rn), air temperature (T1), VPD, and wind speed (u), by using multiple datasets (GLEAM3.0a, EartH2Observe ensemble, GLDAS2.0-Noah and MERRA-Land). The methods and datasets used in this study is similar to a previous study (Li. et al., Journal of Hydrology, 2021) by the same author except this manuscript extends previous study in China to global.

This study is more like a numerical sensitivity exercise, suffers from methodological methodological flaws and does not provide insights to understand ET trend and its contributing factors.

Major comments：

1. The Budyko equation assumes that precipitation is the only water supply for ET. At global scale during the study period (1980-2010), many regions have experienced long-term trends in groundwater storage.  For example, in many regions (e.g., the North Plain in China, the High Plain in US, the northern India) where groundwater is used for agricultural irrigation, the depleted groundwater provides an additional source for ET.  In this study, both the analytical framework (Budyko equation) and some of the datasets (e.g., GLDAS2-Noah) do not capture groundwater dynamics.  Therefore, this study only investigated the climatic factors on ET trend and cannot provide a full picture of ET trend.  Even if the ET trend caused by groundwater is captured (e.g., by the remote sensing based GLEAM ET product), this manuscript may mistakenly attribute ET trend caused by groundwater to climatic factors.
2. The parameter w in Budyko equation in Equation 1 is obtained by regression using each set of data product (Line 7-8). I assume that the authors repeat the regression four times using the four sets of P, PET and ET data.  The parameter w is usually associated with land surface characteristics (e.g., land use, vegetation).  However, this study assumes the parameter w is static.  Therefore, the trends of ET caused by land surface characteristics are neglected.
3. The parameter w is more sensitive to regression in arid climate than in humid climate based on Budyko Equation 1. Therefore, without a detailed study of w, the ET trend analysis in this study may be biased for different climate zones.  In addition, as this study uses four sets of data, it is unclear how w’s obtained from each data set are different from each other.
4. It is a bit confusing on the control experiment setup for sensitivity analysis. The impact of a contributing factor trend on ET trend is analyzed by the difference using 1980 data and the 1980-2010 average (Line 30-34).  As there is inter-annual variability in the climate foricngs, why comparing the 1980-year data to 1980-2010 average would reflect the true trend.  For example, if a pixel has a decreasing trend in P during 1980-2010 and a dry year in 1980 (i.e., P in 1980 is below average), the experiment setup then would predict an opposite increasing P trend. Therefore, I am not sure if choosing a different year (e.g., 1981) would lead to different results on the trend analysis.