General comment:

Runoff change attribution is an interesting research topic. In this manuscript, the authors proposed a novel framework to quantitatively evaluate the contributions of climate change and human activities to runoff changes in the Lancang River Basin. It provides an optional method for attribution of runoff changes on watershed scale, which is valuable. Generally, the manuscript is well organized and presented interesting results, but the description of the innovation of the study should be enhanced. 

My specific comments:

# The authors did not present the advantages of the proposed method clearly. Please clarify what’s the innovation of this study.

# Line 171. I suggest the authors clarify the reasons for selecting the dataset in the study (GMSWU). Is it because of the higher accuracy of the data set, difficult to obtain other field data or some other reasons?

# Line 181: It is not "avoid the common phenomenon of ‘equifinality for different parameters’ in hydrological simulation", in fact, the phenomenon of "equifinality for different parameters" in hydrological simulation is unavoidable. Suggest to revise it to "reduce the impact of "equifinality for different parameters" in hydrological simulation".

# Line 245, If multiple break points are detected for the annual runoff time series, how to select the break points and divide the whole period to natural period and impacted period? The authors should clarify this in the text.

# Line 290-303, The proposed method of quantifying the contributions of climate change and human activities to watershed runoff changes may not guarantee that the sum of the contributions equals to 100% (e.g., equations 6 and 7). This is to say, there is an intersection between climate change and human activities. It is recommended that the authors should give explanations about this issue in the discussion section.

# Line 475, Why the authors present the normalized runoff process of the Yunjinghong station? Please clarify the reasons.

# Section 4.4, In this section, the authors compared the results of the new quantitative framework proposed by the manuscript with two simpler methods. What’s the advantage of the proposed method over the two methods? please clarify.

# Fig 15, How did you produce land use maps? Please clarify the satellite data that you used. The presented land use maps in 1980, 2000, 2010 and 2015 look like similar, no obvious land use changes can be identified. Figure 15 shows 6 types of land use while table 9 shows 7 types. They should be consistent. 

Technical corrections:

# In addition, the format of the listed references is not uniform. For example, some references have DOI information, but some do not (Lines 831, 846, 915, etc).

# If some nouns have been abbreviated in the manuscript, please use the abbreviation after the first occurrence and keep consistent through out the whole text. For example, in line 13, please replace “contribution rate” with “CR”. The same problem exists in other noun expressions in the manuscript. Please review the full text in detail and make consistent revisions (Line 16: human activities, Line 19: Lancang River Basin, etc.).

# Table 5, The titles of the hydrometeorological elements in the table should be consistent. At the moment, some of them are full names, and some of them are abbreviations.

The authors evaluated the attribution of climate change and human activities to streamflow variations with a posterior distribution of hydrological simulations. The contribution distribution has been evaluated in many hydrologic fields using the different methods, however, the posterior distribution has been rarely considered. The author tried to provide a solution to evaluate the attribution of climate change and human activates considering the simulation uncertainty.

1. Line 65-70, the second typed of method seems to be no difference with the third type of method. For example, I can understand that controlling human activities should be to change the climate factor like the second method simulating multiple scenarios by changing one impact factor. What is the nature difference for the two methods, please describe more clearly.
2. Line 91-94, as you stated, Farsi and Mahiouri(2019) has also analyzed the uncertainty of hydrological simulations in the process of quantifying the CR of CC and HAs to streamflow changes, however, you thought they only constructed the posterior distribution of the contribution rates of climate change and human activities, and did not specify the accurate contribution rates. I don’t think so, Farsi have provided the PDF of contribution rates which can tell high-probability contribution rated clearly. Therefore, what is the innovation for this study differing from the previous studies, not averaging the contribution rates with high-probability NSE.
3. Line 344-345. The plant-avaiable water coefficient is set to 0.5, why? It is just to match the result of your proposed method? In addition, the contribution method should be not your finding, is it?
4. L440, the sensitivity analysis method is not describe clearly, SUFI-2 is a optimal method, and how it conduct parameter sensitivity analysis, what is its relation with Latin hypercube sampling
5. About “equifinality for different parameters” of hydrological simulations. The author selected some experiments with high NSE larger than 0.75 to construct the posterior histogram frequency distribution (PHD) of the contribution rate of climate change and human activities to streamflow changes, and then quantify the contribution rates with higher probability. I think it does not solve the “equifinality for different parameters”. The simulations with higher probability still exist the “equifinality for different parameters”. To exclude it, the uncertainty (pdf) of parameter should be analyzed to screen out abnormal parameter values. After that, the contribution rates with higher probability are really results excluding the “equifinality for different parameters”. these suggestion may be added into the discussion section in revised manuscript.