The paper was reviewed by six reviewers to which the authors have responded and extended the text. Removing the overly positive and negative reviews, leaves 4 balanced ones of which one called for rejection and three for major revisions and a second round of review.

A general undertone of the reviews was a concern about the methodology and the interpretation in your work. The responses did not address the essence of some of these concerns. Before we proceed any further with this paper I would like therefore to ask that the authors respond very clearly to the following 5 key questions:

(1) The water balance equation dS/dt = P + I – E – O (Eq. 5) is applied to the basin and the anomalies of the different components are quantified by the ratio C in Eq. 6. The 4 hydrological components are listed (magnitude and anomaly) for the 9 studied droughts in Table 2 and the ratio C in Table 3. In Table 2, the term dS/dt = P + I – E – O does not add up the number reported there (nor does the anomaly) (I tested the first drought only). Furthermore, the ratio C reported in Table 3 cannot be reconstructed from the authors Eq. 6 in the text. Although the terms do add up to 1 (as they should), their magnitude is computed in some other way. Please explain.

(2) The relative contributions of each hydrological component in Table 3 as reported indicate high variability between droughts. A key statement and conclusion of the authors is that “overall for all drought cases, 49% of the total water anomaly of the water budget came from reduced inflow, 24% from increased outflow and 18% from decreased precipitation”. I imagine this was computed from the total integrated hydrological component contributions of all droughts. How (see (1) above)? Consider that each individual drought had very different contributions, so how meaningful is this average contribution? Is it accurate and correct to summarize the conclusions on the contributions of each term to droughts in this way? Please explain.

(3) Connected to the above question, I iterate one of the reviewer’s statements that you have not answered properly: the contribution of the hydrological components during a drought are a sort of contribution to drought maintenance, not to the actual emergence (start) of a drought. In other words, a lack of precipitation in the months preceding a drought may be more important for the drought begin and nor precipitation deficiency during a drought, which may not be reflected at all in your analysis. Is this correct? Please explain.

(4) It is stated that mean annual evapotranspiration is computed from the “difference between precipitation and streamflow”. Which streamflow? for the entire basin or inividual subbasins? If evapotranspiration for the Lake Basin is estimated from streamflow data of the subbasins, then the relation in Fig 4b is trivial, even if it is only for individual droughts. Please explain.

(5) The study is based on identifying droughts from the monthly lake level data L, but it analyzes the contributions of the individual hydrological components from the general water balance equation dS/dt = P + I – E – O. On the average, the water balance equation is closed because mean annual E is chosen such that mean dS/dt = 0 (see (4) above). Throughout the paper there is no direct assessment of the performance of the simple water balance equation for the basin. Comparing lake level data (dL/dt) and dS/dt would clearly given an indication how applicable the model/data are to study the posed problem, even though the lake level-volume rating curve may be complicated. Right now there is a disconnection between using L to define lake droughts and then other data (precipitation, streamflow) to define the water balance in those drought periods. In other words, if you defined droughts from the storage in the system S by integrating dS/dt in time, would you come up with the same droughts as you did with L? If not, how meaningful are then the drough period analyses of the hydrological components. Please explain.

The Chief Editor, prof. E. Zehe is monitoring the progress of this paper. We will wait for your explanations to the question above and then decide how to proceed with the current revision.

Peter Molnar

Dear authors,

Thank you for responding to my questions. Despite making corrections to the calculations, your analysis results remain irreproducible to me. I will stay with question (1) on my list. You have corrected the water balance equation dS/dt = P + I – E – O in Table 2, however the relative contribution for each water component with the new equation (7b) does not make sense, you are making a double sum over k from m to n and then from m to j. What does this mean? I still cannot reconstruct how you obtain Cx in Table 3 with equation (7b) using data from Table 2. Before I send out your revised paper to the reviewers for their assessment, I need you to explain to me step-by-step what exactly you are computing, best with an example of drought 1, so that I can follow your reasoning with your numbers. Thank you.

I apologize for the inconvenience if I am not seeing something obvious. But I cannot take the risk of having erroneous computations in a HESS paper and asking colleagues to spend time on reviewing results that are not correct.

Best regards, Peter Molnar

Dear Authors,

Attached you will find the three referee reports of your revised manuscript. Two of them are quite positive and one questions the application of the SLI concept. In your revision please take special care to address the concern raised in Report 1 regarding the verification that the monthly lake levels submit to a Gamma distribution. Since this reviewer suggested a rejection of your paper, you have to address and answer carefully this concern and provide evidence the reviewer is asking for.

Best regards, Peter Molnar

Dear Author,

You have responded to reviewer 1 questioning the fit of the Gamma distributions to lake levels. I have two questions here that I need to verify with you:

(a) Is the description of the use of SLI in Equation (1) according to your correct? In the SPI context of McKee et al. (1993) the Lij are Gamma distributed and transformed to a standard normal distribution variate Z by mapping the same exceedance probability. This distribution by definition then has E(Z)=0 and Var(Z)=1. If Lij's in Equation (1) are already transformed (or normalized as you write in 207), then removing the mean and dividing by the standard deviation above is pointless. Can you please explain this operation correctly.

(b) Reviewer 1 asked you to demonstrate the fit of Lij's to the Gamma distribution. You did this with the new Figure 2. I inspected closely this Figure and I have to say I do not see any variability from the Gamma distribution in the data at all. Are you sure you are doing the right thing here? I have never seen lake level data so perfectly fit a Gamma distribution for every month. Please be so kind to check this for me.

Peter Molnar

Dear Authors,

In the submitted version you have not replaced the corrected Figure 2. In the text to that figure you need to say that the Gamma distribution did fit the data reasonably well.

I have language edited your text to make it more readable. My suggestions are attached. Play close attention to articles and text clarity. There are also some places where repetition is excessive. I have highlighted those.

Please make these final edits to your document.

Peter Molnar