|I thank the authors for addressing my comments. I must admit that I find the manuscript and their reply a bit difficult to follow, in particular because of grammatical errors - I apologise in advance if some of the points I raise below are the result of a misunderstanding. I am concerned that future readers will also find this study difficult to read and understand, which is unfortunate at this stage of the review process.|
In the abstract, the authors now state:
1. “when using raw GCM outputs, streamflow-based weights better represent the mean hydrograph and reduce more biases of annual streamflow than the weights calculated using climate variables”. This is an interesting and potentially impactful result. But where is it shown? I cannot find a Figure or Table supporting the above statement. Table 3 shows the entropy of weights computed using temperature and precipitation, and Figure S1 the weights based on temperature and precipitation, but as I understand it, this does not demonstrate the benefits of computing weights using streamflow instead of temperature/precipitation. In their reply to my comment, the authors refer to P13, L14-18, which is part of the discussion and does not refer to any Figures or Tables in the manuscript. Please clarify. Please also note that the above sentence is not grammatically correct.
2. “when applying bias correction to GCM simulations before driving the hydrological model, the streamflow-based unequal weights do not bring significant differences in the multi-model ensemble mean and uncertainty of hydrological impacts, since bias-corrected climate simulations become rather close to observations.” As also stressed by reviewer 2, this is expected, as by construction, bias-correction forces climate simulations to look like observations, i.e. artificially reduces differences between them (e.g., Hakala et al., 2018), thereby making it difficult to differentiate between good and poor models. Hence, on its own, this result does not justify publication in my view.
The authors conducted additional analysis based on a pseudo-reality experiment to explore the consequence of model weighting under future climate. I thank them for their effort. The results are consistent with those based on current climatic conditions: bias-correction makes it very difficult to distinguish between good and poor models, and there are essentially no benefits to implement a weighting scheme, as the weighting schemes considered do not reduce biases more than an equal weight strategy (see Figures 9d-f and Figures 10d-f). Again, I believe that sequentially applying bias-correction and model weighting based on bias-corrected simulations is a flawed approach.
Furthermore, I am concerned that the results of this study might be misinterpreted. Readers might believe that there are no benefits to use weighting when the climate simulations have been bias-corrected. It is possible, however, that there are benefits of combining bias-correction and model weighting, but I would argue that the model weighting should be done using other criteria than those considered for bias-correction. For instance, the model weighting could reduce the influence of (or exclude) models with erroneous behaviour (for instance climate models creating snow towers, as it is the case for EURO-CORDEX members) or are unable to capture key processes such as atmospheric rivers or atmospheric patterns (e.g., NAO). This could potentially constrain the ensemble (e.g., Padrón et al., 2018) and could be a successful application of model weighting and bias correction. But this would require substantial additional analysis.
Overall, I think that the paper lacks a clear vision and a clear message. In their title, the authors ask “Does the weighting of climate simulations result in a more reasonable quantification of hydrological impacts?” But then they do not clarify what they mean by “reasonable”…
Hakala, K., Addor, N. and Seibert, J.: Hydrological modeling to evaluate climate model simulations and their bias correction, J. Hydrometeorol., 19, 1321–1337, doi:10.1175/JHM-D-17-0189.1, 2018.
Padrón, R. S., Gudmundsson, L. and Seneviratne, S. I.: Observational Constraints Reduce Likelihood of Extreme Changes in Multi-Decadal Land Water Availability, Geophys. Res. Lett., doi:10.1029/2018GL080521, 2018.