|The revised version of the manuscript is a clear improvement to the original version. The authors carefully prepared this revised version and complemented this with a detailed response letter, were they justify and explain their changes and arguments to the different comments raised. Namely they also invested considerable efforts in revising the figures and tables which definitely provide an added value now. In my view, this is now close to be acceptable, I only have a few, probably minor points that still need to by adjusted in my view.|
My first general comment in the initial review was referring to the terminologies around hazard and susceptibility. Related adjustments in the new manuscript regarding this point are good and helpful, namely the consistent renaming of what was initially called “GLOF hazard parameters” or “diagnostics of GLOF potential” to now “predictors of GLOF susceptibility” is a clear improvement. The only point where I am not yet convinced is the statement in the conclusions (L481-482): “In any case, the widely adapted notion that (rapid) lake growth may be a predictor of impending outburst remains poorly supported by our model results”. In my view lake growth is not mainly considered as an outburst predictor in these assessments (i.e. a proxy for a high outburst susceptibility), but rather as a predictor of increased downstream hazard potential due to increased potential flood volume and thus hazard magnitude. I might be wrong, but since this statement is also used in the abstract, I would ask the authors to more clearly elaborate from the cited publications about these assessment schemes, if any why lake area changes are really considered as predictors of lake outburst susceptibility (rather than indicators of increased potential flood intensities downstream). This could be done, for instance, around L122 where it now says “and many studies have emphasised the role of growing lakes on GLOF susceptibility (Bolch et al., 2011; Prakash and Nagarajan, 2017; Rounce et al., 2016)”.
The second aspect where I am not yet 100% convinced is the point on using catchment area instead of lake area. The authors provide convincing arguments in their response letter and include this also in the revised manuscript (catchment area representing the potential of intense rainfall runoff and snow melt, as well as being invariant at the investigated time scales). But then I wonder why catchment area has not been used in all the models instead of lake area. Or why is this aspect more important for the glacier-mass balance model and the monsoonality model, an less for the elevation dependent and the forecasting model? In my view, a few more words on that would be needed, probably in the paragraph ending on L130.
Then also the reasoning of selecting “lake-area change” as a predictor (Table 2) as well as the response to my detailed comment #4 (lake depth, not area is determining the hydrostatic pressure on the dam) do not convince me yet. Most glacial lakes, once they have a certain size, mainly increase their area (and thus volume) but not their depth while the grow further. It is true that there are depth and volume estimation rules based on lake area, but the uncertainties of these approaches are huge, and they are based on empirical relationships (of static conditions) and do not involve any dynamic considerations, not any physical processes. If I think of any larger moraine dammed lake in the Himalayas that is growing due to glacier retreat at its upstream end (and thus increasing its area and volume), I would not expect a significant increase of lake depth. L119-121 and the reasoning in Table 2 should be reconsidered in my view.
Finally two more details on Table 1:
In the new Table 1, the predictors “glacial lake area” “lake-area change” should be moved from “lake characteristics and dynamics” to the “potential triggering mechanisms” group in my view. As the authors mention in the text and the response, lake area (increase) is a proxy for (increased) probability of the lake being impacted by an upstream mass movement.
Further, I think it is very unfortunate no reference is given for the “summer precipitation” predictor (last line in Table 1). Not that I could make a good suggestion, but having no reference here contradicts the study description (L72-73) “[…]we tested how well some of the more widely adopted predictors of GLOF susceptibility and hazard fare in a multi-level logistic regression […]”. Either some reference(s) should be added to this last line in Tab. 1, or a statement on the summer precipitation predictor should be made around this statement in the introduction.
The first three points require at least a few minimal adjustments in the manuscript. The details regarding Table 1 might also be explained and justified in a response if the authors prefer not to consider this in the manuscript.