This paper presents a valuable case study of a fascinating high magnitude event, namely the anomalous precipitation evet during the 2022 monsoon that led to substantial flooding across the lower reaches of the upper Indus River. The authors provide insight into the impacts of the flood on this high mountain region by evaluating a series of topographic indices, alongside precipitation datasets, to derive stream power proxies mapped in relation to various attributes of the flood. They find that the high precipitation volumes delivered during the event, alongside unusually high temperatures, led to substantial snow and glacier melt, contributing additional flow to the runoff generated by the high precipitation. As a result, very high stream powers were generated at multiple steep locales along the stream network, and these are assessed in relation to measures of channel change inferred from satellite-derived vegetation metrics.

The paper focuses on highlighting the combination of factors that led to the high (inferred) stream powers and therefore its main utility lies in the recognition of flood generation processes in this environment. This is important and understandably, therefore, the paper pays less attention to how the generated stream powers may (or may not) correlate with channel changes inferred from the satellite images. That is, the linkages between the stream powers generated and the channel responses are a story left largely to another analysis. The paper nevertheless represents an important contribution to our knowledge of high mountain flood generation processes and will be of interest to the journal's readership. Of particular interest is the authors' finding that for this event the atmospheric instabilities responsible for delivering the extreme precipitation and temperature anomalies were elevation dependent, which has important implications for evaluating the controls on, and risks posed by, similar events in the future.

The paper is well illustrated and clearly written, and I have few suggestions for further revisions. Nevertheless, the authors may wish to consider the following points:

1) The authors contextualise the flood event, but at L86 they discuss the flooding in the *Lower* Indus valley, seemingly attributing the flooding there to extreme rainfall. This is perhaps a contested view - in the lower basin could much of the flooding have been the result of the poorly maintained canal network - were flood discharges (at peak) generated in the *lower* basin unusually high or not?

2) the authors rely on the use of the CHIRPS precipitation dataset. Readers may value an assessment of any evidence that could support the reliability of CHIRPS in the study region.

3) The authors may wish to comment in more detail on the results of Figure 8, where it is clear that the model precipitation data do not capture the full variability present in the 'observed' data. What is it about the model behavior that means the model seemingly does not represent the high precipitation variability? (Likewise for Figure 9 for the runoff data)

The manuscript "Geomorphic imprint of high mountain floods: Insight from the 2022 hydrological extreme across the Upper Indus terrain in NW Himalayas" addresses an important and timely topic in the field of fluvial geomorphology. The authors' attempt to analyze the impacts of an extreme flood event in a complex mountainous terrain is commendable, and their multi-faceted approach, combining geomorphic analysis, remote sensing, and advanced statistical techniques, demonstrates ambition and creativity in tackling this challenging subject. The study's strengths lie in its comprehensive consideration of multiple factors influencing flood response, including precipitation, temperature, snowmelt, and pre-existing landscape characteristics. The authors' use of various data sources and their attempt to link large-scale climatic drivers to local geomorphic changes is noteworthy. However, despite these positive aspects, the manuscript suffers from several critical methodological flaws that significantly undermine the validity and reliability of its findings. These issues span multiple aspects of the study, including data resolution and quality, analytical techniques, and interpretation of results. The lack of adequate pre- and post-flood comparisons, insufficient validation of remotely sensed data, problematic application of causal analysis, and inadequate error analysis and uncertainty quantification are particularly concerning. Given the severity and pervasiveness of these methodological shortcomings, I regrettably must recommend the rejection of this manuscript. The following detailed comments outline the specific issues that led to this decision, along with suggestions for how the authors might address these problems in future work.

1. Inadequate data resolution and quality: The authors rely heavily on 30m SRTM DEM data (Line 149) for their geomorphic analysis. This resolution is insufficient for accurately capturing the fine-scale topographic changes expected from a single flood event in a complex mountainous terrain. High-resolution LiDAR or drone-derived DEMs (sub-meter resolution) would be necessary for this type of analysis. The authors mention using several datasets (Lines 148-160) but lack specificity about which data is used for each geomorphic index, and how these data were applied. At what resolution was all the data transformed for each index?
2. Inadequate pre- and post-flood comparisons: The authors attempt to use MODIS-derived indices (NDWI, NDSI, EVI) for change detection (Lines 158-160). However, their approach has several limitations :a) Temporal resolution: The authors don't specify the exact dates of the pre- and post-flood images used. Given that MODIS provides daily or 8-day composite products, the selection of these dates is crucial and could significantly affect the results.b) Spatial resolution: MODIS data typically has a spatial resolution of 250-1000m, which may be too coarse to capture detailed geomorphic changes in a complex mountainous terrain.c) Lack of quantitative analysis: The authors present qualitative descriptions of changes in these indices (Lines 350-356) but fail to provide a rigorous statistical analysis of the changes. For example, they could have conducted a pixel-by-pixel comparison and presented statistics on the percentage of area showing significant changes. d) Limited interpretation: While changes in vegetation indices can indicate flood impacts, the authors don't adequately address how these spectral changes relate to specific geomorphic processes or landforms. They make broad inferences about channel morphology changes (Lines 357-360) without directly linking spectral changes to field-observed geomorphic features.e) Absence of complementary data: The use of optical indices alone is insufficient for comprehensive flood impact assessment. The authors could have strengthened their analysis by incorporating other remote sensing data, such as SAR imagery for flood extent mapping or high-resolution optical imagery for detailed change detection. The authors rely heavily on remote sensing indices (e.g., EVI, NDWI) to infer geomorphic changes (Lines 354-363). However, they provide no ground-truthing or field validation of these inferred changes. Without this validation, the reliability of their interpretations is questionable. The SAR/Landsat data can serve as a proxy in most cases
3. Problematic causal analysis: The authors employ the PCMCI (Peter and Clark Momentary Conditional Independence) algorithm for causal discovery among hydro-climatic variables (Lines 244-270). While this is an advanced method for time series analysis, its application in this study has several significant issues:a) Assumption of causal sufficiency: The PCMCI method assumes that all relevant variables are included in the analysis. However, the authors don't justify their selection of variables (temperature gradient, rainfall gradient, and anomalous change indicators) as a comprehensive set for describing the complex geomorphic system. Important factors like soil moisture, vegetation cover, or tectonic uplift rates are not considered, potentially leading to spurious causal relationships.b) Linear relationships assumption: The authors use the ParCorr linear independence test (Line 261), which assumes linear relationships between variables. However, geomorphic and hydrological processes often exhibit non-linear behaviors. The authors don't address this limitation or justify why a linear approach is appropriate for their data.c) Temporal resolution mismatch: The authors use a maximum 2-day lag period (τmax = 2) for their analysis (Lines 267-268). This short-term focus may miss important longer-term causal relationships in the geomorphic system, which can operate on much longer timescales. d) Lack of robustness testing: The authors don't present any sensitivity analysis or robustness checks for their causal discovery results. It's crucial to test how the identified causal relationships change with different parameter choices (e.g., varying τmax or significance levels). e) Interpretation issues: The authors present their causal graph (Fig. 10) without adequately explaining how to interpret the results in the context of geomorphic processes. They don't clearly link the statistical relationships found to physical mechanisms of landscape change. f) Temporal scope limitation: The analysis is limited to the July 1 to August 31, 2022 period (Lines 265-266). This narrow timeframe may not capture the full range of causal relationships relevant to the flood event, especially considering potential antecedent conditions or delayed effects.
4. Unsupported stream power calculations: The authors' novel approach to calculating stream power (Lines 204-224) incorporates precipitation data, but they fail to validate this method against established stream power calculation techniques or field measurements of actual stream power during the flood event.
5. Insufficient error analysis and uncertainty quantification: The authors fail to adequately address uncertainties in their analysis, particularly in their Random Forest modeling (Lines 227-243). Specific issues include: a) Model performance metrics: No information is provided on standard evaluation metrics such as R-squared, RMSE, or Mean Absolute Error for the Random Forest predictions. b) Validation strategy: The authors don't specify their model validation approach (e.g., k-fold cross-validation, hold-out validation set).c) Feature importance: While they mention variable importance (Lines 372-380), they don't provide quantitative measures or visualizations of feature importance. d) Sensitivity analysis: There's no exploration of how model results change with different parameter settings or input variables. e) Propagation of uncertainties: The authors don't discuss how uncertainties in their Random Forest predictions might affect subsequent analyses, such as the causal discovery or stream power calculations.
6. Inappropriate temporal scale: The focus on only July-August 2022 (Lines 148-151) is too narrow to capture the full geomorphic impact of the flood event. This timeframe doesn't account for potential delayed landscape responses or longer-term geomorphic adjustments.
7. Insufficient error analysis and uncertainty quantification: The authors fail to adequately address uncertainties in their analysis, particularly in their Random Forest modeling (Lines 227-243). Specific issues include: a) Model performance metrics: No information is provided on standard evaluation metrics such as R-squared, RMSE, or Mean Absolute Error for the Random Forest predictions. b) Validation strategy: The authors don't specify their model validation approach (e.g., k-fold cross-validation, hold-out validation set).c) Feature importance: While they mention variable importance (Lines 372-380), they don't provide quantitative measures or visualizations of feature importance. d) Sensitivity analysis: There's no exploration of how model results change with different parameter settings or input variables. e) Propagation of uncertainties: The authors don't discuss how uncertainties in their Random Forest predictions might affect subsequent analyses, such as the causal discovery or stream power calculations
8. Lines 331-363: The discussion of spatial distribution of hydro-climatic anomalies could be strengthened by including a statistical analysis of the relationships between different variables. Given the spatial nature of the remote sensing data, the authors should consider employing spatial statistical methods such as:

Geographically Weighted Regression (GWR): This technique could be used to explore the spatially varying relationships between precipitation anomalies and other variables like temperature, snowmelt, and runoff. GWR would allow the authors to identify how these relationships change across the study area, potentially revealing important local variations in flood response.