The paper is suitable for NHESS. Unfortunately is not acceptable in present form. There are the following main deficiencies:

• In the abstract authors should resume what they did instead of writing a sequence of sentences in which the reader gets lost. If the writer is right, they should write that starting from a forecasting (WRF) corrected using data of 12 meteorological stations 4 distributions combining…………were used.
• The abstract and the introduction seems two separate topics. In the abstract a forecasting model and 4 four logistic models were used combining precipitation and slope, while in the introduction a mesoscale logistic model is used. At lines 86-92 it is written that a mesoscale model provides precipitation data that are corrected with the data of the stations and combined in a logistic model with geomorphological features.
• Authors should explain that rainfall data are obtained computing rainfall by means of a mesoscale model . The authors should also explain what is it a mesoscale model because the reader could not know it.

Line 4. “The  models  were  forced  by  corrected  simulations  of  precipitation  and  geomorphological  features.”  Which models?

Lines 21 “What is it AUC?

Please considers also the references of Tiranti et al. (2014), Devoli and Tiranti, (2018), Cremonini et al. (2018), Piciullo et al. (2020). Moreover the use of models has also be tested in early warning systems against debris flows (Sattele et al., 2015, Bernard and Gregoretti, 2021).

Lines 90-91 “A  database  of  previous  RIL  was  studied (Gomez-Cardenas & Garrido-Urzua, 2018), divided into calibration subsets with subsequent validation of the method” Unclear sentence

Line 102 “which  allowed represent” poor English form

Line 107: what is it a mesoscale? Please explain.

Line 119 “corrected simulations of precipitation” substitute it with “modeled and corrected precipitation data”

Quantities, S, P and E of equations (3) and (4) must be explained in the text.

Line 148  Perhaps “Therefore” would better than “Finally”

Lines 156-158  “The stations were compared in the uncorrected simulation showing (~0.26-0.49) to medium (~0.32-0.67) correlation values by Pearson and Spearman coefficients.” Unclear sentence

Line 245 “a low uncertainty precipitation representation” should be substituted “ “precipitation representation characterized by a low uncertainty“

Line 260  It is “becomes”

Line 262 “The bias-correction using meteolab improved the precipitation representation to compared with weather stations (Figure 4).” Unclear sentence

Line 273  “has a complex topography that triggers precipitation events” How topography can trigger a precipitation? Perhaps the complex topography influences…..

Line 283  “slope memory approach”  what is it? Slope is it relative to the terrain morphology? Please explain

Line 294   “The Andes in one of the most propensity zones to be affected by intense precipitation product of climate change.” Unclear sentence

Lines 294-295 “Moreover, the complex topography needs a high temporal resolution to reproduce the precipitation variability of the Southern Andes.” Meaningless sentence

REFERENCES

Bernard, M. Gregoretti, C. (2021) The use of rain gauge measurements and radar data for the model-based prediction of runoff-generated debris-flow occurrence in early warning systems. Water Resources Research, 57(3), doi: 10.1029/2020WR027893

Cremonini, R., Tiranti, D., Cremonini, R., Sund, M., Boren, B. (2018) The Weather Radar Observations Applied to Shallow Landslides Prediction: A Case Study From North-Western Italy. Frontiers in Earth Sciences, 6, 1–12. https://doi.org/10.3389/feart.2018.00134

Devoli, G., Tiranti, D. (2018) Comparison of landslide forecasting services in Piedmont (Italy) and Norway, illustrated by events in late spring 2013, Natural Hazard and Earth System Science, 18, 1351–1372

Piciullo, L., Tiranti, D., Pecoraro, G. Cepeda, C.M., Calvello, M. (2018). Standards for the performance assessment of territorial landslide early warning systems. Landslides 17, 2533–2546 (2020). https://doi.org/10.1007/s10346-020-01486-4

Sättele, M., Bründl, M., & Straub, D. (2015). Reliability and effectiveness of early warning systems for natural hazards: Concept and application to debris flow warning. Reliability Engineering & System Safety, 142, 192–202. https://doi.org/10.1016/j.ress.2015.05.003

Tiranti, D., Cremonini, R., Marco, F., Gaeta, A. R., & Barbero, S. (2014). The DEFENSE (debris Flows triggEred by storms nowcasting system): An early warning system for torrential processes by radar storm tracking using a Geographic Information System (GIS). Computers & Geosciences, 70, 96–109. https://doi.org/10.1016/j.cageo.2014.05.004 

Review of NHESS-2021-317 by Fustos et al.

Overview

The manuscript (MS) deals with the development of a mathematical tool useful for setting up a landslide early warning system (LEWS) in the Southern Andes, Chile. The methodology combines bias correction of precipitation products and a model for estimating the probability of landslide triggering. The topic is within NHESS and the Special Issue. Language and structure of the MS is acceptable but should be improved. I think the MS has potential for publication, but the current version needs major revisions, mainly because some important methodological aspects need to be explained more clearly and thus it is difficult to understand how scientifically sound are the results.

Specific comments

• Abstract needs to be improved: the various sentences are not adequately linked, so it is difficult to understand what is being done in the MS
• Methodology: it should be better explained how the results are supposed to be used within a LEWS. How is rainfall supposed to be used as input to the developed models to produce a warning? Which is the value of probability for which a warning should be issued?
• Section 3.3: When computing ROC sensitivity and specificity how do you treat observed landslides? I mean, observed landslides are point features, while the output of your model is spatially distributed: how is the comparison between the two done? Is a buffer considered around observed landslide points, or you just take the value at the cell including the point?
• Figure 8 and 9, model 1 and model 4 have basically the same performance. This means that Seven-day precipitation does not add much information. Perhaps the authors should think and comment on this
• Table 2 – Model 4 is the only one combining 3 explanatory variables. Why do not investigate also all the other possible combinations of 3 and 4 variables?
• Figure quality should be improved (all)
• 4.2 is mainly a list of the calibrated parameters for the logit and probit models. Perhaps revise but creating a table with the parameters’ values, while the text comments the table
• Section 5.1 Precipitation accuracy, and about uncertainty in general: the paper may benefit from a more complete literature overview on this point: see, in addition to cited papers e.g.:  https://doi.org/10.1016/j.geomorph.2015.04.028 , https://doi.org/10.1016/J.GEOMORPH.2014.06.015 https://doi.org/10.5194/hess-21-4525-2017 https://doi.org/10.1016/j.jhydrol.2015.10.010 https://doi.org/10.1016/j.geomorph.2016.11.019 https://doi.org/10.1016/j.geomorph.2017.02.001 https://doi.org/10.1007/s11069-018-3508-4 https://doi.org/10.1007/s11069-019-03830-x
• L 134: you use only 20 – 30 % of the data for calibration. Why not an higher percentage?
• It is unclear how you select 57 landslide events from the available 4,987 RIL. 57 events are a quite few, according to the literature (see, e.g. DOI: 1007/s10346-021-01704-7; https://nhess.copernicus.org/articles/21/2125/2021/)

Technical corrections

I have annotated the manuscript with some technical corrections (See attachment).