Comment - In general: the symposium deals with open source software and so I think that the GMapCatcher software (the only one to be open source among those cited in the paper) should be better described and, possibly, pro and cons in it use must be highlighted. I would ask to the authors to extend the text in this sense, even if this will require to remove some other parts from the manuscript to remain into 1500 words. Moreover, I suggest to add a sentence that put in evidence how the open source software helps the interoperability.
Response - Thanks for the useful comment. GMapCatcher description has been extended and another reference to a second Open Source complete suite for downloading and managing high resolution freely available imagery has been added (SASPlanet). It is right that these suites are the only Open Source software mentioned but the main focus of the paper is on the exploitation of freely available high resolution images for land characterization and potential risk assessment. In this sense the work wants to come closer to the broadly intended Open (Free) Data world, showing a potentially interesting application for land management purposes. The combination of data gathering by using Open Source software and analyses on freely available data is in this way presented as a tool to leverage for multi-temporal analysis and risk assessment on selected areas of interest.
Comment - row 47: I suppose it is monthly average temperature. Please specify.
Response - thanks for the comment, monthly average temperature has been specified
Comment - row 60: Which is the date of the Bing Images? it is 2006? it is important since you are comparing them with a 2006 orthophoto.
Response - Bing images (now HERE company) are dated 2014. Notwithstanding the time span, the majority of the eroded areas and the extent of the vegetated ones has remained approximately similar. In any case, to avoid bias in the analysis the effectiveness of the approach has been evaluated from a qualitative point of view at catchment scale and from a quantitative point of view on a shallow landslide that remained unchanged during the surveyed period. Furthermore, for the selected area, same illuminating condition can be appreciated and no important shadows are affecting the two images. This, in combination with ad hoc training areas sets for the orthophoto, including or ignoring the role of the shadows, has helped also for an assessment of the performance of the approach independently from the influence of the shadows on the final results.
Comment - row 61: please change ""free"" to ""Open Source"", since people can understand ""free as a beer"" and not "free as a speech".
Response - Changed, thanks, I agree, free speech would be the best conveyed message but it is often a hardly conveyable message.
Comment - row 85: Did you use the training data to evaluate the specificity and sensitivity or did you prepare also a validation dataset? Moreover, I think it is interesting to provide the values of specificity and sensitivity for satellite and orthophoto? Moreover: the specificity measures the true negative rate. I suppose that, in your model, the negatives are represented by ""no-vegetation"" and the positives by ""vegetation"". It can be useful to clearly declare this.
Response - Thanks for the useful comment, we modified the work carrying out an analysis of the performance. Validation has been carried out on a 1000 m2 shallow landslide with no shadows. Furthermore, several ad hoc training areas including or excluding shadows both in vegetated and rock areas (including separate scenarios) have been used but, notwithstanding this, the effectiveness of the satellite based image analysis has always shown a far better matching. The performance related to correctly identified vegetation-covered areas was not evaluated since the result in this case is very much related to the training sets used for the orthophoto based classification (including or excluding shadows). The important feature to be noticed is in our opinion the discriminant power of the satellite-based classification especially in shadowed areas. Shadowed areas in the upper rocky part of the catchment are correctly classified as no-vegetation and shadowed areas under vegetation are correctly classified as vegetation. This discriminant effectiveness is not reproducible even by using ad hoc training areas, when considering the orthophoto.
Comment - row 97: probably "a" must be removed
Response - Removed, thanks.
Comment - row 102: please insert somewhere ""(Figure 5)"
Response - insert in text (Figure 2 to 5)
Comment - row 108: please add a dot ""."" at the end of the sentence.
Response - Added, thanks.
Comment - row 110: may be the authors want to change ""between"" with ""in"".
Response - Changed, thanks.
Comment - row 111: maybe I'm wrong but, IMHO, the discrepancy really depends on the incidence of the light during the shots.
Response - Thanks for the suggestion, an integrative analysis has been added. The influence of the light has been assessed by using different training areas sets for the orthophoto, including or ignoring shadowed areas (both under vegetation and in the upper rocky portion) and quantifying the performance on a 1000 m2 evenly illuminated shallow landslide area. The performance of the satellite based image has revealed always far better than the orthophoto and shadows seem only to play a role on the wrong classification of the vegetated areas, not on the wrong classification of landslide illuminated areas. This behavior can be ascribed probably to the process of creating and sharpening of the satellite image that, within the RGB bands, is maybe carrying some extremely useful NIR information, but this is only a guess.
