**5. Conclusions**

The contribution of terrestrial photography for the definition of the relation between the Fractional Snow Cover and the spectral behavior of the surface is a major issue. Ground-based cameras represent a valuable proxy of data useful for investigating the snow cover extension over a long period. From this perspective, terrestrial photography can be used as ancillary information and it supports the integration among different multispectral remotely sensed datasets. The availability of an automated procedure useful for the discrimination between snow and not-snow covered surfaces can support the analysis of large datasets. The selected approach based on Spectral Similarity was compared with supervised methods and with the Blue Thresholding procedure on a training dataset. Considering the supervised methods as a reference, the Spectral Similarity approach showed better performance

in estimating the snow cover area. Furthermore, expanding the dataset to a 10-year terrestrial image record, the algorithm increased the capability to estimate the Fractional Snow Cover under a larger range of conditions compared to the state-of-the-art method. The integration with three different satellite snow products (Landsat, MODIS and GlobSnow) highlighted the potentiality to define a site-specific relation and threshold useful for isolating the snow cover area from remotely sensed data. Finally, the support provided by terrestrial photography enhanced the possibility to detect artifacts associated with clouds and shadows.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2076-3263/9/2/97/s1, Figure S1: Example of comparison between output obtained by automated classifications on a terrestrial image, Table S1: Training dataset composed by FSC estimated using supervised and automated methods (Figure 4a), Table S2: Complete dataset composed by FSC estimated using only automated methods (Figure 4b), Table S3: Complete dataset composed by FSC estimated using satellite products (Figure 5).

**Author Contributions:** R.S. (Roberto Salzano) and R.S. (Rosamaria Salvatori) worked on the algorithm concept, performed the analysis and prepared the manuscript; M.V. supported the definition of the image dataset, the data analysis and the interpretation of results; G.G. and B.C. provided the access to the Swiss Data Cube consequently making remotely sensed data available; L.I. performed the classification of images using supervised methods.

**Acknowledgments:** This work has been carried out within the ESSEM COST Action ES1404 and the authors would like to acknowledge the European Commission "Horizon 2020 Program" that founded the ERA-PLANET/GEOEssential project (Grant Agreement no. 689443). English revision and spell-check by Lena Rettori. We wish to acknowledge three anonymous reviewers that really helped to improve the quality of the paper.

**Conflicts of Interest:** The authors declare no conflict of interest.
