I would say that they're non-sense values, and you could have a problem with your input data. In the link http://blogs.esri.com/esri/arcgis/2010/10/27/understanding-curvature-rasters/ you can find an expanation about curvature values. I don't work with ArcGIS and don't know which units does it use, but in others GIS, as GRASS, the curvatures are expresed as 1/radius, where radius is usually expressed in metres, as you can see here: https://grass.osgeo.org/grass70/manuals/r.slope.aspect.html.

So I would suggest you to revise your raster data, maybe it is not in the proper units/format for using that functions.

You can explore this subject in detail in "Geomorphometry in GRASS GIS", by Hofierka, J., Mitasova, H., Neteler, M. (2009) [in this link: http://www.sciencedirect.com/science/article/pii/S0166248108000172]

Thanks a lot Santiago.

Best regards.

Another piece of software extremely suitable for digital geomorphometry is SAGA GIS (saga-gis.org) that originated from a software specifically written for that purpose. In SAGA, you can calculate different curvatures (e.g. plan/horizontal, profile/vertical, tangential, etc...) based on a number of different ways to produce the functions fit to the DEM (e.g. Zevenbergen and Thorne 1987 etc). I would agree with Santiago that your values are likely nonsense values; the curvature in SAGA is also expressed with the unit 1/m, referring to the local radius of curvature.

Cheers !

I have to say I'm working on an extremely extended area. Whenever I try to load this raster (.tiff) with SAGA, it gives me back an error warning me that the file is abnormally big to be processed. Anyway, I think that the range of strange values obtained using GIS could be explained by the high geomorphological heterogeneity of the landscape I'm focusing on. I tried to clip, from the curvature layer, some points dispersed along the coastal belt, which is more homogeneous. In this case I had a range of sense values (dividing these by 100 as suggested in many GIS forum).

Thanks

Thanks to you for sharing valuable knowledge.

I would like to second the answer from Manuele Bazzichetto and make this result clearer. The values from ArcGIS tools should be divided by 100. The tool documentation says this (in an obtuse way) but the 'How Curvature Works' makes it more clear that they multiply values by 100 to make the numbers 'more friendly'. http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-analyst-toolbox/how-curvature-works.htm . The ET tools, which are based on ESRI tools, also specify this in a more understandable format. http://www.et-st.com/et_surface/userguide/Raster/ETG_RasterCurvature.htm