Interpolation issues on the margins of retreating glaciers

[adehecq]

This is more a glaciological issue than a code-related issue.

The normalized regional hypsometric interpolation is great for filling large gaps. However, for dramatically retreating glaciers, interpolating the elevation change can create artifacts near the margins of the glacier where the elevation change is bounded by the ice thickness. Here is an example over Aletsch glacier from @erikmannerfelt's results with the Swiss TerrA images. The left figure shows the interpolated dh, while the right figure shows the results obtained with topographic maps. One can see that the elevation change is less negative near the margins of the tongue in the right image, which is not reproduced by the interpolation.

An idea could be to look at the relationship "old elevation" vs "new elevation" as opposed to "elevation change" vs "new elevation". Then the elevation change can be calculated as "new elevation" - "interpolated old elevation", which would never be larger than the ice thickness.