a post which describe an interesting glaciological application of Otsu’s optimal thresholding, a method in digital image processing with which we can segment a gray-level image into two or more classes by finding the best threshold level between the two.
I stumbled upon the method a few years ago when working on a study in which I needed an objective and reproducible method for segmenting satellite images of glaciers into accumulation and ablation areas. After much trial and error, and in spite of its simplicity, Otsu’s optimal thresholding outperformed all other methods for classification, including sophisticated approaches as ISODATA and spectral unmixing.
For the uninitiated, the problem is as follows: on a glacier, the snow line is the limit between snow (above) and ice (below). It is a dynamic feature whose altitude changes according to the weather and seasons. On temperate glaciers, those where the ice is everywhere at the pressure melting point, conventional glaciological wisdom tells us that the snow line at the end of the ablation season (usually the summer) can be considered a good approximation to the equilibrium line altitude, that is, the altitude at which the surface mass balance is zero. This approximation is not valid for subpolar glaciers because superimposed ice develops right down the firn field, effectively extending the accumulation area without appreciable optical contrast. The relevance of this question lies in the fact that glaciologists are usually interested in the relation between climate and glacier mass balance, and therefore mapping snowline altitudes at the end of the summer has always an intrinsic value. At least conceptually, the solution seems simple: find a cloud-free image taken at a suitable time of the year (end of summer, minimal seasonal snow cover) and just map the snow line on it.
Bq. By Hernan de Angelis. More info at