Linking Precipitation Size and Velocity Distribution with Snowpack Height Changes

The snowpack is a fundamental element of the cryosphere and understanding its dynamics is crucial for regions where runoff  is the main source of freshwater. Snowpack variations in very short periods of time can have important security and logistical consequences. On the other hand, snowpack height measurements are complex due to its high spatial variability. The thermodynamic and physical processes that the snowpack undergoes are complex and are dominated by meteorological forcings which are also complex, specially in mountain regions. The most important forcing in terms of snowpack height variation is precipitation. It is well known how precipitation in the form of rain decreases the height of the snowpack almost immediately, while precipitation in the form of snow,  increases its height. The problem is that precipitation occurs with a variety of populations of phases, so this mixed precipitation makes the conversion between precipitation and snow height increase not straightforward.  Disdrometers are instruments capable of determining the size and speed at which precipitation falls very precisely. The population of different sizes and terminal velocities is known as the spectrogram. This map of velocities and sizes makes it possible to estimate the phase since their terminal fall velocities of rain and snow are very different. These instruments are very useful to determine the intensity of mixed precipitation and are widely spreaded in airports, highways and mountain areas. In this work we analyse the possibility of developing a relatively simple algorithm that from the size and velocity distributions detected by a disdrometer we could predict the variation of the snowpack in the next few hours. Several techniques have been tested in this work, some of them simple correlations. But the method that really outstanded was the one based on a reduction of the dimensions of the spectrograms applying a principal component analysis which is then used to search analogue situations. Although the available data is still very small, the results encourage to refine this technique when more data will be available in the next winters.