Discrepancies between two measurements and two model approaches for liquid water flow in snow

Thursday, 18 December 2014
Lino Schmid1, Nander Wever1, Achim Heilig2, Charles G Fierz1 and Michael Lehning3, (1)WSL Institute for Snow and Avalanche Research SLF, Davos Dorf, Switzerland, (2)University of Heidelberg, Institute of Environmental Physics, Heidelberg, Germany, (3)EPFL Swiss Federal Institute of Technology Lausanne, Lausanne, Switzerland
Liquid water flow in snowpacks is a complicated process to measure or to simulate in snowpack models, although it is important for assessing, for example, soil moisture variations, streamflow discharge or wet snow avalanche formation. The measurement site Weissfluhjoch (WFJ) is equipped with instruments recording meteorological conditions and snowpack properties, including a snow lysimeter and an upward looking ground penetrating radar (upGPR). The upGRP, among other capabilities, is able to track the progress of the melt water front through the seasonal snowpack at WFJ, whereas the snow lysimeter only records liquid water runoff from the snowpack. The 1 dimensional physics-based snowpack model SNOWPACK has recently been extended with a solver for Richards equation, which provides a demonstrable improvement in simulating snowpack runoff, especially on the hourly time scale, when compared to a simpler bucket-type approach. Here, we compare the two measurement methods and the two snowpack simulations for four snow seasons with respect to the progress of the melt water front through the snowpack and snowpack runoff. We show that in the studied period, snowpack runoff in the melt season starts before the arrival of the melt water front at the bottom of the snowpack as detected by the upGPR. This discrepancy is in the order of several days to 1-2 weeks. The agreement between measured and modeled snowpack runoff is higher, although modeled snowpack runoff is still lagging several days from observed runoff, depending on the used water transport scheme. This demonstrates that the early start of snowpack runoff is likely associated with the existence of preferential flow paths. The modeled progress of the melt water front is faster than observed in the upGPR data. This contributes to a better predicition of the onset of snowpack runoff, but may have consequences for the representation of the internal snowpack in the model. The study highlights the extreme difficulties in understanding and modeling liquid water flow in snow, but suggests that existing model approaches satisfactorily predict the onset of snowpack runoff in the melt season.