With the growing importance of the earth’s climate and how it is evolving over recent decades, glaciers, since they are good indicators of predominant climate conditions, are of growing interest for scientific research. One of the most meaningful variables describing a glaciers development is its mass balance. However, its calculation usually involves a good amount of field work and gathering measurement data. To simplify the process, a new approach combining weather forecast data with snowpack modelling was developed. This allows for mass balance calculations in very remote areas without the need for lengthy and expensive field work.
The goal of this study was, to assess whether combining a high-resolution snowpack model with a high-resolution numerical weather prediction model would make a viable and accurate approach for glacier mass balance modelling. As a high-resolution snowpack model, SNOWPACK was used, while the meteorological forecast data was provided by COSMO-1. The study was conducted on Hochjochferner, a glacier on the border of Austria and Italy. For verification, data of an Automated Weather Station (AWS) situated on the surface of Hochjochferner was available, as well as snow stake measurements from 6 different locations on Hochjochferner. Before SNOWPACK was run with COSMO-data, it was tested using the AWS data and compared to the snow height measurements by the AWS and ablation measurements at the snow stakes. All modelling was done over the course of the hydrological year 2014/15, spanning from 01/10/2014 to 30/09/2015. As a reference, a manual glacier mass balance done with a combination of stake data and terrestrial photography was used.
At first, SNOWPACK was run at the location of the AWS, to find out the optimal model setup and compare it to the in-situ snow height and ablation measurements. At the end of the ablation season, SNOWPACK had modelled 0.16 m more ablation compared to the stake measurements, which translates to an error of 3.6 %. As a second step, all meteorological parameters necessary to run SNOWPACK were extrapolated from the AWS location to all stake locations. This was done with the help of MeteoIO, a library capable of processing meteorological input and output data. SNOWPACK was then run at these locations, to determine whether the extrapolations produce realistic input data that can be used to force SNOWPACK, and to compare the results to the snow stake measurements. Afterwards, the AWS data was extrapolated to 19 virtual mass balance points distributed over the whole glacier surface, dividing it into elevation bands with an equidistance of 50 m. Again, SNOWPACK was run at these points, and a glacier mass balance was calculated off of the results. The obtained mean specific mass balance overestimated ablation by 450 mm w.e., an error of 20.6 %. As the final step, COSMO-data was introduced to force the model. Meteorological data was extracted from the model for the 6 grid points closest to the glacier, and interpolated to the location and elevation of the desired points. Again, it was first tested at the AWS location, before running the model at the 6 snow stake locations and, finally, at the 19 virtual mass balance points to obtain a glacier-wide annual mass balance. The resulting mean specific mass balance overestimated ablation by 1040 mm w.e., an error of 51 %.
After comprehensive error analysis, it can be said that the reason for this error can most likely be found in the COSMO-data itself, namely a positive bias in relative humidity, coupled with a negative bias in temperature. The model chain was run once more using the same settings, only with substituting humidity data provided by COSMO with the humidity data measured by the AWS. This resulted in a much lower mean specific mass balance error of 11.9 %. Thus, although the resulting error being significant at first, it seems that much more accurate results can be gained with minor adjustment of certain parameters, to compensate for an underlying systematic error. Which means that, after more testing and analysis preferably at different research sites, this approach could indeed make a viable and accurate alternative for glacier mass balance calculation, and considerably ease the process as well as lower the amount of work typically required.