How does homogenization of snow measurements impact snow climatology in the Alps

Snow is not only a significant economic factor but also hazard for mountainous regions like the Alps. Therefore, measuring snow in such regions has long tradition. Moreover, accurate and timely information on the snow cover is of increasing interest for numerical weather prediction models, for hydrological models and for climate monitoring. Hom4Snow focus on long-term daily measurements of new snow and snow depth from Austria and Switzerland as the only source for documenting long-term snow changes. The high density of observations in the Alps enables capturing the large diversity coming from the topography. Like with any other measurement series there is a high probability that long-term snow series are affected by temporal inhomogeneities through changes in the station location, observer personal or measurement practices. Such inhomogeneities could heavily affect spatio-temporal trends, which also could unfavourably affect the work of decision makers or scientific users. Therefore, the proposed project aims at developing not only innovative methods for the homogenization of long-term snow series, but also in demonstrating the impact of such adjustments on trends and climatologies. A unique data set of parallel snow measurements from different elevations and climate regions will allow comparing the sensitivity of new snow height, snow depth and its derived parameters on possible inhomogeneities. The compilation and analysis of the parallel snow measurements and the development of advanced algorithms to fill gaps in new snow height and snow depth data series will be the main task of a PhD student. These steps will help another PhD student to compile a maximum number of high quality reference series, to apply break detection procedures and to develop sophisticated methods for adjusting the inhomogeneities. The collaboration effort of the partners will not only increase the number of long-term high quality snow series in Austria and Switzerland, it will also allow quantifying the uncertainty of trend statements concerning snow measurements. This is of increasing relevance for dimensioning problems (snow load estimates), estimates of the damage potential from snow for insurance economy, assessing future planning from skiing resorts, environmental impact assessments and mitigation measures. Moreover, the results will allow ranking the different snow parameters (e.g. sum of new snow height, max snow depth, snow days, snow duration, date of snow disappearance, etc.) in regard to their sensitivity to inhomogeneities. These findings will help to choose the least affected parameter in case homogenization is not possible, as it is the case in many regions with only very few measurement stations.

The knowledge about climate-induced changes of snow in the Alps is of manifold importance in practice. Water management, ecology, agriculture or tourism in the Alps depend on snow. In order to analyze and assess the changes of snow, one depends on reliable model results and measurements. Satellite data can only cover a much shorter time span and in mountainous regions still have major technical problems in recording snow. Long-term measurements at stations on snow in the Alps mainly refer to total and new snow depth. Even though these measurements are in principle easy to perform, they carry the potential of errors and changes in the measurement series, which can be of the same order of magnitude or even larger than the climate-induced changes. To detect these errors (breaks) in the measurement series and to correct them as close as possible to reality, corresponding to the natural conditions (which is called homogenization), was the challenge and research question of the Hom4Snow project. While quite extensive experience on homogenization is available for air temperature measurements or precipitation measurements, it was still new scientific field for total and new snow depth before the Hom4Snow project. Building on the knowledge of air temperature and precipitation, the methods developed there were tested for snow by Hom4Snow. However, snow has some specifics that need to be taken into account during homogenization. For example, the correction of the snow depth also leads to a change of the snow cover duration, which may not show a break in the time series. Therefore, taking these requirements into account and developing them into concrete guidance for all snow researchers was a challenge of Hom4Snow. In collaboration with colleagues from Switzerland, the project was able to clearly demonstrate the relevance of homogenization for snow measurements, treating this separately for different snow parameters (e.g. mean snow depth, maximum snow depth, snow cover days). Furthermore, an improved method for the correction of breaks and for the addition of failures in the snow measurement series was developed. This newly developed method is not based on an average change in snow depth but corrects the individual classes of the frequency distribution individually. Thus, especially extreme snow depths (extremely low, extremely high) can be corrected more realistically. However, especially the extreme snow depths have a very great importance in practice, as they are responsible for extreme snow loads on buildings, avalanche events or restrictions for infrastructure. Finally, Hom4Snow was also able to show which errors and thus false statements can result from not taking homogenization into account.