Snow avalanches are one of the most significant natural hazards of mountainous regions. They occur not only outside of civilisation, but also in places inhabited by humans. The release of an avalanche is a non-negligible risk. Recent disasters, such as the Galtür Avalanche in Tyrol in 1999 with 39 deaths, clarify the danger of avalanches for human beings. However, not only residents are at risk; every year outdoor athletes fall victim to avalanches as well. The demand, and also the possibility of outdoor activities is steadily increasing, and as a consequence, the number of outdoor athletes is rising as well. One sector which is strongly influenced by this enhancement is the branch of mountain activities, which includes winter sports such as skiing, snowboarding, ski touring or snowshoeing. While skiing and snowboarding on declared and prepared slopes are relatively safe, so-called free-riding can be risky. Offside the prepared slopes, many dangers can occur. This includes hidden stones, trees stumps, steep and icy downhills and of course the release of avalanches. Appropriate risk evaluation is necessary to reduce the risk of an avalanche disaster. In the state of Salzburg, the responsible Avalanche Warning Service (AWS) supplies an actual avalanche risk assessment twice a day. This evaluation is applied on six separated and non-overlapping zones. Previously these zones were defined conceptually by experts from the avalanche warning service in Salzburg. Semi-automated regionalization approaches have not been used until now. However, they could offer new possibilities to define such regions. In recent decades, countless clustering algorithms have been developed, but only a slight portion allow spatial relations to be considered, and only a handful can model multi-dimensional phenomena. The Object-Based Image Analysis (OBIA) related multiresolution segmentation is such a regionalisation method. While this approach was initially intended for remote sensing purposes, and more specifically for image segmentation, it turned out that it could also be applicable for regionalisation purposes in the context of the so-called ‘geon concept’. SKATER, a recently developed spatial clustering method for vector data, could be an appropriate alternative to the raster-based multiresolution segmentation.

Aim of the Study

So far the avalanche risk assessment zones in the state of Salzburg are defined expert based by specialists from the Central Institute for Meteorology and Geodynamics (ZAMG), which acts as the avalanche warning service on behalf of the regional government of Salzburg. Due to the difficulty of modelling appealing, complex real-world phenomena, GIS-based semi-automated methods for delineating these zones have not been integrated until now. However, new concepts allow the incorporation of such methods, whereby the consideration of expert knowledge is still a primary criterion. The geon concept which acts as a framework of regionalisation can be used for this purpose. Routinely, the OBIArelated multiresolution segmentation is used for aggregating relevant factors. Furthermore, a newly presented spatial clustering algorithm, called SKATER, is also a conceivable option for this purpose. The aim of this study is, on the one hand, to delineate new GIS-based avalanche risk assessment zones for the state of Salzburg by applying the geon concept. Therefore, strong influencing factors should be identified, and weighted, based on expert opinions, for further data aggregation. On the other hand, the used regionalisation methods, multiresolution segmentation and SKATER are described and compared to identify assets and drawbacks of both methods.

Objectives

The objectives of this study can be enumerated as follows:

  • To develop a weighted indicator framework for the delineation of avalanche risk assessment zones
  • To model assessment zones for the state of Salzburg
  • To compare the functional principle of the used regionalisation methods
  • To compare the model outcomes in order to identify advantages and disadvantages of the used regionalisation methods

Research Questions

The five research questions which are dealt with in this study are as follows:

  1. Which criteria are determinants for modelling avalanche risk assessment zones?
  2. How could assessment zones be modelled computationally?
  3. Which indicators are crucial for modelling avalanche risk assessment zones?
  4. What are the principle differences between the used regionalisation methods?
  5. Are there significant varieties in the model outcomes and do advantages or disadvantages exist in the regionalisation approaches?

manualvssemi-01

workflow-01

mean-01

finalmap-01-01