R. Frauenfelder1*, B. Schneider2**, B. Etzelmüller3
1 Department of Geography, University of Zurich, Switzerland
* now at: Norwegian Geotechnical Institute, Postboks 3930 Ullevaal Stadion, 0806 Oslo, Norway
Telephone: +47 22 02 19 87
Fax: +47 22 23 04 48
2 Department of Earth Sciences, University of Basel, Switzerland
** now at: ilu AG, Zentralstrasse 2a, 0610 Uster, Switzerland
Telephone: +41 (0)44 944 55 54
Fax: +41 (0)44 944 55 66
3 Department of Geocsiences, University of Oslo, Blindern, 0316 Oslo, Norway
Telephone: +47 22 85 72 29
Fax: +47 22 85 72 30
Rockglaciers are landforms originating from periglacial talus (‘talus-derived’ rockglaciers) and/or glacier-transported debris, mostly from lateral and terminal moraines (‘morainederived’ rockglaciers). The occurrence of talus-derived rockglaciers (Fig.1) is influenced primarily by climatic and topographic preconditions (cf., e.g., Barsch 1996). They are found in areas characterized by specific topographic attributes; they occur, for example, within a certain altitudinal band, favour certain slope aspects, require a particular slope, and need a rock-contributing headwall above them (cf., e.g., Frauenfelder et al. 2003, Janke and Frauenfelder 2007).
Modelling of the entire rockglacier bodies applying simple geomorphometric approaches is difficult as these landforms vary considerably in form and size. There are, however, areas within each talus-derived rockglacier that possess specific characteristics similar to all these forms: in the so-called rockglacier root zone (RRZ) the accumulated debris is triggered to creep (cf. Barsch 1996). This zone is located within or at the end of a concave landform where debris can accumulate (see Fig. 1, area in the upper right corner).
Recent advances in statistical modelling triggered a considerable amount of studies to model periglacial features (e.g., Brenning 2009, Etzelmüller et al. 2001, Etzelmüller and Sulebak 2003, Hjort et al. 2007, Hjort and Luoto 2008, Hjort and Marmion 2009, Luoto and Hjort 2004, 2006, Marmion et al. 2009, to name just a few). While some of these studies apply sophisticated statistical modelling schemes, for example, generalized additive models, support vector machines, boosting, etc., our contribution reports about the application of a simpler approach. We present an exploratory approach to the analysis of relationships between multiple explanatory variables and a presence/absence response variable for the modelling of RRZ distribution. Our approach essentially consists of the study of histograms and descriptive statistics, and the construction of heuristic probabilistic and deterministic classification rules.
The focus in this contribution is on the distribution of talus-derived rockglaciers (i.e. RRZs of talus-derived rockglaciers), and the explanations in the following text refer, by implication, to these forms only.