An examination of the snow and avalanche hazard on the Milford Road, Fiordland, New Zealand
Degree GrantorUniversity of Canterbury
Degree NameDoctor of Philosophy
Avalanches pose a significant natural hazard in many parts of the world. Worldwide the hazard is being managed in a number of new and traditional methods. In New Zealand, the Milford Road, Fiordland, has a significant avalanche problem which has been managed by the Transit New Zealand Milford Road Avalanche Programme since 1984. This avalanche programme has generated a database of all avalanche occurrences and associated meteorological parameters for the time period 1985 to 2002. Elsewhere around the world, similar and more extensive data sets have been used to examine a wide variety of aspects in relation to the snow cover, avalanching and avalanche hazard. The availability of the Milford Road database has provided the opportunity use new and traditional approaches to examine many aspects of avalanching including; the trends in and relationships with the snow and avalanche regime, evaluation of the avalanche hazard, statistical forecasting of avalanches and the visualisation of avalanche occurrence information in a GIS. Statistical and graphical examination of the inter-annual variation in the snow and avalanche regime revealed relationships between the snow depth, avalanche occurrences and atmospheric circulation similar to those found elsewhere around the world, but not previously examined in New Zealand. Furthermore, the analysis resulted in strong correlations despite using a database significantly shorter than those used elsewhere. Atmospheric circulation types that bring strong winds and precipitation were found to be highly significantly correlated with avalanche occurrences and snow depth. Avalanche occurrences were more highly correlated with atmospheric circulation than snow depth was, reflecting the strong maritime avalanche climate. Risk evaluation was undertaken using two approaches, the avalanche hazard index (AHI) and the probability of death to individuals (PDI) method. The present avalanche risk was compared to a theoretically uncontrolled avalanche regime, using 2002 traffic volumes for AHI and PDI. The AHI analysis highlighted the reduction in the AHI resulting from the control programme, and the significantly lower AHI when compared to Rogers Pass, B.C., Canada. The PDI analysis using equations modified to allow for a range of consequences indicated that the Milford Road is similar in risk to roads in Switzerland, but is far more accessible, with fewer closed days. A new equation for PDI, which accounted for waiting traffic was derived, and suggested that the calculated risk was high and unacceptable compared to standards applied to other hazards. Statistical forecasting using classification tree analysis has been successfully applied to avalanche forecasting in other climatic settings. This study has applied an extension to this technique through 10-fold cross validation to permit classification of an avalanche day in this direct action maritime climate. Using varying misclassification costs two classification trees were generated. The tree that used only wind speed and wind speed and precipitation combined in a temperature sensitive wind drift parameter obtained an overall accuracy of 78%, with correct prediction for an avalanche day at 86%. These predictor variables are considered to be the fundamental controls on avalanche forecasting in this climate, and coincide with important variables inferred from the atmospheric circulation analysis. Following the investigation of various methods for the creation of a high resolution digital elevation model (DEM), a GIS was used for the visualisation and examination of avalanche occurrences. Similar to other studies, qualitative and quantitative analysis of the spatial distribution in terms of aspect of avalanche occurrences was undertaken using the GIS. Colour coding of occurrences highlighted the influence of two storm directions, while an excess ratio showed the clear influence of aspect on avalanche occurrences in relation to two dominant storm directions, avalanche size and avalanche paths. Furthermore, the GIS has many applications for operational forecasting, teaching and the maintenance of institutional memory for the avalanche programme.