A whole-system approach to a fast-changing risk

Should all development be banned in high-risk zones, or should these areas instead be provided with protective infrastructure? The twin demands of development and safety mean that planning decisions are always hard-fought in mountain areas. Decisions concerning avalanche risk can cost millions of euros, but their cost can also be counted in human lives. The evaluation of this risk demands complex research at widely differing scales and in different domains, from the study of the physical processes involved in avalanches to monitoring the ever-changing patterns of hazards, vulnerability and exposure that combine to produce it. 

A research unit to investigate all aspects of avalanche risk

1970 was a difficult year for French mountain communities. On February 10, an avalanche engulfed a UCPA holiday-centre chalet in Val-d’Isère, leaving 39 dead. Then, two months later, on the Assy plateau, a landslide formed of a mixture of mud, snow and water buried a sanatorium, causing 72 further deaths. 
These two tragedies acted as a wake-up call. The French authorities realised that the country had few defences against natural events such as these, events that were both recurrent and hard to predict. They asked the CTGREF¹ to set up a snow study service, the ETNA Research Unit on Torrential Erosion, Snow and Avalanches at INRAE. ETNA became the Institute of Environmental Geosciences (IGE) joint research unit IGE in January 2023. The newly fledged ETNA service’s principal mission was to build a better understanding of avalanche risk. “In its early days, the unit was chiefly interested in mapping avalanche locations and in developing and improving protective structures”, Florence Naaim, IGE’s deputy director, tells us as she looks back over the unit’s development. “We then began to study the snow itself, looking more closely at the physics of avalanches. Nowadays we also take account of the potential impacts, the vulnerability of those at risk, and we factor in all the physical and social mechanisms that contribute to the risk. So, we don’t just concentrate on the hazard, we look at the risk in the round, adopting a whole-system approach.” 

This methodology recognises the systemic nature of risk and the fact that each of its three dimensions – hazards, vulnerability and exposure – evolves differently. It is an approach that the Sendai framework (2015)², in particular, has worked hard to promote. Indeed, to be able to predict the harms caused by an event, we must understand how several different factors (such as climate, topography, the spatial practices of societies and their development choices) can combine to produce a disaster, calling for a vision that is both comprehensive and interdisciplinary. The current profile of the research team reflects this need – the researchers, engineers and technicians who make up its thirty members hail from different disciplines, so that specialists in physics, modelling and measurement work alongside statisticians and historians…

Observe, model, quantify  

The tradition of recorded observations in this field goes back to the creation in 1899 of the French Permanent Avalanche Survey (EPA), still jointly managed with the French National Forest Office and the Ministry for the Ecological Transition. At first, the survey’s approach was largely founded on the natural sciences, but it expanded in the 1970s to include research in physics and process modelling. In the early years of the millennium, in order to improve risk quantification, designate zones where risk could be considered acceptable, and optimise mitigation strategies, the Institute’s researchers coupled deterministic models of avalanche spread with the statistical data provided by the EPA. This resulted in increasingly accurate probabilistic numerical models designed to evaluate “exceptional avalanches” (with a 100-year return period). Working first at the scale of a single avalanche path, the modelling was then scaled up to take in entire ranges, making it possible to evaluate the hazards for paths that were either poorly documented or unrecorded. This probabilistic hazard analysis, carried out first at local and then at regional scale, has enabled us to characterise the variability of avalanche activity at different spatio-temporal scales. It helps us to understand why certain ranges are more active than others, taking account of their topography and locations, and to use meteorological “forcing” to explore the links between avalanche activity and weather. 

Meanwhile, methods were also developed to model vulnerability to the impacts of an avalanche, examining how vulnerabilities in both the built environment and those who live in avalanche zones could contribute to the risk. These models were then coupled to the quantitative hazard models, enabling individual risk models to be calculated that could be expressed in mortality figures or destruction levels, and protective structures to be optimised. Further parameters drawn from the human and social sciences are now adding to the sophistication of the models. 
This fresh input means that expert historians can, for example, help to make better use of archive materials to feed current models. Likewise, historical analysis of development decisions taken by businesses and of their perceptions of risk can be incorporated into our understanding of current vulnerabilities and exposure factors, enabling us to act appropriately. 
We might be forgiven for assuming, therefore, that avalanche risk is under control. Sadly, this is far from the case. At present, the socio-environmental systems in our mountains are undergoing a process of very rapid flux brought about by global change (to the climate, society, etc.), meaning that the risks, too, are changing.

The many impacts of a changing climate

New patterns of avalanche risk threaten high mountain communities, despite the overall decrease in snow cover.

To deal with this, we first needed to understand how climate change is affecting the hazards themselves. The research teams began by studying changes in avalanche activity at climate-process scale. This enabled them to show that a series of very harsh winters produced a peak in avalanche activity in France at the beginning of 1980 and that, from the mid 1980s onwards, avalanches at low and mid altitudes then became less frequent and smaller, with a greater proportion of wet snow avalanches. Looking to the future, work conducted jointly with the French meteorological service’s Centre for Snow Studies using the IPCC projections (see p. 9) made it possible to demonstrate that current trends are set to intensify. Thus, for example, a 20 to 30% overall reduction in avalanche activity is projected during the 21st century as the consequence of drastic losses in snow cover. We should nevertheless note that avalanches at high altitudes do not follow this trend. Indeed, at these altitudes, extreme snowfall may cause the number of avalanches to increase for a time before it falls again, with a continued chance of major dry and cold snow avalanches occurring, such as the 1999 Montroc avalanche in the Chamonix valley. In that event, 14 chalets were swept away and 12 people were killed. The team is currently working to improve the accuracy and detail of future projections of avalanche patterns, incorporating their number, intensity, location and seasonality into the calculations. 

The impacts of tourism and buildings 

The pace of change for other determinants of avalanche risk is no slower, indeed, it may even be faster. As the climate heats up and agro-pastoral activities are abandoned, we are seeing a rapid reafforestation of the slopes. Meanwhile, the footprint of development continues to expand overall, spurred on, in particular, by the growth in winter tourism. Brand new avalanche protection structures are going up alongside existing structures that may be showing signs of wear, sometimes for lack of maintenance, producing a shifting picture of overall exposure levels. Research is ongoing to reveal the workings of this complex process (which can vary greatly from one local context to another), pulling together qualitative and quantitative analysis at different spatio-temporal scales. It has been possible to show, for example, that in very high alpine valleys, the risk to buildings and their occupants would appear to have risen in the past few decades, as their exposure increases to a continued hazard. By contrast, at lower altitudes, risk levels are falling due to a combination of reduced snow cover and the gradual regeneration of trees along avalanche flow paths, sometimes all the way up to the avalanche starting zones. Quantitative analysis has helped us to demonstrate that it may be possible to adapt to these changes by strengthening exposed buildings and managing our forests for protective purposes, gaining greater control of risk levels.
This, then, is the new pattern of avalanche risk that threatens high mountain communities, despite the overall decrease in snow cover. Stakeholders in such areas are being asked to develop an adaptive approach in their long-term investment strategies, but to do this they must have access to detailed and reliable information. The whole-system approach and increasingly detailed studies produced by the research community will help these decision-makers to develop a detailed understanding of the changing nature of the threat they face.