Permafrost distribution in the French Alps : statistical modelling using rock glacier inventory
Résumé
long with a rising temperature trend registeredin the past decades along the European Alps, severalsignals of permafrost degradation have been reportedin the same region, e.g. rock glacier displacementacceleration and thickening of the active layer. Per-mafrost degradation is alarming due to its role in ter-rain destabilization and, in particular, ice-rich debrisslopes are of great concern due to their proximity toinhabited areas. Therefore, in the past years increas-ing efforts have been made either by the scientificcommunity and local authorities in understanding theprocesses and hazards linked to the alpine permafrostdegradation.The first step to manage the risk associated withpermafrost degradation is to know its spatial extent.Since alpine permafrost does not manifest itself on thesurface, exception made for rock glaciers, it is not suit-able for automatic mapping. Therefore distributionmodels are required to infer permafrost occurrence.Traditionally, permafrost distribution models take ad-vantage of rock glaciers as evidence of permafrostto extrapolate statistically its occurrence based oncorrelation with climate variables. Although severalstatistical models have been made at the whole alpinerange scale, only one permafrost distribution map,calibrated on the Combeynot massif and validated onthe Vanoise and Mercantour massifs, was designedspecifically for the French Alps. However, permafroststatistical modeling using rock glacier inventories hasbeen recognized to be biased depending on the datacharacteristics, making permafrost maps inaccurateat local scale when applied to specific areas. Thisis due to the fact that rock glaciers extents are notonly influenced by climate, but also by their dynam-ics, a complex function of local topography, geologyand ice content in the landform. Therefore, the valid-ity of current models in the French Alps is, at least,questionable.To address this issue, the service Restauration desTerrains en Montagne (RTM) in collaboration withthe Institut de Géographie Alpine (IGA) realized thefirst complete rock glacier inventory of the FrenchAlps. The inventory, which counts more than 2600features, spacing from the Mont Blanc Massif to theMediterranean Sea, represents an extremely valuablemodeling dataset. The present study aims to exploitthis inventory in order to assess permafrost distribu-tion in the French Alps. As a first step, the perform-ance of already existing models is evaluated. Rockglaciers are used as validation dataset and the modelsare tested for spatial bias, i.e. their performance isassessed in function of the area they are applied on.Then, a statistical model is specifically developedfor the French Alps. At first, the correlation betweenrock glaciers and geomorphological local parameters,as lithology and feeding area, is investigated by per-forming a statistical analysis. This allows to studyand model climate control on rock glaciers activityand boundaries, avoiding biases introduced by topo-graphic and geological local conditions. Climatic datainclude air temperature, solar radiation and precipita-tion. Due to the extent of area investigated, constantlapse rates are not applicable. In particular, geograph-ical gradients in air temperature and precipitation areinvestigated and spatial interpolation of observationsis performed.Different modeling strategies are investigated, com-pared and discussed. The rock glacier inventory ispartitioned to be used either as training and valida-tion dataset. Other data, as BTS and GST, are usedas validation where available. Finally, the model per-forming best on the validation set is selected. A landuse map is used to select debris covered areas on theFrench Alps and permafrost occurrence is modeled.The resulting permafrost distribution map is expectedto be a valuable decision support tool in permafrostdegradation related issues