Circulation patterns related to debris-flow triggering in the Zermatt valley in current and future climates

•This paper analyses joint occurrence of debris flows (DF) and daily precipitation amounts.•High probabilities of DF occurrence are found for rainfall exceeding the 95th percentile.•These conditions were related to synoptic situations as modelled by the RCM HIRHAM.•RCMs indicate shifts and changes in synoptic conditions triggering DF.

The principal objective of this study was to investigate the types of large-scale meteorological situations that are conducive to the precipitation and temperature conditions most likely to trigger debris flows in the Zermatt valley, Switzerland, under current and future climates. A two-dimensional Bayesian probability calculation was applied to take account of uncertainties in debris-flow triggering. Precipitation quantities exceeding the 95th percentile of daily precipitation amounts were found to have a significantly higher probability to coincide with observed debris flows. A different relationship exists for extreme temperatures, however. Southerly air flows, weak horizontal pressure gradients over Europe, and westerly flows are mostly associated with observed debris flows and 95th precipitation percentile exceedances. These principal flow directions are well represented in the regional climate model (RCM) HIRHAM control simulations for events exceeding the 95th precipitation percentile and the 30th temperature percentile. Under the IPCC A2 emission scenario, westerly and southerly flows are mostly responsible for these precipitation and temperature conditions under the hypothesis of slow adaptation to climate change (HS1/HC1). Under the hypothesis of rapid adaptation to climate change (HS1/HS1), southerly flows and weak horizontal pressure gradients are likely to gain in importance. In both scenarios for the future, southeasterly flows are among the principal flow directions responsible for the joint exceedance of the 95th precipitation percentile and the 30th temperature percentile, while these were absent in observations and the control simulation.