Discussion and Conclusions of the same research paper:
"The top seven melt extremes have systematically occurred under high global radiation, with large turbulent heat fluxes taking part in the 2000s extremes (Figure 4a). Except for 1986, 1989, and 1990, which belong to low turbulent fluxes conditions of Cluster#2 and are explained by very high global radiation alone, melt extremes are related to uncommon high turbulent fluxes in association with strong global radiation (Cluster#3). Most melt extremes are associated with low winter accumulation, except 2003 and 2004 (Figure 4b). Long‐wave irradiance is not associated with extreme melt, except to some extent in 2003. This extreme summer melt, which is well simulated (0.13 mwe of model‐data discrepancy), is physically explained by the combination of the highest energy fluxes in long‐wave irradiance and sensible and latent heat over seven decades. The 2003 simulated SEB (Surface Energy Balance) has a deviation of +62 W m−2 from the seven‐decade average. This deviation comes from large deviations in latent heat (+17 W m−2), long‐wave irradiance (+15 W m−2), global radiation (+14 W m−2), and sensible heat (+11 W m−2). Deviations in sensible heat and long‐wave are linked to the +2.5°C temperature anomaly observed over the 2003 melting season. The deviation in global radiation is related to the low 2003 summer cloudiness (8% below average). The change in turbulent latent heat flux was due to much lower snow and ice sublimation conditions on this specific summer. Those conditions were slightly reversed into atmospheric moisture condensation conditions (positive latent heat, Figure S11c), providing additional heat for melt at the glacier surface. The 2003s melt extreme was nevertheless limited by the winter balance (Figure 4b; in the range of the seven‐decade average) that provided a significant amount of snow at the beginning of the season, thereby reducing melt by negative feedback from the albedo.
Summing up, our present results demonstrate that glacier melt follows extreme value statistics if nonstationarity is accounted for, detrending raw observations from two‐decade long‐term trends in averages that shift distributions. Around this long‐term trend, extreme melt anomalies are distributed along an upper‐bounded Weibull‐type extreme value statistic law. The mean seasonal energy fluxes associated with these melt intensities are reconstructed from a SEB model. Melt deviations and extremes are controlled by three independent drivers: (1) the winter balance determining the amount of snow at the beginning of the melting season, (2) the global (short‐wave) radiation giving rise to the largest melt deviations and required for melt extreme occurrences, and (3) the latent heat flux that is controlled by air moisture. Sensible heat is involved in extremes but is a flux connected to latent heat (through wind speed) and global radiation (through air temperature). The long‐wave irradiance, varying only slightly and systematically anticorrelated with the net short‐wave balances, is not involved in melt extremes.
In light of Thibert et al. (2013), nonstationarity is explained mostly by the lengthening of the ablation season observed since the mid‐1980s and also by snow and ice melt intensification in the core of the melting seasons. Regarding the longer ablation seasons, positive feedback from the albedo change due to longer ice versus snow ablation is the main factor. Altitude lowering of the glacier surface accounts here for less than 16% (Thibert et al., 2013) of the trend, but some potential changes in ice albedo cannot be ruled out (Oerlemans et al., 2009). A remarkable finding is that the long‐term melt intensification is mainly driven by the latent heat flux increase (+17 W m−2) due to higher air moisture and less snow/ice sublimation, tending to cancel this systematic sink of energy in the SEB. The long‐wave irradiance rise (+5 W m−2) is the second factor in melt intensification. It is related to a larger forcing of +3.6 W m−2 per decade as assessed by SAFRAN data and consistent with the +2.5 W m−2 per decade reported at global scale for the 1990s and 2000s by Ohmura (2012) (Text S9). We expect the long‐term trend in melt attributable to the drift in latent heat and long‐wave fluxes to continue due to the projected rises in air moisture, greenhouse gases, and higher air temperatures. Projected earlier snowmelt (Musselman et al., 2017), as already reported for seasonal snow cover (Durand, Giraud, et al., 2009), and snow over the glacier accumulation area (Thibert et al., 2013) associated with the lengthening of the ablation season in autumn may increase the ice melt duration and enhance melt from albedo positive feedback. Under increased atmospheric water vapor (Santer et al., 2007), despite more‐or‐less unchanged relative humidity (Ingram, 2002), snow/ice sublimation together with the associated energy sink of latent heat will be much more limited, providing more energy for melt in the energy balance. More frequent record breaking of glacier melt values should be expected from these upward shifts in SEB averages. Whether future record breakings constitute extreme deviations from averages could be inferred analyzing trends and carrying out GEV analyses. Moreover, potential changes in melt extreme properties cannot be ruled out as already established for temperatures (Schär et al., 2004). For this, a peak over threshold model (Katz et al., 2002) should be tested in place of the GEV approach which supposes steady state for extremes.
