Critical zone modelling for alpine catchments : towards a calibration-light model of snow hydrology

The current operational hydrological modelling of mountainous catchments mostly relies on conceptual and semi-distributed models, which are calibrated based on historical discharge measurements. As a consequence, in spite of their good performance in seasonal river runoff prediction, they often fail to project the impact of climate change on the hydrological cycle over decades. This is due to a limited representation of physical processes and their inability to simulate water paths and their modification.

To overcome such limitations, we applied the data-intensive and calibration-light critical zone model ParFlow-CLM, to a highly instrumented mid-elevation catchment (0.15 km² area between 1950 and 2150 m.a.s.l) close to the Lautaret Pass, in the French Alps. Our setup showed promising results with good correlation when compared to the observed discharge (KGEnp = 0.91), consistent evapotranspiration compared to local Eddy-Covariance measurements and realistic snow disappearance patterns. This simulation served as the proof of concept towards the feasibility of physical-based critical zone hydrological modeling in alpine terrain. It highlighted the necessity of a careful redistribution of the locally observed meteorological forcing including solid precipitation and incoming radiations. It also pointed out the necessity of well simulating the snow aging and albedo to represent streamflow regimes all along the snow period. It was achieved through manually incorporating the impact of grain growth, refreezing and dust in the albedo parameterization (snow age) within the current CLM version of ParFlow (CLM 3.5).

In this presentation we will introduce our strategy for a calibration-light model of mountain watersheds by developing realistic but data-parsimonious strategies of data collection and processing. Specifically, we aim at taking into account the impact of complex topography on meteorological forcing (shading, radiation incidence angle, wind acceleration, snow redistribution, altitude gradient). In this framework we will compare several land surface snow schemes (CLM3.5, CLM5, Crocus and MORDOR) using ESM-SnowMIP data. This will help to quantify the improvement expected for moving from CLM3.5 to CLM5 and compare it further with a complex snow scheme (Crocus) and an advanced degree-day approach (MORDOR snow module). Implementation of CLM5 in ParFlow will also enable the representation of dynamic vegetation processes. This will be the first step towards a functional critical zone modeling for a mid-elevation alpine catchment, which will allow the reanalysis and projection of hydrological conditions with minimum calibration.