Should altitudinal gradients of temperature and precipitation inputs be inferred from key parameters in snow-hydrological models?

Abstract

This paper evaluates whether snow-covered area and streamflow measurements can help assess altitudinal gradients of temperature and precipitation in data-scarce mountainous areas more efficiently than using the usual interpolation procedures. A dataset covering 20 Alpine catchments is used to investigate this issue. Elevation dependency in the meteorological fields is accounted for using two approaches: (i) by estimating the local and time-varying altitudinal gradients from the available gauge network based on deterministic and geostatistical interpolation methods with an external drift; and (ii) by calibrating the local gradients using an inverse snow-hydrological modelling framework. For the second approach, a simple two-parameter model is proposed to target the temperature/precipitation–elevation relationship and to regionalize air temperature and precipitation from the sparse meteorological network. The coherence of the two approaches is evaluated by benchmarking several hydrological variables (snow-covered area, streamflow) computed with snow-hydrological models fed with the interpolated datasets and checked against available measurements. Results show that accounting for elevation dependency from scattered observations when interpolating air temperature and precipitation cannot provide sufficiently accurate inputs for models. The lack of high-elevation stations seriously limits correct estimation of lapse rates of temperature and precipitation, which, in turn, affects the performance of the snow-hydrological simulations due to imprecise estimates of temperature and precipitation volumes. Instead, retrieving the local altitudinal gradients using an inverse approach enables increased accuracy in the simulation of snow cover and discharge dynamics while limiting problems of over-calibration and equifinality.