Impacts of subgrid temperature distribution along elevation bands in snowpack modeling: Insights from a suite of Andean catchments

The implementation of elevation bands is a popular strategy to account for topographic heterogeneities in snowpack modeling. Here, we characterize the implications of subgrid temperature distribution along elevation bands through numerical experiments in nine mountainous basins of the Andes Cordillera, central Chile. Specifically, we analyze outputs from the Variable Infiltration Capacity model with six different setups: no elevation bands (i.e., flat grid cells; benchmark model) and elevation bands with vertical discretizations of 1,000, 750, 500, 200, and 100 m. The analyses are conducted in a wet period (April/1982-March/1987), dry period (April/2010-March/2015) and a climatological period (April/1982-March/2015). The results show that adding elevation bands yields little variations in simulated monthly or daily streamflow; however, there are important effects on the partitioning of precipitation between snowfall and rainfall, snowmelt, sublimation, and the spatial variability in 1 September snow water equivalent (SWE), suggesting a form of model-structure equifinality. Vertical temperature distribution generally yields less basin-averaged snowmelt and more (less) catchment-scale sublimation across water-limited (energy-limited) basins. Further, the implications of subgrid temperature distribution vary with the analysis period: fluxes are more affected during the wet period, while variations in 1 September SWE are more noticeable during the dry period. In general, the effects of topographic temperature distribution are reduced with increasing vertical discretization and can differ among catchments. Finally, the grid cells that yield the largest sensitivities to vertical discretization have relatively more humid conditions, large intra-annual variations in the water/energy budget, lower mean altitude, elevation ranges > 1,000 m, and steep slopes (> 15 degrees). Plain Language Summary Spatially distributed computer-based models are widely used to make predictions on water availability. In mountainous areas, it is common to distribute air temperature using elevation bands in modeling units with complex topography; however, the effects of the selected number of bands and/or elevation range on model results have not been assessed in detail. We use a suite of diverse Andean basins to document how the vertical distribution of air temperature along elevation bands affects the simulation of the water cycle at different spatial scales. Our results show that, although the incorporation of air temperature variability has little effects on the simulation of discharge at the basin outlets, similar results can arise from different spatial distributions of rainfall, snowfall, snowmelt, sublimation, and maximum annual accumulation. The implications of adding elevation bands may vary with the climate conditions (i.e., wet/dry) of the analysis period. Finally, we identify climate seasonality, mean altitude, elevation range, and slope as the key variables that should be examined carefully to decide where (i.e., which grid cells) the choice of elevation band configuration should be made with more caution.