Influence of microscale in snow distributed modelling in seminarid regions

Resumen

This work focuses on the importance of the microscale snow distribution in the modelling of the snow dynamics in semiarid regions. Snow over these areas has particular features that further complicate its measuring, monitoring and modelling (e.g. several snowmelt cycles throughout the year and a very heterogeneous distribution). Most extended GIS-based calculation of snowmelt/accumulation models must deal with non-negligible scales effects below the cell size, which may result in unsatisfactory predictions depending on the study scale. This study proposes the joint use of physically- based distributed snowmelt-accumulation modelling and remote sensing observation datasets to study the subgrid variability of snow distribution and its effects on the snow modelling at the watershed scale. The study has been carried out in Sierra Nevada Mountains, southern Spain, where the highest submit of the Iberian Peninsula can be found close to the seaside, which results in a sharp gradient of climate conditions associated to topography. The typical Alpine climate in the mountains is modulated by the subtropical conditions at the coast, with occurrence of snowfall usually from November to April in altitudes greater than 2000 m, and successive cycles of accumulation and snowmelt during the season. Terrestrial photography data, an alternative and economical remote sensing information source whose scales can be adapted to the studied processes requirements, has been employed to study the snow dynamics at the subgrid scale (30 x 30 m). Snow cover area and snow depth datasets were obtained from terrestrial images in a pilot study area during 2009-2013. These dataset was employed to define the subgrid variability by means of depletion curves with two different approaches. As a first step, different snow depletion curves proposed by other authors were tested at the study area. The observations were included in a data assimilation scheme, an Ensemble Transform Kalman Filter, in the energy and mass balance equations of the snow model. The results identified the need for selecting a particular depletion curve parameterization depending on the succession of accumulation-melting cycles. Secondly, based on the former results, these datasets were directly employed to define parametric depletion curves at the pilot area. A flexible sigmoid function was found to satisfactorily reproduce the observed trends of the snowmelt effects on the snow cover area at the cell scale, but different values resulted for the sigmoid’s parameters depending on the different snow states found: 1) cycles with a high accumulated snow depth which came from a metamorphosed snow; 2) cycles with great snow depth preceded by short accumulation phases; 3) cycles with low accumulation that occur in the cold season; and 4) cycles with low snow depth values which take place during spring. Moreover, an unique expression for the accumulation curve was also proposed. These results confirm the need of different parameterization to represent the physical variability in the accumulation-melting cycles in semiarid regions. Furthermore, this selective curve improves the model performance when compared to the results previously obtained with the data assimilation scheme. Finally, the proposed parameterization was tested at the watershed scale, at the Guadalfeo River Basin, at the southern face of Sierra Nevada. Snow cover area distributed results from the model simulations were assessed with Landsat TM and ETM+ observations (30 x 30m spatial resolution), consisting of snow cover maps at the area from an endmembers spectral mixture analysis of the Landsat imagery. These maps had previously been validated from snow cover maps at higher spatial resolution (10m x 10m) obtained from terrestrial photography in a monitoring hillside in the area. The results showed a significant agreement between observed...