 |
| Titel |
Scaling precipitation input to distributed hydrological models by measured snow distribution |
| VerfasserIn |
Christian Voegeli, Michael Lehning, Nander Wever, Mathias Bavay, Yves Bühler, Mauro Marty, Peter Molnar |
| Konferenz |
EGU General Assembly 2016
|
| Medientyp |
Artikel
|
| Sprache |
en
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 18 (2016) |
| Datensatznummer |
250135052
|
| Publikation (Nr.) |
 EGU/EGU2016-15849.pdf |
|
|
|
|
|
| Zusammenfassung |
| Precise knowledge about the snow distribution in alpine terrain is crucial for various
applications such as flood risk assessment, avalanche warning or water supply and
hydropower. To simulate the seasonal snow cover development in alpine terrain, the spatially
distributed, physics-based model Alpine3D is suitable. The model is often driven by spatial
interpolations from automatic weather stations (AWS). As AWS are sparsely spread, the data
needs to be interpolated, leading to errors in the spatial distribution of the snow cover
- especially on subcatchment scale. With the recent advances in remote sensing
techniques, maps of snow depth can be acquired with high spatial resolution and vertical
accuracy.
Here we use maps of the snow depth distribution, calculated from summer and winter
digital surface models acquired with the airborne opto-electronic scanner ADS to preprocess
and redistribute precipitation input data for Alpine3D to improve the accuracy of spatial
distribution of snow depth simulations. A differentiation between liquid and solid
precipitation is made, to account for different precipitation patterns that can be expected from
rain and snowfall. For liquid precipitation, only large scale distribution patterns are applied to
distribute precipitation in the simulation domain. For solid precipitation, an additional small
scale distribution, based on the ADS data, is applied. The large scale patterns are generated
using AWS measurements interpolated over the domain. The small scale patterns are
generated by redistributing the large scale precipitation according to the relative snow depth
in the ADS dataset. The determination of the precipitation phase is done using an air
temperature threshold.
Using this simple approach to redistribute precipitation, the accuracy of spatial snow
distribution could be improved significantly. The standard deviation of absolute snow depth
error could be reduced by a factor of 2 to less than 20 cm for the season 2011/12. The
mean absolute error in snow depth distribution could be optimized by choosing
reasonable large scale precipitation patterns. For inter-annual scaling, the model
performance could also be improved, even when using an ADS dataset from a different
season.
We conclude that using ADS data to process precipitation input, complex and
computation power intensive effects such as snowdrift due to wind, can be substituted and
modelling performance of spatial snow distribution is improved. |
|
|
|
|
|
|