 |
| Titel |
Spatial analysis and statistical modelling of snow cover dynamics in the
Central Himalayas, Nepal |
| VerfasserIn |
Johannes Weidinger, Lars Gerlitz, Jürgen Böhner |
| Konferenz |
EGU General Assembly 2017
|
| Medientyp |
Artikel
|
| Sprache |
en
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 19 (2017) |
| Datensatznummer |
250151625
|
| Publikation (Nr.) |
 EGU/EGU2017-16364.pdf |
|
|
|
|
|
| Zusammenfassung |
| General circulation models are able to predict large scale climate variations in global
dimensions, however small scale dynamic characteristics, such as snow cover and its
temporal variations in high mountain regions, are not represented sufficiently. Detailed
knowledge about shifts in seasonal ablation times and spatial distribution of snow cover are
crucial for various research interests. Since high mountain areas, for instance the Central
Himalayas in Nepal, are generally remote, it is difficult to obtain data in high spatio-temporal
resolutions. Regional climate models and downscaling techniques are implemented
to compensate coarse resolution. Furthermore earth observation systems, such as
MODIS, also permit bridging this gap to a certain extent. They offer snow (cover) data
in daily temporal and medium spatial resolution of around 500 m, which can be
applied as evaluation and training data for dynamical hydrological and statistical
analyses. Within this approach two snow distribution models (binary snow cover and
fractional snow cover) as well as one snow recession model were implemented
for a research domain in the Rolwaling Himal in Nepal, employing the random
forest technique, which represents a state of the art machine learning algorithm.
Both bottom-up strategies provide inductive reasoning to derive rules for snow
related processes out of climate (temperature, precipitation and irradiance) and
climate-related topographic data sets (elevation, aspect and convergence index) obtained by
meteorological network stations, remote sensing products (snow cover - MOD10-A1
and land surface temperatures - MOD11-A1) along with GIS. Snow distribution is
predicted reliably on a daily basis in the research area, whereas further effort is
necessary for predicting daily snow cover recession processes adequately. Swift changes
induced by clear sky conditions with high insolation rates are well represented,
whereas steady snow loss still needs continuing effort. All approaches underline
the technical difficulties of snow cover modelling during the monsoon season, in
accordance with previous studies. The developed methods in combination with
continuous in situ measurements provide a basis for further downscaling approaches. |
|
|
|
|
|
|