 |
| Titel |
Comparison of various remote sensing snow products in a distributed hydrological model |
| VerfasserIn |
Tomasz Berezowski, Jarosław Chormański, Okke Batelaan |
| Konferenz |
EGU General Assembly 2014
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 16 (2014) |
| Datensatznummer |
250089963
|
| Publikation (Nr.) |
 EGU/EGU2014-4177.pdf |
|
|
|
|
|
| Zusammenfassung |
| With the development of remote sensing, more and more data series with spatially
distributed snow cover become available. These data can be obtained for free, from many
sources varying in spatial and temporal resolution, the length of the time series
and the method of acquisition (VIS-NIR or microwave sensors). A popular use of
remotely sensed snow distribution data is in hydrological modelling. However, a
suitability test of different remote sensing snow products for hydrological models was so
far not conducted. In this work, some of the most common remote sensing snow
products (MOD10A1, IMS , GLOBSNOW and AMSR-E_DySno) are used as input
data in the WetSpa distributed hydrological model. Each of the snow products has
different properties and is based on different algorithms, which makes the analysis
interesting and multidimensional. The area of research is the Biebrza River catchment -
located in north-eastern Poland, comprising approximately 7000 km2. Biebrza is
a natural river with a snow melt regime, making it very suitable for this kind of
analysis.
In total 6 modelling scenarios were conducted (4 with remote sensing data, 1 standard
approach – temperature threshold for snow accumulation and melting, 1 based on snow data
from meteorological stations). Each model was calibrated against discharge with the Shuffled
Complex Evolution (SCE) algorithm. The calibration was repeated three times
for each model to make sure that the global optimum was found. The calibration
and validation periods were both 3 years long. The next stage was a comparison
with the GLUE uncertainty analysis for each of the models, on a shorter, one-year
period.
The best model in terms of Nash-Sutcliffe efficiency and r2 was using the MOD10A1
data; however, the models using GLOBSNOW SWE and the standard approach received
similar scores. In terms of the model bias the best results were obtained for the IMS and
MOD10A1 data. Nevertheless, the lowest root mean square error was found for the models
using the standard approach and the MOD10A1 data. The GLUE analysis shows
that models using remote sensing data have lower uncertainty than those using the
standard approach, as the remotely sensed driven models need less global parameters. |
|
|
|
|
|
|