![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
Potentials and limitations of seasonal runoff predictions for Swiss mesoscale basins |
VerfasserIn |
Simon Schick, Ole Rössler, Rolf Weingartner |
Konferenz |
EGU General Assembly 2014
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 16 (2014) |
Datensatznummer |
250092319
|
Publikation (Nr.) |
EGU/EGU2014-6653.pdf |
|
|
|
Zusammenfassung |
Prediction of long-term runoff (i.e. month, season, year) is a valuable information for
decision-makers in hydropower industries, water resources management and inland
water transports. Common approaches for seasonal runoff forecasts can mainly be
categorized by the integration or disintegration of numerical climate predictions. While
the former is a quite new development in this research field, the latter has been
introduced already in the 1950s for Swiss basins using a linear regression model and
several case studies can be found. Nevertheless scientific literature lacks an overview
concerning spatial as well temporal differences of seasonal runoff predictability in Swiss
basins.
In this study we applied a simple partial least squares regression model to predict seasonal
runoff and evaluated how this approach performs across different discharge regimes and over
the year. Here, we defined season as a time window of 91 consecutive days with
arbitrary position within the calendar year. Furthermore, the quartiles of these 91 daily
runoff values were choosen as the target values and temperature, precipitation and
runoff prior to the forecast date as predictors. Hence, the model itself does not make
any assumptions about future weather and climate – the forecasts are based on
the disposition of the specific basin at the date of forecast and assume a memory
effect caused by interactions of water storages such as soil, groundwater, lakes and
snow.
Seasonal runoff forecasts for 24 Swiss mesoscale basins (100-2000 km2) were then
analyzed to estimate temporal and spatial differences in goodness of prediction. We show that
model skill varies strongly through the calendar year. In spring and autumn we observe best
model performance, whereas for summer the prediction benefit is smaller relative to the
discharge regime as reference. On the other hand spatial differences in goodness of prediction
were much smaller – alpine catchments show best predictability by trend, because of snow
accumulation during winter and delayed melting in spring. In addition the study showed the
importance of lakes for seasonal runoff predictions. Integrating seasonal climate
predictions in a next step is expected to increase model skill, but will remain a
challenging task because of the low predictability of climate over Central Europe. |
|
|
|
|
|