|
Titel |
Generating spatial precipitation ensembles: impact of temporal correlation structure |
VerfasserIn |
O. Rakovec, P. Hazenberg, P. J. J. F. Torfs, A. H. Weerts, R. Uijlenhoet |
Konferenz |
EGU General Assembly 2012
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 14 (2012) |
Datensatznummer |
250061622
|
|
|
|
Zusammenfassung |
Precipitation is the most dominant input term determining the hydrological response at the
catchment scale. For both scientific and applied hydrological studies one is interested to
have a sound spatial estimate of precipitation and its uncertainty. Sound spatially
distributed rainfall fields including a proper spatial error structure can be obtained by
conditional simulation, which unlike kriging interpolation does not only provide the best
local estimate, but generates realizations which match the sample statistics and
can also be conditional on the neighbouring estimates. Both interpolations and
conditional simulations have their origin in the spatial domain and do not primarily
take into account the temporal evolution of the spatial field. Nevertheless, the large
temporal variability of precipitation is important to be considered together with its
spatial variability within the whole ensemble. This can be achieved using spatial
conditional simulations which are made conditional on previous simulations back in
time.
Synthetic and real world experiments are carried out within the hilly region of Belgian
Ardennes. 27 hourly rain gauges are available within the simulation domain with an area of
10 000 km2. Precipitation fields were simulated on a grid with 10 x 10 km2 raster resolution.
The analysis tested the uncertainty in the simulated fields based on 1) the number of previous
simulation hours, on which the new simulation is conditioned, 2) the advection speed of
the rainfall event, 3) the size of the catchment considered and 4) the rain gauge
density within the catchment. The results show that for typical advection speeds
of >20Â km/h no uncertainty in terms of across ensemble spread (expressed using
coefficient of variation) is added to simulated precipitation fields by conditioning it on
more than one or two previous simulations. Moreover, by halving the observation
network, i.e. using 14 rain gauges, the uncertainty in simulations increases only
slightly. |
|
|
|
|
|