 |
| Titel |
Basin-scale runoff prediction: An Ensemble Kalman Filter framework based on global hydrometeorological data sets |
| VerfasserIn |
Harald Kunstmann, Christof Lorenz, Mohammad Tourian, Balaji Devaraju, Nico Sneeuw |
| Konferenz |
EGU General Assembly 2016
|
| Medientyp |
Artikel
|
| Sprache |
en
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 18 (2016) |
| Datensatznummer |
250128288
|
| Publikation (Nr.) |
 EGU/EGU2016-8267.pdf |
|
|
|
|
|
| Zusammenfassung |
| In order to cope with the steady decline of the number of in situ gauges worldwide, there is a
growing need for alternative methods to estimate runoff. We present an Ensemble
Kalman Filter based approach that allows us to conclude on runoff for poorly or
irregularly gauged basins. The approach focuses on the application of publicly available
global hydrometeorological data sets for precipitation (GPCC, GPCP, CRU, UDEL),
evapotranspiration (MODIS, FLUXNET, GLEAM, ERA interim, GLDAS), and water storage
changes (GRACE, WGHM, GLDAS, MERRA LAND). Furthermore, runoff data from the
GRDC and satellite altimetry derived estimates are used. We follow a least squares prediction
that exploits the joint temporal and spatial auto- and cross-covariance structures of
precipitation, evapotranspiration, water storage changes and runoff. We further consider
time-dependent uncertainty estimates derived from all data sets. Our in-depth analysis
comprises of 29 large river basins of different climate regions, with which runoff is predicted
for a subset of 16 basins. Six configurations are analyzed: the Ensemble Kalman Filter
(Smoother) and the hard (soft) Constrained Ensemble Kalman Filter (Smoother).
Comparing the predictions to observed monthly runoff shows correlations larger than 0.5,
percentage biases lower than ± 20%, and NSE-values larger than 0.5. A modified
NSE-metric, stressing the difference to the mean annual cycle, shows an improvement of
runoff predictions for 14 of the 16 basins. The proposed method is able to provide
runoff estimates for nearly 100 poorly gauged basins covering an area of more than
11,500,000 km2 with a freshwater discharge, in volume, of more than 125,000 m3/s. |
|
|
|
|
|
|