 |
| Titel |
Reduction of predictive uncertainty in estimating irrigation water requirement through multi-model ensembles and ensemble averaging |
| VerfasserIn |
S. Multsch, J.-F. Exbrayat, M. Kirby, N. R. Viney, H.-G. Frede, L. Breuer |
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| ISSN |
1991-959X
|
| Digitales Dokument |
URL |
| Erschienen |
In: Geoscientific Model Development ; 8, no. 4 ; Nr. 8, no. 4 (2015-04-29), S.1233-1244 |
| Datensatznummer |
250116287
|
| Publikation (Nr.) |
 copernicus.org/gmd-8-1233-2015.pdf |
|
|
|
|
|
| Zusammenfassung |
| Irrigation agriculture plays an increasingly important role in food supply.
Many evapotranspiration models are used today to estimate the water demand
for irrigation. They consider different stages of crop growth by empirical
crop coefficients to adapt evapotranspiration throughout the vegetation
period. We investigate the importance of the model structural versus model
parametric uncertainty for irrigation simulations by considering six
evapotranspiration models and five crop coefficient sets to estimate
irrigation water requirements for growing wheat in the Murray–Darling Basin,
Australia. The study is carried out using the spatial decision support system
SPARE:WATER. We find that structural model uncertainty among reference ET is
far more important than model parametric uncertainty introduced by crop
coefficients. These crop coefficients are used to estimate irrigation water
requirement following the single crop coefficient approach. Using the
reliability ensemble averaging (REA) technique, we are able to reduce the
overall predictive model uncertainty by more than 10%. The exceedance
probability curve of irrigation water requirements shows that a certain
threshold, e.g. an irrigation water limit due to water right of 400 mm,
would be less frequently exceeded in case of the REA ensemble
average (45%) in comparison to the equally weighted ensemble average (66%).
We conclude that multi-model ensemble predictions and sophisticated model
averaging techniques are helpful in predicting irrigation demand and provide
relevant information for decision making. |
| |
|
| Teil von |
|
|
|
|
|
|
|
|