Giessener Elektronische Bibliothek

GEB - Giessener Elektronische Bibliothek

Reduction of predictive uncertainty in estimating irrigation water requirement through multi-model ensembles and ensemble averaging

Multsch, Sebastian ; Exbrayat, J. F. ; Kirby, M. ; Viney, N. R. ; Frede, Hans-Georg ; Breuer, Lutz


Originalveröffentlichung: (2015) Geoscientific Model Development 8(4):1233-1244 doi:10.5194/gmd-8-1233-2015
Zum Volltext im pdf-Format: Dokument 1.pdf (1.186 KB)


Bitte beziehen Sie sich beim Zitieren dieses Dokumentes immer auf folgende
URN: urn:nbn:de:hebis:26-opus-119713
URL: http://geb.uni-giessen.de/geb/volltexte/2016/11971/


Sammlung: Open Access - Publikationsfonds
Universität Justus-Liebig-Universität GieĂźen
Institut: Institute for Landscape Ecology and Resources Management
Fachgebiet: IFZ Interdisziplinäres Forschungszentrum für Umweltsicherung
DDC-Sachgruppe: Biowissenschaften, Biologie
Dokumentart: Aufsatz
Sprache: Englisch
Erstellungsjahr: 2015
Publikationsdatum: 25.02.2016
Kurzfassung auf Englisch: 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.
Lizenz: Lizenz-Logo  Creative Commons - Namensnennung