Premium
Probabilistic multi‐model ensemble predictions of nitrogen concentrations in river systems
Author(s) -
Exbrayat JeanFrançois,
Viney Neil R.,
Frede HansGeorg,
Breuer Lutz
Publication year - 2011
Publication title -
geophysical research letters
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.007
H-Index - 273
eISSN - 1944-8007
pISSN - 0094-8276
DOI - 10.1029/2011gl047522
Subject(s) - probabilistic logic , environmental science , nitrogen , ensemble forecasting , eutrophication , statistical model , drainage basin , econometrics , hydrology (agriculture) , computer science , statistics , mathematics , ecology , geology , chemistry , geography , machine learning , geotechnical engineering , cartography , organic chemistry , nutrient , biology
Because of an increasing human pressure on the naturally balanced nitrogen cycle, eutrophication‐driven hypoxia and corresponding dead zones have multiplied over the last years. Models are used to predict the nitrogen balances and develop mitigation scenarios for such systems. Due to the very complex interaction of water and nitrogen fluxes, no single model structure has been commonly adopted to describe the fluxes best and modeling results often differ substantially. Relying on a single model prediction only can be therefore highly uncertain. Here we illustrate the potential advantage of using a probabilistic multi‐model ensemble approach in comparison to predictions of each of its members. Evaluation of corresponding skills and potential economic values to correctly predict a 2 mg N/L target of total nitrogen concentration in the water flowing out of a catchment in south‐west Western Australia is conducted. Results show that the ensemble has always more skill and almost always more potential economic value than any of its four members and that it therefore constitutes a more reliable choice in the decision‐making process.