%0 Journal Article
%9 ACL : Articles dans des revues avec comité de lecture répertoriées par l'AERES
%A Ryan, A. G.
%A Regnier, C.
%A Divakaran, P.
%A Spindler, T.
%A Mehra, A.
%A Smith, G. C.
%A Davidson, F.
%A Hernandez, Fabrice
%A Maksymczuk, J.
%A Liu, Y.
%T GODAE OceanView Class 4 forecast verification framework : global ocean inter-comparison
%D 2015
%L fdi:010066110
%G ENG
%J Journal of Operational Oceanography
%@ 1755-876X
%M ISI:000368117600007
%N 1
%P S98-S111
%R 10.1080/1755876x.2015.1022330
%U https://www.documentation.ird.fr/hor/fdi:010066110
%> https://www.documentation.ird.fr/intranet/publi/2016/02/010066110.pdf
%V 8
%W Horizon (IRD)
%X As part of the work of the GODAE OceanView Inter-comparison and Validation Task Team (IV-TT), 6 global ocean forecasting systems spread across 5 operational oceanography forecast centres were inter-compared using a common set of observations as a proxy for the truth. The `Class 4' in the title refers to a set of forecast verification metrics defined in the MERSEA-IP/GODAE internal metrics document (Hernandez 2007), the defining feature of which is that comparisons between forecasts and observations take place in observation space. This approach is seen as a departure from other diagnostic approaches such as analysing model trends or innovation statistics, and is commonly used in the atmospheric community. The physical parameters involved in the comparison are sea surface temperature (SST), sub-surface temperature, sub-surface salinity and sea level anomaly (SLA). SST was measured using in-situ observations obtained from USGODAE, sub-surface conditions were compared to Argo profiles, while sea level anomaly was measured by several satellite altimeters courtesy of AVISO. The 5 forecast centres involved in the project were Met Office, Australian Bureau of Meteorology, Mercator Ocean, Environment Canada and NOAA/NWS/NCEP. Combining Met Office, Mercator Ocean and Environment Canada forecasts into a mixed resolution multi-model ensemble produces estimates of the ocean state which have better accuracy and associativity properties for SST, SLA and temperature profiles than any individual ensemble component.
%$ 032 ; 020 ; 126