HR: 0800h
AN: PP51C-1345    [Abstracts]
TI: Sea Surface Salinity reconstruction in Fiji during the last century from Multi-Proxies of Coral Skeleton using Neural Network: preliminary results.
AU: Naveau, P
EM: Philippe.Naveau@lsce.saclay.cea.fr
AF: LSCE, Domaine du CNRS, Gif sur Yvette, 91198 France
AU: * Juillet-Leclerc, A
EM: Anne.Juillet@lsce.cnrs-gif.fr
AF: LSCE, Domaine du CNRS, Gif sur Yvette, 91198 France
AU: Blamart, D
AF: LSCE, Domaine du CNRS, Gif sur Yvette, 91198 France
AU: Correge, T
AF: IRD, BPA5, Noumea, 000000 New Caledonia
AB: In contrast to the Equatorial eastern pacific where sea surface temperature (SST) anomalies are maximum during an ENSO, the western tropical pacific is essentially affected by sea surface salinity (SSS) variability which could be a powerful indicator of ENSO during the past. Coral skeleton is regarded as the best archives to record oceanic conditions over the last centuries. Unfortunately, up to now, it seems difficult to decipher SST and SSS reconstructions even from multi proxies time series. Midway Tahiti and Darwin, Fiji islands are key areas to record SSS. Located at the southern warm pool edge, this zone is submitted to heavy precipitation due to South Pacific Convergence Zone (SPCZ) during La Ni¤a and salty subtropical waters are advected westward during El Ni¤o. This situation has been well documented from salinity data collected by ships between 1976 and 2000, but SSS measurements are sparse in historical database, specially in this area. By combining geochemists and biologists understanding we recognize the processes involved in the coral skeleton deposit (cf BG13). We infer that the response to environmental forcing is embedded with biological reactions and geochemical records are not linear in the time. This result is obtained by using a neural network on multiproxies (carbon and oxygen isotopic ratio, trace elements and density) to reconstruct SSS and SST during the last century in a coral head collected in Yasawa (Fiji). The proxies necessary for estimating SSS and SST are chosen by heuristic variable selection (HVS); SSS is reconstructed from 4 proxies while for SST 7 proxies are used. Thus, the first step of this mathematical treatment is to train our data set during the period 1961-1997. Then, the reconstructions are derived. Over the last century, the interannual SSS and SST variability's respectively range from 1% to 1øC. A comparison with SOI shows that as it has been observed from the last decades, during the whole XXth century, SSS is higher and SST lower during El Ni¤o. The SSS increase is due to reduced precipitation and also the displacement of cold and salty water by a zonal advection.
DE: 4808 Chemical tracers
DE: 4870 Stable isotopes
DE: 4215 Climate and interannual variability (3309)
DE: 4522 El Ni¤o
SC: Paleoceanography and Paleoclimatology [PP]
MN: 2004 AGU Fall Meeting