HR: 17:00h
AN: PP34A-05    [Abstracts]
TI: Independent Temperature and Salinity Reconstruction from Coral Skeleton
AU: * Juillet-Leclerc, A
EM: Anne.Juillet@lsce.cnrs-gif.fr
AF: Laboratoire des Sciences du Climat et de l'Environnement, Campus du CNRS, Gif sur Yvette, 91189, France
AU: Thiria, S
AF: LOCEAN, Universite Pierre et Marie Curie, Paris, 75005, France
AU: Peron, C
AF: Laboratoire des Sciences du Climat et de l'Environnement, Campus du CNRS, Gif sur Yvette, 91189, France
AB: Massive coral skeleton offers the best-suited material for reconstruction of tropical climate during the last century. Indeed, growth rate of some species is fast enough to provide high-resolution (monthly) sampling. The formation of a big colony may cover continuously several decades, even, several centuries. Chronology is made easy by annual growth layers. Finally, aragonite, which composes this skeleton, presents several proxies, such as isotopes and trace elements. However, the aragonite deposit is biologically controlled and all the proxies are influenced by both environmental and biologic factors. By example, temperature and light may affect both oxygen isotopes as well as Sr/Ca. Thus, these two parameters are difficult to separate from seasonal records. Such an effect may be neglected for annual averages or filtered records. Indeed, the difference of irradiation recorded during two consecutive years is much limited than between winter and summer (see PP24). It is the reason that annual and monthly reconstitutions will be conducted separately. In addition, it seems that other factors are the causes of the high variability shown by all the proxies. The common factor affecting them is related with metabolism. We suppose that proxies being measured from a powder collected from homogeneous material are fractionated by external factors through the same biologic filter during the time, but each of them differently because incorporated in the mineral by different ways. This is the reason that we used neural network (NN), which learns the behavior of several proxies submitted to one forcing during a known period. Then, the complex relationship recognized by neurons between the different proxies is used to "predict" the forcing during the past. The relationship between external factor and proxies remains hidden, but this could not be used for other colonies, even for other sampling from a same head. By this way, it is possible to calibrate temperature and salinity and thus to reconstruct separately, with two different networks, the two variables and thus reduce errors. A similar operation may be performed for seasonal variations. Results obtained from a coral core harvested in Fiji from seven proxies for SST and four for SSS, indicated that SST and SSS are modified at each strong El Nino during the last century. The optimization of the use of this statistical treatment and the estimation of the error will be discussed.
DE: 0473 Paleoclimatology and paleoceanography (3344, 4900)
DE: 1616 Climate variability (1635, 3305, 3309, 4215, 4513)
DE: 3374 Tropical meteorology
DE: 4916 Corals (4220)
SC: Paleoceanography and Paleoclimatology [PP]
MN: 2007 Fall Meeting