Impact of proxies and prior estimates on data assimilation using isotope
ratios for the climate reconstruction of the last millennium
Abstract
In climate reconstructions by data assimilation, the sensitivities to
both proxies and prior estimates need to be taken into account because
models are uncertain and proxies are limited spatiotemporally. This
study examines these sensitivities using multiple climate model
simulations and different combinations of proxies (corals, ice cores,
and tree-ring cellulose). Experiments were conducted based on an offline
data assimilation approach. These experiments show annual variations in
the global distribution of surface air temperature and precipitation
range from 850 to 2000. The results indicate that standard deviations of
surface air temperature and precipitation amount during the entire
period differ by up to 50% due to prior estimates. Experiments with
different types of proxies show that the El Niño-like distribution of
positive anomalies in the central to eastern tropical Pacific can be
reproduced adequately in experiments with corals, but not in experiments
without corals. The correlation coefficient of the NINO.3 index from
1971 to 2000 between experiments with corals and the Japanese 55-year
Reanalysis (JRA-55) were 0.79 at maximum, while the correlation
coefficient between experiments without corals and JRA-55 were 0.20 at
maximum.