Deuterium excess in Greenland snow: Analysis with simple and complex models

International audienceA simple Rayleigh-type isotope model, typical of those used to develop algorithms for extracting climatic information from stable water isotope paleodata, is evaluated against the more complex and presumably more reliable calculations of a general circulation model (GCM) fitted with isotope tracer diagnostics. The evaluation centers on an analysis of how the temperature T e of an oceanic moisture source affects the deuterium excess d of Greenland precipitation. The annual Te-d relationship derived from the GCM diagnostics is largely reproduced by the simple isotope model when the latter is properly initialized. This, coupled with the fact that the GCM itself reproduces observed isotope behavior, suggests that the simpler model's atmospheric calculations are indeed adequate for isotope studies. Furthermore, the GCM results support the idea, originally developed with the simpler models, that polar deuterium excess values contain information on meteorological conditions at distant evaporative sources. 1. Introduction The stable isotopes of water, HDO and H}sO, have been measured in ice cores and other paleowaters in varying concentrations. Through a detailed analysis of current isotope concentration fields, isotope/climate relationships have been derived which allow the extraction of paleoclimatic temperatures from paleowater measurements (see Jouzel et al. [1997] for a recent review). A related isotopic quantity, deuterium excess, is now being used to infer additional paleoclimatic information. Deuterium excess d was defined by Daansgaard [1964] as d = •D-8•sO, where • indicates a permil deviation from the corresponding isotope ratio in standard mean ocean water (SMOW). The factor 8 comes from the meteoric water line, which defines the locus of modern precipitation in a 5D/5•80 plot [Craig, 1961]. Using a simple evaporation model and a Rayleigh-type precipitation model, Merlivat and Jouzel [1979] inferred that the deuterium excess of precipitation primarily depends on the mean relative humidity above the evaporative (oceanic) source for the moisture. Jouzel et al. [1982] then interpreted the reduced glacial d values (relative to modern values) in an East Antarctic core as a reflection of higher relative humidity over the oceanic areas providing moisture for Antarctic precipitation. Johnsen et al. [1989] pointed out that d is also significantly affected by the temperature of the moisture source and examined d variations with respect to absolute (rather than relative) Paper number 98JD00274. 0148-0227/98/98JD-00274509.00 humidity, leading Daansgaard et al. [1989] to interpret the abrupt d change at the termination of the Younger-Dryas in Greenland's Dye 3 core in terms of a rapid retreat of sea-ice cover. The dual importance of humidity and temperature at the evaporative source has also been recognized for Antarctica [Petit et al., 1991; Ciais and Jouzel, 1994; Ciais et al., 1995; Fisher, 1991]. The Rayleigh or Rayleigh-type distillation models (herein-after often referred to as "simple isotope models") usually applied in these studies essentially model isotope behavior within isolated air masses transported poleward from an ocean source. The idealized paths traversed by these air masses are determined by prescribed initial and final states for temperature and pressure. The simple isotope models account for the interplay between cloud microphysics and the fractionation processes occurring at each phase change of the water. They cannot, however, account for the complexity of dynamical processes that lead to the formation of precipitation. Furthermore , Jouzel and Koster [1996] recently showed that the standard approach used in these models for specifying the initial isotope contents within the air parcels introduces a systematic bias that can significantly affect the simulated relationships between deuterium excess and evaporative source conditions. An alternative approach to studying global water isotope behavior is to incorporate the isotopic cycles into an atmospheric general circulation model (GCM), which does simulate the dynamical complexity of the atmosphere and which avoids the noted initial conditions bias in the simple Rayleigh-type models. Isotope tracer diagnostics have been incorporated into at least four different GCMs [Joussaume et al., 1984; Jouzel et 894