Stable water isotopes in tropical South American ice-cores
Numerous ice-cores were recovered from mid-latitude glaciers, e.g. as reported by Schotterer et al. . Tropical icecores detain a highly valuable information, since they record the climate variability of the inter-tropical belt. The latter is proved to be particularly sensitive to global climate variations. Nevertheless, only high altitude sites can offer no-melting conditions in spite of intense zenith radiation at tropical latitudes. The requirements are met at tropical latitudes in the South American Andes only, where summits commonly exceed 6000 m. Table 1.2 summarises the locations of 6 tropical and 1 sub-tropical ice-core drilling sites in the Andes. Several ice-cores records extend back to 20 000 years B.P., i.e. they comprise the transition from the last glacial maximum (LGM) to the present Holocene inter-glacial stage [Pierrehumbert, 1999, Thompson et al., 1995, 1998, 2000, Ginot, personal communication]. The no-melting condition is essential for preserving the isotopic signal in the ice-core archive. The energy balance on high altitude / low latitude glaciers show that significant sublimation of ice occurs. However, Stichler et al.  reports that bulk sublimation is a non-fractionating process, hence it does not affect the isotopic signal.
The interpretation of the SWI as a climate proxy is even more complex for tropical ice-core than polar ones: unlike the latter, the site of precipitation evaporation is highly variable, which later undergoes recycling (transpiration) by the vegetation and condensates dominantly in convective instead of stratiform clouds. Figure 1.2 illustrates the location of ice-core drilling sites. The origin of precipitation at the drilling site is inferred from streamlines based on the mean seasonal moisture advection. An extensive description of the applied method is given in Sturm et al. [2005a, cf. following chapters]. Although being very close to the Pacific ocean, most of the Andean precipitation during the rainy season (austral summer, from December to February) originally evaporate over the Atlantic ocean, and are later recycled over the Amazon basin. In order to calibrate the SWI signal in the tropical Andes in term of climate proxy [Edwards et al., 2002], a network of precipitation sampling stations was established across Bolivia by Françoise Vimeux [Vimeux et al., -]. This calibration reveals that local temperature has hardly any control on the isotopic composition of precipitation. Yet the proximity of Andean ice-cores to the tropical Pacific is reflected in the SWI signal. Bradley et al.  reports a good correlation between the SWI signal and sea-surface temperature (SST) anomalies in the tropical Pacific, home of the El Niño Southern Oscillation (ENSO) phenomenon. How can these contradicting observations be explained ?
Atmospheric modelling of the stable water isotope cycle
Rayleigh-based SWI distillation models [Dansgaard, 1964] are unable to account for observed variations of the isotopic signal in tropical ice-cores. One the other hand, SWI enabled general circulation models shed a new light on the climate controls of the regional isotopic signal. Vuille et al. [1998, 2003b,a] reports that inter-annual variability is dominated by ENSO mode. The authors argue that the regional SWI patterns, ultimately recorded in the Andean ice-cores, are remotely controlled by SST anomalies in the tropical Pacific. Yet this statistical evidence is not consistent with the regional circulation patterns over South America. Hoffmann , Hoffmann et al.  underline the similarity of the SWI signal in all tropical ice-core records. Accordingly, the authors define an Andean Isotope Index (AII) that reinforces the importance of circulation patterns east of the Andes, including soil-atmosphere interactions across the Amazon basin. The actual control of the SWI signal in the tropical ice-cores is still a controversial issue. The poor representation of orography notably restricts the representativeness of GCM simulations for the particular case of high altitude records. At a horizontal resolution commonly used in GCMs (3.75 ±, i.e. » 420 km), the smoothed Andean cordillera culminates at 1500 masl.
How well are meso-scale features then represented, when the actual relief exceeds 6000 m above the Altiplano at 4000 masl ?
Facing this limitation, the need of SWI simulations at a finer resolution was patent. This gave rise to the present PhD project: incorporating stable water isotope diagnostics in a regional circulation model, in order to account both for synoptic and meso-scale controls of the SWI signal over South America and eventually in the tropical ice-cores. The present thesis illustrates the successive steps required to address this initial question, each of which constitute one chapter.
Implementation of SWI diagnostics. The technical implementation of SWI diagnostics in the regional circulation model REMO constitutes a significant part of the present work, despite the similarity between the physical parameterisationscheme of REMO and the SWI enabled GCM ECHAMiso. Technical details are presented in chapter 2 Presentation of REMOiso on page 15. Bärbel Langmann (Max Planck Institute for Meteorology – MPIfMet), Georg Hoffmann (Laboratoire des Sciences du Climat et de l’Environnement) and Ralf Podzun (MPIfMet) contributed to this work.
Initial validation of REMOiso over Europe. Following the different steps of SWI implementation in REMO, we conducted an initial validation of the isotope module over Europe. REMO has been extensively applied over this domain. The characteristics of the European climate are well reproduced by REMO. Furthermore, Europe features a dense coverage of reliable SWI measurements. These results are presented in chapter 3 Validation over Europe on page 29. They are published in a special issue of the publication Hydrological Processes, consecutive to the IAEA
Symposium on isotope hydrology [Sturm, Hoffmann, Langmann, and Stichler, 2005b].
Mean annual cycle over South America. In accordance with the general objectives of the present PhD, REMOiso’s study domain was shifted from Europe to South America. No thorough evaluation of the model skills is available in the South American climatic context. Hence a pre-requisite consists evaluating to which extend REMO is able to reproduce typical characteristics of the South American climate. The representation of the annual SWI cycle is then assessed on the basis of monthly SWI measurements. This study further provides the opportunity to test the sensitivity of the simulated SWI signal to horizontal resolution. REMOiso is compared to ECHAMiso at T30 and T106 resolution. These results constitute chapter 4 South America isotope climatology on page 61. They are currently submitted for publication in the Journal of Climate [Sturm, Hoffmann, and Langmann, 2005a].
The South American monsoon system recorded by SWI. Once the validity of REMOiso over South America is established, we focus on the dominant intra-seasonal mode during the rainy season (austral summer – December to February). Precipitation features a dipole pattern between the Paraná and Nordeste regions. This bimodal behaviour is also captured in the SWI, which further reveals the meteorological mechanisms responsible for the latter.
The South American monsoon system offers a comprehensive framework for the explanation of the precipitation dipole. This process-based analysis provides new elements for the interpretation of the SWI signal in Andean icecore records. These results are presented in chapter 5 South American monsoon system on page 95, to be submitted for publication in the Journal of Geophysical Research [Sturm, Vimeux, and Krinner, 2005c].
Conclusion and outlooks. The logical links between the successive research stages listed above are highlighted in chapter 6 on page 127. Furthermore, we suggest scientific questions, which the application of REMOiso would be suitable for. These include the recent implementation of nudging techniques to address specific case studies with REMOiso. Further improvements of parameterisation of fractionation processes are also proposed.
The appendixes briefly present sideline projects conducted with REMOiso. Appendix A on page 135 is a technical memorandum on the computation of mean sea-level pressure (SLP) for regional climate models. Appendix B on page 143 shows preliminary results by REMOiso at sub-polar and polar latitudes: the study domain comprises Canada and Greenland. This work was performed in collaboration with Jean Birks (Uni. of Waterloo, Canada). Appendix C on page 153 introduces the international project IPILPS (Isotopes in the Project for Inter-comparison of Land-surface Parameterisation Schemes). REMOiso provided forcing conditions at three locations (Manaus – Brazil, Neuherberg – Germany, Tumbarumba – Australia) for various SWI enabled land-surface schemes. The validation was conducted with Matthew Fischer (Australian Nuclear Science and Technology Organisation – ANSTO) and Kendal McGuffie (Uni. of Technology, Sydney), under supervision of Ann Henderson-Sellers (ANSTO).
Table of contents :
1.1 From climate change to paleo-climatology
1.2 Stable water isotopes: a major climate proxy
1.3 Stable water isotopes in tropical South American ice-cores
1.4 Atmospheric modelling of the stable water isotope cycle
2 Presentation of REMO
2.1 Structure of the regional circulation model REMO
2.1.1 Rotated horizontal grid
2.1.2 Vertical discretisation
2.1.3 Partial REMO flow diagram
2.2 Physical parameterisation and isotopic fractionation
2.2.1 Vertical diffusivity
2.2.2 Cloud microphysics
2.2.3 Land-surface scheme
2.3 Pre-processing of REMO boundary files
2.3.1 Invariant soil libraries
2.3.2 Time-dependant boundary conditions
2.4 Practical considerations
3 Validation over Europe
3.1.1 Empirical analyses and identification of isotope ’effects’
3.1.2 Modelling the isotopic composition of precipitation
3.2 Description of REMOiso
3.2.1 Standard climatic settings for REMO
3.2.2 Isotopic settings in REMO
3.3 Results and discussion
3.3.1 Mean annual d18O in precipitation
3.3.2 Monthly time series at GNIP-IAEA stations
3.3.3 Precipitation event records at GSF stations
3.4 Conclusions and outlooks
3.4.1 Summary of REMOiso performances
3.4.2 Downscaling of isotopic precipitation
3.5 Spectral nudging of REMOiso over Europe
3.5.1 Limits in comparing GCM to station observations
3.5.2 GCM nudging
3.5.3 Spectral nudging of RCM
3.5.4 Case study in February 1983
4 South America isotope climatology
4.2 Model and data
4.2.1 Model description and experiment set-up
4.2.2 Gridded precipitation observations and isotopic measurements
4.3 Evaluation against observations
4.3.2 Atmospheric circulation
4.3.3 Isotopic composition of precipitation
4.4 Climatic interpretation of the water isotope signal
4.4.1 The simulated altitude effect
4.4.2 Continental gradient across the Amazon basin
4.4.3 Evolution of the isotopic composition of moisture along its trajectory
4.5 Conclusion and perspectives
5 South American monsoon system
5.2 Model and methods
5.2.1 REMO Experiment with climatological SST
5.2.2 Empirical Orthogonal Functions: why and how ?
5.3 The Paraná – Nordeste dipole
5.3.1 Oceanic versus continental SACZ
5.3.2 Upper level atmospheric motion
5.3.3 Low level atmospheric motion
5.4 Stable water isotopes: an integrated proxy of the SAMS
5.4.1 d18Oand moisture trajectory
5.4.2 Regional relevance of station measurements
5.5.1 Synoptic forcing of the SAMS
5.5.2 Role of the low-level jets (LLJ)
5.5.3 South Atlantic SST feedback on the SACZ
5.5.4 ENSO modulation of the SAMS
5.6 Summary and conclusion
6 Conclusion and perspectives
6.1 Summary of studies with REMOiso
6.1.1 Validation over Europe
6.1.2 South America isotope climatology
6.1.3 South American monsoon system recorded in stable water isotopes
6.2 Further applications of REMOiso
6.2.1 Nudged simulation over South America
6.2.2 Collaborative projects: Canada, Tibet, polar regions
6.3 Suggested development of REMO’s stable water isotope module
6.3.1 Adaptation of boundary conditions for paleo-simulations
6.3.2 Soil hydrology improvements for isotopic catchment studies
6.3.3 Coupling to an improved SVAT model : MECBETH
A SLP computation
B REMO over Canada
C IPIPLS forcing fields