Paleoclimate studies commonly interpret the stable water isotope ratio in precipitation as a proxy for temperature based on observed present day spatial and temporal correlations. The temperature-isotope relationship results both from the greater efficiency of isotopic fractionation at low temperatures and from the preferential condensation of the heavy isotopic species during moisture transport, which tend to coincide for the present day climate. The Melbourne University General Circulation Model is used to simulate an alternative scenario where these two effects are opposed, with idealized continents located in such a way that moisture transport occurs from cooler to warmer latitudes. It is found that the positive correlation between isotopic ratio and temperature breaks down for cases with oceans limited to mid- and high latitudes, indicating that the observed present day correlation between isotopic ratio and temperature is not invariant, but depends on the specific distributions of continents and resulting moisture transport.
 The ratio of oxygen isotopes (18O/16O) in water is given as δ18O relative to RVSMOW, the ratio in Vienna Standard Mean Ocean Water. Annual average δ18O is observed to be correlated with surface temperature in the extratropics [e.g., Dansgaard, 1964]. The isotopic ratio in meteoric water varies due to fractionation during evaporation and condensation. According to the Rayleigh distillation model, a moist air parcel undergoes condensation with the preferential removal of the heavy isotopic species during transport, resulting in progressively isotopically lighter vapor and precipitation. While the efficiency of isotopic fractionation is temperature-dependent, the observed spatial temperature-δ18O correlation also results from the “rainout” of the heavy isotope during transport, in the mean, towards cooler mid- and high latitude sites.
 A number of modeling studies have addressed the question of whether the present day temperature-isotope relationship holds under substantially different climate conditions, such as those which existed at the Last Glacial Maximum (LGM). For example, the modeling study of Charles et al.  found that changes in moisture transport and source regions for Greenland at the LGM may have produced an isotopic response independently of temperature changes. Noone  modeled the isotopic response to LGM climate over Antarctica and found that changes in the position and intensity of midlatitude cyclones produced a change in the isotopic rainout of moisture during transport to Antarctica. Such studies indicate that it is important to understand the fundamental controls on the extent of isotopic depletion in order to determine the robustness of the temperature-isotope relationship for a range of climate conditions.
 In this study we report on experiments which illustrate conditions under which the present day isotope-temperature relationship breaks down. A set of idealized experiments using the Melbourne University atmospheric General Circulation Model (MUGCM) with isotopic tracers are used to investigate the isotopic response to conditions where moisture transport and isotopic rainout proceed from cooler to warmer latitudes. In order to produce such a change in atmospheric circulation, the present day continents are replaced with continents centered on the equator extending in zonal bands with a range of meridional extents. The absence of a tropical ocean means that all precipitation at tropical latitudes results from moisture transported from (cooler) midlatitude and polar sources. The resulting temperature-isotope relationship is compared with the observed present day relationship.
2. Model Description and Experiments
 The MUGCM is a spectral atmospheric model which is used here at R21 resolution (3.3° lat × 5.6° lon) with nine vertical levels. The model incorporates a stable water isotope tracer scheme as described by Noone and Simmonds . This scheme represents the isotopically heavy species of water in parallel to “normal” water in the model's hydrologic cycle, with the inclusion of the appropriate equilibrium and kinetic isotopic fractionation during phase transitions. The scheme implemented in the MUGCM is similar to those described by Jouzel et al.  and Hoffmann et al. . The model has been found to produce a reasonable simulation of the global distribution of isotopic ratios in precipitation for the present day climate [Noone and Simmonds, 2002].
 Present day annual average sea surface temperatures (SSTs) from Reynolds  were used to construct global SST boundary conditions using the zonally averaged Southern Hemisphere SSTs repeated symmetrically in the Northern Hemisphere. At southern polar latitudes and where the zonal average SSTs were below zero, the SST values were set to zero. Over land, all continents had zero topography. The model was run with constant SSTs over the seasonal cycle, resulting in approximately meridionally symmetrical annual solar input.
 Five different continental distributions were constructed: an aqua-planet (AQUA), and zonal continental bands centered at the equator extending from 10°S–10°N (L10), from 25°S–25°N (L25), from 45°S–45°N (L45) and from 60°S–60°N (L60). In the simulations with modified continents, all other regions were designated as ocean with the same SSTs as prescribed in the AQUA simulation. The model was initialized from an isothermal, dry state with soil moisture isotopic ratios initialized to RVSMOW, and ocean surface isotopic ratios set to a constant value of RVSMOW. The model was run for twenty years for each scenario, with the final ten years of each simulation analyzed.
3.1. Simulated Climate and Isotopic Distribution
 The zonal averages of annual mean surface temperature, precipitation and δ18O for each scenario are shown in Figure 1. The surface temperatures over land in the L10 and L25 experiments are slightly warmer than the (fixed) AQUA SSTs at the same latitudes, while in the L45 and L60 experiments the surface temperatures are cooler over land than the AQUA SSTs. Despite the different gradients, in all simulations the surface temperature decreases meridionally from the equator to the poles. The atmospheric circulation for the L45 and L60 experiments is also substantially altered, with the subtropical high and subpolar low pressure regions shifted polewards, indicating a reorganization of the atmospheric overturning circulation.
 The changes in surface fluxes and atmospheric circulation are associated with altered zonal precipitation distributions. In the AQUA, L10 and L25 experiments, precipitation maxima are located at the equator and in the midlatitudes at around 40°. The highest precipitation totals are seen at the equator in the L10 experiment where the presence of the narrow band of land enhances convergence. The L25 experiment has less tropical precipitation due to reduced evaporation over the land-covered tropics, with maxima in the midlatitudes. In the L45 and L60 experiments, the reduced evaporation due to land coverage in the tropics and midlatitudes leads to low precipitation amounts at these latitudes, with maxima over the polar ocean moisture source regions.
 The complex response of atmospheric circulation and hydrology to the prescribed surface boundary conditions is reflected in the different zonal average δ18O signals produced. The greatest isotopic depletion in the AQUA experiment is seen at the equator and at around 60° latitude, where local temperatures are cool and midlatitude source moisture has undergone upstream isotopic rainout. The highest isotopic values are seen in the dry subtropical regions and at the poles where precipitation is low and surface temperatures are not substantially cooler than the midlatitudes.
 A similar isotopic distribution to AQUA is seen for the L10 experiment, with a slight equatorward shift in the maximum δ18O values. A higher δ18O minimum occurs at the equator due to warmer local temperatures as well as reduced rainout of tropical and subtropical moisture. The L25 experiment has a similar zonal isotopic distribution to AQUA polewards of 25°, with a local δ18O maximum at the equator where local precipitation and isotopic distillation is lower than for the AQUA and L10 experiments.
 In the L45 experiment, relatively depleted isotopic ratios are found inland from the region of maximum precipitation near the continental boundary. In the tropics, more enriched isotopic ratios reflect the low precipitation amounts locally and during transport. Secondary isotopic enrichment during reevaporation of falling condensation is also expected to be significant inland where atmospheric relative humidity is extremely low. The δ18O values are not calculated for the L60 experiment in the tropics where precipitation is negligible. The L60 experiment produces a δ18O maximum over the warmest ocean surface at around 70° latitude, with increasing depletion towards the equator reflecting isotopic rainout during transport.
3.2. Temperature-Isotope Relationships
 The relationship between surface temperature and precipitation δ18O for the five idealized scenarios is compared with the temperature-isotope slope from a simulation of present day climate (CONTROL) using monthly SSTs from Reynolds  and the standard continental distribution in Figure 2. The control climate simulation produces a close to linear temperature-isotope slope for annual average temperatures below approximately 15°C in agreement with observations [Dansgaard, 1964]. As in observations, the modeled temperature-isotope relationship breaks down in the tropics and the isotopic ratio in precipitation is more strongly correlated with the local amount of precipitation than with surface temperature.
 The temperature-isotope slope for the AQUA experiment is also positive for surface temperatures below around 20°C, with the exception of the scatter over the polar regions of constant SSTs. In the tropics, the relationship is reversed due to the greater role of local precipitation amount in determining the extent of isotopic depletion. The slope for the L10 and L25 experiments is similar to AQUA, with a positive temperature-isotope slope at temperatures below approximately 20°C and a negative slope in the tropics for L10. In the L25 experiment, the larger scatter at high temperatures reflects the competing influence of increased rainout with inland (equatorward) transport and reduced rainout in the tropics due to lower precipitation amounts.
 In the L45 experiment the linear temperature-isotope relationship breaks down, reflecting the influence of midlatitude synoptic variability as well as the sensitivity of isotopic ratios to local humidity and precipitation conditions for very small precipitation amounts. This case may be interpreted as one in which isotopic distillation is not primarily controlled by large-scale precipitation or temperature, in contrast to the present day climate. The L60 experiment produces a complete reversal of the observed δ18O/temperature slope, with δ18O values decreasing towards the warmer tropics. This relationship cannot be due to the temperature-dependence of fractionation (which can only produce a positive δ18O/temperature slope), and therefore must result from the progressive isotopic distillation of moisture evaporated from the polar oceans.
4. Discussion and Conclusions
 This study has demonstrated that the assumption of a positive correlation between surface temperature and the isotopic ratio of precipitation is not valid under all conditions. The imposition of land covering the tropics forced net moisture transport to occur in the opposite direction to the present day climate, producing progressive moisture loss and isotopic depletion with equatorward transport to warmer latitudes. While precipitation amount was an important control over the isotopic ratio where the moisture source was local, the role of rainout during transport was the main cause of the negative temperature-isotope slope seen in the simulations with polar and midlatitude oceans.
 It was seen that the isotopic ratio of precipitation was not a simple function of temperature in any of the experiments. While the surface temperature decreased from equator to pole in all cases, the isotopic ratio did not follow this distribution uniformly in any simulation. Instead, isotopic depletion was determined by a combination of local precipitation, upstream moisture rainout and surface temperature. The temperature-isotope slope was positive only for those experiments with a local tropical and midlatitude moisture source (AQUA, L10 and L20). Where extratropical precipitation resulted from transport of polar moisture towards the equator, the temperature-isotope slope broke down (L45) or reversed to a negative slope (L60), indicating that local temperature was not the primary control on isotopic depletion in those cases.
 The positive temperature-isotope relationship in observations reflects the association of isotopic depletion with poleward moisture transport towards cooler latitudes in the present day global climate, in addition to the greater efficiency of isotopic fractionation at cooler temperatures. The direct application of present day empirical temperature-isotope relationships to past climate may therefore not be justified under conditions where changes in moisture transport produce an isotopic response independently of temperature variability. Such changes may occur, for example, due to altered continental distributions on geological time scales, or changes in the sea ice distribution around Greenland or Antarctica on glacial time scales. In such cases, it may be necessary to employ additional paleoclimate techniques to reconstruct the complete atmospheric circulation rather than assuming the validity of present day temperature-isotope regressions.
 David Noone and Richard Wardle are gratefully acknowledged for useful comments on the manuscript.