Résumé
Climate models exhibit a wide spread in their projections of regional precipitation changes in the tropics. Here we explore whether the precipitation isotopic composition (delta (super 18) O) in tropical archives (e.g. Andean and Tibetan ice cores) could be used to assess which precipitation change projection is the most credible. Two conditions need to be met for this to be possible: (1) tropical archives of precipitation delta (super 18) O robustly record past precipitation changes at the regional scale; (2) mechanisms at play for past regional precipitation changes are relevant for future precipitation projections. These two conditions are examined using simulations of 11 past and future climates (including last glacial, mid-holocene, last interglacial, doubling and quadrupling of greenhouse gases) with the isotopic general circulation model LMDZ. Sensitivity tests to the horizontal resolution (with high-resolution simulations down to 50km) and to the model physics are also analyzed. Simulations suggest that at paleo time scales, delta (super 18) O in Andean and Tibetan ice cores record the precipitation amount in upstream regions. As precipitation increases in the Amazon or in Northern India, delta (super 18) O in Andean or Tibetan ice cores decrease. This relationship stems from the fact that tropical deep convection depletes the low-level water vapor during their transport to the ice cores. This relationship is robust among all climates. However, it is sensitive to the model physics, and thus needs to be evaluated using present-day data. In addition, temperature has some impact in high altitude regions, especially when those are represented with a high resolution. Simulations suggest that at paleo time scales, the regional precipitation responds to changes in latitudinal gradients and patterns of sea surface temperature (SST). For example, the precipitation in the Amazon increases all the more as the Southern Atlantic warms more than the Northern Atlantic. This precipitation response to SST is robust among all climates, both past and future. However, this precipitation response is sensitive to the model physics, in particular to the extent to which cloud feedbacks amplify changes in the large-scale circulation. The sensitivity of the precipitation response to the model physics impacts the simulated ice core delta (super 18) O. Therefore, this study suggests that ice core delta (super 18) O could help evaluate the credibility of simulated regional precipitation changes in response to SST changes, and thus to assess the credibility of projected precipitation changes.