Résumé
The oxygen isotope signature of leaf water is used to trace several processes at the soil-plant-atmosphere interface. During photosynthesis, it is transferred to the oxygen isotope signature of atmospheric CO 2 and O 2 , which can be used for reconstructing past changes in gross primary production. The oxygen isotope signature of leaf water additionally imprints leaf organic and mineral compounds, such as phytoliths, used as paleoclimate and paleovegetation proxies when extracted from sedimentary materials.Numerous experimental and modelling studies were dedicated to constrain the main parameters responsible for changes in the 18O of leaf water. Although these models usually correctly depict the main trends of 18O-enrichment of the leaf water when relative humidity decreases, the calculated absolute values often depart from the observedones by several ‰. Moreover, the 18O of leaf water absorbed by plants is dependent on the 18O value of meteoric and soil waters that can vary by several ‰ at different space and time scales. These added uncertainties make our knowledge of the parameters responsible for changes in the 18O of leaf water and phytoliths flawed.Changes in the triple oxygen isotope composition of leaf water, expressed by the 17O-excess, are controlled by fewer variables than changes in 18O. In meteoric water the 17O-excess varies slightly as it is weakly affected by temperature or phase changes during air mass transport. This makes the soil water fed by meteoric water and theatmospheric vapour in equilibrium with meteoric water changing little from a place to another. Hence the 17O-excess of leaf water is essentially controlled by the evaporative fractionation. The latest depends on the ratio of vapor pressure in the air to vapor pressure in the stomata intercellular space, close to relative humidity. Leaf waterevaporative fractionation can lead to 17O-excess negative values that can exceed most of surficial water ones. Here we present the outcomes of several recent growth chamber and field studies, for the purpose of i) refining the grass leaf water and phytoliths 18O and 17O-excess modelling, ii) assessing whether the 18O and 17O-excess of grass leaf water can be reconstructed from phytoliths, and iii) examining the precision of the 17O-excess of phytoliths as a new proxy for past changes in continental atmospheric relative humidity. Atmospheric continental relative humidity is an important climate parameter poorly constrained in global climate models. A model-data comparison approach, applicable beyond the instrumental period, is essential to progress on this issue. However, there is currently a lack of proxies allowing quantitative reconstruction of past continental relative humidity. The 17O-excess signature of phytoliths could fill this gap