Variations in the isotopic composition of precipitation are determined by fractionation processes which occur during temperature- and humidity-dependent phase changes associated with evaporation and condensation. Oxygen stable isotope ratios have therefore been frequently used as a source of palaeoclimate data from a variety of proxy archives, which integrate this signal over time. Applications from ombrotrophic peatlands, where the source water used in cellulose synthesis is derived solely from precipitation, have been mostly limited to Northern Hemisphere Sphagnum-dominated bogs, with few in the Southern Hemisphere or in peatlands dominated by vascular plants. New Zealand (NZ) provides an ideal location to undertake empirical research into oxygen isotope fractionation in vascular peatlands because single taxon analysis can be easily carried out, in particular using the preserved root matrix of the restionaceous wire rush (Empodisma spp.) that forms deep Holocene peat deposits throughout the country. Furthermore, large gradients are observed in the mean isotopic composition of precipitation across NZ, caused primarily by the relative influence of different climate modes. Here, we test whether δ18O of Empodisma α-cellulose from ombrotrophic restiad peatlands in NZ can provide a methodology for developing palaeoclimate records of past precipitation δ18O. Surface plant, water and precipitation samples were taken over spatial (six sites spanning >10◦ latitude) and temporal (monthly measurements over one year) gradients. A link between the isotopic composition of root-associated water, the most likely source water for plant growth, and precipitation in both datasets was found. Back-trajectory modelling of precipitation moisture source for rain days prior to sampling showed clear seasonality in the temporal data that was reflected in root-associated water. The link between source water and plant cellulose was less clear, although mechanistic modelling predicted mean cellulose values within published error margins for both datasets. Improved physiological understanding and modelling of δ18O in restiad peatlands should enable use of this approach as a new source of palaeoclimate data to reconstruct changes in past atmospheric circulation.