Three sedimentary sequences from a coastal wetland behind a sand barrier, on the west coast of d'Urville Island, New Zealand, were examined using a multi-proxy approach, including sedimentological, geochemical and microfossil (diatom) analyses, with the chronology established using radiocarbon and 210Pb dating. Data show that a brackish lagoon started developing 7300 year ago after formation of the sand barrier in Moawhitu. This was followed by periods of alternating wetland encroachment and open water, displaying spatial variations, until a peatland was established about 1300 cal. yr BP. The wetland was then partially drained in the early 1900s, leading to compaction of the peat near the surface. A coarse layer containing gravel, sand and shells, with a sharp lower contact, in the northern area of the wetland, is attributed to an overwash about 2500–3000 yr BP, most probably a tsunami generated by the rupture of a local or regional fault. The high-resolution continuous record obtained with XRF core scanning revealed a geochemical signature (Ca and S) for the overwash 600 m inland in the middle area of the wetland, while it was absent from the southern site 1.1 km from the shore. This suggests that, except at the northern end of Moawhitu, the sand dune acted as an effective barrier preventing any sediment from overtopping 2500–3000 yr BP, with only a geochemical evidence marking the extent of seawater inundation. No sedimentological or geochemical evidence could be found in the wetland for the 15th century tsunami that had destroyed almost an entire community in Moawhitu, as recounted in Māori oral tradition (pūrākau), although pebbles at the surface of the dune are likely to be linked to this event. Thus, the sand dune appears to have again acted as an effective barrier for the overwash in the 15th century. However, our study suggests that the area might have been impacted by more than one tsunami in the last 3000 years. It also shows that a high-resolution continuous XRF record can provide the evidence for short-term changes (catastrophic or not) that did not leave any clear sedimentological signature, thus providing a better insight of environmental changes in any depositional environment.