The importance of explicitly modelling sea-swell waves for runup was examined using a 2D XBeach short wave-averaged (surfbeat, “XB-SB”) and a wave-resolving (non-hydrostatic, “XB-NH”) model of Roi-Namur Island on Kwajalein Atoll in the Republic of Marshall Islands. Field observations on water levels, wave heights, and wave runup were used to drive and evaluate both models, which were subsequently used to determine the effect of sea-level rise and extreme wave conditions on wave runup and its components. Results show that specifically modelling the sea-swell component (using XB-NH) provides a better approximation of the observed runup than XB-SB (which only models the time-variation of the sea-swell wave height), despite good model performance of both models on reef flat water levels and wave heights. XB-SB has a bias of −0.108 – 0.057 m and scatter index of 0.083–0.639, whereas XB-NH has bias of −0.132 – 0.055 m and 0.122–0.490, respectively. However, both models under-predict runup peaks. The difference between XB-SB and XB-NH increases for more extreme wave events and higher sea levels, as XB-NH resolves individual waves and therefore captures SS-wave motions in runup. However, for even larger forcing conditions with offshore wave heights of 6 m, the island is flooded in both XB-SB and XB-NH computations, regardless the sea-swell wave energy contribution. In such cases XB-SB would be adequate to model flooding depths and extents on the island while requiring 4–5 times less computational effort.