Different imputation methods to fill existing gaps in wave data propagated from offshore buoys are compared in terms of their effect on modelled 9 yr morphodynamic evolution. To do so, the Q2Dmorfo model is applied to the Llobregat delta site (south of Barcelona, Spain) where an accurate reference set of wave data is available at the model boundary. The first strategy relies on available wave hindcasts to build transfer functions based on correlations between hindcast and reference datasets. The derived transfer functions are used to fill several synthetic gaps (artificially generated), and the model results are compared with those obtained with the reference data. Despite the effort towards reducing the hindcast biases, the modelled final bathymetries still have significant errors in bed level. The second strategy involves a purely statistical method (predictive mean matching) that has the virtue of conserving the probability distributions of the reference dataset and the correlation among variables. This seems to be a key factor since the bathymetries obtained are more accurate than those of the hindcast-based methods, below decimetric scale.