It is investigated how the quality of the temporal interpolation of lateral-boundary coupling data for limited-area models (LAMs) can be improved or kept under control, while increasing the data transfer between the coupling and the coupled model onldy marginally. This problem is approached from the point of view of a user of a LAM who depends on coupling data that is received from elsewhere. To this end, the interpolation is corrected using the higher-order time derivatives of the fields. It is shown that in practical situations, these time derivatives can be estimated by a one-time-step integration, well enough to be useful. This allows the user of a LAM to formulate a criterion to decide operationally in which situations the quality of the linear interpolation will be unacceptable. The idea has been implemented in a numerical weather prediction limited-area model ALADIN, and has led to a substantial reduction of the errors of a forecast of one of the French Christmas storms in December 1999.