I suggest to represent the river with a 3rd kind Boundary Condition (Cauchy-BC). A Cauchy-BC applies a pre-defined reference head combined with a conductance parameter (transfer-rate).
Then, you could impose several time-series of the water levels as reference heads. Usually, measured time-series are available from local gauges. To regionalize water levels along the river you could use a 1D interpolation. The 1D linear interpolation is a method for interpolating linearly along line features provided by a map.
Floodwaves are damped by natural or artificial retention. To mimic the impact of floodwaves and retention on the groundwater you could take a time-dependent interpolation into account in addition to the spatial 1D-interpolation along the river.
However, this approximation is pure mathematical interpolation. If you are interested to represent the “real” physics of retention (including hysteresis) I suggest to couple FEFLOW with another numerical simulator, which is capable to solve the shallow water equations (St-Venant equation). In this context, an interface between FEFLOW and MIKE11 exists.