Signatures of submarine reefs near Heligoland in the North Sea were observed in airborne radar images recorded at L-, C- and X-bands on 14 November 1990 during rather high wind speed of 9 ms-1. Predictions from various models of the imaging mechanism were compared to these observations. One of the models is the so-called weak hydrodynamic interaction theory (WHIT) model. It is fully two-dimensional in position as well as wavenumber space, so any surface current variation can be handled. Also more sophisticated scattering models than first-order Bragg scattering can be included. The model contains a number of parameterizations for the roughness length, the equilibrium wave height spectrum and the relaxation rate as well as different forms for the local relaxation source term. In the model intercomparison, the WHIT model performed not very well. It is shown here that this is due to the choice of the radial relaxation rate. In a sensitivity analysis it is shown that also the form of the relaxation source term is important. A linear source term may lead to unrealistically high positive hydrodynamic modulations (up to 50 dB) at some positions over the reefs for waves with a wavelength of about 0.6 m. Such effects do not occur in quadratic or cubic source terms, which are therefore to be preferred. The parameterizations chosen for the roughness length and the angular relaxation rate have little influence on the model results. Also shoaling may be neglected. A scattering model based on first iteration of the Stratton-Chu equation gives results similar to that of an improved two-scale model. When compared to the observations, good agreement is obtained at L-band, but at C- and X-bands the model underestimates the modulations. A number of possible causes is discussed, but additional data are needed to settle this question.