Abstract. The simulation of aggregation of ice particles is critical for precipitation prediction, but still a major challenge. Its simulation requires assumptions about numerous parameters, many of which are either not well known or difficult to represent accurately in bulk microphysics schemes. However, knowing the sensitivity of aggregation to various simplified assumptions can help to identify critical parameters. By comparison with suitable observations, these critical parameters can even be constrained. We investigate the sensitivity of the model variables, and the modeled multi-frequency and Doppler radar observables to different parameters in a two-moment microphysics scheme. Therefore, we revise hydrometeor parameters by using a recently published dataset of particle properties, modify the formulations of the aggregation process (which allows using an area-based differential sedimentation kernel) and update other ice microphysical parameters in the scheme such as the sticking efficiency Estick and the shape of the size distribution. Overall, particle properties, definition of the aggregation kernel, and size distribution width prove to be most important, while Estick and the cloud ice habit have less influence. Finally, we run multi-week simulations with the most promising parameter combinations. The statistical comparison between real and synthetic observables shows a reduction in the velocity and snow particle size. With this study, we show a possible way to revise processes in microphysical schemes by using statistics of detailed cloud radar observations.