The evaluation of the trajectories of plumes rising into a crossflow is relevant for the correct modelling of pollutant dispersal in the atmosphere. Plume rise models are therefore an important part of several dispersion models. There is a general consensus on the formulation of plume rise through solution of equations describing the conservation of energy and momentum in the plume closed with an empirical formulation of the entrainment rate of fresh air into the developing plume. However, there are differences in the entrainment coefficients found in different studies. In this work the trajectories of sixty plumes, simulated at small scale in a towing water tank, have been analysed in order to test the performance of widely used dispersion codes and to find statistically the best entrainment coefficients for the different models. The plumes simulated cover a wide range of scenarios from pure jets to buoyant plumes developing in both neutral and linearly stable stratified crossflows. A new analytical model for stable crossflows, representing an extension of an existing model, has been presented and tested. Results show that the entrainment coefficients are different for neutral and stable crossflows, especially for approximate analytical codes. In contrast, the coefficients of the integral model seem to be less sensitive to the stability of the crossflow. The entrainment coefficients found by the fitting of the analytical models are significantly lower than the measured spread rate of the plumes. In neutral crossflows the generalised Briggs model and the integral model give statistically similar performances. In stable crossflows the new analytical model as well as the integral model are able to predict the oscillation of the plumes around their equilibrium height; however, there is an underestimation of both the oscillation frequency and the downwind position of the maximum height. The use of an added mass coefficient allows, with almost the same entrainment coefficients, improved prediction of the oscillation frequency and of the maximum rise position. Measured plume height oscillations are more strongly damped than predicted ones.