38 In conclusion, it can be stated that when the dispersion of a tracer is simulated, a significant part of the variability between models is due to water circulation and that model agreement increases if the same circulation data is applied by all models. In the case of 137Cs, when water–sediment interactions are included, there are significant differences between models, with calculated concentrations in sediment expanding over several orders of magnitude. Nevertheless, agreement between models for water is similar to the tracer case. Each model used its own description for water–sediment interactions, as well as its own set of parameters for describing such processes. The next step is to homogenize the description of water–sediment interactions, using equivalent parameters in all models. This is the main objective of Exercise 3 which is described in the following section. Moreover, some additional harmonization between models was carried out which consisted of using the same topographic data and horizontal and vertical di?usion coe?cients. FIG. 19. Time series of radionuclide concentrations for surface water at points P1, P2 and P3 for Exercise 2, tracer dispersion. I/K-L denotes the IMMSP/KIOST model running in a Lagrangian framework.