37 3.3.4. Tracer dispersion Results for the tracer dispersion exercise are presented in Figure 19 where it can be observed that the agreement between models has been significantly improved. The shapes of the signals are much more similar than in Exercise 1. Results are within one order of magnitude with the exception of both the Sisbahia and NTUA models. The NTUA model is highly dispersive, since radionuclides reach point P3 essentially instantaneously after release, as can be observed in the lowest panel of Figure 19. Although there are times and locations where some significant di?erences between models still remain (see for example point P1, some 30 days after 11 March 2011), most of the variability has been removed by use of the same water circulation when the transport of a tracer is simulated. 3.3.5. 137Cs dispersion The modelling of 137Cs dispersion, which includes interactions with sediment, was also carried out for Exercise 2. To simplify the problem, it was assumed that bed sediments are uniform over all of the model domain and are composed entirely of fine material (clays) with a mean size of 10 µm. A uniform porosity of 0.6 was assumed and, finally, the thickness of the bed sediment which interacts with water was set to 10 cm. With this homogenization it is assured that di?erences between model outputs are due to intrinsic factors of the models, but not to input data. Thus, hypothetical, but realistic, values for some parameters may be used. Time series of 137Cs concentrations for surface water, bottom water (deepest water layer, in contact with the seabed) are provided at points P1 to P3, as described above. The results of these experiments are summarized in Figures 20–22. In the case of surface water, results from all models are similar, as per the tracer exercise. The reason for this is that surface water does not feel the presence of the bed sediment, especially when water depth increases. Exceptions are again the Sisbahia and NTUA models. Signal arrival at point P3 produced by the KAERI and JAEA models is in very good agreement, and the NTUA model produced very high concentrations too quickly. In the case of bottom water, very low concentrations were calculated by all models at point P1 (see Figure 20). At P2, which is close to the Fukushima release point, higher concentrations were calculated in the bottom water, with the signal being similar for most models. The majority of the variability between models now occurs for activity concentrations in sediment. At point P2 (see Figure 21), for instance, results vary over several orders of magnitude. In general, the JAEA model tends to produce lower concentrations in sediments than the other models. The NTUA model produces significant concentrations in bottom water and sediment at point P3, while zero concentrations are calculated by the other models. Maps showing the computed distribution of 137Cs in surface water and sediment are presented in Figures 23 and 24, respectively, for the JAEA and Sisbahia model as examples. The di?erence in scale of model domains makes a direct comparison di?cult but, in general, it can be seen that the behaviour of the radionuclide patch is very similar, even in the case of sediment. In this case, there is an extension of contaminated sediment to the south of Fukushima and also along the shore of the Bay of Sendai to the north. These radionuclides appear as discrete spots in the case of the JAEA model, due to its Lagrangian nature, while a continuous patch is produced by the Sisbahia model.