26 several orders of magnitude. A very noisy signal is produced by all models at P2 and this is attributed to the very rapidly changing water speeds and directions, as is described below. In the case of point P3 (see Figure 14), it is interesting to note that the arrival of the signal is similar for most models, i.e. approximately 40 days. Again, predictions expand over several orders of magnitude and di?erences between particle-tracking (i.e. KAERI, JAEA PT) and finite di?erence techniques (i.e. JAEA FDM, I/K-E) are clearly apparent in this figure. Finite di?erences introduce artificial (numerical) di?usion and thus, once the signal has arrived at the location of interest, a continuous line is obtained for the time series of concentrations. However, numerical di?usion is not introduced by particle-tracking methods. These models give a concentration above zero at a given location only if there is at least one particle there. Consequently, periods with zero concentration may alternate with periods during which some finite concentration is computed, as can be seen in the results of KAERI model (see Figure 14). As an example, maps showing the distribution of the tracer for surface water at the end of the simulation for some models are presented in Figure 15. The two simulations of JAEA show similar dispersion patterns since they are using the same water circulation. Nevertheless, the more di?usive character of Eulerian models is clearly shown. In the case of the I/K model the dominant transport is directed to the northeast, while the KAERI model produces a rather isotropic pattern around Fukushima. All models predict a contaminated band along the coast, although with di?erent values for the concentrations. A better agreement between the outputs of dispersion models is expected if the same water circulation is used by each model and this is carried out in Exercise 2. A previous step is to define which is the most appropriate hydrodynamic model to be used and a quantitative comparison between hydrodynamic models was carried out for this purpose and is described in Section 3.3.2. Essentially, calculated sea surface temperature fields have been compared with fields obtained from satellite observations, and differences between them have been evaluated.