R. Peridfiez et al. 
Water circulation 
Physical transport 
(advectio/diffusion) 
Bio-geochemical 
processes 
Environmental Modelling and Software 122 (2019) 104523 
Hydrodynamic 
model 
Eulerian 
model 
Water currents 
Lagrangian 
model 
Water fluxes =: Box 
model 
Dynamic 
model 
Equilibrium 
model 
Suspended sediments 
and sedimentation rates 
Sediment | 
transport model: 
Observations 
Observations 
Radionuclide concentrations 
- Water 
- Sediment 
Biota 
Fig. 7, Scheme showing the models required to simulate the dispersion of radionuclides in the marine environment. 
from land in offshore dispersion events, The relevant role of winds in 
the shelf region was highlighted by means of sensitivity analysis, using 
as well an Eulerian dispersion model for !37Cs, carried out by Miyazawa 
st al. (2012). In this sense, sensitivity analysis showed that a tuning 
öf the wind drag coefficient was required for a better reproduction of 
'137Cs measurements (Bailly du Bois et al., 2014). 
In addition to pure advection-diffusion simulations, as those cited 
above, the transport of Fukushima radionuclides through drifter data 
and statistical methods was evaluated as well (Rypina et al., 2014). 
One of the key problems in the marine dispersion modelling is 
Ihe determination of radionuclide sources, There were several major 
sources of radionuclide contamination to the marine environment due 
(Oo the FDNPP accident: (i) atmospheric deposition of radionuclides 
onto the sea surface, (ii) direct release of radionuclides into the ocean; 
(iii) releases from land via river and coastal runoff; (iv) groundwater 
release, The first two sources dominated during the first year after the 
accident. However, later ongoing groundwater and river releases were 
ı0cally important, A feasible method for determining the source term 
is to combine radionuclide measurement data and advection--diffusion 
models (“inverse modelling”). A number of atmospheric transport mod- 
als using different tracer inversion algorithms were used to estimate 
deposition onto the ocean surface (see review in SCJ (2014)). Scenarios 
of direct release in the ocean were constructed using monitoring data 
‚n the vicinity of FDNPP to scale computations (Kawamura et al., 2011; 
Fsumune et al., 2012, 2013). The estimated total direct releases of 1?7Cs 
were 4 PBq (Kawamura et al., 2011) and 3.5 PBq (Tsumune et al,, 
2012, 2013). These scenarios were used in several subsequent studies 
(e.g. Tsumune et al., 2013; Kawamura et al., 2014; Tsubono et al., 
2016; Maderich et al., 2014a,b; Bezhenar et al., 2016). A total direct 
release of 5.1-5.5 PBq, using a four-step inverse approach based on 
‘he measured!?7Cs activity south and north outlet channels of FDNPP, 
was also estimated (Estournel et al., 2012). The 27 PBq direct release 
estimate by Bailly du Bois et al. (2012) was based on interpolated 
monitoring data in a 50-km area around FDNPP and the environmen- 
cal half-time for it, which was deduced from observations. However, 
;his source term was considered to be significantly overestimated by 
Dietze and Kriest (2012). Inverse estimation of direct releases based 
on the Green function approach (Enting, 2002) was also carried out 
by Miyazawa et al. (2013). An inversion method based on minimizing 
the differences between model and eruise data was applied by Rypina 
et al. (2013) to estimate releases. Corresponding total direct release 
was 16.2 PBq. The total release of !37Cs from FDNPP harbour was 
sstimated by Kanda (2013) as 2.25 PBq. This value was comparable 
with estimates of Kawamura et al. (2014) and Tsumune et al. (2012). A 
3.6 TBq y“! continuous underground leak of contaminated water from 
FDNPP was also suggested by Kanda (2013). This value was confirmed 
by comparison of modelling results and measurements within an area 
with 15 km radius around FDNPP in the period 2012-2015 (Maderich 
et al, 2014a,b; Bezhenar et al., 2016). According to Kanda (2013), total 
river flux of!?7Cs in Fukushima, Ibaraki and Miyagi prefectures in 2012 
was 1.56 TBq y-}. 
All modelling studies mentioned above (which does not try to be 
an exhaustive list) had the common feature that !?7Cs was treated 
as a conservative radionuclide which did not interact with sediments. 
The first models including !97Cs contamination of bed sediments were 
described by Periäßez et al. (2012) and Min et al. (2013). In the first 
zase a local study was carried out, covering only the coastal region of 
Japan. A larger domain was considered in the second paper. In both 
cases, calculated and measured 197Cs concentrations in bed sediments 
were compared, Also, water-sediment interactions were described in 
a dynamic way in both studies. Adsorption by bottom sediments was 
considered by other authors as well (Choi et al., 2013; Misumi et al., 
2014; Higashi et al., 2015). All these papers agree on the fact that 
significant adsorption occurs in the first months after the accident, 
nost of radionuclides staying on the sea bed once they were adsorbed, 
which may be indicative of a two-step kinetics. Later, a box model 
(POSEIDON-R) was used to perform a radiological assessment of the 
accident in the period 2011-2040 (Maderich et al., 2014a). This box 
model included not only adsorption to sediments, but also the transfer 
of radionuclides through the marine food web and subsequent doses to 
humans, The benthic food chain was included in this model (Bezhenar 
et al., 2016). The simulation results indicated a substantial contribution 
of the benthic food chain in the long-term transfer of !37Cs from 
contaminated bottom sediments to marine organisms. 137Cs levels in 
coastal biota in the area near FDNPP were reconstructed by Tateda 
et al. (2013) using a circulation model (Tsumune et al., 2012) to