Large quantities of radiocesium (12–62 PBq) were released into the environment by the Fukushima Dai-ichi Nuclear Power Plant (FDNPP) accident in March 2011 (Stohl et al., 2012; Chino et al., 2011). Accordingly, this accident hit the maximum level (i.e. 7) on the International Nuclear and Radiological Event Scale (INES) (IAEA, 2013a). Airborne and ground contamination surveys demonstrated that the contamination was the highest (i.e. initial 137Cs levels > 100 000 Bq m-2) in a plume extending to the northwest of the FDNPP covering an area of ∼3000 km2 (Kinoshita et al., 2011; Chartin et al., 2013; Yasunari et al., 2011). Although many radioactive substances were released into the environment by the FDNPP accident, radiocesium (i.e. 134Cs and 137Cs) presents the most serious risk to the local population over the medium to long term as it was emitted in very large quantities and it has a relatively long half-life (i.e. 134Cs – 2 years; 137Cs – 30 years) (Steinhauser et al., 2014). The highest contamination radiocesium levels observed in Fukushima (> 185 kBq m-2) are similar to those recorded in the area impacted by Chernobyl fallout, although on a much smaller land surface area in Japan (1700 km2) compared to eastern Europe (29 400 km2) (Steinhauser et al., 2014).

Numerous investigations have been conducted by Japanese and international researchers to improve our understanding of the fate of radiocesium in the Fukushima region (for a review, see Evrard et al., 2015). In general, radiocesium sorption mechanisms were characterised (Fan et al., 2014; Nakao et al., 2015) and their fluxes measured in riverine systems draining the main radioactive plume (Nagao et al., 2013). Land use (Koarashi et al., 2012) and soil properties (Nakao et al., 2014) were shown to control the migration of radiocesium in soils. Accordingly, the fate of this contaminant was intensively investigated in forest ecosystems (Gonze and Calmon, 2017) and cultivated landscapes (Yoshimura et al., 2016), which are the two main land uses in the fallout-impacted region. Typhoons and other major rainfall events were also demonstrated to drive soil erosion and sediment migration processes, thus directly influencing post-fallout radiocesium dynamics (Chartin et al., 2017).

Between 2011 and 2018, there were 578 published studies with the keywords “radiocesium” and “Fukushima” in the Scopus database (Fig. 1). Approximately 90 % of these articles were published by Japanese scientists, demonstrating the extensive research effort conducted by the national scientific community in Japan on the processes occurring in this post-accident context. Since the second half of 2013, remediation activities started to be implemented under the supervision of the Japanese authorities to decontaminate soils. These activities have significantly affected the spatial and temporal redistribution of radionuclides in the area impacted by fallout from the FDNPP accident. As decontamination is now completed in many regions and more than 50 scientific studies have been conducted on different aspects of these operations (Fig. 1), synthesising the results obtained by this applied research is important for the scientific community. Of note, this review will not synthesise non-peer-reviewed reports published by the Japanese authorities, although numerous resources are available on the official websites of multiple Japanese ministries (Table 1).

Although radiocesium is mainly transported in particle-bound form (i.e. through sorption and fixation to micaceous clay minerals) in the Fukushima fallout-impacted area (Konoplev et al., 2016), dissolved radiocesium was found in numerous environmental compartments, primarily during the immediate post-accidental phase (Yoshimura et al., 2014). As most of the dissolved radiocesium migrated through these landscapes immediately after the FDNPP accident, this literature review will focus on particulate radiocesium. Furthermore, as 134Cs and 137Cs were emitted in equivalent proportions into the environment in March 2011, with an initial 134Cs/137Cs activity ratio of ∼1 (Kobayashi et al., 2017), this review will focus primarily on 137Cs owing to its longer half-life and thus greater risk to the local population over the medium to long term. Although knowledge has been gained on radiocesium transfers (Ivanov et al., 1997) and ecosystem remediation (Santschi et al., 1990) after the Chernobyl accident, the circumstances in which Fukushima and Chernobyl accidents occurred are very different (Steinhauser et al., 2014). Moreover, the contrasting environmental conditions prevailing in the fallout-affected areas of Japan and eastern Europe (Konoplev et al., 2016) complicate the direct comparison of the fate of radionuclides and the effectiveness of potential remediation measures in both regions. Accordingly, the goal of this review is to examine the remediation strategies and their effectiveness for particulate bound 137Cs in Japan.

This literature review will be divided into five main sections. First, the spatial extent of the decontaminated zone and the schedule of these remediation activities will be outlined. Second, the remediation strategies in different environments (i.e. farmland, river, forests) will be presented along with a summary of their cost effectiveness. Third, the impacts of remediation activities on dosimetry will be summarised. Fourth, the initiatives to manage the large volume of waste generated by remediation will be discussed. Fifth, major research questions and requirements to guide the future management of Fukushima fallout-impacted areas will be identified and presented. The objective of this review is to provide a synthesis of the remediation lessons learnt in Japan following the FDNPP nuclear accident, which are fundamental in light of the potential for a similar situation to occur somewhere on Earth in the future (Christoudias et al., 2014).

The FDNPP suffered major damage following an earthquake and a tsunami that occurred on 11 March 2011. At this time, Units 1, 2 and 3 of the power plant were operational, and they suffered a series of major failures (Burns et al., 2012). The main resulting radionuclide emissions that affected Japanese landscapes occurred on 15 March 2011. Rainfall and snowfall that occurred on 15 and 16 March resulted in the formation of a radionuclide plume on soils located to the northwest of the power plant, up to ∼70 km from the site (Yasunari et al., 2011). The inhabitants living in areas coinciding with this plume were progressively evacuated in spring and summer in 2011, and it took time to find long-term housing solutions for these populations (Asanuma-Brice, 2012).

In November 2011, the Japanese government adopted the Act on Special Measures Concerning the Handling of Pollution by Radioactive Materials (Japanese Ministry of the Environment, 2011b) in order to reduce the impact of radioactive substances from the FDNPP accident on human health and the environment (Yasutaka and Naito, 2016). In support of this act, decontamination guidelines were released by the Japanese Ministry of Environment in December 2011 and updated in 2013. These guidelines outlined the methods for surveying and quantifying the levels of contamination and the way to prepare these areas targeted for remediation (Japanese Ministry of the Environment, 2013). A decontamination roadmap (Policy for Decontamination in the Special Decontamination Area) was implemented in January 2012 under the direct supervision of the Japanese government.

According to the decontamination roadmap, the remediation programme had to be implemented in “special areas” where targets were set for the exposure of the public to external dose rates in order for residents to return to their day-to-day lives (Yasutaka and Naito, 2016). Achieving pre-accident radiation levels is not the objective; rather the effectiveness of decontamination will ultimately depend upon the land use and the air dose of each particular area.

Two zones were delineated with different strategies for remediation (Fig. 2), based on different values of dose equivalent (i.e. the biological effect of ionising radiation) expressed in the Sievert (Sv) and its sub-units. First, Special Decontamination Zone (SDZ) are areas located within a 20 km radius of the FDNPP or areas where the cumulative dose 1 year after the accident was expected to exceed 20 mSv yr-1. The SDZ covers 11 municipalities (1117 km2) where residents were evacuated after the FDNPP accident in 2011. The central government of Japan is responsible for remediation works in the SDZ. Second, Intensive Contamination Survey Areas (ICAs) refer to 102 municipalities from 8 prefectures with ambient dose rates exceeding 0.23 µSv h-1 (equivalent to 2 mSv yr-1), designated as ICAs by the Ministry of Environment on 28 December 2011 (Mori et al., 2017). The area of the ICAs is 8 times greater than the SDZ (Yasutaka and Naito, 2016). In particular, the decontamination methods and target areas for remediation in the ICAs differ from those of the SDZ with decontamination activities for the ICAs conducted by local governments with support from the central government. In total, the SDZ and ICAs cover a surface area of 8953 km2 with a population that was not evacuated after the accident of ∼1.7 million (Yasutaka and Naito, 2016).

In the literature, there is debate regarding the need to initiate decontamination so quickly after the FDNPP accident (e.g. Yasutaka et al., 2013a). Delaying decontamination could allow for the natural decay of radioisotopes and thus significantly lower the costs of achieving radiation exposure targets. For example, Munro (2013) estimated that the optimal delay for implementing remediation activities was in the range of 3–10 years after the accident, with an optimal delay of 8.8 years.

The effectiveness of decontamination was assumed to strongly vary depending on the remediation method and the initial radiation dose rates prior to decontamination. Different remediation techniques were proposed depending on the land use and the zone (i.e. SDZ vs. ICA). Yasutaka et al. (2013a) and Yasutaka et al.  (2013b) compared the impact of four scenarios of decontamination, including two very unlikely options (i.e. minimal and maximal scenarios), in terms of effectiveness and cost according to the results of demonstration tests conducted by JAEA. These results were updated in a more recent publication (Yasutaka and Naito, 2016). Only the two scenarios following the guidelines provided by the Japanese government are assessed in this review. As such, the results of both the minimal and maximal remediation options are therefore not discussed. In the first scenario, 5 cm of topsoil was removed from 50 % of agricultural land in the SDZ, where 137Cs concentrations exceeded 5000 Bq kg-1 and replaced with a 5 cm layer of “clean” soil (measure A1; Table 2). In the remaining 50 % of agricultural land with 137Cs concentrations below 5000 Bq kg-1, topsoil was replaced with subsoil (measure A3). Ploughing with zeolite and potassium (measure A4) was adopted for agricultural land in ICAs where the annual additional effective dose exceeded 1 mSv yr-1. In the second scenario, measure A1 was applied to all cultivated land. Similar measures (e.g. Table 2) were included in both scenarios to decontaminate forested areas, roads and houses. The total decontamination cost for implementing these remediation measures varied between JPY 2 and 5.1 trillion (EUR ∼16–41 billion), with JPY 1.3–2 trillion (EUR ∼10–16 billion) for the SDZ and 0.7–3.1 (EUR ∼6–25 billion) for the ICAs. Although the area where decontamination has been implemented in the ICAs (922–3330 km2) covers a surface 3 to 11 times larger than that of SDZ (295 km2), the decontamination costs for the SDZ and ICAs are in the same order of magnitude.

The decontamination programme includes a variety of other activities on top of the actual on-site remediation works, including the transport of waste; the volume reduction of waste; and the temporary, the interim, and final storage of decontamination waste and removed soil in containers. Depending on the set of measures implemented in the field, the cost of the remediation works will therefore be highly variable. A synthesis of the unit costs for waste management and storage is provided in Table 3 (Yasutaka and Naito, 2016). As shown by Yasutaka et al. (2013b), the quantity of waste generated when decontaminating agricultural land varies considerably depending on the decontamination method used. Consequently, differences in the quantity of waste generated resulted in large differences between agricultural land decontamination methods and the costs associated with storage containers, temporary storage sites and interim storage facilities.

According to the latest available figures from the Japanese Ministry of the Environment (2019b) at the end of 2018, the volume of soil waste generated in the SDZ was 9 100 000 m3 with a remediation cost of approximatively JPY 1.5 trillion (EUR ∼12 billion). In the ICAs, the latest figures available for March 2018 showed that 7 900 000 m3 of waste soil were produced with a remediation cost of approximately JPY 1.4 trillion, equivalent to EUR ∼11 billion (Japanese Ministry of the Environment, 2019b). This amount corresponds to ∼8 % of the total annual expenses of the European Union (EUR ∼134 billion in 2017) (European Union, 2018).

Two parameters are assessed before authorising evacuees to return home: the prevalent dose rate and the cumulative dose. Importantly, background radiation levels need to be incorporated into this assessment. In the Fukushima Prefecture, background dose rates before the FDNPP accident were estimated to be ∼0.04–0.05 mSv h-1 (National Institute of Advanced Industrial Science and Technology, 2011).

Individual external radiation doses (mSv d-1) may not be directly related to outdoor air doses (mSv d-1) as people do not spend 24 h a day outside. When people are inside, the distances from the sources of radiation are greater and walls generate a shielding effect (IAEA, 2000). In Japan, when the Ministry of the Environment estimated daily individual external effective dose rates, it was assumed that people spent 8 h outdoors and 16 h indoors, with the indoor air dose rate being 40 % of the outdoor air dose rate (Japanese Ministry of the Environment, 2013). Based on these assumptions, the external radiation dose rate is 60 % of the air dose rate. Several researchers have estimated external conversion coefficients based on data provided by the Ministry of the Environment (Yasutaka and Naito, 2016).

Figure 5 compares the annual and individual dose rates that the global population may be exposed to in order to help facilitate a comparison with levels in the FDNPP fallout-impacted region. In Japan, a long-term dosimetric target of 1 mSv yr-1 was adopted by the Nuclear Emergency Response Headquarters. Accordingly, a guidance value of 0.23 µSv h-1 was proposed to achieve the target by implementing decontamination measures. In particular, areas with ambient dose rates exceeding this value were defined as ICAs. This guidance value is based on a simplified deterministic model assuming that inhabitants again spend 8 h outdoors and 16 h indoors (i.e. a shielding factor of 0.4) per day and that the contribution of natural radiation is 0.04 mSv h-1 (IAEA, 2013b). According to Mori et al. (2017), this model has three main challenges. First, the same behavioural pattern is assumed for the entire population. Second, the radiation exposure is assumed to be uniform. Third, conservative assumptions are adopted when converting the ambient dose into an effective dose. For instance, the time spent outside is assumed to be 8 h, which is more than anticipated for the majority of the population and likely results in an overestimation of the actual measured doses (Nomura et al., 2015).

Although this approach is effective for the immediate post-accidental context, more sophisticated approaches are required to estimate doses over the longer term. Therefore, a probabilistic method that accounts for spatial variations (i.e. houses, workplaces and other environments) in the contamination and for inter-populational variations (i.e. indoor workers, outdoor workers, pensioners) in behavioural patterns was developed by Mori et al. (2017). For this approach, the 95th percentile doses for outdoor workers were above 1 mSv yr-1 in 25 of the 59 municipalities in Fukushima Prefecture (1–35 mSv yr-1). In particular, the doses to more than 90 % of the outdoor workers in Okuma town, Futaba town, Tomioka town, Namie town and Iitate village were over 1 mSv yr-1. Furthermore, the 95th percentile doses for indoor workers were above 1 mSv yr-1 in Okuma town, Futaba town, Tomioka town, Namie town and Iitate village. If people return home in these municipalities, it is possible that they would be exposed to doses exceeding 1 mSv yr-1 for all population groups. However, the results indicate that the same behavioural patterns and contamination levels should not be assumed for all inhabitants nor all municipalities. Based on the different behaviour of the local population, the 95th percentile doses of indoor workers and pensioners in 53 of the 59 municipalities were below the dosimetric target of 1 mSv yr-1 (0.026–0.73 mSv yr-1) (Mori et al., 2017). Radiation dose rates were also measured among different types of workers taking part in professional activities in the village of Kawauchi, where the annual doses of foresters (range: 0.7–1.9 mSv yr-1) were not significantly higher than those of farmers (0.7–1.5), builders (0.6–1.5), office workers (0.5–1.5) and unemployed individuals (0.5–1.7). In contrast, decontamination workers (0.5–7.1) were found to have significantly higher dose rates (Orita et al., 2017).

The workers involved in decontamination activities were often directly exposed to internal irradiation through inhalation, which is much more difficult to measure than the external irradiation. Accordingly, 83 people who worked in highly contaminated areas where surface 137Cs deposition density was over 100 kBq m-2 were enrolled in a study (Tsubokura et al., 2013). Using a database on internal exposure from the Hirata Central Hospital in Fukushima Prefecture, data were compiled on age, gender, body weight, equipment used in decontamination activity, total working period, duration between the final working day and the day of an examination, and 134Cs and 137Cs body burden. Hirata Central Hospital was also equipped with a permanent whole-body counter with detection limits of 300 Bq per individual for both 134Cs and 137Cs measurements following a 2 min scan. The levels of internal radiocesium exposure among all the decontamination workers were below the detection limits. No other radionuclides besides natural 40K were detected. No acute health problems had been reported. However, levels of external exposure were not assessed, as individual data on dose rates were not available. This study suggests that the resuspension of radioactive materials may cause a minimal internal contamination during decontamination works (Yamaguchi et al., 2012). Other studies calculated that radiation doses from internal exposure were marginal (Hayano et al., 2013; Tsubokura et al., 2015). As such, remediation efforts should be concentrated on reducing the external exposure of the local population.

According to the decontamination scenarios described in Sect. 3, the reduction in annual individual additional effective dose (ED) for all decontamination scenarios was 1666 person-Sv for the SDZ and 876–1245 person-Sv for the ICAs (Yasutaka and Naito, 2016). Despite the higher reduction rate achieved in the SDZ compared to the ICAs, they remained at the same order of magnitude although the decontamination efficiencies were very different in both areas. This result may be directly attributed to the differences in population density in SDZ and ICAs, with 90 000 inhabitants living in the SDZ in 2010 versus approximately 1.5 million inhabitants living in ICAs exposed to over 1 mSv yr-1. This strong dependence of ED on population densities may lead the authorities to concentrate their remediation efforts in the most densely populated areas. The results obtained also depend on the effectiveness of these decontamination programmes. For instance in ICAs, where approximately 1 million inhabitants reside in areas exposed to 1–5 mSv yr-1, the reduction in annual individual additional ED was much larger in those areas where the full decontamination scenario would be implemented.

From the aforementioned research on river channel decontamination, the external radiation dose was calculated for paths along the river used by children to go to school and the nearby playgrounds used for outdoor activities incorporating an adapted time of exposition (35 h yr-1 for commuting and 24 h yr-1 for outdoor activities) (Nishikiori and Suzuki, 2017). After decontamination of the river channel, radiation dose rates decreased by a factor of approximately 2. These authors stated that the optimal strategy should be to reduce the annual individual additional ED as much as possible for the whole population, while also decreasing high-dose individuals (Yasutaka and Naito, 2016). Indeed, the authorities should not only assess the cost–benefit effectiveness of remediation programmes, they must also consider ethical and social costs (Oughton et al., 2004).

The total estimated decontamination cost would exceed 16 trillion JPY (EUR ∼128 billion) if all forested areas exposed to radiation dose rates exceeding 1 mSv yr-1 were decontaminated. However, decontaminating all of the forested areas would not result in a major ED reduction for the average inhabitant (Yasutaka and Naito, 2016). As almost 70 % of the surface area of Fukushima Prefecture is covered with forests (Hashimoto et al., 2012) and forestry is a significant economic activity in the region, future research should prioritise investigating radiocesium dynamics in these regions. In particular, the biological cycling of 137Cs in forests has now been affected by the decomposition of litter where radiocesium was concentrated shortly after the FDNPP accident (Koarashi et al., 2012). Furthermore, the local population in rural areas of the Fukushima Prefecture enjoy satoyama, or the collection of vegetation, including mushrooms, edible wild plants and firewood from forested landscapes (Prand-Stritzko and Steinhauser, 2018; Nihei, 2016). The collection and shipping of mushrooms remains prohibited in the main fallout-affected areas (Fukushima Prefecture, 2018). Furthermore, the low permissible levels for radiocesium contamination in wood (e.g. 40 Bq kg-1 in wood for cooking or eating) will restrict the use of this commodity for at least several decades in the region (Ohashi et al., 2017). In addition, approximately 1800 workers are employed by the forest industry in the region (Yasutaka and Naito, 2016). For many of the local inhabitants, the forest, the satoyama and its harvest are inseparable from their daily lives.

Forest sources were also shown to deliver a significant proportion of contaminated material to the river systems draining the fallout-impacted region. The analysis of deposited particulate matter collected in three fallout-contaminated coastal catchments between November 2012 and November 2014 demonstrated that forest sources supplied a mean of 17 % (standard deviation – SD – 10 %) of the sediment transiting these river systems (Laceby et al., 2016). Huon et al. (2018) obtained similar results through the analysis of sediment cores collected between November 2014 and April 2015 in a dam reservoir draining fallout-impacted cultivated and forested landscapes, with the latter supplying a mean of 27 % (SD 6 %) of the material deposited in the lake. These conclusions were validated through an analysis of a larger number of sediment samples (n=400) collected in coastal river systems in the Fukushima region over a longer time period (from November 2011 to November 2017), where a mean of 24 % (SD 21 %) of the material transiting these systems was modelled to be derived from forested landscapes (Evrard et al., 2019). Cumulatively, these results demonstrate that forested landscapes represent a potential long-term source of particulate contaminated matter that likely will require diligent management for the foreseeable future.

In cultivated landscapes where the remediation activities were concentrated, the main question is whether or not to restart agricultural production. The removal of the topsoil layer concentrating the radiocesium, the replacement of this material with crushed granite extracted from local quarries and the final mixing of the entire profile to prepare the soils for re-cultivation raises several important questions. For example, to what extent will the residual radiocesium in the soil be transferrable to the plants cultivated on these soils? How will the crushed granite, which was homogenised into the soil, affect the soil's fertility? Recent research showed that potassium fertilisation is required to maintain productivity when restarting cultivation after decontamination (Kurokawa et al., 2019). Indeed, as was demonstrated in the current literature review, the reopening of the region after the completion of remediation activities represents a unique situation in history, coupled with unprecedented challenges that require further ongoing investigations.

Although previous dosimetric studies demonstrated that currently the internal exposure of both the local population and the decontamination workers remains minimal, both internal and external exposures of these groups should be studied over longer temporal periods to help understand the long-term impacts of this accident on exposed population groups. More research is also required to understand the fate and dynamics of other longer-lived radionuclides in the Fukushima region including radiocarbon (Paterne et al., 2018; Povinec et al., 2016; Xu et al., 2016), plutonium and uranium isotopes (Jaegler et al., 2018; Zheng et al., 2013; Steinhauser, 2014) as they may be persistent in the environment even though many were emitted only at the trace and ultra-trace levels.

The quick and early decision of the Japanese authorities to decontaminate FDNPP fallout-impacted landscapes was unprecedented. Decontamination activities were rapidly implemented in agricultural and residential areas covering a surface area of ∼9000 km2. These remediation activities produced ∼20 million m3 of soil waste in less than 6 years (2013–2019) with an approximate cost of JPY 3 trillion (EUR ∼24 billion). The strategy of removing the surface layer of the soil concentrating 137Cs was shown to be effective in cultivated land when the strategy was applied at a catchment scale to avoid the supply of mobilised contamination from the headwaters. The main current challenges are associated with the treatment and the transport of this waste to the interim storage facilities for the next ∼30 years that are being built near the FDNPP. The re-cultivation of the soils after decontamination also raises several concerns. In particular, more information is required regarding soil fertility after decontamination and the potential transfer of the residual 137Cs to the plants cultivated on decontaminated fields.

The risks of internal and external radiation dose exposures of the decontamination workers and the local population to exceed the target of 1 mSv yr1 appeared to be low during the early post-accidental phase. However, dosimetric monitoring programmes should be carried out to confirm this result over the longer term, particularly after the local population returns to the region, as a risk of internal contamination remains if these inhabitants consume local food. Furthermore, as ∼75 % of the surface exposed to the highest 137Cs fallout levels in the Fukushima Prefecture are covered with forests where decontamination was not implemented, the potential long-lasting contribution of radiocesium to the river systems draining these mountainous, forested landscapes exposed to typhoons should be investigated. The behaviour and the dynamics of longer-lived radionuclides such as plutonium isotopes remains poorly documented and they should also be studied in the future as they may persist in the environment on long timescales even though they were emitted at trace and ultra-trace levels.