Towed radiation survey system
Two NaI (Tl) scintillation detectors were used for the towed radiation survey system: a small version (dimensions: 3″f × 3″H) and a large type (dimensions: 3″f × 6″H). Fig. 5 shows an overview and a photo of the towed radiation survey system. This system is 8 m long and has an external diameter of 0.145 m. The towed radiation survey system has enough underwater weight (115 kg) to maintain contact with the seafloor at an operational speed of 2 knots14. The spectrometer was calibrated to measure the gamma-ray spectrum between 0.8 and 1.8 MeV over 1024 channels and has a resolution of 7.0% at 0.662 MeV. The radiation detector is covered with a rubber hose to reduce its risk of snagging and protect it from abrasion and impact damage as it follows the undulations of the seafloor.
Data acquisition
The gamma-ray spectrum was acquired every second, and the temperature and depth were recorded by a depth sensor at the same time. The readings were synchronized with time via the global positioning system (GPS) receiver installed on the ship. The spectrum and the GPS data (date, time, latitude, longitude, and height above ellipsoid) were recorded every second. The actual detector position was estimated from the water depth, length of supplied wire, and GPS position data from the ship. The survey lines for each campaign are shown in Fig. 6. The intervals between the survey lines was set as approximately 1–2.5 km in consideration of the location of the river mouth and the seafloor characteristics, such as silt and sand. These monitoring data were obtained by the NRA and the National Maritime Research Institute as part of the national radiation survey project18.
For data cleansing, two criteria were set on the basis of the detector count rate. One criterion was less than the background count rate, which was obtained at 10 m locations from the sea surface at 20 m water depth. The reason of these data was that the detector was away from the seafloor. The other criterion was the lack of a radiocesium region-of-interest (ROI) peak in the obtained gamma spectrum caused by a high count rate. At the time of data acquisition, these data were deemed to indicate the pile-up loss of signal treatment or effects of detector noise. These data were excluded from the analysis.
Gamma spectrum analysis
In this study, the gamma spectrum at the seafloor was analyzed according to the method suggested in an International Atomic Energy Agency report19. A typical gamma spectrum collected at the seafloor is shown in Fig. 7. The counting peak of radiocesium was found in this gamma spectrum. The following sources of background radiation influence radiocesium counting: (1) the natural nuclides in the detector crystal and the structure of the detector system and (2) the natural nuclides in the bottom sediment.
Regarding the first disturbance factor, the background counting rate (CBG) was obtained at 10 m locations from the sea surface at 20 m water depth. An example of the background spectrum is shown in Fig. 7. For the second disturbance factor, the gamma spectrum was obtained from an area with low-level radiocesium contamination 300 km from the FDNPP. The maximum 137Cs concentration in the sediment at this location was 0.34 Bq kg−1. This value was negligible compared with the approximately 100–300 Bq kg−1 of 40 K concentration in this region. Spectrum data of 600 s measurement time (which was set to reduce counting errors) were obtained at ten points around this area. Detailed information about this natural background data is shown in Fig. S2. As indicated in Fig. 7, we set more than 1300 keV as the ROI for the natural background. With use of this natural background spectrum (which is deducted the CBG from the natural background spectrum), the index of the natural background (IBG) was set to the ratio of the total counting rate (Call’), and more than 1300 keV was set as the ROI for the natural background (Cnat’) of this natural background spectrum (IBG = Call’/ Cnat’). The counting rate of radiocesium was calculated as
This following typical equation is applicable to calculation of the DL:
where nb is defined as (CBG + Cnat IBG), k is the standard deviation of the Gauss distribution (k = 3), ts is the measurement time, and tb is the background measurement time.
Sediment sample
For calibration of the small detector used in the towed radiation survey, 39 sediment core samples (length 10–40 cm) were collected from the area of the towed radiation survey in 2014, as shown in Fig. 8. The sampling location was selected among the survey lines for the towed radiation survey. The samples were collected with an HR-type core sampler (Rigo Co. Ltd., Saitama, Japan), sectioned into 5 cm layers, dried, and measured for radiocesium activity concentration (on the wet basis: Bq kg−1wet) via gamma spectrometry using a p-type high-purity germanium detector (Seiko EG&G Co. Ltd., Tokyo, Japan) that was calibrated by standard gamma sources. For calibration of the large detector used in the towed radiation survey, 150 sediment grab samples (weight 0.5–1 kg) were collected from the area of the towed radiation survey in 2020, as shown in Fig. 8b. The sampling location was selected around an intersection of the survey lines for the towed radiation survey. The samples were collected with an Ekman-Berge bottom sampler (Rigo Co. Ltd., Saitama, Japan), portioned, dried, and measured for radiocesium activity concentration (on the wet basis: Bq kg−1wet) via gamma spectrometry using a p-type high-purity germanium detector. For validation of the towed radiation survey, 21–59 sediment core samples (length 30–40 cm) were collected from the area of the towed radiation survey between 2014 and 2019 (Table 1). These samples were treated and measured in the same manner as were the calibration samples.
Calibration
For conversion from counting rate to 137Cs concentration, two types of towed radiation survey systems were calibrated using field data obtained off the coast of Fukushima. In 2014, a towed radiation survey with a small detector was conducted within the survey lines, as shown in Fig. 8a. The radiocesium measurement result from the sediment surface (0 cm) to 10 cm of the sediment core sample, which was collected from the survey lines, was averaged for comparison with CCs, which was averaged in 100 m at the center of the sampling location. A scatter diagram of the radiocesium measurement result from the sediment surface to 10 cm of the sediment core sample and CCs is shown in Fig. 8b. In this figure, the actual sample measurement and CCs correlate well from 30 Bq kg−1 to 3500 Bq kg−1. This approximate straight line inclination was defined as a conversion factor (CF: 0.877 cps Bq−1 kgwet). In 2020, a towed radiation survey with a large detector was conducted within the survey lines, as shown in Fig. 8c. The radiocesium measurement result from the sediment surface (0 cm) to 10 cm of the grab sample collected at an intersection of the survey lines was prepared for comparison with CCs. We selected a grab sample instead of a core sample because many samples are needed due to the low radiocesium concentration in the sediment. The CCs and radiocesium concentration of the actual sample in the selected grid (marked by red line in Fig. 8c) were averaged and compared. A scatter diagram of the radiocesium measurement result in the sediment surface of the grab sample and CCs is shown in Fig. 8d. In this figure, the actual sample measurement and CCs correlate well from 30 Bq kg−1 to 750 Bq kg−1. This approximate straight line inclination was defined as a conversion factor (CF: 1.01 cps Bq−1 kgwet).
In addition, the 40K calibration of the detector was conducted by substituting Cnat for CCs in the above calibration methods. A scatter diagram of the Cnat of the large detector towed radiation survey is shown in Fig. S3. In this comparison, the 40K measurement data of the grab sample and averaged Cnat (red line in Fig. 8c) were compared. The 40K distribution map was created using this parameter, as shown in Fig. S3.
The concentration of 137Cs was calculated and corrected with the physical half-life of radiocesium (134Cs: 2.0652 years, 137Cs: 30.167 years) and the ratio of both cesium (134Cs/137Cs) on the accident date (March 15, 2011). The detector energy responses of 134Cs and 137Cs were assumed to be equal. With use of these parameters and Eq. 2, the DLs of the small and large detectors were determined to be 8.9 and 5.0 Bq kg−1, respectively.
The radiocesium in the sediment more than 10 km off the coast of Fukushima is spread approximately 10 cm from the sediment surface, as indicated by considerable field data6, 20, 21. The radiocesium depth profile in nearshore and estuary areas is complex; a radiocesium peak has been observed at a depth of over 50 cm7. Towed radiation survey cannot evaluate all types of radiocesium depth profiles. In this study, the averaged radiocesium concentration from the sediment surface to 10 cm depth was the target of assessment.
Mesh averaging and interpolation
Mapping was performed by supplementing unmeasured areas via two methods. The first was a simple technique for the mean in a 1 km × 1 km mesh. The gamma spectrum data in the mesh were summed, and the counting rate of each ROI (Cnat, CCs, and Call) was calculated. Afterward, these ROIs were converted to 137Cs concentrations as averaged data in the mesh. This map was created using the software ArcGIS (Environmental Systems Research Institute Inc., California, USA).
The second method was interpolation of the survey results. Various methods have been proposed for interpolation, such as kriging and spline approaches. In this study, the inverse distance weighting (IDW) method, wherein weights are assigned to the values of measurement points linearly and in inverse proportion to distance, was applied to the survey data. The IDW method is easy to use in analyzing large amounts of data because of the simplicity of its parameter setting19. This interpolation processing was conducted using ArcGIS. The spatial resolution of the resulting contour map was 250 m × 250 m, and the maximum distance of the search radius was 360 m.
Offshore soil distribution of Fukushima
Tsuruta et al. (2017) created a geological map of the Ukedo River estuary, which is located 7 km north of the FDNPP7. In the national project covered by the present study, a bathymetric survey of approximately 500 km2 at Fukushimaoff-shore was conducted using the technique of Tsuruta et al. (2017)7. A bathymetric survey using sonic prospecting was performed to visualize the characteristics of the seabed sediment from 2013 to 2020. PDR-1300(W) (SenbonDenki Numazu, Shizuoka, Japan; depth range of up to 2 m) and Sonic 2024 (R2SONIC, Texas, USA; sectors deeper than 2 m) echo sounders were selected for the bathymetric survey. A bathymetric map with a resolution of 2–3 m was produced using the commercial software Marine Discovery (Ocean High Technology Institute Inc., Tokyo, Japan). Sonic prospecting was conducted using a System 3000 (L-3 Klein, New Hampshire, USA; depth range of up to 3 m) and a 2000-DSS (EdgeTech, Massachusetts, USA; sectors deeper than 3 m), and it involved a full-coverage side-scan sonar survey and a sub-bottom profile survey. The sonic prospecting data were processed using the software CleanSweep (Oceanic Imaging Consultants, Hawaii, USA). Side-scan sonar mosaics with a 0.5 m resolution were obtained. The backscatter intensity value of these mosaics can be used to semi-qualitatively identify the seafloor type (e.g., rock or sediment). Seabed sediment sampling was performed at 18 locations by using a Smith-McIntyre grab sampler (Rigo Co. Ltd., Saitama, Japan), and the sediment types, namely, silt (including clay), sand (fine, medium, and coarse), and granule, were confirmed via visual observation. The backscatter intensity values were compared with the results of seabed sediment sampling and the geological map, and the seafloor types in the investigation area were predicted. The areas with high backscatter intensities correspond to bedrock. The remaining areas, which were considered to be the distribution areas of the seabed sediment, were qualitatively classified into three grain-size regions on the basis of the backscatter intensity features: coarse sand–granule, fine–medium sand, and silt. The major factors used for this classification are as follows: coarse sand–granule (high and spotted backscatter intensity), fine–medium sand (medium–low and pervasive homogeneous backscatter intensity), and silt (including clay; very low and pervasive backscatter intensity).
The sub-bottom profile data were collected using a Bathy-2010 (SyQwest, Osaka, Japan) up to 3 m depth, and using an EdgeTech 2000-DSS in sectors deeper than 3 m. These data were processed using the Marine Discovery 8 software package (Ocean High Technology Institute Inc.). The sub-bottom profilers were roughly divided into bedrock and sediment areas according to the intensity of reflected waves. The thickness of the seabed sediment overlying the bedrock was estimated on the basis of an interpretation of the boundary between the seabed sediment and bedrock.