Sediment and ¹³⁷Cs transport and accumulation in the Ogaki Dam of eastern Fukushima

The Ogaki Dam Reservoir (figure 1) supplied water to paddy fields in the lower Ukedo River Basin in eastern Fukushima Prefecture before the nuclear accident. The Ogaki Dam is located 22 km upstream from the Ukedo River mouth. The catchment area of the Ogaki Dam is 111 km², which is about one-quarter of the total Ukedo River Basin (420 km²). The reservoir water level was lowered from 170 to 140 m above sea level after the accident to protect the dam structure against subsequent earthquakes and to ensure against collapse. As the upper part of the Ukedo River Basin is substantially more contaminated than the arable land on the coast, movement of radioactive cesium from the reservoir is a major concern in the considerations to reutilize the dam to support arable farming in the downriver basin.

Zoom In Zoom Out Reset image size

Recently, Iwasaki et al (2015) simulated sediment-sorbed ¹³⁷Cs in the Abukuma River and they estimated that 3.3 TBq of ¹³⁷Cs was discharged to the ocean during a flood in September 2011. This value is comparable to the annual ¹³⁷Cs discharge from the Abukuma River of 3.0 TBq calculated by our previously developed soil and cesium transport (SACT) model (Yamaguchi et al 2013, Yamaguchi et al 2014). Thus, typhoons are the most significant influence on sediment migration for river basins in the region. We simulate sediment movement in the Ogaki Dam Reservoir during two typhoon floods: i) for a flood that occurred between 15–17 September 2013 and ii) for a hypothetical 'average' yearly flood based on ten-year averaged data for floods in this river basin (Yamaguchi et al 2014, Kitamura et al 2014).

For the present study, we use the combined river flow and riverbed deformation solver Nays2D of the iRIC (International River Interface Cooperative) package. Nays2D was developed to predict the distribution of sediments and radioactive materials in dam reservoirs in a two-dimensional (2D) configuration (Shimizu 2003, Wongsa and Shimizu 2004). It is a simulation code for calculating unsteady 2D plane flow and riverbed deformation using boundary-fitted coordinates within general curvilinear coordinates (Shimizu et al 2012).

The absorption of cesium by sediments is strongly affected by particle size, surface area, mineral composition (Jackson and Inch 1983, Tanaka and Yamamoto 1988), and organic matter content. We considered three types of sediment, namely clay, silt, and sand, and modeled particles with representative diameters of 0.001, 0.01, and 0.1 mm for each sediment grade, respectively. The surface area proportionality of cesium absorption in these sediment types was accounted for by the method outlined by Yamaguchi et al (2014). In summary, we assumed that the total volume of ¹³⁷Cs fallout was instantly absorbed into surface soil, and assumed that the ¹³⁷Cs inventory in each fraction is given by ${{C}_{n}}=22.1\cdot S_{{\rm sp}}^{0.60},$ where C_n is the ¹³⁷Cs concentration (mBq g⁻¹) and S_sp is a specific surface area (m² g⁻¹) (He and Walling 1996).

For this study, we required calculations for sediment and cesium transport over a relatively large area of the reservoir and over a long period of time. We therefore parallelized the Nays2D code to run the Japan Atomic Energy Agency (JAEA) supercomputers and simulated 201 × 201 grids covering a 2 km × 2 km area. Each cell size was 10 m × 10 m and the time step of simulation was 0.2 s.

This study utilizes previously published results from two other watershed simulation models. Results from the SACT model (Yamaguchi et al 2013, Yamaguchi et al 2014) were used to inform boundary conditions to initiate the Nays2D simulation of the average annual flood (see section 2.3). Results for sediment discharged from the reservoir over the course of a flood from the time-dependent one-dimensional degradation and migration model (TODAM) (Kurikami et al 2014) were used to check the Nays2D simulations against an independent simulation method.

Neither SACT nor TODAM were able to determine accurately the locations where contaminated sediment tends to accumulate in the reservoir. As pointed out by Yamaguchi et al (2014), SACT cannot properly treat sediment deposition on reservoir and riverbeds, and TODAM is a one-dimensional (1D) code, hence it cannot determine locations of sediment deposition across the width of the reservoir. The key advantage of Nay2D is that it can model these processes; this is why using Nays2D was necessary for this study on using Ogaki Dam Reservoir management as a countermeasure for environmental protection.

The administrator of the Ogaki Dam, the Tohoku Regional Agricultural Administration Office (TRAAO), has been monitoring the waters entering the reservoir from two tributaries (at the Hirusone and the Yaguno monitoring stations, figure 1) and the water discharged at the reservoir exit since September 2012. The river discharge rate, the sediment concentration, and the radioactive cesium concentration in the water are being archived to understand their behavior through the Ogaki Dam Reservoir (Tohoku Regional Agricultural Administration Office (TRAAO) (2014)). The river discharge rate is monitored by pressure sensors and the sediment concentration and radioactive cesium concentration are analyzed by taking water samples.

Between September 2012, when TRAAO commenced their monitoring program, and July 2014, there have been two heavy rainfall events caused by typhoons, in September 2013 and October 2013 respectively (Kurikami et al 2014). During these events, higher concentrations of suspended sediment and radioactive cesium were detected at all three monitoring posts. The September 2013 typhoon was the stronger typhoon and the resulting concentrations of suspended sediment and radioactive cesium in the water were highest in this case. Our modeling approach in this study was to begin by simulating the sediment migration behavior during and after this rainfall event. The input data to the Nays2D simulation of the September 2013 flood were TRAAO time series of water and sediment inflow into the reservoir. Concentrations of sediment inflow at the Hirusone and the Yaguno monitoring stations during this flood are shown in figure 2. The performance of the simulation method was checked by comparing temporal results for the sediment concentration exiting the reservoir with field monitoring data.

Zoom In Zoom Out Reset image size

We compliment the simulation for this large flood (September 2013) with simulations for a hypothetical flood representative of the average yearly flood that occurs in the basin. These simulations are based on ten-year averaged data for the water inflow rate to the reservoir during flooding periods (Ministry of Land, Infrastructure and Transport (MLIT) (2013), Yamaguchi et al 2014, Kitamura et al 2014). SACT model results were used as boundary conditions for sediment inflow rates to the reservoir. The simulation conditions were a flood occurring over 40 h with an average reservoir inflow rate of 35 m³ s⁻¹. After 40 h, the average inflow rate decreased to 2 m³ s⁻¹, representing ambient river conditions. For the average yearly flood, we simulated two reservoir water levels (140 and 170 m above sea level).

The total length of the simulations for the September 2013 and average yearly flood with 140 m reservoir water level was 10 days. For the average yearly flood with the higher reservoir water level (170 m), the length of the simulation was increased to 80 days to allow settlement of suspended sediments in the reservoir.

There is reasonable agreement between the simulation predictions for the sediment concentration in the reservoir outflow and the observed values (upper panel of figure 3). The quantities of sediment discharge decrease over time because of the decreasing sediment inflow to the reservoir as the flooding abates. The simulation results over-predict the sediment concentration about the peak flow rate for the flood, and tend to under-predict the concentration as the flooding subsides. Also shown in figure 3 are the predictions for the ¹³⁷Cs loading of the water along with the observed values. The simulations reproduced the ¹³⁷Cs concentrations for the peak in the flooding well, but underestimated the values at later stages. The simulation results are therefore suitable for understanding the qualitative behavior of the bulk of contamination transport through the reservoir, which occurs during the peak of the flooding.

Zoom In Zoom Out Reset image size

We next evaluated the average amount of sediment deposited onto the reservoir bed, the suspension in the reservoir waters, and the discharge from the reservoir for the typical yearly floods. Currently, the water level in Ogaki Dam Reservoir is set at a low level (figure 1) and we considered this case first. The simulation results show sand that enters the reservoir is deposited almost immediately onto the reservoir bed. The sand deposition thus occurs where river tributaries enter the reservoir. The amount of sand that remains suspended in the reservoir water or that has the opportunity to be discharged downstream is negligible.

In terms of silt, most of this sediment type deposits on the reservoir bed over the course of the flood. Silt deposition occurs across the entire length and width of the reservoir bed after a lag period of a few hours, in which it is suspended in the reservoir water (figure 4(a)). Due to this lag period, the suspended silt has the opportunity to migrate across the length and width of the reservoir, and a small percentage (15%) is discharged downstream.

Zoom In Zoom Out Reset image size

In the case of clay transport, the results are substantially different. The primary tendency is for clay to remain suspended in the reservoir water and ultimately to be discharged downstream (figure 5(a)). In contrast to sand and silt, only about 30% of the clay entering the reservoir deposits onto the reservoir bed. The residence time for the suspension of clay in the reservoir water is thus substantially longer. In comparison to the characteristic suspension time for silt of a few hours, clay remains suspended in the reservoir water for over 200 h after the 40 h flood event had subsided.

Zoom In Zoom Out Reset image size

Based on these results, we estimate an annual quantity of sediment deposition on the Ogaki Dam Reservoir bed between 7.0 × 10³ to 8.0 × 10³ t. JAEA has a field survey group that is undertaking measurements to calculate the annual sediment deposition in the dam so these simulated values can be checked more completely in the future. However, their provisional results agree with the simulation result presented here within one order of magnitude.

The average yearly flood is two times larger in terms of the water inflow rate to the reservoir, and this results in a six times larger inflow of sediment to the reservoir over the course of the flood (table 1). The ¹³⁷Cs inflow is only four times larger for the average yearly flood than the September 2013 flood because, relatively, the larger flood transports more sand and radio-cesium has a lower affinity for binding to larger-grained sediments.

In both cases, sand deposits quickly on entering the reservoir, and sediment outflow from the reservoir is dominated by clay and silt. The sediment outflow in the case of the average yearly flood is six times greater than the September 2013 flood, whereas the ¹³⁷Cs outflow is five times greater. As the deposited sediments contain a higher fraction of sand, in the case of the average yearly flood, ¹³⁷Cs deposition is only three times larger than the September 2013 flood.

We next considered the effect of the reservoir water level on sediment transport. In a second set of simulations for average yearly flood conditions, the height of the water level exiting the dam was increased from 140 to 170 m above sea level. While the transport behavior for sand was basically unchanged, different results were obtained for both silt and clay transport. The bulk of silt deposits onto the reservoir bed (figure 4(b)), and only a very small portion discharges downstream from the reservoir, inset of figure 4(b).

For clay, the tendency to deposit onto the reservoir bed was much higher than in the case of the lower reservoir water level. The simulations showed a substantial settling period that lasted over 50 days after the two-day flood event subsided. On completion of the simulated period (about three months), 70% of the clay that entered the reservoir during the flood period had deposited onto the bed of the reservoir, figure 5(b). The remainder of the clay sediments passed downstream from the reservoir and a smaller portion remained suspended in the reservoir waters.

When figures 5(a) and (b) are compared, we observe that the residence time of clays suspended in the reservoir water increases by a factor of about 10 when the reservoir water level is increased by 30 meters. Such a large increase was not found for silt (see figures 4(a) and (b)), as it deposits quickly onto the reservoir bed for both reservoir water heights. A large reduction in the mass of silt and clay discharged from the reservoir was seen in both cases. Similar results were found using the 1-D TODAM code (Kurikami et al 2014), and qualitative agreement between the two sets of results can be observed in table 2.

We finally calculated the ¹³⁷Cs discharge from the dam and deposition onto the reservoir bed. Table 3 displays the values for both high and low reservoir water levels. When the reservoir water height is 140 m, the ¹³⁷Cs outflow from the reservoir was found to be four times larger than when the water level is 170 m. The corresponding ¹³⁷Cs deposition within the reservoir was found to be one-third lower for the 140 m reservoir water level than for the 170 m level. Because the clay and silt portions of the sediment play a dominant role in radioactive cesium transport (Yamaguchi et al 2013, Yamaguchi et al 2014, Kitamura et al 2014), and because our simulations indicate that by raising the height of the reservoir water level the clay outflow from the reservoir is substantially reduced, these results suggest that reservoir management can provide a countermeasure against radio-cesium migration downriver.

We next determined the locations where sediment tends to accumulate on the bed of the Ogaki Dam Reservoir. Figure 6 shows the temporal behavior of deposited sand, silt, and clay in the reservoir in the simulations for the low reservoir water level (140 m). Sand deposits on the bed close to where it enters the reservoir, whereas silt deposits across the length and width of the reservoir. For clay, the amount of deposition is minor during the initial stage of flooding due to its tendency to remain in suspension. However, deposition gradually increases in the hours after the flood, and occurs across the length of the reservoir and primarily toward the dam. Observations by TRAAO identified relatively high concentrations of radio-cesium on the bed in the lower downstream portion of the reservoir (Tohoku Regional Agricultural Administration Office (TRAAO) (2014)), which is in agreement with our simulated results.

Zoom In Zoom Out Reset image size