In March 2011, the accident at the Fukushima Daiichi Nuclear Power Plant (FDNPP) after the Great East Japan Earthquake resulted in the release of radioactive contaminants to the atmosphere and environment. In October 2015, Lawrence Berkeley National Laboratory (LBNL) and Japan Atomic Energy Agency (JAEA) began collaborating under an agreement between the U.S. Department of Energy (US DOE) and JAEA. Their primary objective is to support and enhance JAEA’s research into the environmental restoration of the Fukushima area.
(1) Multiscale Data Integration
Radiation measurements and monitoring in the region around the Fukushima Daiichi NPP have been performed continuously since the accident (Mikami et al., 2015; Saito et al., 2015). Such mapping efforts are essential to protect the public, guide decontamination efforts, estimate the amount of decontamination waste, and accommodate residents upon their return.
It has become clear that discrepancies exist among the many available data survey types – particularly in terms of measured dose-rate values, even when collected at the same times and locations. This is due to the degree to which accuracy varies by data type, or to the fact that each data type has a different support scale (i.e., support volume, resolution).
Recently, Wainwright et al. (2017) developed a Bayesian hierarchical modeling approach to integrate multiscale datasets (i.e., car, walk, and airborne surveys). They also created an integrated high-resolution map of air dose rates at 1 m above ground surface across the region. The maps have been distributed within the Japanese cities and used for planning decontamination efforts, resident return, and to optimize monitoring strategies for long-term monitoring.

Figure 1. (a) Original datasets (the car and walk survey data over the airborne data), (b) integrated radiation dose-rate map within the evacuation zone, and (b) the map of standard deviation, representing the uncertainty in the estimation. The red polygons are the evacuation zone extent, and the black contour lines are the threshold of 20mSv/yr and 50mSv/yr. In (a), the car and walk survey can be seen as lines and dots of lower-dose values compared to the surrounding air survey data. The unit is log10-microSv/hr.
Wainwright, H. M., Seki, A., Chen, J., & Saito, K. (2017). A multiscale Bayesian data integration approach for mapping air dose rates around the Fukushima Daiichi Nuclear Power Plant. Journal of environmental radioactivity, 167, 62-69.
(2) Predictive Modeling for Evaluating Remediation Tradeoffs

Figure 2. Land use and land cover (LULC) base case (a) and three forest management scenarios (c-d).
The long-term impact on local communities neighboring FDNPP may be substantial due to contaminated watersheds near the plant site, with radioactive cesium being the largest potential contributor to dose. While many former residents wish to return to re-establish communities, there is a lack of understanding and predictive capability for the transport of radioisotopes at the catchment to watershed scale. This is particularly true in the forested regions of the Fukushima Prefecture where the bulk of contamination still exists today. Forest de-contamination, however, has potentially harmful impacts, since it removes essential ecosystem components such as soil plants, and/or leaf litters.
Recently, Woodburn et al. (2015) applied an integrated hydrological model to understand radionuclide transport in complex forest ecosystems with the goal of providing informed, quantitative risk management decisions with guidance for local residents and stakeholders. Integrated hydrologic watershed model simulators can be used to produce increasingly high resolution and high fidelity hydrologic models capable of capturing the nonlinear dynamics of a watershed system. The researchers demonstrated different forest-thinning scenarios in terms of surface water runoff and erosions.
Siirila-Woodburn, E. R., Steefel, C. I., Williams, K. H., & Birkholzer, J. T. (2015, December). An Integrated Hydrologic Modeling Approach to Cesium-137 Transport in Forested Fukushima Watersheds. In AGU Fall Meeting Abstracts.
(3) Data Management and Model-Data Integration

Figure 3. Schematic showing architecture of the brokering service developed for the Fukushima data management and data-model integration.
Extensive environmental characterization and monitoring have been underway since the FDNPP accident. Diverse datasets – including radiological, geological, geochemical, geophysical, microbiological, hydrological, and meteorological data – are required to map the radiation dose rate and cesium soil contamination and to predict the future extent of contamination. There has been a tremendous effort to organize and archive these extensive datasets.
To further enhance the data management and integration in the Fukushima environmental restoration, LBNL has developed a data brokering service architecture that can connect distributed data sources via web services, which enables dynamic retrieval of data and universal access by any authorized client (Figure 3). The broker is designed to provide unified access to a diverse set of data sources and data types by connecting to those data sources in real-time and transforming the data streams to provide an integrated view.