UC Berkeley

Since the Fukushima Daiichi Nuclear Power Plant (FDNPP) accident in 2011, radiation measurements and monitoring have been conducted continuously. Radiation air dose rate datasets have been archived extensively in this area. There are several different types of measurements: fixed-point measurements, walk surveys, car surveys, airborne surveys, and monitoring posts. They have different spatial coverage, footprints, and uncertainty. Currently, the monitoring program is expected to transition to long-term monitoring after ten years’ monitoring. The challenge of long-term monitoring is to build a cost effective and sustainable strategy for minimizing the cost associated with the number of monitoring locations or sampling, while maximizing the ability to meet the objectives of long-term monitoring.

This study aims to develop the long-term radiation monitoring strategies after the FDNPP accident. In this dissertation, we tackle three key challenges: (1) multiscale spatial data integration, (2) monitoring optimization, and (3) spatiotemporal data integration. First, we developed an efficient algorithm for integrating the multiscale data sets; the algorithm is based on Kriging to estimate the dose rates for unobserved locations. Secondly, we developed a strategy and an algorithm to optimize the monitoring post placement and their number. This strategy is designed in order to reduce the number of sensors while capturing spatial heterogeneity. The algorithm is based on Gaussian process model to capture and estimate the heterogeneity of air-dose rates across the domain. Lastly, we built a Kalman-filter based algorithm combined with Gaussian Process Model to predict the spatial–temporal distribution of radiation dose rates. We expect that these methods will have valuable contributions for the long-term monitoring in the Fukushima region, but also for the preparation for the future nuclear accidents.