Citizen monitoring during hazards: validation of Fukushima radiation measurements

Citizen-led movements producing scientific hazard data during disasters are increasingly common. After the Japanese earthquake-triggered tsunami in 2011, and the resulting radioactive releases at the damaged Fukushima Daiichi nuclear power plants, citizens monitored on-ground levels of radiation with innovative mobile devices built from off-the-shelf components. To date, the citizen-led Safecast project has recorded 50 million radiation measurements worldwide, with the majority of these measurements from Japan. The analysis of data which are multi-dimensional, not vetted, and provided from multiple devices presents big data challenges due to their volume, velocity, variety, and veracity. While the Safecast project produced massive open-source radiation measurements at specific coordinates and times, the reliability and validity of the overall data have not yet been assessed. The nuclear disaster at the Fukushima Daiichi nuclear-power plant provides a case for assessing the Safecast data with official aerial remote sensing radiation data jointly collected by the governments of the United States and Japan. This study spatially analyzes and statistically compares the citizen-volunteered and government-generated radiation data. An assessment of the Safecast dataset requires several preprocessing steps. First, it was necessary to convert the data from the Safecast ionized radiation sensors since they were collected using different units of measure than the government data. Secondly, the normally occurring radiation and decay rates of cesium from deposition surveys were used to properly compare measurements in space and time. Finally, the GPS located points were selected within overlapping extents at multiple spatial resolutions. Quantitative measures were used to assess the similarity and differences in the observed measurements. Radiation measurements from the same geographic extents show similar spatial variations and statistically significant correlations. The results suggest that? actionable scientific data for disasters and emergencies can be inferred from non-traditional and not vetted data generated through citizen science projects. This project provides a methodology for comparing datasets of radiological measurements over time and space. Integrating data for assessment from different Earth sensing systems is paramount for societal and environmental problems.