Assessment of the Dynamic Exposure to PM2.5 Based on Hourly Cell Phone Location and Land Use Regression Model in Beijing

The spatiotemporal locations of large populations are difficult to clearly characterize using traditional exposure assessment, mainly due to their complicated daily intraurban activities. This study aimed to extract hourly locations for the total population of Beijing based on cell phone data and assess their dynamic exposure to ambient PM2.5. The locations of residents were located by the cellular base stations that were keeping in contact with their cell phones. The diurnal activity pattern of the total population was investigated through the dynamic spatial distribution of all of the cell phones. The outdoor PM2.5 concentration was predicted in detail using a land use regression (LUR) model. The hourly PM2.5 map was overlapped with the hourly distribution of people for dynamic PM2.5 exposure estimation. For the mobile-derived total population, the mean level of PM2.5 exposure was 89.5 μg/m3 during the period from 2013 to 2015, which was higher than that reported for the census population (87.9 μg/m3). The hourly activity pattern showed that more than 10% of the total population commuted into the center of Beijing (e.g., the 5th ring road) during the daytime. On average, the PM2.5 concentration at workplaces was generally higher than in residential areas. The dynamic PM2.5 exposure pattern also varied with seasons. This study exhibited the strengths of mobile location in deriving the daily spatiotemporal activity patterns of the population in a megacity. This technology would refine future exposure assessment, including either small group cohort studies or city-level large population assessments.

Global Dynamic Exposure and the OpenBuildingMap - A Big-Data and Crowd-Sourcing Approach to Exposure Modeling

<p>The substantial reduction of disaster risk and loss of life, a major goal of the Sendai Framework by the United Nations Office for Disaster Risk Reduction (UNISDR), requires a clear understanding of the dynamics of the built environment and how they affect, in the case of natural disasters, the life of communities, represented by local governments and individuals. These dynamics can be best understood and captured by the local communities themselves, following two of the guiding principles formulated by the UNISDR: "empowerment of local authorities and communities" and "engagement from all of society". The two lead to societies increasing their understanding of efficient risk mitigation measures.</p><p>Our Global Dynamic Exposure model and its technical infrastructure build on the involvement of communities in a citizen-science approach. We are employing a crowd-sourced exposure capturing based on OpenStreetMap (OSM), an ideal foundation with already more than 375 million building footprints (growing daily by ~150,000), and a plethora of information about school, hospital, and other critical facilities. We are harvesting this dataset with our OpenBuildingMap system by processing the information associated with every building in near-real-time. We are enriching this dataset in a truly big-data approach by including built-up area detection from remote sensing with satellite and radar imagery combined with different sources of road networks, as well as various open datasets and aggregated exposure models that provide relevant additional information on, buildings and land use. </p><p>A task of such a scale does not come without challenges, particularly in matters of data completeness, privacy and the merging and homogenizing of different datasets. We are thus investing a large effort on the development of strategies to tackle these in a transparent and consistent way.</p><p>We are fully automatically collecting exposure and vulnerability indicators from explicitly provided data (e.g., hospital locations), implicitly provided data (e.g., building shapes and positions), and semantically derived data, that is, interpretation applying expert knowledge. The latter allows for the translation of simple building properties as captured by OpenStreetMap users or taken from open datasets into vulnerability and exposure indicators and subsequently into building classifications as defined in the Building Taxonomy 2.0 developed by the Global Earthquake Model (GEM) and in the European Macroseismic Scale (EMS98). A task of such a scale does not come without challenges, particularly in matters of data completeness, privacy and the merging and homogenizing of different datasets. We are thus investing a large effort on the development of strategies to tackle these in a transparent and consistent way. With our open approach, we increase the resolution of existing exposure models minute by minute through data updates and step by step with each added building, as we move forward from aggregated to building-by-building descriptions of exposure. </p><p>We expect the quality of near-real-time estimates of the extent of natural disasters to increase by an order of magnitude, based on the data we are collecting. We envision authorities and first responders greatly benefitting form maps pinpointing the greatest trouble spots in disasters and from detailed quantitative estimates of the likely damage and human losses.</p>