Danijel Schorlemmer

OpenStreetMap (OSM) is the largest crowd-sourced mapping effort to date, with an infrastructure network that is considered near-complete. The mapping activities started as any crowd-sourced information platform: the community expanded OSM anywhere there was a collective interest. Initial efforts were found around universities or hometowns of mappers. Events, such as natural disasters can also trigger a major update. The recent earthquakes in Turkey and Syria lead to a massive contribution by the Humanitarian OSM Team (HOT) of more than 1.7 million buildings in the region in less than a month after the event1. This type of activities result in a map that is of non-uniform completeness, with some areas having all building footprints in, while other areas remain incomplete or even untouched. Currently, with 550 million footprints, OSM identifies between a quarter and half of the total building footprints in the world, if we estimate that there are around 1-2 billion buildings in the world. A global view on the local completeness of buildings in OSM did not yet exist. Unlike other efforts, that only look at a subset of OSM building data (Biljecki & Ang 2020; Orden et al., 2020; Zhou et al., 2020), we have used the Global Human Settlement Layer (GHSL) to estimate completeness of the entire dataset. The remote sensing dataset is distributed onto a grid of approximately 100x100 meter tiles. In each tile of the grid, the built area of GHSL is compared to the total area of OSM building footprints. The computed ratio is measured against a completeness threshold that is calibrated using areas that were manually assessed. Using information derived from remote sensing datasets can be problematic: GHSL does not only measure building footprints: it includes any human-built structures, including infrastructure and industrial areas. Next to that, due to sub-optimal input data or failing algorithms, the dataset is not of the same quality as the crowd-sourced data in OSM in areas that are complete. Even with these limitations, a comprehensive global completeness assessment is created. The assessment should not be used as ground truth, but rather as reflection on the OSM building dataset as is and as a guideline for priorities for the future. Statistics on regional completeness can be created and the quality of GHSL could be assessed on countries that are considered to be complete, such as France or the Netherlands.

None

Other people: Laurens Oostwegel

The substantial reduction of disaster risk and life losses, 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 it affects, in case of natural disasters, the life of communities, represented by local governments and individuals. The framework states that communities participating in risk assessments should increase their understanding of efficient risk mitigation measures. Earthquakes are threatening many regions in the world with constantly increasing risk due to rapid urbanization and industrialization. Earthquakes do not kill people, buildings do. Thus, the main threat of earthquakes comes from building damage and collapse. To improve resilience and preparedness, we need to estimate the risk, the possible damage of buildings and the related human and financial losses. This requires not only the position, size and class of buildings, but also the reconstruction value and the number of people inside the building at any time. For this, exposure models are used that translate the physical earthquake hazard to building damage, human and financial losses. Exposure models usually describe the built environment of administrative regions as groups (aggregates) of different building classes and their frequency. We present our open, dynamic, and global approach to describe, model, and classify every building on Earth with the greatest level of detail. Our model is based on the building data from OpenStreetMap and engineering information from open exposure models, combining these two sources to a building-by-building description of the exposed assets. We retain the aggregated descriptions where the building coverage in OpenStreetMap is incomplete and describe every building separately where building data is available. Due to the near-real-time computations of our model, it directly profits from the growth of OpenStreetMap and with about 5 million buildings added each month (or approx. 2 per second), the areas of incomplete coverage are constantly shrinking, making way for our building-specific exposure model. Here, we introduce shortly the earthquake phenomenon, how it affects the built environment, why a high level of detail is necessary for useful assessments of the impact and the consequences of earthquakes, how OpenStreetMap and other open data helps us to achieve this goal and how communities can benefit for the model for their own risk assessments.

None

To assess the possible human and financial losses of earthquakes and to estimate the long-term earthquake risk that many people on Earth are exposed to, detailed knowledge of buildings is paramount. This encompasses not only the position, size, and type of buildings, but also the reconstruction value and the number of people inside the building at any time. Using OpenStreetMap data and further open data, we are implementing an open, global, dynamic, purely algorithmic, and reproducible exposure model for the probabilistic description of the aforementioned parameters for every building on Earth, growing and changing with every edit in OpenStreetMap.

Earthquakes are threatening many regions in the world with constantly increasing risk due to rapid urbanization and industrialization. To improve resilience and preparedness, we need to estimate the risk of earthquakes with the greatest possible detail. For this, exposure models are used that translate the physical earthquake hazard to building damage, human and financial losses. The level of detail in the risk model directly depends on the level of detail in the exposure model. So far, exposure models are usually described as aggregated building-type descriptions for larger geographic areas, from city districts to even larger administrative units. We present our new open, dynamic, and global exposure model based on OpenStreetMap that does not stop at administrative boundaries but rather attempts to classify and describe every building on Earth with the greatest level of detail. Our open-source model extracts all possible exposure indicators, i.e. footprint shape, number of stories, occupancy type, shape of the roof, etc. and combines the OpenStreetMap data with other open datasets if available. Using these indicators, the model assesses in a probabilistic way the possible building classes, the number of people inside the building depending on the time and day, and the reconstruction value. In areas with incomplete building coverage, the classical aggregate-based exposure models are combined with our model to deliver a probabilistic description of the entire building stock. To achieve a better spatial distribution of buildings in areas of incomplete coverage, we estimate the likely locations of buildings through remote sensing using again open data only, mainly Sentinel-I radar data. Due to the near-realtime computations of our model, it directly profits from the growth of OpenStreetMap and with about 5 million buildings added each month, the areas of incomplete coverage are constantly shrinking, making way for our building-specific exposure model.

What are the perspectives and challenges around the new wealth of building data in OpenStreetMap?

OpenStreetMap (OSM) is constantly expanding into emerging fields, particularly around building datasets. Through its participatory character, OSM provides an incomplete, but global coverage of buildings, nowadays already outnumbering roads. Their distribution and properties are the best indicator to population density, economic activity patterns, and risks of natural hazards. Thus, OSM constitutes a perfect base to provide answers to many related questions of various stakeholders from communities, governments, private sector and research. We present an overview of how worldwide building data is used and will be in the near future. We will guide through our latest projects, reaching from the creation of risk models, via automated building extraction from satellite sensor data to building completeness estimates, and set these developments in context with other community-driven activities. We report on the challenges of managing global building data, and combining attributes from other open and public domain big datasets towards rich handling of building data in and around OSM. A special focus is on the ways how this new information can be made useful to support mappers to achieve higher completeness of building data.

Other people: Felix Delattre