Using open-access data for the development of industrial building portfolios

Earthquake impacts to buildings of industrial use have been particularly relevant in recent events, highlighting the importance of the need for detailed modelling of direct and indirect losses from this type of structure. In Europe, in particular, evidence of satisfactory earthquake response goes hand in hand with reports of collapse of precast buildings, which is the most common structural typology used in industrial buildings. In Italy, specifically, the seismic sequence that struck the region of Emilia Romagna in 2012 caused approximately three quarters of the precast concrete industrial buildings designed with no seismic to suffer significant damage, with one quarter of the total presenting partial or total collapse of the roof. Examples of the damage observed in this event are presented in the figure below.

damage_industrial
Damage in industrial buildings in the Emiglia-Romagna region after the 2012 seismic event

In the context of earthquake loss estimation of industrial buildings, not only is census data or other types of cadastral information extremely difficult to obtain at any level of regional aggregation, spatial disaggregation methods typically applied in residential building portfolios are also not suitable. As a result, alternative approaches need to be explored, particularly with respect to sources of spatial-based building data.

The IMAGE 2000 & CORINE Land Cover project (EEA-ETC/TE, 2002) was launched in 1994 by the European Environmental Agency (EAA) and the Joint Research Centre (JRC) of the European Commission. Its main objective was to provide an up-to-date land cover database, as well as information regarding general land cover changes in Europe between 1990 and 2000. These datasets were derived from satellite imagery and other ancillary data such as aerial photos, digital elevation datasets and hydrology models. As a result, CORINE classifies the European territory into 44 classes at a spatial detail comparable to that of a paper map on a scale of 1:100,000. Amongst the total set of identified land cover classes, the industrial and commercial units are of interest for this study, as illustrated below and further presented in detail.

Corine
Non-residential areas (black) provided by CORINE for 36 European countries

OpenStreetMap (OSM) is a collaborative project founded in 2004 by the University College London, with the aim of creating a free geographic database of the entire world. Its launch marked a new approach to gathering geodata, made possible by the increasing proliferation of GPS devices amongst private users and by the availability of web-based mapping services. In the case of exposure datasets for earthquake loss assessment specifically for industrial buildings, several georeferenced features provided by OSM are of interest. Available information regarding building footprints and location, building height and land use areas are of increasing usefulness, especially when combined with additional open-access data such as the CORINE initiative. In this study, data concerning building location and footprint, height, area, and land use are used as input for the proposed methodology to derive industrial exposure data.

OSM
Industrial building footprints (blue) provided by OSM for 36 European countries

The methodology begins with the identification of  the non-residential areas containing undifferentiated industrial and commercial units from CORINE. Then, the method identifies polygons (building footprints) classified as industrial buildings from OSM in the previously selected CORINE areas. This process allows the calculation of the density of built-up area of industrial structures in the regions marked in the CORINE dataset as industrial or commercial. Unlike CORINE which has a uniform coverage, OSM is still incomplete in several regions in Europe, which means that the densities calculated in complete regions are used in the areas where the OSM data is still insufficient.

The outcome of this process is the total industrial built-up area in a given region. Here, the average industrial building density in the region was estimated to range from 500 m2/km2 or less in countries such as Austria, Croatia, Finland, Sweden and Turkey; to values of 1000m2/km2 to 1500m2/kmin France, Germany, Denmark, Italy, Czech Republic, Hungary, Poland, Portugal, Romania, Switzerland and the United Kingdom. Overall maximum industrial building densities of approximately 5000m2/km2 and 7000m2/kmare found in Belgium and the Netherlands, respectively. These indicators represent the ratio between total area of industrial buildings in a given country and the area of its territory. The following figure illustrates the aggregation of results at the first administrative level in each country.

Europe
Illustration of industrial “exposure at risk” in Europe. The solid contour lines represent areas where PGA values equal or higher than 0.05 g (green), 0.15 g (orange), and 0.30g (red) are expected to be exceeded for a return period of 475 years

We qualitatively evaluated the level of risk in Europe through the assessment of what is herein referred as “exposure at risk”. To this end, the probabilistic seismic hazard results provided by the SHARE project were employed in order to determine the proportion of industrial areas subjected to ground motion intensities for specific return periods of exceedance. The results presented in the following figures allow the identification of the European areas in which particular attention should be devoted during vulnerability and loss assessment studies. These figures rank the top 10 medium to high seismicity countries based on the total industrial areas subjected to different ground motion intensities (PGA) at the 475-year hazard level. The first figure shows the ranking in absolute values of total area exposed while the latter shows the ranking by area exposed normalized by the total area of industrial buildings in each country.

Total_area
Total area of industrial assets (for the top 10 countries) exposed to PGA values of 0.15 g (grey) and 0.30 g (black) or higher for a return period of 475 years
Relative_area
Same as the previous figure, but normalized by the total area of industrial buildings in each country

This study was performed under the PRODEC – Doctoral program in Civil Engineering from the University of Porto (Portugal), in collaboration with GEM. For more details, please contact the doctoral candidate at costa.sousa@fe.up.pt or the integrated risk team at integrated_risk@globalquakemodel.org.

.

Advertisements