EGU22-2526
https://doi.org/10.5194/egusphere-egu22-2526
EGU General Assembly 2022
© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Building typologies for Norway: a case study for Oslo using machine learning

Federica Ghione1,2, Steffen Mæland3, Abdelghani Meslem1,4, and Volker Oye1,2
Federica Ghione et al.
  • 1NORSAR, Department of Applied Seismology, Kjeller, Norway
  • 2University of Oslo, Department of Geosciences, Oslo, Norway
  • 3SimulaMet, Department of Holistic Systems, Oslo, Norway
  • 4NMBU, Ås, Norway

To evaluate potential human and economic losses in a seismic risk assessment, it is important to define an exposure model by defining building materials and characteristics. The common approach to develop an exposure model is to have a first overview of the area with Google Earth and to perform extensive fieldwork in representative areas of the city. This procedure is time and cost consuming, and it is also subject to personal interpretation. To mitigate these costs, a Convolutional Neural Network (CNN) is used to automatically identify the different building typologies in the city of Oslo, Norway, based on facade images taken from in-situ fieldwork and Google Street View.

The present article attempts to categorize Oslo’s building stock in five main building typologies: timber (T), unreinforced masonry (MUR), reinforced concrete (CR), composite (steel reinforced concrete) (SRC) and steel (S). This method shows good results for timber buildings with 91% accuracy score, but only 41% for steel reinforced concrete buildings. These variations can be explained by differences in the number of labelled images for each typology, comprising the training data, and differences in complexity between typologies.

This work is the first tentative to identify Norwegian building typologies: based on experts judgement, the five types observed in Oslo can be applicable at national level. In addition, this study shows that CNNs can significantly contribute in terms of developing a cost-effective exposure model.

How to cite: Ghione, F., Mæland, S., Meslem, A., and Oye, V.: Building typologies for Norway: a case study for Oslo using machine learning, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2526, https://doi.org/10.5194/egusphere-egu22-2526, 2022.

Display materials

Display file

Comments on the display material

to access the discussion