Abstract
AbstractThis study analyzes socio-economic demographics (including Geomatic conzoom® segmented demographic variables) as well as building and property registration information as risk factors in relation to the prevalence of residential building fires within 100 m × 100 m cells. The logistic regression model achieved a receiver operating curve (ROC) of 0.74 and a precision-recall curve of 0.12 on the testing dataset. The model identifies 19 significant variables related to the risk of residential fire. The top 5 highest performing variables in our model and their odds ratios are the following: number of people (OR 1.25), Multi/family residence-building type (OR 1.20), number of buildings (OR 1.18), conzoom® Type C—Country/Rural Communities (OR 0.85), construction year (OR 0.87). These results indicate that socio-economic factors play a large role in influencing fire vulnerability within residential areas and can help prioritize resource allocation to reduce fire vulnerability in the identified risk factor groups.
Funder
European Civil Protection and Humanitarian Aid Operations
Copenhagen University
Publisher
Springer Science and Business Media LLC
Reference40 articles.
1. Nordic Fire Statistics (2020) Nordic fire statistics. Retrieved from Building Fires: https://ida.msb.se/nfs#page=e9d6c228-9a34-49f6-b210-41ef30d43002
2. Danish Emergency Management Agency (2018) Dødsbrande i Danmark. Beredskabsstyrelsen, Birkeroed
3. Danish Emergency Management Agency (2003) Mennesker og Brand. Beredskabsstyrelsen, Birkeroed
4. Beredskabsstyrelsen (2016) National strategi for forebyggelse af ulykker og katastrofer. Beredskabsstyrelsen, Birkeroed
5. Vilar del Hoyo L, Isabel M, Vega JM (2011) Logistic regression models for human-caused wildfire risk estimation: analysing the effect of the spatial accuracy in fire occurrence data. Eur J Forest Res 130(6):983–996