Uncertainty in Binned Building Construction Year Data: Comparing EPC and Crowdsourced Datasets

Abstract. The building’s construction year indicates energy performance and retrofit potential, and is used in energy performance ratings. However, in many countries, comprehensive datasets providing the construction date of buildings are unavailable or difficult to access. In the UK, Energy Performance Certificates (EPC) provide the only open dataset on building construction years, at national scale. However, the data are binned, and minimal information is available on data quality. For specific areas of England, construction year data have also been contributed by historians to the Colouring Britain data platform. This study investigates systematic differences between EPC and historian-crowdsourced construction year data for 4,849 buildings in Loughborough, covering 14 EPC bands. To increase comparability of datasets with differing age bands, we test a random forest method to resolve the binning using two feature sets: urban form features, and urban form features plus EPC bands. The median of each EPC band acts as the baseline. The results show that combining urban form features and EPC bands delivers the highest accuracy.