Geometry-informed material intensity reveals considerable intra-archetype material variability of UK housing

Building stock modelling underpins energy and environmental assessments of the built environment. Material Intensity (MI), representing material mass per unit dimension, is vital for bottom-up estimation of building material stocks. However, reliance on sparse or uniform MI data can lead to significant inaccuracies due to intra-archetype variability, often stemming from differing building morphologies. This paper develops a geometry-informed MI (GIMI) method to characterise MI variability using machine learning and morphology features, applied to four materials—brick, concrete, mortar, and stone—in Sheffield, UK. Results indicate that GIMI reduces potential material uncertainties by up to 18% compared to conventional unitary MIs. This approach enhances bottom-up building mass accounting, advancing a circular economy and low-carbon building sector.