ABSTRACT
In this research paper, we establish a physics-based and data-driven methodology for early-stage embodied carbon assessment of reinforced concrete buildings’ structural systems. This approach first develops a rigorous parametric model based on bottom-up engineering calculations to algorithmically create a high-fidelity structural model with a few key inputs. Then it leverages a supervised machine learning model, serving as a surrogate to rapidly predict structural material quantities, thereby allowing real-time early-stage embodied carbon analysis. In addition, this paper investigates the potential of scaling up our methodology to urban building stocks’ data. We deployed our model within an existing urban modeling design tool and conducted a case study of a neighborhood in Lisbon, Portugal. Our results show the benefits of spatially mapping the distribution of embodied carbon in an existing building stock, identifying carbon hotspots, and obtaining a granular estimation of embodied carbon comparable with existing benchmarking studies in Europe.
