dc.description.abstract | Accurate assessment of the built environment stock and its in- and outflows are a key factor to move humanities largest material consuming sector towards a circular economy. Current models do often not take advantage of the increasingly available data for buildings and infrastructure, but there is a clear trend and outspoken desire to move to high resolution bottom-up models to better map the stock and its development. This study reviewed these trends and available data for a case study in Trondheim to synthesize a model to assess the residential building stock. The chosen approach uses buildings as its base unit and predicts material contents and renovation profiles directly, overcoming traditional material intensity coefficients and improving current renovation modelling. Data from the cadastre for all buildings combined with high resolution building information models (BIMs) of a small sample of the stock are used for the predictions using machine learning. The model was then tested on an artificial test stock and its feasibility, accuracy, precision and limitations evaluated based on the results. We showed that the developed model can predict the material flows from demolition and renovation activity sufficiently accurate and precise for several decades. The occurring imprecision and inaccuracies can likely be decreased further in real world case studies when applying more advanced statistical learning techniques and through improvements made in the lifetime estimation. We conclude that this approach shows great potential and should be tested in a case study. | |