Modelling Chinese Urban Residential Stock Turnover Uncertainties Using System Dynamics and Bayesian Statistical Inference

Building stock turnover is one of the key determinants in building energy modelling and policy analysis. Building lifetime is integral to the dynamics of stock turnover. However in China, despite anecdotal claims that urban residential buildings are generally short-lived, there are no official statistics on building lifetime, and empirical data is extremely limited. Official statistics on total floor area of Chinese urban residential stock only exist up to 2006. This paper presents a Bayesian approach to estimate Chinese urban residential building lifetime and characterise the overall stock turnover dynamics for the period of 2007 to 2017. Firstly, the building stock evolution process is described by a system dynamics model in which survival analysis is used to characterise the dynamic interplay between new construction, aging, and demolition of buildings. The uncertainties associated with building lifetime are represented using a Weibull distribution. Secondly, based on this model and official statistics on urban residential floor area up to 2006, a Bayesian probabilistic model is developed to simulate the posterior distribution of Weibull parameters through Markov Chain Monte Carlo (MCMC) technique. As a result, the distribution of building lifetime unconditional on the Weibull parameters is obtained. Further, the posterior distributions of Weibull parameters, along with official statistics on annual new construction, enable the estimate of stock turnover in the form of posterior predictive distribution for the period of 2007 to 2017. This Bayesian modelling framework, and its results in the form of probability distributions of annual total stock and underlying age-specific sub-stocks, can provide the basis for further modelling and analysing policy trade-offs of embodied-versus-operational energy consumption and carbon emissions of buildings in the context of sector-wide decarbonisation.