Abstract Manager

COVID-19 has impacted many domains of our life including the way we work, travel and live. Integrated urban models (IUMs) present a fitting platform to test the impacts of such socio-economic shocks as IUMs simulate population, their transportation and land use-related decisions, and their interactions. However, the development and implementation of such large-scale models suffer many challenges including behavioural representation and computational requirements. In this line of investigation, this paper presents a next generation agent-based integrated urban model, STELARS - Simulator for Transportation, Energy, and LAnd use for Regional Systems. This paper specifically presents the microsimulation of the residential relocation component and testing it to represent the COVID-19 housing market. Theoretically, the model is motivated by the life history-oriented perspective – assuming that households become active for making decisions such as relocate residences in anticipation or following a life event (e.g., birth of a child), otherwise remain inactive. Furthermore, residential relocation has been conceptualized as a four-stage decision process where households first decide to move (i.e., mobility decision), then search for locations, assess housing price and finally move to a location. To address the relocation behavioural dynamics of the households’, advanced methodologies have been developed. For example, a hazard-based duration modelling technique has been used to capture the continuous time dimension of stay at a location for the mobility component; whereas, a logit methodology has been used to model the choice of location as a discrete choice problem. Consistent with the theoretical and behavioural modelling methods, STELARS adopts an event-based approach utilizing a hybrid of continuous time and discrete microsimulation techniques. For example, mobility has been simulated as a continuous time decision whereas location choice follows a discrete simulation technique. The programming code has been developed using Python. The model has been implemented for the 100% population of the Central Okanagan region of British Columbia, Canada for the years of 2011-2021. The runtime for 10 simulation years was ~8 hours. The multi-year validation results suggest a reasonably satisfactory representation of the observed behaviour. For example, during the COVID-19, the model predicts a higher price for larger dwelling units in suburban areas – which reflects the need for additional spaces for home office during the pandemic. STELARS adds the capacity to integrated urban models to test for socio-economic shocks, which is a suitable application for such large-scale models given the wide-ranging longer-term impacts of the pandemic and the fitting scope of urban models.