Integrating Land-Use and Travel Demand in the Belgian Activity-Travel Simulation (BATS) Model: A Case Study from Wallonia

Travel behaviour research aims to understand how people build their travel choice sets, including essential elements such as travel purpose, departure time, travel distance, travel mode, activity location, and duration. This paper presents an integrated land-use and travel demand activity-travel scheduling model, extending the BATS (Belgian Activity-Travel Simulator), an agent-based activity-travel demand model based on Markov chain simulation. First, we describe the preparation of points of interest (POI)-based land-use data stemming from the Wallonia geoportal of Belgium. The POI-based land-use data (PICC) provide detailed building address points, footprints, linear road elements, and area centroids. Second, we introduce an activity-based microsimulation scheduler of twenty-four-hour travel in the large-scale region of Wallonia across urban and rural areas. The scheduling is based on the 2010 Belgian Household Travel Survey (BELDAM) and enhanced by explicitly considering the POI-based land-use data. We include all types of activities recorded in the travel survey data and categorize POIs accordingly. Apart from PICC, we integrate the geographic dataset representing the Walloon tourist offers to create a comprehensive POI-based land use as the potential activity locations. The relationship between land use categories and travel purposes is defined. Finally, we validate the simulation results by leveraging mobile phone-based origin-destination (OD) matrices, focusing on aggregate inbound trips considering similar spatial and temporal resolution. After the simulation, daily activity-travel plans for the population in Wallonia are completed with socio-demographics and the corresponding activity-travel choice sets. Our framework demonstrates the feasibility of integrating activity-travel scheduling with POI-based land use, validated using external mobile phone OD matrices. The comparison of daily mean attractions indicates a strong positive relationship. The Pearson correlation coefficient decreases slightly, while the R-squared score of the hourly mean attraction decreases significantly.