Improving transportation planning tools for the Global South: applying advanced modelling frameworks to address data issues and accommodate behavioural complexities

Transport planning in the Global South faces unique challenges, particularly regarding data reliability and behavioural complexities. Nevertheless, since the transport planning literature is predominantly based on findings from high income countries, the planners and policymakers in the Global South often have to adopt transport models originally developed for the Global North. However, such models often overlook the behavioural dynamics and contextual factors specific to the Global South. This oversight can lead to uncertainties in results, and consequently, in the planning decisions. The primary barriers to developing effective transport planning tools in this context include the absence of reliable data sources, a lack of thorough investigation into behavioural complexity, and the absence of a framework for modelling choices in the Global South. This thesis aims to address some of the challenges related to data scarcity, behavioural complexity, and unique contextual factors in travel behaviour modelling by focusing on two distinct cities in the Global South: Dhaka, Bangladesh, and Concepción, Chile.
One of the main challenges in developing robust behavioural models, particularly in the Global South, is the absence of comprehensive and up-to-date mobility data. While conventional survey data can offer valuable sociodemographic insights, they often lack the necessary information for developing robust behavioural models, such as preferred departure time, arrival time, availability of mode, and travel time at other times of the day and for different mode. Moreover, emerging data sources like Microsoft Bing Maps and the Google Maps API, which offer reliable real-time mobility data in the Global North, often have more uncertainty and hence prove insufficient in the Global South. To complement travel diary survey data, there is increasing interest to utilise passive data sources such as mobile phone CDR, GSM, and GPS data. However, many models developed using passive data lack validation for policy implementation due to a lack of appropriate ground truth data.
To overcome these data challenges, this thesis investigates the application of emerging data sources to enhance the limited data available in the global south, particularly to develop advanced discrete choice and agent-based micro-simulation models. A new method is proposed to estimate travel time for alternative time periods using the Google Maps API and stated travel times. This estimation enables the development of mixed logit-based departure time choice models using survey based revealed preference (RP) data for car commuters in Dhaka, Bangladesh. The developed models facilitate the inference of required statistics, such as preferred departure time, sensitivity to travel time which are crucial for implementing appropriate strategies (e.g., peak spreading policy) to reduce traffic congestion.
In situations where travel time data for unchosen modes (e.g., public transport, auto-rickshaw, motorcycle) are unavailable, a congestion matrix is formulated based on car travel times obtained from Google Maps data. This matrix enables the development of a mode choice model using RP and SP data to generate travel demand for future scenarios, such as the implementation of Bus Rapid Transit (BRT) in Dhaka. Implementing this model in an agent-based simulation platform (MATSim) enables the simulation of potential BRT demand scenarios in the context of Dhaka, facilitating a more comprehensive assessment of future transport scenarios. Additionally, this thesis generates a sample of representative agents’ activities, time choice, and movement trajectories on Dhaka’s existing transportation system. The simulated scenario is then used to evaluate the accuracy of three mainstream passive data sources (i.e., GPS, CDR, and GSM data) in inferring spatiotemporal trajectory information such as stay location, travel distance, and departure time. The results underscore the feasibility of leveraging emerging data sources for the development of advanced behavioural models. The framework can be extended in future for evaluating the implications of the margin of errors in estimating value of travel time savings and other welfare measures based on these passive data sources.
Furthermore, in order to better capture the behavioural dynamics (e.g., correlation, satiation, random heterogeneity) prevalent in the Global South, the thesis emphasises the importance of extending state-of-the-art models (i.e., departure time choice, activity choice, and time use decision). A comparison between the state-of-the-art models and other candidate models is carried out to examine the impacts of correlation, role of satiation, and activity start time on choice prediction. By employing a new polynomial functional form, the thesis investigates the correlation between outbound and duration choice, exploring its influence on time-of-day preference. Furthermore, it examines utility preferences that vary over time using time-dependent utility and upper bound time constraints based on activity start time, while jointly modelling activity type, start time, and duration. Finally, it examines the impact of satiation and correlation among alternatives on the prediction accuracy of activity duration using different activity models in the context of Concepcion, Chile. The results emphasise the importance of thoroughly examining behavioural dynamics within the context of the Global South and the need to improve and expand the existing state-of-the-art models to better capture the specific behavioural dynamics in the Global South.
The findings of the thesis are expected to be useful for the planners and policymakers in the Global South in the selection of appropriate data sources and modelling frameworks for improving the current state of practice in transport planning.