The rapid urbanisation that Asia is in the midst of has brought attendant economic benefits from economies of scale, but has also meant that cities are choking. The rising levels of traffic lead to significant productivity losses that come with being stuck in traffic, whilst also increasing pollution levels with greater emissions. Efficient transportation planning and management requires spatio-temporal data of high frequency. In cooperation with the Office of Transport and Traffic Policy and Planning of Thailand and Pulse Lab Jakarta (which is leveraging a data partnership with Grab), the GIZ Data Lab initiated work to explore the feasibility of ride hailing data to both inform transportation policy and planning, as well as to develop proxy measures of air-quality at high spatio-temporal resolution for the Bangkok metropolitan region.
The initial research shows that ride-hailing data could potentially provide a range of needs for transportation planning and management, both in terms of data points as well as in terms of modelling such as traffic speeds, nowcasting congestion, and macroscopic traffic flow modelling. It also suggests that even with a low number of actual sensors for calibration, models developed through the use of ride-hailing data coupled with other data sources including those from satellite imagery, could potentially infer air quality at high spatial and temporal resolutions. Immediate practical application will require further research to improve robustness of the model, in particular to include other data sources such as traffic counting, hourly air quality, land use, population and on-the-field observation. Further study is recommended in selected areas of a city, based on government priority and the availability of primary data sets.
Main data source: Anonymous Grab Car Data covering December 2018, March 2019, April 2019 and September 2019.
Link to report
Macroscopic traffic flows can provide useful insights needed for transportation planning and modelling. These insights allow planners and modelers to understand important properties related to traffic flow, particularly on possible factors contributing to congestion. Macroscopic traffic flows are useful for short-term forecasting, providing the ability to understand and calculate, amongst others, average travel times, mean fuel consumption, etc. Traditional methods for developing macroscopic traffic flows depend on data collected through field surveys and observations, which are time consuming to conduct, but also have limited spatio-temporal coverage, and can be fairly subjective. Commercial products for getting those information, including Google Maps and Waze are popular and widely used but considered costly for city wide analysis. A trip distribution model is an alternative approach using transportation theory as a proxy to infer the traffic flows. The model promises lower cost, the need for fewer data, and huge spatio-temporal coverage but suffers from localization constraints, which need city-specific calibration. This research explore the possibility of using Grab data to assess the performance of four trip distribution models to proxy traffic flows in Bangkok. These four models include: i) a gravity model with exponential decay; ii) a gravity model with power law distance decay; iii) a radiation model; and iv) a radiation-extended model. Results are calibrated with actual traffic activity inferred from Grab data to prove the feasibility. The radiation model shows better correlation score (ρ = 0.5) with inferred actual traffic flows as compared to the others.
Road-speed profiling refers to the process of computing the expected, median, upper-bound and lower-bound (within a confidence interval) speeds of road segments in a city at a given time of the day. Road speed profiling is used to measure point-to-point travel time estimation, build advanced traffic information and management systems that could eventually help ease congestion given the constraints in terms of building new infrastructure. Aggregating road speed is a challenging task, because vehicle speeds are inherently stochastic. Traditionally, road-speed profile is calculated based on field survey and observation data, with limited spatio-temporal coverage. This research explores ways to improve road-speed profile accuracy, by developing road-speed profiles for nine major roads in Bangkok in 15 minutes intervals in a day using one month ride-hailing data provided by Grab. The road speed profiling shows promising results that reflect mobility patterns between normal days and major event days.
Traveling returning to the central
Traveling away from the central
Traffic congestion is one of the most important problem domains for transportation planning and management. The causes can be complex and sometimes random. The prediction of urban traffic congestion has emerged as one of the most important research topics related to advanced transportation systems. In the field of traffic flow forecasting, a number of models and methodologies have been put forward for the improvement of the existing model. This research aims to nowcast traffic congestion using the following predictors: historical congestion, road-speed profile, population count, district characteristics and time seasonality. The model is able to capture hourly seasonality and daily seasonality with limitation during heavy congestion.
Accurate estimates of human exposure to inhaled air pollutants are necessary for a realistic appraisal of the risks these pollutants pose and for the design and implementation of strategies to control and limit those risks. This estimation, except in an occupational setting, is usually based on measurements of pollutants concentrations in outside air, recorded with outdoor-fixed site monitors. From a public health perspective, it is important to determine the population exposure - the aggregate exposure for a specified group of people. This research aims to explore ways of harnessing four months observation data to quantifying population exposure to air pollution. Using Land Use Regression (LUR), this research infers daily air quality in Bangkok at 1kmx1km level. Bangkok Metropolitan has 13 official ground air quality sensors, although limited in number and coverage, data from these sensors are sufficient for preliminary validation. The best model shows r2 = 0.6 inference performance.