Travel demand modeling in a bus transit network: an approach focusing on spatially dependent data

• Ano de defesa: 2023
• Autor(a) principal: Marques, Samuel de Franca
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: bus stop; count data; dados de contagem; dados faltantes; demanda por transporte público; geographically weighted regression; krigagem; kriging; missing data; ponto de parada; regressão geograficamente ponderada; transit ridership
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/18/18144/tde-06082024-120025/

 Boarding and alighting per bus stop modeling, along bus lines, plays a fundamental role in Transport Public network planning, in addition to contributing to transit-oriented development. However, in this case, the variables of interest (boarding and alighting) show four characteristics that have implications for the modeling process and/or may affect the estimates\' results. They are: (1) spatial dependence; (2) asymmetry; (3) the trips occur along the transport network; and (4) limited sample. As these peculiarities are overlooked by classical modeling and the scientific literature has not addressed them so far simultaneously, the main objective of the present study was to model transit ridership at the bus stop level, including, in the estimating process, the four aforementioned characteristics. As specific objectives, we analyzed the effect of explanatory variables, the effects of using network or Euclidean distances, the amount of missing data, and the sampling type. The text is divided into seven chapters, and four of them are articles, whose contents address one or more specific objectives separately and/or simultaneously. Two classes of spatial models were proposed, geostatistical interpolators and geographically weighted regressions, and a database comprising eight lines in the city of São Paulo (Brazil) was used as a case study. The following conclusions were achieved: models that consider asymmetry and spatial dependence should be prioritized over the ones that overlook these characteristics, as well as the multivariate models over the univariate ones. Public policies toward increasing public transport usage should focus mainly on four groups of explanatory variables: sociodemographic characteristics, bus network coverage, street layout and land use. The spatial models proved to be able to estimate the volume of boardings and alightings in unsampled points accurately, solving the problem of a lack of stop-level transit ridership data. Despite the network distance approach not contributing significantly to improving the models\' prediction power, this type of distance may better represent the relationship between the transit ridership and its intervening factors. Prioritizing the use of network distances in the spatial modeling of boardings and alightings is recommended. In addition, a balanced sample on predictor data and well-spread in the geographic space might be preferred to accurately estimate missing stop-level ridership data. Therefore, the present research adds important methodological and practical contributions to the urban planning associated with sustainable transport.