Confiabilidade do tempo de viagem: uma análise exploratória em vias urbanas de Fortaleza

Abstract

Travel Time Reliability (TTR) is a critical component of the quality of urban transportation systems, yet its dynamics and influencing factors remain underexplored in Brazilian cities. The main objective of this master's dissertation was to characterize TTR on arterial corridors in Fortaleza-CE, exploring the infrastructure, operational performance, and land use factors that potentially affect reliability. Utilizing a database with over 1.2 million trips from 35 urban road sections in 2019, the methodology involved the analysis of the statistical distribution of travel times, the selection of reliability indicators, and the exploration of possible determining factors. This last stage compared parametric linear models with Generalized Additive Mixed Models (GAMMs) to capture complex and possibly non-linear relationships. The results revealed the complexity of traffic variability behavior: although the Lognormal distribution was the best-fitting model (52% of cases), no probability model was adequate in 37% of the scenarios. The factor exploration stage showed the statistical invalidity of the linear models, reinforcing the non-linearity hypothesis. The application of GAMMs demonstrated superior performance (Adjusted R2 of 0.65 for the Buffer Time Index - BTI, and 0.97 for the Travel Time Index - TTI) and allowed for a better understanding of the relationships. The main finding was the non-linear effect of speed on both BTI and TTI, with diminishing reliability and efficiency gains at higher speeds. Additionally, the analysis showed that roads with wider lanes were consistently associated with a worsening of both reliability (higher BTI) and efficiency (higher TTI). It is concluded that TTR in an urban environment such as Fortaleza's is a multifaceted phenomenon, governed by complex interactions that challenge linear assumptions. The study offers a characterization of reliability for important corridors in Fortaleza and proposes a method that suggests the use of non-linear models for the analysis of this phenomenon.