Abstract
Abstract
Household travel surveys (HTS) remain essential for transportation planning but are susceptible to recall and proxy-reporting biases that can systematically omit short, discretionary, and intermediate trips. This study develops a Bayesian core–satellite framework that uses insight from smaller, GPS-based smartphone travel surveys to identify and correct missing trips in a large conventional HTS while preserving population coverage and socioeconomic detail. The framework is demonstrated using the 2022 Transportation Tomorrow Survey in the Greater Toronto and Hamilton Area, supplemented by multi-day smartphone surveys conducted in 2023 and 2025. Bayesian models estimated from HTS are treated as priors and sequentially updated using GPS observations. Parametric Bayesian models estimate missing non-skeleton trip counts and distances, while non-parametric Bayesian models reconstruct trip purpose, start time, and mode. Primary and proxy respondents are modelled separately to account for differences in reporting errors. Posterior predictive diagnostics show close agreement between observed and simulated trip-count and distance distributions from the updated models. Among mobile primary respondents, correction increases the mean trip rate from 3.05 to 3.90 trips per person per day, while the proxy rate increases from 2.64 to 3.63. Corrected trip rates agree with rates found in the GPS-based data. The reconstructed trips increase representation of discretionary activities and shorter trips while retaining the underlying temporal, distance, and modal structure of the core survey. The framework provides a probabilistic approach for integrating passive mobility data with conventional HTS datasets to improve travel statistics used in transportation planning and demand modelling.