Existing studies have indicated that the vehicle trajectories derived from Newell's car-following model (NCM) fail to capture driving behavior heterogeneity, resulting in considerable emission estimation errors. This study investigated the situation-dependent heterogeneity of car-following behavior, based on field vehicle trajectories in Beijing, and proposed a multidimensional stochastic Newell car-following model (MSNCM) incorporating three stochastic parameters: random response time, speed-dependent critical jam spacing, and speed difference- and spacing-dependent acceleration. The comparison between the field data and numerical simulations of the NCM and MSNCM shown that the MSNCM performed well in generating realistic vehicle trajectories for emission estimation. The relative errors of the emission factors derived from the field and the MSNCM simulated trajectories were 0.26%, 0.91%, 1.37%, and 0.25% for CO₂, CO, HC, and NOx, respectively, which represented reductions of approximately 15%-46% compared with the traditional NCM.