Artificial neural networks (ANNs) are useful alternative techniques in modelling the complex vehicular exhaust emission (VEE) dispersion phenomena. This paper describes a step-by-step procedure to model the nitrogen dioxide (NO₂) dispersion phenomena using the ANN technique. The ANN-based NO₂ models are developed at two air-quality-control regions (AQCRs), one, representing, a traffic intersection (AQCR1) and the other, an arterial road (AQCR2) in the Delhi city. The models are unique in the sense that they are developed for ‘heterogeneous¹’ traffic conditions and tropical meteorology. The inputs to the model consist of 10 meteorological and 6 traffic characteristic variables. Two-year data, from 1 January 1997 to 31 December 1998 has been used for model training and data from 1 January to 31 December 1999, for model testing and evaluation purposes. The results show satisfactory performance of the ANN-based NO₂ models on the evaluation data set at both the AQCRs (d = 0.76 for AQCR1, and d = 0. 59 for AQCR2).