Modeling air pollutant emissions from Indian auto-rickshaws: Model development and implications for fleet emission rate estimates

Abstract

Abstract Chassis dynamometer tests were conducted on 40 Indian auto-rickshaws with 3 different fuel–engine combinations operating on the Indian Drive Cycle (IDC). Second-by-second (1 Hz) data were collected and used to develop velocity-acceleration look-up table models for fuel consumption and emissions of CO 2 , CO, total hydrocarbons (THC), oxides of nitrogen (NO x ) and fine particulate matter (PM2.5) for each fuel–engine combination. Models were constructed based on group-average vehicle activity and emissions data in order to represent the performance of a ‘typical’ vehicle. The models accurately estimated full-cycle emissions for most species, though pollutants with more variable emission rates (e.g., PM 2.5 ) were associated with larger errors. Vehicle emissions data showed large variability for single vehicles (‘intra-vehicle variability’) and within the test group (‘inter-vehicle variability’), complicating the development of a single model to represent a vehicle population. To evaluate the impact of this variability, sensitivity analyses were conducted using vehicle activity data other than the IDC as model input. Inter-vehicle variability dominated the uncertainty in vehicle emission modeling. ‘Leave-one-out’ analyses indicated that the model outputs were relatively insensitive to the specific sample of vehicles and that the vehicle samples were likely a reasonable representation of the Delhi fleet. Intra-vehicle variability in emissions was also substantial, though had a relatively minor impact on model performance. The models were used to assess whether the IDC, used for emission factor development in India, accurately represents emissions from on-road driving. Modeling based on Global Positioning System (GPS) activity data from real-world auto-rickshaws suggests that, relative to on-road vehicles in Delhi, the IDC systematically under-estimates fuel use and emissions; real-word auto-rickshaws consume 15% more fuel and emit 49% more THC and 16% more PM 2.5 . The models developed in this study can be used to further explore the impact of varying vehicle activity patterns on emissions in efforts to manage air quality and mitigate air pollution exposure and air pollution related health impacts.