| Emission factors are fundamental measures for developing an emission inventory, that supports air quality management decisions and help to design control strategies. Motor Vehicle Emission Simulator (MOVES) by US Environmental Protection Agency (EPA) is a well-known tool which defines 23 operating mode identification bins (OMID) based on the vehicle specific power (VSP), speed, acceleration and idling state. As vehicular emissions have been considered a complex function of several factors, this study has developed statistical models to quantify the impact of pavement roughness on emission indexes of CO2, CO, HC, NOx, and Fuel Consumption (FC), by approaching the MOVES (OMID) binning method. Real road driving test has been conducted at Austin, El Paso, Houston, and San Antonio in Texas. A Portable Emission Measurement System (PEMS) was used to measure the emissions from a dedicated test vehicle's vent-pipe, and the corresponding pavement roughness data (IRI) were collected by Roadroid app, covering almost 1,068 miles of Texas roadways. The matched emission and IRI data were classified into 22 OMID bins (except bin 1 for idling). These paired data were employed to develop two emission models: 1) Polynomial Regression Model and 2) Random Trees model for each OMID bin. Several indexes have been considered to evaluate model results.;The modeled output from the Polynomial Regression Model showed a uniform nonlinear relationship with higher value of emission factors in 100% OMID were associated with the lower range of IRI, like the observed data. The Co-efficient of determination (R2) value for 80% of the bins were greater than 0.50 and the normalized root-mean square value ranged from 1- 16.66%, which quantify a quite good fit with the observed value. Later a non-linear model of random forest has been statistically showed that almost 62% of OMID has predicted IRI as the most important predictor in emission factors estimation. The model prediction also showed a good fit with the input data as the validation errors ranged between 1-5% for all indexes. With the statistical results of this study, agencies can consider the emission factors associated with pavement roughness to improve their control strategies and more accurate emission inventory. |
