01367691

Component

This paper has empirically assessed spatial transferability of automobile-specific carbon dioxide (CO2) emissions model and evaluated the prediction performance of four transfer methods. The transfer methods evaluated include Naïve, Bayesian Updating, Joint Context Estimation, and Combined Transfer Estimator. Furthermore, the study investigated the effect of sample size on model’s transferability. The study used the 2009 National Household Survey (NHTS) conducted by the U.S. Department of Transportation. In this application, San Antonio region data was used in conjunction with Austin-San Marcos data to predict the observed CO2 emissions in Austin-San Marcos in Texas. Overall, the results show that CO2 emissions model could be transferred from one region to another region and yield reasonable prediction performance for urban cities with comparable characteristics. The Combined Transfer Estimator method gave superior predictive performance measures than other methods. Practically, Combined Transfer Estimator is rather intractable to implement because it requires additional steps for computing covariance matrices and transfer bias. In light of this, Joint Context Estimation method should be used in lieu of Combined Transfer Estimator because of its simplicity and produces reasonable prediction performance. The results also highlight the potential and importance of regions with smaller sample size using data from another region in formulating CO2 emissions models for evaluating different transportation investments. In turn, this would minimize a need for large data collection efforts. These results, however, may not produce robust results when applied to region-pairs that are somewhat disparate, or for regions where there is a higher proportion of non-automobile travel.

This paper was sponsored by TRB committee ADC70 Transportation Energy

01362476

12-4251

English

23p

2012

Transportation Research Board 91st Annual Meeting

Digital/other

Figures; References; Tables

Automobile travel; Carbon dioxide; Exhaust gases; Mathematical prediction; Spatial analysis; Urban areas

Spatial transferability

Austin (Texas); San Antonio (Texas)

Energy; Environment; Highways; Planning and Forecasting; I15: Environment

Transportation Research Board Annual Meeting 2012 Paper #12-4251

TRIS, TRB

Feb 8 2012 5:23PM