Uncertainty Analysis On Vehicle Specific Power Distributions For Emission Estimates At Link Level

Traffic emissions have become one of the main sources of the air pollutions in urban areas.In recent years,policy-makers concerning emission reduction raise two new requests for traffic emission modeling:(1)traffic emissions are included in the total emission inventories in order to make policies from the planning level,and thus,traffic emissions need to be coupled with other emission sources both spatially and temporally.(2)Results of emission estimates are sensitive to traffic condition parameters,in order to quantify correctly the performance of microscopic emission reduction programs related to traffic management and control.Traditional emission models,like the MOBILE,use driving cycles to represent on-road vehicle activity for emission estimates,and thus,is not applicable any more.In 2009,the US Environmental Protection Agency released the new traffic ermission model,the Motor Vehicle Emissions Simulator(MOVES)and superseded the MOBILE.MOVES uses the Vehicle Specific Power(VSP)distributions to represent on-road vehicle activity for emission estimates at the link level for high time resolution.However,it is expensive and time consuming to collect massive vehicle trajectory data for developing real-world VSP distributions.Thus,in the default case,MOVES still uses MOBILE driving cycles for developing VSP distributions and thus,it increases significant uncertainties in the results of emission estimates.Over the past couple of years,Beijing collected hundreds of millions of vehicle trajectory data points for light duty vehicles(LDVs),buses,and trucks and developed a comprehensive database of VSP distribution maps in order to model on-road vehicle activity for emission estimates.This study addresses a couple of crucial issues raised during the development of the VSP distribution maps for LDVs and proposes some comments in order to solve these problems.The conclusions are helpful for the development,application and optimization of the traffic emission models which are based on the VSP distributions.The methodologies used in this study are important references in the research field of uncertainty analysis on traffic emissions.Main findings are summarized in six aspects:(1)There is a one-to-.one mapping relationship between the vehicle type-,facility-and speed-specific VSP distribution and the on-road vehicle average activity.Thus,it can be achieved to employ the VSP distributions,link speeds and traffic volumes for high-resolution traffic emission estimates.(2)The VSP distributions are convergent as traffic volume increases.The results are unbiased to employ the VSP distributions for real-world traffic emission estimates.The confidence upper limits of the emission estimated errors decrease as traffic volume increases and their relationship can be described by a power function.(3)The VSP distributions are consistent both temporally and spatially.The emission estimated errors related to the temporal and spatial differences of the VSP distributions are less than 3%.(4)The VSP distributions for professional driving behaviors and non-professional driving behaviors are significantly different.Thus,this study proposes to develop the VSP distributions for professional and non-professional driving behaviors separately,which can decrease emission estimated errors by approximately 13%.(5)The VSP distributions for intercity expressways and urban expressways ares are ignificantly different,and the VSP distributions for arterial roads and local streets are also significantly different.Hence,for emission model structure,this study proposes to employ the road types of"intercity expressways/urban expressways/arterial/local" instead of"restricted access/unrestricted access",which decreases emission estimated errors by approximately 14%.(6)For real-world link emission estimates,traffic volume,substituting the VSP distributions for professional driving behaviors for the VSP distributions for non-professional driving behaviors,and employing the road types of"restricted access/unrestricted access" all increase uncertainties significantly.In the case study,they increase emission estimated errors by 37%,5%and 7%,respectively.