The Transportation & Air Quality Research Group TRAQ Modelling GHG Emissions in the GTHA Using Link-Based Operating Mode Distributions as a Proxy for Driving Behaviour An Wang, Christos Stogios, Yijun Gai, James Vaughan, Gozde Ozonder, SeungJae Lee, Daniel Posen , Eric J. Miller, Marianne Hatzopoulou
Research Objectives Generate a lifecycle emission inventory for transportation in the GTHA Incorporate “uncertainty” within emission inventory Quantify the effects of electric vehicles on total emissions 2
Methods and data 3
Methods and data Based on MOVES and GREET with local inputs Vehicle registration data from MTO Various scenarios for electricity generation in Ontario Hybrid approach using distributions of emission factors derived from AIMSUN and applied within EMME 4
Methods and data Based on MOVES and GREET with local inputs Vehicle registration data from MTO Various scenarios for electricity generation in Ontario Hybrid approach using distributions of emission factors derived from AIMSUN and applied within EMME 5
MTO vehicle composition (2016)
Methods and data Based on MOVES and GREET with local inputs Vehicle registration data from MTO Various scenarios for electricity generation in Ontario Hybrid approach using distributions of emission factors derived from AIMSUN and applied within EMME 7
Electricity generation scenarios Ontario electricity generation mix obtained from the Independent Electricity System Operator (IESO) Four electricity generation scenarios: 1. Current Ontario mix: 61% nuclear, 23.7% hydro, 8.4% gas/oil, 6.2% wind, 0.3% biofuel, 0.3% solar 2. All fossil mix: 100% natural gas 3. Only dispatchable source mix: 73% hydro, 26% gas/oil, and 1% biofuel 4. Solar and wind mix: 95.3% wind and 4.7% solar 8
Methods and data Based on MOVES and GREET with local inputs Vehicle registration data from MTO Various scenarios for electricity generation in Ontario Hybrid approach using distributions of emission factors derived from AIMSUN and applied within EMME 9
Two types of EFs in MOVES Motivation for hybrid approach Two types of EFs in MOVES By gram per unit time Based on single operating mode; Requires instantaneous speed to estimate emissions Based on default driving cycles (built-in operating mode distribution); Only requires average speed to estimate emissions Various drive cycles can occur at the same average speed By gram per unit distance A distribution of emission factors for each average speed makes more sense 10
Proposed approach 11 Select n random roads to run microsimulation Sample of roads includes arterials, ramps and freeways For link i in each road type, generate an operating mode distribution and calculate total emissions 𝐸 𝑖 based on instantaneous speeds Generate EF of link i: 𝐸𝐹 𝑖 = 𝐸 𝑖 𝑉𝑀𝑇 𝑖 , 𝑉𝑀𝑇 𝑖 refers to the vehicle miles travelled on link i Generate distributions of EFs by average speed bins Test area: City of Toronto With instantaneous speed: Link-level total emission; Link-level EF; Link average speed; 11
AM-Peak Distribution of EFs 12
PM-Peak Distribution of EFs No freeway data 13
Midday Distribution of EFs No freeway data 14
Evening Distribution of EFs No freeway data 15
Results 16
Daily GHG Emissions in GTHA 95% of total emissions 17
Comparison with Public Transit Green bar illustrates the range of private vehicle emission intensities Colored boxes illustrate the ranges of public transit emission intensities 18
Spatial distribution of emissions AM Midday 19
Daily Fuel-Cycle Stochastic Inventory From single number to distribution Daily Fuel-Cycle Emissions (gram) 20
Scenario 1: Current electricity mix Base case Scenario 1: Current electricity mix 21
Base case Scenario 2: Natural Gas 22
Scenario 3: Dispatchable sources Base case Scenario 3: Dispatchable sources 23
Base case Scenario 4: Renewables 24
BUT: Deep reductions in traffic emissions may not be achieved without large investments in public transit Emission intensities for electric vehicles are: 12.8, 130.1, 37.2, and 2.8 g CO2eq/km for passenger cars 14.6, 149, 42.6, and 3.2 g CO2eq/km for passenger trucks w with electricity mix 1 to 4 respectively Current Ontario mix All fossil mix: 100% natural gas Only dispatchable source mix: 73% hydro, 26% gas/oil, and 1% biofuel Solar and wind mix: 95.3% wind and 4.7% solar Emission intensity for transit (right now with mostly transit buses which are diesel fuelled) is 18.5 g/PKT (daily average) 25
Further improvements 26
Limitations of previous approach Distributions of emissions that we developed for each average speed should be refined Std. Dev. of speed, delay, other measures? There is no need to group links by average speed, why not a cluster approach? 27
28 First step: hybrid simulation Hybrid simulation is composed of two parts: full network mesoscopic simulation and sample area microsimulation. (see next page) 28
Full network mesoscopic + sample microsimulation 29
30 Clustering is based on meso traffic variables: average speed, standard deviation of speed, average delay, standard deviation of delay, Road capacity Segment density Vehicle kilometers travelled After finishing hybrid simulation, two parts of data will be used in the learning step: (see next page) 30
Clustering result: This shows the cluster result. 31
The cross validation contains random process, so here in CLEVER, cross-validation will be repeated for 100 times. (next page) 32
Distributions of EFs of each cluster 33
34 In full network, mesoscopic data are available. Assign a cluster to each segment in the full network Apply representative EF and estimate total emissions 34
Downtown Toronto GHG estimation: Hybrid VS microscopic model 67.4 - 74.3 tons Microscopic estimation: 70.06 tons Mesoscopic estimation: 49.04 tons 35