Transportation emissions forecasting

[eroten]

Close compare county level emissions, VMT from different sources over time, including

• StreetLight
• MnDOT/WisDOT
• EPA NEI
• EPA LGGIT
• EPA MOVES

[eroten]

MOVES Q&A notes

Met with Dennis and Charles G. 6/20/24

MOVES for the TPP is an inventory run of what the total emissions will be in year ___, they do an Inventory run.

However, with a Rates run, you running (for all the inputs we have), give us the rates/factor without having to re-run the entire MOVES model. Rates give emissions by speed bins by link type. Then apply that rate to the Volume on those links. Speeds are involved, so they are speed specific. Setting up a Rates run is an exercise in itself - it takes configuration and evaluation.

One thing that could cause uncertainty is how Mark Filipi got the numbers and are we able to recreate them/how much documentation we have on how it was made.

EPA NEI differences:

• One difference might be StreetLight data in and of itself - we don't have super great documentation on StL.
• NEI is using HPMS data, whereas it seems like StreetLight is HPMS (HPMS has its own issues/assumptions regarding tying trip length with COUNTS)
• Compare VMT by functional class coming from HPMS with VMT by functional class from the models. Look at lane miles and come up with a conversion factor. Model does better on all functional classes (model network is based on OSM). New network requires reconciliation and establish a base network before comparing.
• Discuss with Jim Henrickson at MnDOT to get more detail on HPMS and local count data.

Isolating on Hennepin County

• Compare NEI VMT and our VMT estimates, MnDOT VMT estimates. What is the base activity dataset used in the EPA NEI?

Also compare state emissions dataset to national emissions datasets

The energy vs. transportation emissions issue is ongoing discussion.

NEXT:

• Compare model outputs with trip lengths, VMT, all the things
• What is the data source for vehicle volume? AADT/VMT estimates, or more direct values from loop detectors?
• Vehicle classification breakdowns
• Geography aggregation levels available
• Data pieces needed to complete a MOVES run
• What kind of/does MOVES include options for emissions mitigation strategies? What are they?
• How do we run our local instance of MOVES? Where do the outputs live? Can I access them?

[eroten]

Meeting 2, 6/28/24

Met with Charles G and Dennis.

• What is the data source for vehicle volume? AADT/VMT estimates, or more direct values from loop detectors?
  • VMT and speed data come out of the model (right now TourCast, will be ActivitySim)
  • TourCast - speed and volume are outputs from model assignment. It generates trips at the TAZ level, then assigned to networks, then comes up with condition related speeds, travel time, down to the links
  • the network is an abstraction of the highway system represented by nodes and links. Each intersection is a node, nodes are connected by links.
  • It includes some local roads, but not all local streets. It would include all the interstates, higher level facilities (arterials, collectors), and local streets. Abstracting local roads to TAZ. centroid connectors connect zone to the network. All trips coming out of each TAZ use a connector to get onto the regional network. Centroid centers represent locals.
  • Generalized network optimization algorithms (shortest path, least cost, etc), iterative. It also will divert traffic based on volume and road capacity. Will continue distributing out until no single path has an advantage over another (by travel time) -> convergence achieved. From the model, you get travel times, volume, and speed (distance by time) for each link. This is all modeling an average weekday across 11 time periods (AM Peak, Midday, PM Peak, Evening/Overnight).
• The inputs to MOVES are hourly. They ask for the COUNTY average speeds for different facility types for each of the 24 hours. To go from TourCast to MOVES, assign the average speeds of the time period to each of the hours.
• Speeds from links are weighted average (speed weighted by volume/vehicle hours traveled)

Outputs from TourCast (county level)

• Travel times on highway networks at county level
• Volumes on highway networks county level
• Speeds on highway networks county level

TourCast estimates the number of people who are likely to take bike and ped trips, but we don't do anything with that data. 1-2% biking, 2-3% transit. It does show up, but it is not very granular and we don't know under what scenario those would need to change to make a difference in MOVES outputs. TourCast is doing a behavioral/logistic regression to find the percent of people who will walk/bike/transit. TAZs can be very large, and its difficult to see how small changes in pedestrian/bike infrastructure would effect things. Transit infrastructure will show up more.

MOVES does forecast forward, mostly based on speeds and volumes/vehicle miles traveled (which come from travel demand model, with different scenarios). The TourCast also forecasts the population of private vehicles, but the distribution of vehicle types (EV, light duty, small cars) is NOT there.

Items to compare

• Volume

Questions

• Can MOVES go down to the city level? Not really, lots of defaults only available at the county level that we use. The defaults can be used for city level, but get tricky when cities straddle multiple counties. If you were to just scale down the county level to the city level, there would need to be some activity measure to scale it down (not area). MOVES is tabular, some people even use MOVES for project level.
• For Minneapolis, you would need vehicle population/registration data, fuel usage

NEXT:

• Compare model outputs with trip lengths, VMT, all the things
• What is the data source for vehicle volume? AADT/VMT estimates, or more direct values from loop detectors?
• Vehicle classification breakdowns
• Geography aggregation levels available
• Data pieces needed to complete a MOVES run
• What kind of/does MOVES include options for emissions mitigation strategies? What are they?
• How do we run our local instance of MOVES? Where do the outputs live? Can I access them? A: J:/ drive. Liz has requested access from IS

TODO

• Update EPA NEI comparison section to clarify scopes and show a correlation instead of a direct comparison

MOVES4CheatsheetOnroad.pdf

[eroten]

Forecasting methods

Essentially, we are going to use the forecasted VMT from the 2050 TPP model run. Whether we calculate emissions directly from MOVES or some other way, the VMT values will be there

[eroten]

Meeting 3, 7/3/24

• R scripts take the model output network and find speeds -> Liz will create a GitHub repo with these scripts for tinkering/ongoing development. https://github.com/Metropolitan-Council/tpp_2050_scripts
• In the network, each link gets tagged as urban, rural, (off-network indicates construction, non-roads). Restricted access = highways, because some vehicles cannot get on highways
• we are still using vehicle registration data from 2015. MPCA is still working on converting the registration data to a MOVES-acceptable format.
• We are using MOVES4 in the TPP 2050 model
• we are using defaults for meteorology, fuel,
• we are also using speed distributions, road type,
• we can use a batch file to run multiple counties at once,

[eroten]

Meeting 4, 1/28/25

Met with Dennis, Charles, Rachel

• best way to get links to CTUs, clip the links to CTU
• if links are split, then still use the same VOLUME, but link length will be shorten
• multiply volume by link length to get VMT
• that way we will be able to separate out NHS, with assigned group. Make local roads assigned to CTU, but potentially pull out NHS roads

Are model links the same as OSM

• mostly, but there have been some edits adding in projects, process to convert was not flawless, there are some sections that don't have coverage that were built up by hand
• Shared streets ID, more complicated version, but our project cards work from link ID. it was used in conversion from OSM to model network (Network Wrangler). Number of lanes was challenging, and the data is not perfect, but should be okay for this exercise.
• Network that travel model uses is a straight-line network. Distances come off of true shape with a GIS shapefile. This will come in with edge cases where links in the loaded network for clipping with CTUs.
• Within a TAZ, population is assumed to be in the centroid of the TAZ. But some local streets will have no volume because there is only a few centroid connectors. They also assume that destinations are all in the centroids. Path of travel is normal in between, but there may be places with no volume. Some suburban areas look weird (ie Woodbury)
• With VMT for MOVES, there is a conversion process to go from links to local streets, assigns some centroid proportion to local streets. When we do the computations, remove centroid connectors (no capacity constraints). We should look at lowest functional class of local street.
• at the CTU level, the inconsistencies in the network should be okay. For every edge case, there might be another that compensates for it 🤷. Bias is mostly random, but look out for road type inconsistencies.
• the work of assigning who is responsible for roads/intersections is at the public works level

Practicality, getting hands on data

• one hour time periods for AM and Pm
• use shapefiles from link provided by Charles
• Liz will create a GitHub based repo that pulls the links and clips them to the CTU level (potentially utilizing some of the r-arcgis packages)
• also need to Compare the MnDOT VMT by CTU with the model
• Networks vary spatially based on time period because of managed lanes!
• Charles will send more metadata on column names, dataset specifications
• Liz will send charles MnDOT VMT at CTU level

[eroten]

Meeting 5, 2/10/2025

• questions about how those will relate to functional class
  • getting from assign group to functional class is harder
  • this would require some mapping, and conflation
  • would be good to do some mapping and compare to mndot assessments
  • will also compare model network coverage to MnDOT network coverage
  • comparing with ghg.sp dataset, they probably used assumptions to get down to transit, walk, CTU level VMT data.