Although the technical analyses ERG performed during the course of this work are sometimes specific to the Houston and Kansas City metropolitan areas, we believe that the broadest benefit of this study will be in applying lessons learned in transitioning from traditional MOBILE6 modeling and inventory planning inputs to using MOVES effectively. With that in mind, we have included in this report not only analysis of model inputs and outputs, but also documentation of the process of running MOVES, example model runs that users can execute on their own computer systems, inventories they can compare their results against, and analysis of the ramifications of transitioning to the new MOVES system.
In the section that follows, we present important issues dealt with during the course of this project, with an emphasis on how users can deal with these issues themselves. We also discuss the study questions originally proposed in the project Work Plan.
In order to effectively transition from use of MOBILE6 to MOVES in developing regional emissions analyses, users must not only understand how to use MOVES to develop inventories, but also keep in mind several issues arising from characterization of data that is input to both models. We list some of these significant concerns below.
Because VMT is perhaps the single most important variable in calculation of emissions, it is crucial to understand both the process of converting VMT into a format usable by MOVES, and the pros and cons of using various VMT bases. In this study, ERG explored the usage of two separate VMT types. The first was based on VMT derived from a travel demand model (TDM), and the second on VMT summarized in Highway Performance Monitoring System (HPMS) format. Users may have sources of VMT available to them in other formats as well, which may require the development of spreadsheets or other software tools in order to process the VMT for use in modeling.
Difficulties and Sensitivity to Emissions
Running exhaust emissions and evaporative hydrocarbon emissions calculated in MOVES are directly proportional to the VMT input to the model. In addition to translation of raw VMT from MOBILE6 to MOVES format, users must also be careful to properly allocate VMT across various speed, age, and road type profiles to ensure that calculated emissions are truly representative.
In the case of VMT formats used in this study, the travel model-based VMT process generally provides a more detailed VMT basis, but it also requires a greater degree of data manipulation and processing on the part of the user. Use of HPMS-based VMT, meanwhile, requires less upfront data processing, but does require a separate MOVES model run, as described in Section 2.1, to arrive at a default source type distribution required for use in EPA's VMT converter tool. The use of HPMS based VMT also necessitates the independent development of road type distributions appropriate to the modeled area of interest.
Further, the VMT converter tools provided by EPA, while effective in calculating inputs for the MOVES model, are prone to potential user error due to their complexity. Such user errors, if unchecked, can cascade throughout the created model inputs files.
To ensure that VMT is appropriately developed for input to MOVES, we recommend that users take advantage, to the extent possible, of any existing MOBILE6 files (such as age distributions, speed distributions, VMT by hour, and so forth) during VMT pre-processing. Any software developed by users to process VMT and its associated distributions should be extensively quality checked to ensure that VMT transformations are being applied correctly.
Users should also make themselves familiar with the wide variety of converter tools that EPA has developed and made available. Our methodology for using these tools on both the TDM and HPMS VMT basis is discussed above in Section 2.2.
The MOVES model has been built from the ground up to utilize emission factors in units of grams per hour of operation. This is a distinct difference from the grams per mile basis used in MOBILE6. In addition, the MOVES model produces emission estimates for emissions processes that could not be accurately produced by the MOBILE6 model, such as extended idling emissions. In the development of MOVES, EPA transitioned away from the standard MOBILE6 vehicle classes (based on gross vehicle weight rating, or GVWR) and instead characterizes vehicles by source type based on their expected mode of operation. However, the emission factors contained within the model are still primarily aggregated on a vehicle class/weight basis. This requires the MOVES model to map the source types to vehicle classes internally in what is called the Source Bin Generator.
Difficulties and Sensitivity to Emissions
This change in calculation basis presents challenges to users on a number of fronts. The first is proper conversion of VMT and associated distributions. Fortunately, EPA has provided tools and documentation to assist users in transition from MOBILE6 vehicle types to MOVES source types for total VMT, speed distribution, age distribution, and vehicle populations. The second challenge this transition presents to users comes in the form of development of appropriate I/M program inputs. This is discussed in further detail in the next subsection.
The third challenge manifests itself in comparison of model outputs between MOBILE6 and MOVES. In particular, vehicle types HDDV2B through HDDV5 inclusive (along with their gasoline counterparts), considered heavy-duty in MOBILE6, map to MOVES source types 31 (passenger trucks) and 32 (light commercial trucks) in EPA's converter tools and associated guidance. These source types are often considered "light-duty" when analyzing aggregated MOVES outputs. Because of this, difficulties arise when attempting to directly compare MOBILE6 outputs by vehicle type with MOVES outputs by source type.
An example of such a difficulty in presented in Section 5.2, where a first glance at VMT output from the models indicates a disparity between light-duty and heavy duty vehicle activity. This disparity turns out to be a result of the vehicle classification scheme, and not necessarily indicative of any calculation errors in the model.
Because the vehicle/source type characterization issue is present on a number of fronts when transitioning from MOBILE6 to MOVES, users will do well to keep it in mind at all points of the process: in preparing and converting model inputs, in model execution, and in comparison and analysis of model outputs. Proper use of EPA's provided converter tools, as well as the methodology for model inputs development presented in EPA's Technical Guidance, will be helpful in developing VMT by source type.
If users elect to model areas affected by local Inspection and Maintenance (I/M) programs, they need to take care to ensure that not only is the I/M representative of the area being modeled, but also that inputs are formatted properly, and conversions from MOBILE6 vehicle to MOVES source types, as alluded to above, are applied in the correct manner (if appropriate).
Difficulties and Sensitivity to Emissions
I/M program information input to the model will reduce calculated model emissions by varying degrees, depending on the type of program implemented, model year coverage, and other factors. If users elect to utilize MOVES I/M defaults for the area of interest, they should first be aware that not all such defaults are accurate, or even present at all, in the model. At the very least, users should carefully examine the defaults and make changes to the I/M program as appropriate.
Secondly, as mentioned above, one of the challenges of applying I/M programs in MOVES is that MOBILE6 vehicle types do not map precisely with MOVES source types. In particular, MOVES source types 31 and 32 can encompass MOBILE6 source types LDGT1 all the way up to HDGV5. As stated in Section 3.10.6 of the Technical Guidance, "Inspection and Maintenance (I/M) programs entered in MOVES can only be applied to source types. However, as discussed in section 3.10.2, this association of I/M programs and source type may be inconsistent with state I/M program regulations that define I/M programs by the vehicle weight classes. MOVES source types are composed of several vehicle weight classes and, therefore, applying I/M benefits to the entire MOVES source type may be inappropriate." Users should be aware that they may run into similar difficulties when trying to apply existing I/M programs information across source types in MOVES.
When making changes to an existing I/M program, users need to be aware that this is a tricky process in which errors can easily be introduced. In particular, it is critical to change the useIMyn field to N for existing records in the IM table, and alter the IMProgramID field for any new or updated records. ERG's methodology for preparation of I/M inputs was discussed previously in Section 2.3.
With respect to proper application of I/M programs across MOVES source types, EPA recommends in section 3.10.6 of the Technical Guidance the calculation of a compliance factor based not only on compliance rate and waiver rate (which both provided in MOBILE6 inputs), but also on regulatory class coverage. Calculating the regulatory class coverage is now fairly straightforward, as the information in Table A.3 of the April 2010 Technical Guidance Appendix (Gasoline I/M Regulatory Coverage Adjustments) provides the proper conversion factors, along with example calculations.
There are a couple of methods for obtaining vehicle population data for input to MOVES, which is used in calculation of vehicle start and extended idle emission components in the model. We ultimately used by-county registration data provided by the Texas Department of Transportation, but if such data is not available, users can go through the process of developing populations surrogates based on VMT, as described in detail in section 2.3 of this report, as well as the EPA Technical Guidance.
Difficulties and Sensitivity to Emissions
Vehicle start emissions and extended idle emissions calculated in MOVES are directly proportional to the vehicle population input to the model. Although usage of registration data to develop population is suggested by EPA in the Technical Guidance, users should understand that the vehicles registered in a county for a given point in time do not necessarily correspond to the VMT driven over that same period in the county. Using the counties we modeled as an example, commuters living in Fort Bend county may drive a significant portion of their vehicle miles in Harris county. In this case, using registration data as a population surrogate may lead to overestimation of start and evaporative emissions in Fort Bend county, while underestimating those same emissions in Harris County.
While a better alternative for determining representative source populations does not currently exist, this is an important issue for users to be aware of. Ultimately, users must be confident that whatever population data they choose to input to the model is accurate.
For this study, ERG used the first non-draft version of the MOVES model (MOVES2010) released by EPA in late December 2009. The EPA has released subsequent versions of the model since that time, and will continue to update and improve the model going forward into the future. However, in the course of their modeling efforts, users may, like we did, discover discrepancies in model outputs that only be explained by problems in MOVES.
While using MOVES2010, ERG found two bugs in model calculations during the course of this study, both mentioned previously in Section 2.3. One of these bugs was related to fuel type IDs and associated oxygenate volume percentages. The other bug we found has been documented in the latest version of the MOVES Errata, and involved user creation of new fuel formulation IDs for input to MOVES. When a user creates and imports new fuel formulation IDs, MOVES does not currently "zero out" the market share of existing fuel formulation IDs.
Difficulties and Sensitivity to Emissions
It is difficult to speculate on how general calculation errors in the model might effect emissions output. In fact, some bugs that users come across may not have an effect on emissions at all, but rather relate to user interface difficulties, Java execution errors, or other issues. In our case, the first bug we found (related to fuel type IDs) led to underestimation of VOC emissions, underestimating them by approximately an order of magnitude 30 The second bug we encountered, involving user creation of fuel formulation IDs, led to a pollutant overestimation of approximately two to four times in model calculations.
Careful examination of outputs in each case led us to contact EPA staff, who provided us with temporary workarounds for each bug while adjustments were made to the model itself. Other potential issues are explained in the currently available version of the MOVES Errata 31. Users should carefully examine these Errata before performing a MOVES analysis, and be ready to report to EPA any potential issues identified in their own outputs for correction in future versions of the model.
In this section we attempt to answer the questions originally proposed in the Work Plan submitted at the beginning of this study. These questions included the following:
What is the best approach for creating an emissions inventory with MOVES based on activity data obtained from a travel demand model (TDM)?
We believe the best approach for developing a MOVES emissions inventory using TDM activity data, as described in detail in Sections 2 and 3, involves a number of steps. First, methodical preparation of TDM VMT data, consistent with EPA's Technical Guidance, must be undertaken, in which VMT, VHT, and associate activity distributions are derived with an eye on inputs to EPA's multiple Converter Tools. Second, these Converter Tools must be properly utilized to arrive at inputs that can directly input into the MOVES County Data Importer itself. Third, analysis and processing of other MOVES inputs, including vehicle populations, I/M program data, fuel characterization, and meteorological data should be appropriately developed. Fourth, execution of MOVES with appropriate model options is required. Fifth, care must be taken in processing of model output, whether using the MOVES GUI or external SQL queries, to ensure emissions summaries are representative. Finally, at all stages of the modeling process, effective QC and QA procedures should be developed to minimize errors and ensure accurate production of model outputs.
What is the best approach for creating an emissions inventory with MOVES based on activity data obtained from the Highway Performance Monitoring System (HPMS)?
The best approach for developing a MOVES emission inventory based on HPMS activity data is fairly similar to that described above for a TDM activity basis. It differs in that the HPMS VMT data requires a different type of pre-processing prior to integration with EPA's Converter Tools; all other modeling steps are the same. Effective pre-processing of HPMS data involves application of default MOVES source type distribution to the data, as described above in Section 2.1. It also necessitates the development of road type distributions outside of the EPA Converter Tools.
What will be the likely impacts on an emission inventory from developing and implementing different drive cycles derived from real world testing data within MOVES?
In this study, the effects of implementing drive cycles in MOVES, based on real world testing data collected in Kansas City, varied considerably depending on which county was being modeled. The drive cycles we developed were based on largely urban driving data as discussed in Section 4, and applying these drive cycles to more rural counties in the Houston-Galveston area (as exhibited by fewer VMT driven in those counties) produced emissions that may not be representative. On the other hand, emissions from Harris County, the most urban of the counties modeled (with the corresponding largest daily VMT modeled), did not vary much at all. We believe that any alternate drive cycles developed and used in MOVES should closely match the type of driving activity for the area being modeled for emissions to be considered representative.
Apart from the impact of altering the drive cycles themselves, it is worthwhile to note that the effort required to produce representative drive cycle based on real world driving data is substantial, and may be prohibitive for some users. For this reason, we anticipate that most users will elect to use the default drive cycles provided as part of MOVES for their analyses.
What will be the likely impacts on an emission inventory after conversion from a MOBILE6 basis to a MOVES basis?
As presented in Section 5, this study shows that users can expect a considerable increase in modeled NOx emissions when transitioning from MOBILE6 to MOVES. This is consistent with other known comparisons of modeled MOBILE6 and MOVES results. 32 33 Our study also shows smaller increases of both CO and VOC when transitioning from MOBILE6 to MOVES. Other studies that modeled these pollutants have had mixed results on this point (TTI modeled decreased CO and increase VOC, while EPA has found decreased VOC results). Beyond the expected changes in emissions, users should be aware that the level of effort and computing time required to successfully execute and process a MOVES model run relative to MOBILE6 has increased substantially.
30 MOVES tables of interest related to this bug include etohbin and fuelsubtypeID.
32 TTI's Production of MOVES On-Road Mobile, Link-Based Emissions Estimates and Document Preparation Technical Note, submitted to TCEQ in July 2010.
33 Update on EPA's Motor Vehicle Emissions Simulator MOVES2010, presented at CRC by J. Koupal in March 2010