MOVES-Matrix and Distributed Computing For Region-Level Line Source Dispersion Analysis

MOVES-MATRIX MODELING TOOLS

Investigators:

Dr. Randall L. Guensler, Georgia Tech     

Dr. Michael O. Rodgers, Georgia Tech

Haobing Liu, Ph.D. Student, Georgia Tech

Daejin Kim, Ph.D. Student, Georgia Tech

Project Overview:

This dissertation research is building build a framework for regional-level microscale dispersion analysis using MOVES-Matrix, AERMOD, and distributed computing. Liu, et al. (2017) introduced an advanced framework for line source modeling, resulting in a huge savings in computing cost compared to traditional methods. However, this previous research ran into computational limitations due to the limited number of links allowed in CALINE4 for a single run. As California Department of Transportation suggested, it is possible to run CALINE4 repeatedly to cover the entire road network; however, combining the results from individual results is challenging. Although there are no limitations on running AERMOD with respect to the number of links in the roadway network, as network size increases, AERMOD run times increase on-linearly. While it is also possible to perform multiple AERMOD runs with relatively small-size networks (similar to the CALINE4 approach), dividing the entire network for use in AERMOD while minimizing the running is challenging. In addition, the preparing of the extensive model inputs required for regional-level analysis is a huge effort (e.g. road geometry information, locating receptors, etc.).

This research proposes advanced techniques to efficiently prepare the extensive input datasets needed for line source modeling over an entire metropolitan area and will address the technical issues associated with expanding the line source dispersion model from project-level to regional-level. For example, dispersion model simulation time increases exponentially with the number of roadway links and receptors in the simulation. Therefore, setting an appropriate number of receptors is required. In addition, estimated pollutant concentrations are determined by geographical elements such as the distance between the roadway links and receptors. Hence, the number of receptors, placement of receptors, and the density of receptors in a gridded approach (used to develop pollutant concentration profiles across an area) impact model accuracy as well as model run time. In this context, most of previous research conducting regional-scale and sub-regional scale pollutant dispersion analysis (e.g., Zhai, et al., 2019; Wu, 2018; Zhai, et al., 2016; D’Onofrio, et al. 2016; Guensler, et al., 2008; Kall, et al., 2008; Guensler, et al., 2000) considered gridded-receptors with fairly low resolution for metropolitan scale analysis (e.g., at 200 meter by 200 meter resolution). Previous regional work has been useful to identify overall trends in pollutant concentration levels across an area; however, the resultant pollutant concentration results are necessarily biased, to some degree, depending on the geographical locations of link sources and receptors. For example, gridded-receptors placed without consideration of the geographical locations of nearby roadway links tend to be greatly influenced by a specific roadway link in proximity to the receptors. As such, setting proper locations and resolution of receptors is important to produce elaborate pollutant concentration profile. In this context, this research will propose a methodology to determine proper receptor locations and resolution, considering the model precision and available computational resources.

As discussed above, processing dispersion analysis for a large-scaled area requires a huge amount of processing time and computational resources. The preliminary analysis conducted for the metropolitan Atlanta area indicates that AERMOD running time for processing 203K roadway links and a single receptor takes about 12 hours on average with a regular single desktop computer, which would require 206,000 days (564 years) to process 412K receptors that are placed to cover the whole Atlanta area at 200 meter by 200 meter resolution. To address computational run-time, a previous Georgia Tech study (Guensler, et al., 2008; Shafi; 2008) suggested a methodology to screen roadway link sources from the analyses that have an insignificant contribution to a pollutant concentrations at a particular receptor, based upon mass flux from the roadway and distance to receptor. The proposed methodology produced a significant reduction in computational resources and processing time for a complex project with numerous links, but the research team never conducted field-verification tests to confirm that screening method did not introduce any downward bias in modeling results. Similar screening tools need to be developed for the latest USEPA-approved dispersion models (e.g., AERMOD), assessed for sensitivity, and field-verified using pollutant concentration measurements. Building upon the methodology proposed by Shafi, G. (2008), this research proposes to develop an advanced modeling technique to classify roadway links for use in regional-level dispersion analysis and screen links within the distributed computing environment to optimize regional model run times without negatively impacting model results.

Lastly, this research will consider the impacts of dynamic traffic operations and road grade on pollutant concentrations, rarely considered in previous studies, particularly at the regional-scale. Previous research conducted by Liu (2018) yielded a methodology to generate road grade data using USGS digital elevation model and integrate road grade impacts in MOVES emission rate and AERMOD dispersion modeling (Liu, et al., submitted; Liu, et al., 2019, Liu, et al., 2018). Hence, high-resolution road grade is available for the entire metropolitan area for this research (as well as emission rates that account for the impacts of grade). In addition, the developed modeling framework based on Georgia Tech’s PACE (partnership for an Advanced Computing Environment) computing cluster will allow the research to assess the variability in model-predicted concentration levels across a variety of traffic operations and grades. Once the analysis tools are complete, case studies for three large-scale transportation projects (Atlanta’s I-85 HOV to HOT conversion, the 2018 Northwest Corridor Express Lanes, and the proposed Truck-only Express Lanes) will assess the performance of the model and the impacts of these projects on pollutant concentrations under diverse conditions.

  1. D’Onofrio, D., B. Kim, Y. Kim, and K. Kim. Atlanta Roadside Emissions Exposure Study-Methodology & Project Overview. Atlanta Regional Commission, 2013.
  2. Guensler, R., V. Pandey, D. Kall, G. Shafi, P. Blaiklock, M.O. Rodgers, and M. Hunter. MOBILE-Matrix and CALINE-Grid: Project-Level Conformity Screening and Microscale Air Quality Impact Assessment Tools. Prepared for the Georgia Department of Transportation, Atlanta, GA. Georgia Institute of Technology. Atlanta, GA. June 2008.
  3. Guensler, R., M. Rodgers, J. Leonard II, and W. Bachman. A Large Scale Gridded Application of the CALINE4 Dispersion Model. Transportation Planning and Air Quality IV. A. Chatterjee, Ed. American Society of Civil Engineers. New York, NY. 2008.
  4. Kall, D., V. Pandey, J., and R. Guensler. MOBILE-Matrix and CALINE-Grid: Project-Level Conformity Screening and Microscale Air Quality Impact Assessment Tools. 18th Annual On-Road Vehicle Emissions Workshop, San Diego, CA. Coordinating Research Council. Atlanta, GA. March 2008.
  5. Liu, H. Modeling the Impact of Road Grade on Vehicle Operation, Vehicle Energy Consumption, and Emissions. Doctoral Dissertation, Georgia Institute of Technology. 2018.
  6. Liu, H., M.O., Rodgers, and R. Guensler (submitted). The Impact of Road Grade on Vehicle Acceleration Activity, PM2.5 Emissions, and Dispersion Modeling. Transportation Research Part D: Transport and Environment.
  7. Liu, H., H., Li, M.O., Rodgers, and R., Guensler. Development of Road Grade Data using the USGS Digital Elevation Model. Transportation Research Part C: Emerging Technologies, 92, 2018, pp. 243-257.
  8. Shafi, G. Development of Roadway Link Screening Criteria for Microscale Carbon Monoxide and Particulate Matter Conformity Analyses through Application of Classification Tree Model. Master’s Thesis, Georgia Institute of Technology. 2008.
  9. Wu, Y. Integrated Assessment for Health Effects of Sustainable Transportation Strategies. Doctoral dissertation, University of California, Davis. 2018.
  10. Zhai, X., J.A. Mulholland, M.D. Friberg, H.A. Holmes, A.G. Russell, and Y. Hu. Spatial PM2.5 Mobile Source Impacts using a Calibrated Indicator Method. Journal of the Air & Waste Management Association, 1-13, 2019.
  11. Zhai, X., A.G. Russell, P. Sampath, J.A. Mulholland, B. Kim, U.Y. Kim, and D. D’Onofrio. Calibrating R-LINE Model Results with Observational Data to Develop Annual Mobile Source Air Pollutant Fields at fine Spatial Resolution: Application in Atlanta. Atmospheric Environment, 147, 446-457, 2016.