Calculation of Sensitivity Coefficients for Individual Airport Emissions in the Continental UnitedStates Using CMAQ-DDM3D/PM

Abstract

Previous estimates of aviation contributions to ground-level ozone and fine particulate matter concentrations have either offered domain- and sector-wide estimates or focused on a few airports. Using the decoupled direct method (DDM), an advanced sensitivity analysis module for the CMAQ air quality modeling suite, we calculate per-airport sensitivity coefficients allowing quantification of 66 individual airports’ impact on air quality in the United States. Preliminary results show that these airports, collectively representing about 76 % of aviation activity by fuel burn in the US, are responsible for about 0.04 % of nationwide PM_2.5 concentrations; near-airport concentrations are proportionately much higher. Peak annual average contributions from individual airports vary from 0.018 to 0.0001 μg/m³; secondary PM_2.5 has effects at distances of up to 700 km downwind while primary PM_2.5 affects only the immediate vicinity of the airport. Complete results detailing specific air quality and health impacts of each airport will be presented at the ITM conference in May.

Questions and Answers

Questioner: Peter Builtjes, Talat Oldman

Question: What is the rationale behind using 3000 ft as a kind of average mixing height? In reality the mixing height varies in time and space. How do you allocate take-off emissions in the vertical? Do you consider ground equipment emissions?

Answer: The 3,000 ft emissions cutoff is not meant to correspond to any atmospheric condition; rather, it is an operations-based distinction between landing and takeoff (LTO) operations and cruise operations. The emissions processing sequence is primarily based on the operations mode provided as part of the aviation activity and emissions inventory from the Federal Aviation Administration (FAA)™s Aviation Environmental Design Tool (AEDT). In this dataset, spatiotemporal aircraft activity is provided from which emissions are calculated and gridded (in three dimensions) for use in CMAQ. One attribute of this database is emissions mode, which corresponds to one of ten sequential aircraft operation types, from taxiing to takeoff to climb to cruise, and finally descent to landing. Modes 1–3 correspond to taxi and takeoff modes and assigned to the departure airport, while modes 7–10 correspond to arrival and taxi modes and assigned to the arrival airport. Modes 4–6, corresponding to cruise activity, are not included in the modeling process. These chorded segments by activity mode for each flight path are then gridded into the corresponding CMAQ grid-cells based upon the relevant flight paths that fall in each grid-cell, as described in Baek et al. (2012). The AEDT only provides activity information for aircraft, and thus non-aircraft emissions (such as those from airport ground service vehicles) are not included. Thus, while a hard ceiling of 3,000 ft is built into the system, a soft cap based on aircraft operation mode limits emissions to a much lower altitude. And finally, EDMS (the precursor model to AEDT for preparing airport emissions inventories) uses a value of 3,000 ft as the default mixing height and the vertical extent of aircraft operations during LTO activities, and this provided the key motivation for us to define LTO activity in the vertical. However, the mixing height used in CMAQ predictions is based upon the WRF model, and truly varies in time and space.

Questioner: Talat Odman

Question: Would you consider verifying your results for one large and one small airport by modeling their domain of influence with higher resolution (e.g., 4-km grid size)?

Answer: One of the primary challenges of this work was balancing model runtime (CPU-hours) and quantity of disk space required against domain size. Certainly, a 4-km grid (a grid with 81 times the resolution of our 36-km grid) would provide a higher-resolution idea of the sensitivities caused by aircraft operations, which would give a more spatially-resolved indication of where impacts are located. Arunachalam et al. (2006) gives a good overview of the impact of grid size on model performance, with statistically significant differences between 36- and 4-km grid cells; further in a subsequent airport-specific grid-resolution investigation of air quality and health impacts, Arunachalam et al. (2011) showed that in spite of significant differences in maximum concentrations attributable to aviation emissions due to differing grid resolutions, total population health risks over the entire model domain were largely unaffected by model resolution. Thus, the need for higher model resolution is dependent on the objective of the study.