Modeling and Ambient Monitoring of Air Toxics in Corpus Christi, Texas

1
Modeling and Ambient Monitoring of Air Toxics in
Corpus Christi, Texas
Gary McGaughey, Elena McDonald-Buller, Yosuke
Kimura, Hyun-Suk Kim, and David T. Allen The
University of Texas at Austin Center for Energy
and Environmental Resources Greg Yarwood, Ed
Tai, and Chris Colville ENVIRON International
Corporation Novato, CA John Nielsen-Gammon and
Wenfang Lei Department of Atmospheric Sciences,
Texas AM University College Station, TX
2
Outline

• Background
• Configuration of the Corpus Christi air toxics
  monitoring network
• Development of a conceptual model for TNMHC and
  benzene meteorological conditions, emission
  source areas, and temporal trends (diurnal and
  seasonal)
• AERMOD and CALPUFF results for benzene
• On-going efforts including WRF/CAMx at 1km
  horizontal resolution

3
Background

• Neighborhood-scale monitoring and air quality
  modeling of air toxics is critical for human
  exposure and health risk assessments.
• Since June 2005, The University of Texas at
  Austin (UT) has operated a dense monitoring
  network for air toxics in Corpus Christi that
  will continue at least several more years.
• UT with ENVIRON and Texas AM University are now
  developing neighborhood-scale air quality models.
  

Population of 400,000 (2008) Currently in
attainment with the NAAQS for O3 and PM2.5.
Significant petroleum refining and chemical
manufacturing industries.
4
Corpus Christi Air Monitoring and Surveillance
Camera Installation and Operation Project

• Sulfur compounds, TNMHC, and meteorological
  measurements at all seven UT sites since early
  2005.
• Hourly auto-GC measurements and camera
  surveillance at two sites (Oak Park and Solar
  Estates in blue above).
• Event triggered canisters at five sites (TNMHC gt
  2000 ppbC for 15 min).
• Canister sampling at TCEQ CATMN sites.

5
Locations of Industrial Property Boundaries,
Terminals and Docks In Nueces County Relative to
UT and TCEQ Monitoring Sites
6
Conceptual Model Development Identifying Key
Characteristics of Air Toxics Events

• Investigate meteorological conditions, emission
  source areas, and temporal trends associated with
  TNMHC, benzene and other air toxics.
• Based on data collected during June 2005 May
  2008 by the UT network in additional to
  historical data from the TCEQ CATMN sites.
• Important for selecting time periods of interest
  for air quality modeling, for identifying
  emission source areas, and for understanding and
  targeting conditions that lead to higher
  concentrations of air toxics.
• This analysis will be continued during the
  lifetime of the network.

7
Benzene Concentrations

• TCEQ thresholds for evaluating health effects
• Reference Values (ReVs) Acute and chronic ReVs
  for benzene are currently 1080 ppbC and 516 ppbC.
  
• Effects Screening Levels (ESLs) Acute and
  chronic ESLs for benzene are 324 ppbC and 8.4
  ppbC.
• Average benzene concentrations between June
  2005-May 2008 range from 1.93 (Solar Estates) to
  8.92 ppbC (Huisache).

8
(No Transcript)
9
Benzene Concentrations

• Similar to TNMHC, higher concentrations of
  benzene tend to occur during the night and early
  morning hours and during the fall/winter.
• These results are generally consistent with other
  areas of the US.
• Weekend vs. weekday comparison suggests weak
  trend towards lower concentrations on Sunday
  compared to weekdays.
• Annual benzene trends at CATMN sites indicate
  lower concentrations recently at Huisache and
  Dona Park.

10
Benzene Concentrations

• A Trajectory Analysis Tool was used to identify
  upwind source areas during periods with higher
  benzene concentrations.
• Consistent upwind geographic areas were
  identified during periods with higher benzene
  concentrations, suggesting site-specific
  emissions sources.

Surface back-trajectories for all hours
characterized by a benzene concentration of 30
ppb or greater at the Oak Park monitoring station
during June 2005 - May 2008.
11
Emission Inventory

• TCEQ Photochemical Modeling EI (2000, 2005)
• Same level of source resolution as the State of
  Texas submittals to EPAs National Emission
  Inventory (NEI)
• Used for air quality planning in Texas
• Accounts for rule effectiveness (RE)
• To account for reductions in control efficiency.
• Applied at the SCC/SIC/abatement level by
  geographic regions.
• Primarily affects VOC emissions from flares,
  equipment leak fugitives, external floating roof
  and internal floating roof tanks.
• Results in approximately a 28 increase in VOC
  emissions (6600 tpy to 8500 tpy) in Nueces and
  San Patricio Counties.
• Most detailed chemical speciation of VOC
  emissions
• Needed for responding to regulations in the
  Houston area that target highly reactive VOCs and
  for assessment of control strategies.
• Source-specific profiles originally developed by
  Pacific Environmental Services under contract to
  ENVIRON and Gabriel Cantu at the TCEQ (See Thomas
  et al. Emissions Modeling of Specific Highly
  Reactive Volatile Organic Compounds in the
  Houston-Galveston-Brazoria Ozone Nonattainment
  Area, presented at the 17th Annual International
  Emission Inventory Conference, Portland, OR June
  2008). Profiles are updated continuously.

12
AERMOD and CALPUFF Meteorological Processing

• Chose Oct 1 Nov 30, 2006 (based on results from
  the conceptual model) for initial testing and
  development of the AERMOD and CALPUFF models.
• Used 2005 TCEQ Photochemical Modeling EI
  (stationary point sources only)
• Does not include on-road mobile EI currently
  being developed by ENVIRON
• AERMET was used to process the meteorological
  data collected at the Solar Estates and Oak Park
  (on-site) monitors. Surface parameters (albedo,
  Bowen ratio, and roughness length) were provided
  by the TCEQ for Nueces County. Additional
  surface and all upper air data were from the NWS
  station at Corpus Christi Airport.
• CALMET was used to generate meteorological input
  for CALPUFF.
• 8 UT/TCEQ surface stations, 10 NWS monitors, 1
  upper air station (Corpus Christi Airport), 5
  precipitation NWS locations, 1 NOAA buoy
• CALMET sensitivity tests were performed to yield
  the most acceptable wind fields (subjective
  judgment).
• CALMET options included relocation of buoy closer
  to grid domain, terrain kinematics, smoothing in
  higher layers, high resolution LULC data for
  coastline.

13
AERMOD Maximum Predicted Benzene
Concentrations using Solar Estates and Oak Park
Meteorology
Solar Estates Meteorology Max 31.8 ppb
Oak Park Meteorology Max 43.5 ppb
Distance between Oak Park and Solar Estates 10
km, yet 30 difference in maximum
concentrations. Oak Park meteorology also
predicts higher concentrations further west
compared to Solar Estates meteorology.
14
Maximum Predicted Benzene Concentrations between
AERMOD (with Oak Park Meteorology) and CALPUFF
AERMOD with Oak Park Meteorology Max 43.5 ppb
CALPUFF Max 53.2 ppb, Second Max 47.4 ppb
AERMOD simulated one peak compared to two peaks
for CALPUFF. CALPUFF predicts higher
concentrations nearer to the emissions
sources. AERMOD tended to disperse emissions
further downwind.
15
Comparison of Observed and Predicted AERMOD and
CALPUFF Maximum Daily Hourly Benzene
Concentrations
Daily Maximum Benzene during Oct/Nov 2006 at Oak
Park Observed, AERMOD, and CALPUFF
Daily Maximum Unpaired Benzene Concentrations
Although the modeled values capture the range of
observed concentrations, there is large
variation on any given day.
16
On-going Work Coupling WRF/CAMx at 1km
horizontal grid resolution

• Two WRF simulations (36/12/4/1 km resolution)
• October 16-22, 2002
• 1 km domain
• 10 minute output resolution
• 44 vertical layers
• Lowest 21 layers mapped to
• CAMx (up to 3 km)
• WRF Run 1
• ACM2 PBL scheme
• Noah LSM
• Monin Obukhov scheme for surface layer physics
• WRF Run 2 changes
• Pleim-Xiu surface layer physics
• Added analysis nudging in 36 and 12km domains
  above layer 10 (700m)

17
CAMx Sensitivity Runs with CALMET and WRF (Run 1
and Run 2) Meteorology at Solar Estates

• Inert CAMx run with PiGs
• Emissions from point sources only
• Run 16 CALMET
• Run 19 WRF Run 1
• Run 20 WRF Run 2
• CAMx concentrations were extracted from 200m
  sampling grid over Solar Estates.
• CALMET layer 1 18m (fixed)
• WRF layer 1 17m

• Peak observed daily benzene 6.6 ppb on Oct 17
  (8AM)
• Second daily peak 3.5 ppb on Oct 21 (10AM)
• All runs under-predicted the peaks, but CAMx Run
  20 (WRF Run 2) was closest
• CAMx Run 19 (WRF Run 1) produced un-observed
  benzene spikes (Oct 18)

18
Wind Speed Comparison over Solar Estates
CALMET vs. WRF Run 1 WRF Run 1 vs.
WRF Run 2

• CALMET follows observed wind speeds well due to
  strong obs nudging.
• WRF Run 1 similar or slower than observed winds
  during low-wind speed periods.
• WRF Run 2 wind speeds faster than WRF Run 1 and
  closer to observed.

19
Wind Fields during Peak Observed Benzene at
Solar Estates on October 17
CALMET WRF Run 1
WRF Run 2

• CALMET does not show a clear distinction in winds
  over land and water crop circle effects
• WRF Run 1 has a more distinguished land/water
  interface.
• WRF Run 2 wind direction similar to Run 1, but
  wind speeds are relatively higher
• over land and slower over water.

20
Wind Direction at Solar Estates
CALMET vs. WRF Run 1 WRF Run 1 vs.
Run 2

• CALMET wind direction matches observed winds
  well.
• WRF Run 1 wind direction tracks observed
  reasonably well
• except during the mornings of Oct 17 and 21.
• WRF Run 2 handles the nighttime rotation wind
  shift better,
• especially on Oct 21.
• Benzene higher with WRF Run 2 compared to WRF
  Run1.

21
CAMx/WRF Summary Solar EstatesOctober 16-22,
2006

• Daily peak observed benzene was 6.6 ppb on Oct 17
  (8AM) and 3.5 ppb on Oct 21 (10AM).
• CAMx under-predicted peaks using CALMET and WRF
  Run 1.
• CAMx with WRF Run 2 predicted benzene comparable
  in magnitude to observed concentrations, but
  half a day off in time.
• Meteorological conditions needed for high benzene
  concentrations
• Low Kvs.
• Both WRF runs and CALMET met this criteria.
• Low wind speeds
• Both WRF runs had comparable or slower wind
  speeds compared to observations, but WRF Run 2
  was faster.
• An industrial facility is located ENE of Solar
  Estates.
• On October 17 and 21, WRF Run 1 failed to rotate
  early morning winds clockwise from north to
  southeast.
• WRF Run 2 had a similar problem on October 17,
  but performed better on October 21, resulting in
  relatively higher benzene (3.2 ppb in Run 2 vs.
  0.6 ppb in Run 1 observed 3.5 ppb)

22
Summary

• UT is operating a dense ambient monitoring
  network for air toxics with a lifetime of
  approximately 10 years that includes hourly
  auto-GCs and camera surveillance, threshold
  triggered canister samples and meteorological
  data for the Corpus Christi area.
• Conceptual models of meteorological conditions
  and associated temporal trends and emission
  source regions have been developed for TNMHC and
  benzene.
• Similar to other areas, higher concentrations of
  benzene and TNMHC tend to occur during the night
  and early morning hours and during the
  fall/winter.
• Concentration gradients in benzene exist between
  monitors and trajectory suggests strong
  associations with site-specific emissions
  sources.
• On-going work focuses on the development and
  application of Gaussian and neighborhood-scale
  photochemical grid models.
• Evaluate and compare model performance (AERMOD,
  CALPUFF, and CAMx).
• A primary goal of our work is the development of
  a modeling system that predicts the
  three-dimensional concentrations of selected air
  toxics concentrations (e.g., benzene) at the
  neighborhood (lt 1-km horizontal resolution)
  scale.
• Model results used to
• assess the accuracy of the emission inventories
  and the sensitivity of predictions in the spatial
  patterns of air toxics.
• examine whether the locations of the existing air
  quality monitors captures the locations of
  predicted maximum air toxics concentrations.