This presentation contains preliminary results and data.
-------------------Do not quote, cite, or distribute
without the authors’ explicit permission
Air Quality in the Northern Colorado Front Range (First) Results from FRAPPÉ and DISCOVER-AQ
Gabriele Pfister and Frank FlockeSojin Lee
Atmospheric Chemistry Observations and Modeling Lab (ACOM)National Center for Atmospheric Research (NCAR)
& the FRAPPÉ and DISCOVER-AQ Science Teams
NOAA Seminar, 29 March 2017
Air quality issues in the NFRMA
The Front Range is in non-Attainment for the 8-hour Ozone NAAQS
The Front Range is in non-Attainment for the 8-hour Ozone NAAQS
Air quality issues in the NFRMA
U.S. air overall has become cleaner
Due to emission controls and advanced technology, pollutant emissions in the U.S. have decreased.
Front Range: 22% decrease
2005-2007 2009-2011
Trends in Colorado Surface Ozone and NOx
Source: CDPHE 2015 Annual Report
NOx
Ozone
linear regression lines with confidence limits
Source: Colorado SIP, released December 2016; Correction Method (Reddy and Pfister, 2016) http://raqc.org/technical-support-documents-for-the-moderate-area-2008-8-hour-ozone-standard-state-implementation-plan/
75 ppb
Surface Ozone Trends 2006-2015
Ozone = NOx + VOCs +
Trends in 4th highest MDA8 corrected for weather (exceptional events excluded)
Rocky Flats NREL/Golden CAMP/Denver
Background Ozone Trends
+ Background/Inflow
Cooper et al., 2014
Ozone = NOx + VOCs +
What and where are the relevant sources?
How do these emissions get transported?
How do they get chemically processed?
How much pollution comes into Colorado?
Which are the best ways to improve air
quality?
The Questions
Air quality monitoring (and regulation)
(ground)Observations
Emissions
Model
Most operational measurements of air quality are made
at the surface and cover few pollutants.
The Challenge
• The atmosphere is three dimensional. Pollutants move and mix horizontallyAll processes are time-dependent and many are non-linear ⇒4D
• There is not just NOx and “VOC.” VOC are comprised of hundreds of different species with varying roles in ozone chemistry and are emitted from different sources.
• Reducing VOCs or NOx does not necessarily decrease ozone. Sometimes ozone might increase because ozone chemistry is dependent on the mix of ingredients.
Most operational measurements of air quality are made
at the surface and of only few pollutants. BUT:
In order to adequately simulate Front Range ozone and accurately predict outcomes of possible control strategies, the model must:
• Use accurate emissions (magnitude, speciation, location, diurnal and seasonal variability)
• Distribute emissions into an accurate boundary layer
• Predict wind direction and speed, horizontal dispersion and mixing and vertical mixing/dilution with accurately modeled background air
• Accurately compute the chemical reactions and physical transformations
• Predict future trends
(or anywhere for that matter)
Modeling AQ in the Front Range
Air quality monitoring (and regulation)
(ground)Observations
Emissions
Model
FRAPPÉ (Field Campaign) Study Approach
Observations
compare
Model
Emissions
evaluaterefine
Processes
evaluate
refine
Integration of different platforms and observation strategies
FRAPPÉ & DISCOVER-AQ15 July - 18 August 2014, Colorado Front Range
Funding SourcesFRAPPÉ: State of Colorado / CDPHE
National Science Foundation (NSF)
DISCOVER-AQ: NASA
Others: NOAA, GO3 Project, NPS, EPA
PIs: Gabriele Pfister and Frank FlockeNational Center for Atmospheric Research (NCAR)
PI: James CrawfordNASA Langley
Merging of two major field campaigns
Comprehensive look at air quality in the Northern Colorado Front Range
FRAPPÉ and DISCOVER-AQ
4 aircraft
6 mobile vans
6 enhanced ground sites
Tethered balloons and daily ozone sondes
Enhanced CDPHE network
The Colorado Front Range
• Diverse sources of air pollutants• Separated spatially/temporally in some cases, co-located in others • Different expectations of future growth• Emissions difficult to assess due to variability and high
number of individual sources
• Unique, mountain-driven local meteorology• Drives local mixing and transport• Recirculation of pollutants• A challenge for chemical-transport models
- Non-Attainment Area
● Oil/Gas Wells
- Non-Attainment Area
● Oil/Gas Wells
NASA P-3 flight tracks: • focus on non-Attainment area• repetitive flight pattern to look at
diurnal variability• Vertical profiling over selected
ground sites 1,000agl-18,000asl
NCAR/NSF C130 flight tracks:• larger region with targeted flights • emission sources in the NFRMA and other areas
(urban, O&NG, agriculture, powerplants, …)• Mixing, outflow, upslope events,…
Methane emissions
Methane important greenhouse gas and released by both agriculture and O&G activities
What are the relative contributions of biogenic (landfills and cattle) and thermogenic (oil and gas) sources to methane in the Denver area?
Hydrogen stable isotopic measurements of source and flight samples indicate biogenic sources may be a significant contributor to methane despite the large number of oil and gas production sites – perhaps due to regulation of methane emissions from oil and gas wells in Colorado?
RF05July 31, 2014
85% biogenic49% biogenic
Amy Townsend-Small et al., U Cincinnati (GRL, 2016)
Methane emissions from O&G & agriculture
Methane: Picarro 2311 (RAF/ACOM)Ethane: CAMS (A. Fried, CU)
C-130 flights E of 105.2W and N of 39.8N
Methane emissions from O&G & agriculture
Franco et al., 2016 [~5%]SMOKE/CMAQ similar
XCEL Energy 2014 Jul/Aug Fraction in Distr. NG [7%]NEI (McKeen) similar
Methane: Picarro 2311 (RAF/ACOM)Ammonia:TILDAS(Aerodyne)
C-130 flights E of 105.2W and N of 39.8N
USDA inventory for CAFOs (JEQ, 2010)
Methane emissions from O&G & agriculture
Methane emissions from O&G & agriculture
Rocky Mountain NP
• Transport of Nitrogen from sources in the Front Range damages the fragile ecosystem.
• Nitrogen deposition in the Park is ~15 times greater than the natural background deposition rate.
• Ecosystem health first beganto decline at high elevationson the east side of the Park between 1950 and 1964. ~ 1,000,000 cattle
Methane: Picarro 2311 (RAF/ACOM)Ammonia:TILDAS(Aerodyne)
C-130 flights E of 105.2W and N of 39.8N
USDA inventory for CAFOs (JEQ, 2010)
Methane emissions from O&G & agriculture
Distribution of ammonia (gas phase, C130)
Distribution of ammonium ion (particle phase, C130)
Model setup
CMAQ version v5.2 beta (Deborah Luecken (U.S. EPA)); NCAR ECMWF analysis fields; WRF V3.8.1 met; CB06r3 chem.
Category EPA sectors Source
Mobile Onroad mobile, nonroad mobile, Rail emissions
d01: EPA 2011 mobile emission based on EPA2011v6.3 platform; d02: 2017 Alpine geophysics inventory (Dennis McNally, AG)
Oil and Gas nonpoint oil and gas, point oil and gas emissions
2014 CDPHE emission inventory with spatial, temporal, and chemical speciation profiles from EPA 2011v6.3 platform (Dale Wells, CDPHE)
Other point EGU, point NONIPM EGU: 2014 activity data, point NONIPM: 2017 projection EPA data + spatial, temporal, and chemical speciation profiles from EPA 2011v6.3 platform
Other Nonpoint
nonpoint RWC, nonpoint other, nonpoint agriculture
2017 projection EPA data + spatial, temporal, and chemical speciation profiles from EPA 2011v6.3 platform
Biogenic Biogenic BELD v3.61 with BELD4.1 land use database which incorporates the NLCD
Fires Fires Fire INventory from NCAR (FINN) v1.5
D01 domain 12x12km,D02 domain 4x4 km
Boundary cond. From RAQMS(Brad Pierce, NOAA)
Emissions (a priori)
Front RangeCounty
Mobile emission (onroad + nonroad + rail emission, unit: ton / year)
CO NOx VOC
EPA2011 EPA2014 This study EPA_2017 EPA2011 EPA2014 This study EPA_2017 EPA2011 EPA2014 This study EPA_2017
Adams 57,998 51,070 46,377 42,933 10,440 8,580 5,661 6,297 5,327 4,463 3,554 3,513
Arapahoe 76,235 67,165 61,052 56,709 9,633 8,278 5,416 5,475 7,039 5,939 4,786 4,619
Boulder 37,889 33,483 30,007 28,004 5,015 4,378 2,987 2,845 3,576 3,025 2,486 2,350
Broomfield 7,067 6,335 6,114 4,940 1,167 1,097 665 645 613 540 434 381
Clear Creek 6,528 4,133 3,747 3,922 1,646 1,590 481 919 474 325 234 277
Denver 72,826 72,142 57,794 52,101 11,037 13,497 6,043 6,122 6,538 6,442 4,433 4,041
Douglas 38,159 34,024 32,247 33,298 6,775 6,138 3,920 4,553 3,419 2,885 2,586 2,663
Elbert 5,263 3,398 3,483 3,550 1,325 862 514 820 509 311 298 324
Gilpin 1,295 817 853 776 429 380 289 295 161 104 106 98
Jefferson 72,870 60,689 53,220 52,057 10,184 8,298 5,223 5,646 6,831 5,363 4,487 4,356
Larimer 49,457 38,041 42,838 32,797 7,232 5,419 5,110 4,003 5,081 3,761 3,646 3,109
Park 5,193 3,215 3,227 3,317 780 500 350 408 748 500 479 469
Weld 45,882 34,852 45,809 29,649 9,265 7,632 7,831 5,623 4,475 3,229 3,489 2,680
Total 476,661 409,364 386,767 344,053 74,927 66,649 44,491 43,650 44,790 36,887 31,018 28,881
Emissions (a priori)
Front RangeCounty
Oil and Gas(nonpoint oil and gas + point oil and gas, unit: Ton / year)
CO NOx VOC
EPA2011 EPA2014 This study EPA_2017 EPA2011 EPA2014 This study EPA_2017 EPA2011 EPA2014 This study EPA_2017
Adams 891 695 637 825 1,966 686 1,497 1,765 3,475 1,548 1,173 3,576
Arapahoe 208 249 286 191 515 282 480 455 532 1,112 933 553
Boulder 129 81 49 129 185 40 24 180 1,308 369 233 1,288
Broomfield 42 23 9 42 69 12 3 67 689 135 85 653
Clear Creek 0 0 0 0 0 0 0 0 0 0 0 0
Denver 18 13 2 20 43 9 7 44 133 48 29 146
Douglas 0 0 0 0 0 0 0 0 0 0 0 0
Elbert 46 13 8 49 81 9 8 87 316 73 42 355
Gilpin 0 0 0 0 0 0 0 0 0 0 0
Jefferson 11 0 13 10 6 0 11 5 8 1 6 7
Larimer 58 58 39 59 97 34 21 94 796 415 305 787
Park 0 0 0 0 0 0 0 0 0 0 0 0
Weld 14,159 20,746 21,017 13,640 18,249 17,892 19,783 17,211 123,506 91,709 75,665 120,556
Total 15,562 21,879 22,061 14,965 21,211 18,963 21,835 19,908 130,762 95,409 78,471 127,922
Emission evaluation strategy
Observations (PTRMS; A. Wisthaler, U Innsbruck)Model
Benzene at Platteville
Hourly Averages
Surface Data are essential but evaluation is impacted by grid resolution/nearby sources and local winds.
Note: scale is cut off at 3 ppb
Halliday et al., 2016
Benzene at Platteville
Diurnal Cycle in Surface and Aircraft Benzene
Maximum values up to 30 ppb
Observed Winds (8-18 LT)
~ Model Grid Size
Modeled Winds (8-18 LT)
~ Model Grid Size
Modeled versus observed winds
Ft Collins West
C-130 aircraft - PBL
CAMP
Emission evaluation strategy
Compare measurements with model predicted mixing ratios - but select days and time periods when models represent transport well.
Estimate source contribution to each sample from surrounding grid cells using wind direction and speed
Evaluate absolute concentrations and emission ratio predictions versus measured ratios.
Adjust individual emission sectors, based on data selection.
Emission evaluation strategy
- Identify grid boxes for 10-17 LT < 1km ag where(1) The contribution of the evaluated emission sector is at least 50%(2) Observed and modeled winds are from same sector (10-17 LT, < 1km ag)
- Compare individual samples with modeled concentrations averaged over each set of grid boxes
…
Emission evaluation strategy
Identify grid boxes for 10-17 LT < 1km ag where(1) The contribution of the evaluated emission sector is at least 50%(2) Observed and modeled winds are from same sector (10-17 LT, < 1km ag)
…
Mobile EmissionsNOx
OnG Emissions Benzene
Aircraft Model
Comparison CMAQ to aircraft (<1km agl)
NCAR/NSF C-130NASA P-3Fort Collins
Denver
Aircraft
10-17 am
NOx
Benzene
Model
Aircraft Model
Comparison CMAQ to aircraft (<1km agl)
NCAR/NSF C-130NASA P-3Fort Collins
Denver
Aircraft
10-17 am
Toluene Model
Acetylene
Comparison CMAQ to aircraft (<1km agl)NCAR/NSF C-130
10-17 am
“PAR”Propane
Aircraft ModelAircraft Model
Emission evaluation strategy
(1) Divide NFRMA into three sub-regions, based on to emission sector dominating each sub-region
(2) Statistical model-observation comparison of- absolute concentrations- tracer-tracer ratios (“emission ratios”)
Comparison Model to Aircraft (C-130)
• Ethyne has a low model bias in all of NFRMA, about factor of 2
• Other mobile emission tracers show reasonable absolute agreement in Denver Area
• All mobile emission tracers indicate that traffic emissions are underestimated in the non-Denver urban and oil and gas sectors
Comparison Model to Aircraft (C-130)
We have seen this for NEI 2011 running WRF-Chem.
NEI 2011 emissions provided by Stu McKeen, NOAA
Comparison Model to Aircraft (C-130)
NOxEthane
Benzene Toluene
“PAR” Propane
NOxEthane
Benzene Toluene
“PAR” Ratios to
Propane
Model versus Obs.
All measured concentrations are strongly underestimated by the model
(except Ethane)
But the ratios to Propane look pretty good
(except Ethane, *which is good*)
Current status of emission adjustments
• Ethyne (acetylene) in mobile emissions: ~ double.
• Total mobile emissions (traffic volume?) in Non-Denver NFRMA: increase by a factor of ~3.
• Oil and gas sector emissions for propane and heavier alkanes: increase by at least a factor of 4.
• Oil and Gas sector emissions of Benzene: increase also by at least a factor of 4.
• Oil and Gas sector emissions of Ethane: leave unchanged (though might be on the high side)
• Next model run iteration pending.
Mobile Ground Sample (Weld County)
Alkanes:Propane 150 ppbButane 425 ppbPentanes 700 ppbHexanes 900 ppb
Benzene 65 ppb
“strong smell”
Mobile Ground Sample (Weld County)
Benzene 120 ppb Alkanes: Propane 55 ppbMethane 1.9 ppm Butanes 300 ppb
Pentanes 600 ppb
Model evaluation strategy
Adjust individual individual emission sectors to improve inventories and perform iterative model runs
Surface In-situ to refine inventories (point sources, emission ratios, ....). TBD
Support CTM study through Box Modeling to identify relevant VOCs and characterize chemical regimes (and the model’s ability to represent the chemistry)
Denver Urban5.4 s-1
NCAR/ACOM TOGA group; Rebecca Hornbrook Rebecca Hornbrook
0.1 Commerce City
16.5 s-1
Weld County6.7 s-1
Calculated OH Reactivity by FRAPPÉ Region
Alkane OH Reactivity
Alkane OH reactivities from Commerce City and Greeley Missed Approach are greatest at C5.
Weld region reactivity is centered around C4.
Rifle and Uintah have relatively even reactivities
from C2-C6.
Using alkane data from TOGA and WAS canisters: D. Blake, UC Irvine
NCAR/ACOM TOGA group; Rebecca Hornbrook
• Puts the air parcel in a jar then follows the chemistry• MCM (Leeds Master Chemical Mechanism) plus comparison with lumped mechanisms• No dilution, entrainment, mixing, deposition, etc. • Yields the ozone formation potential of a sample/average• Simulations at 2 pm LT, 40N, clear sky, fixed Temperature (“idealized conditions!”)
BOXMOX: Knote et al., Atmos. Environ, 2015; https://www2.acom.ucar.edu/modeling/boxmox-box-model-extensions-kpp
Chemical Box Modeling (BOXMOX)
C-130 measured
sample
NCAR/ACOM TOGA group; Rebecca Hornbrook
Chemical Box Modeling (BOXMOX)
MCMV3.3MOZART-4MOZART pre-T1MOZART T1RADM2SAPRC99
Denver (NOx(t0) = 8 ppb Comm. City (NOx(t0) = 47 ppb
Weld(NOx(t0) = 2 ppb Weld(NOx(t0) = 10 ppb
Time (hours) Time (hours)
Ozo
ne (p
pb)
Ozo
ne (p
pb)
NCAR/ACOM TOGA group; Rebecca Hornbrook
Chemical box modeling (BOXMOX) – Commerce City
Both mechanisms indicate NOx saturation but MOZART produces much less ozone
Frank Flocke et al., NCAR
Problems found with MCM mechanism – stay tuned for updates
Chemical box modeling (BOXMOX) – Commerce City
Both mechanisms produce ozone much more efficiently without EGU NOx
Better agreement with lower NOx
Emission contribution based on SMOKE/CMAQ
MCM -EGU NOx
MOZ -EGU NOx
-90% NOx
MCM Control
MOZ Control
Problems found with MCM mechanism – stay tuned for updates
Ozone production in WCReduced by 23 ppb (or factor of 10 in 8h) with O&NG emissions removed
Chemical box modeling (BOXMOX) – Weld County
Problems found with MCM mechanism – stay tuned for updates
Chemical box modeling (BOXMOX) – Denver urban
Transportation is still important but does not appear to be the major factor in Denver Metro
Problems found with MCM mechanism – stay tuned for updates
• FRAPPÉ & DISCOVER-AQ analysis is coming along, but there still is work to do. We could use a larger band.
• Emissions are one of the largest source of uncertainty in current understanding and modeling of NFRMA pollution. Being close to extensive (and variable) O&NG activities doesn’t help, but industrial and mobile emissions are also uncertain.
• Transport in the NFRMA is extremely complex. WRF does a reasonable job, but could use improvement for full emission assessment. Tests with observation nudging of FRAPPE observations are pending.
• Box modeling and OH reactivity studies show promise in assessing ozone chemical regimes and sensitivities, but need more explicit model runs (GECKO-A, modified MCM) to accurately test lumped mechanism performance.
• Reactive nitrogen lifetime is a critical factor for chemistry model performance in areas with high reactivity and large NOx sources.
• Long range transport, regional background (increasing in the western US) and (maybe) stratospheric influence can be important for NFRMA ozone (including surface ozone) Current work shows shortcomings in the models in
Summary
THANK YOU
Joint FRAPPÉ and DISCOVER-AQ Science Team Meeting2-3 May, UCAR Center Green, Boulder CO
Air Quality Open House at NCAR Mesa Lab Wed, 3 May 4-7 pm