Synthesis of MRPO and Stakeholder OzonePM2'5 Modeling Results Part III: Findings and Recommendations

1
Synthesis of MRPO and Stakeholder Ozone/PM2.5
Modeling Results Part III
Findings and Recommendations

• T. W. Tesche
• Dennis McNally
• Alpine Geophysics, LLC
• Ft. Wright, KY
• Ralph Morris
• Edward Tai
• ENVIRON International Corp
• Novato, CA
• 23 July 2003

2
Overview of Presentation Part III

• 8. Prospects for 8-hr Ozone PM2.5 NAAQS
  Attainment in 2010
• 9. Interim Findings
• Model Performance Evaluation
• Need for Fine Grid Modeling
• Model Response to Emissions Changes
• Effectiveness of Precursor Controls
• Influence of Grid Resolution on Model Sensitivity
  to Emissions Changes
• 10. Conclusions and Recommendations
• Conclusions
• Modeling Refinements
• Emissions Inventory Refinement
• Attainment Analyses
• 11. Schedule

3
8. Prospects for 8-hr Ozone PM2.5 NAAQS
Attainment in 2010

• Interim Estimates Developed by MRPO
• No calculations performed yet in Stakeholder
  analyses

4
MRPO Estimation of Year 2010 PM2.5 DV Using Proxy
2001 and 2010 Inventories
5
MRPO Estimation of Year 2010 8-hr O3 DV Using
Proxy 2001 and 2010 Inventories
6
9. Interim Findings

• CAMx/M4 Model Performance Evaluation
• Findings from MRPO and Stakeholder modeling
• Performance for 8-hr Ozone
• Performance for 24-hr PM 2.5
• Need for Fine Grid Modeling
• Findings from MRPO Flexi-Nesting at 12 km scale
• Findings from Stakeholder Modeling at 36/12/4 km
  scales
• Model Response to Emissions Changes
• 8-hr Ozone
• 24-hr PM 2.5
• Influence of Grid Resolution on Model Sensitivity
  to Emissions Changes
• 8-hr Ozone
• 24-hr PM 2.5

7
MRPO Findings Model Performance Evaluation

• Ozone on 36 Km Grid
• Peaks overestimated.
• During mid-July in Great Lakes and Ohio Valley
  subregion, afternoon ozone peaks overestimated by
  20-30 ppb.
• Day-to-day (hour-to-hour) variation considered
  good.
• Spatial patterns both okay and not okay.
• PM2.5 on 36 Km Grid
• Agreement between modeled and measured PM 2.5
  judged good but attributed, in part, to
  compensating errors.
• Model overestimates nitrates and sulfates
  (slightly in summer).
• Model underestimates organic carbon and sulfates
  (slightly in winter).

Note The MRPO finding that CAMx/M4
overestimates nitrates slightly in summer is
not presently corroborated by Stakeholder
modeling on any of the three grids 36 km, 12 km,
or 4 km. (See the spatial mean nitrate plots for
the Summer 01 episode at the three individual
grid scales.) This may simply be due to
different procedures for judging results (e.g.,
domain-wide maximum concentrations, spatial mean
plots of averaged predictions observations at
all monitors, etc., These differing viewpoints
need to be examined further.
8
Stakeholder Findings Model Performance Evaluation

• Ozone on 36 Km Grid (summer 01 episode)
• Daily 8-hr ozone peaks are overestimated by 30.3
  on average.
• Daily 8-hr peak ozone concentrations are biased
  high by 17.
• Distribution of 8-hr ozone concentrations
  systematically biased high for virtually all days
  across entire concentration range (true for all
  grid sizes). That is, Q-Q plots on all grids
  reveal systematic overprediction.
• Daily 8-hr ozone gross errors (29.1) are
  reasonable.
• Hourly and daily maximum spatial patterns in
  modeled ozone fields exhibit expected mixed
  (good and problematic) behavior.
• PM2.5 on 36 Km Grid (summer 01 episode)
• Good agreement between modeled and measured PM2.5
  (e.g., an episode mean bias of 6.4) likely stems
  from compensating biases in component species
  (e.g., sulfate, nitrate, OC, EC).
• Model systematically overestimates daily-average
  nitrate (all grid scales).
• Model systematically underestimates daily-average
  organic carbon and elemental carbon (all grid
  scales).
• Model overestimates sulfates on a number of days
  (all grid scales).

9
MRPO Findings Need for Fine Grid Modeling

• Ozone
• 36 km results not good enough.
• 12 km Flexi-Nest results promising. The
  overestimation in afternoon ozone peaks during
  mid-July in the Great Lakes and Ohio Valley
  subregions is reduced to approximately 10-20 ppb.
  
• PM2.5
• Flexi-Nest results on 12 km grid not reported
  yet.

10
Stakeholder Findings Need for Fine Grid Modeling

• 8-hr Ozone
• Peak Prediction Accuracy improves markedly from
  36 km scale to 4 km scale (30.3 ? 21.7).
• Bias substantial improvement at 4 km scale
  (17.0 ? 3.7).
• Gross Error modest improvement at 4 km scale
  (29.1 ? 23.4).
• 24-hr PM2.5
• Peak Accuracy Gross Error improve modestly
  from 36 km scale to 4 km scale (47.3 ? 34.9).
  Note Peak Accuracy over all Stations is
  numerically identical to the Gross Error metric
  when considering 24-hour averages.
• Bias little significant change (6.4 ? -4.9).
• Summary
• For 1-hr and 8-hr Ozone finer grid resolution
  significantly improves model accuracy (bias) and
  precision (error) as higher resolution (smaller
  grid spacing) is used.
• For PM2.5 finer grid resolution appears yield
  smaller gains in model skill compared with ozone
  however, the results of this analysis remain
  tentative, pending further analysis and
  examination of additional episodes.

11
MRPO Findings Effectiveness of Global Precursor
Controls

• PM2.5
• Summer SO2 emissions reductions effective
• Winter NOx emissions reductions effective
• Year-round NH3 would appear to be most effective
  (although annual modeling not yet reported by
  MRPO).
• Ozone
• Not reported yet.

12
Stakeholder Findings Effectiveness of Global
Precursor Controls

• PM2.5
• Current findings are based on only two days (25
  Jun 01 8 Jan 00) accordingly, caution is
  needed in generalizing results.
• SO2 emissions reductions more effective in summer
  than winter
• NH3 emissions reductions (singly or in
  combination with NOx and/or SO2) are most
  effective, winter summer.
• Combined NOx/SO2 controls more effective than
  NOx/VOC controls in winter and probably in
  summer.
• NOx emissions reductions (singly or in
  combination) appear to be more effective in
  winter than summer.
• Only very small PM2.5 disbenefits (0.2 to 0.8
  ug/m3) modeled.

13
Stakeholder Findings Effectiveness of Global
Precursor Controls (concluded).

• 8-hr Ozone
• Heavy VOC control (60) yields greatest peak
  ozone reduction on the day intensively
  examined21 July 01.
• Heavy NOx control (60), with or without other
  precursor reductions, is moderately effective in
  reducing peak ozone.
• Modest (15-30) NOx controls, with or without
  other precursor reductions, is somewhat effective
  on 21 July 01.
• NOx controls produce substantial, localized ozone
  disbenefits (25-75 ppb) in the Lower Lake
  Michigan region and elsewhere.

14
MRPO Findings Effectiveness of Subregional
Precursor Controls

• PM2.5
• Not reported yet.
• Ozone
• Urban VOC and rural NOx appear to be most
  effective.

15
Stakeholder Findings Effectiveness of
Subregional Precursor Controls

• PM2.5
• In the MRPO region, adding heavy NH3 controls
  (60) roughly doubles the effectiveness of heavy
  SO2 or NOx/SO2 controls on peak PM2.5 reduction.
• Wintertime precursor controls less effective in
  MRPO region compared with controls in summer.
• Wintertime SO2 controls in MRPO region appear
  ineffective.
• In the VISTAS region, heavy NH3 controls (60)
  appears to have a similar, though more
  pronounced, effect as in the MPRO region,
  particularly in winter.
• Ozone
• Stakeholder results available but not yet
  examined thoroughly.

16
MRPO Findings Influence of Grid Resolution on
Model Sensitivity to Emissions Changes

• Not reported yet.

17
Stakeholder Findings Influence of Grid
Resolution on Model Sensitivity to Emissions
Changes

• PM2.5
• Generally, the effects of heavy, combined
  precursor controls (NOx60SOx60Amm60) on reducing
  peak PM2.5 concentrations tends to diminish at
  finer grid scales.
• There is some indication, though, that the
  response is non-linear.
• Heavy precursor controls in the MRPO subdomain,
  for example, produce greater PM2.5 reductions on
  the 12 km grid compared with the 36 km grid.
• Current modeling results need to be examined much
  more extensively to establish a more robust
  synthesis.

18
Stakeholder Findings Influence of Grid
Resolution on Model Sensitivity to Emissions
Changes (concluded).

• Ozone
• Generally, the effects of heavy, combined
  precursor controls (NOx60SOx60Amm60) on reducing
  peak 8-hr ozone concentrations tends to diminish
  at finer grid scales
• As with PM2.5, the response may be non-linear
  heavy precursor controls in the MRPO subdomain
  also yield greater ozone reductions on the 12 km
  grid compared with the 36 km grid.
• As expected from basic ozone chemistry
  principles, the magnitude of the NOx disbenefit
  effect substantially increases as finer
  horizontal grid scales are used.

19
10. Conclusions and Recommendations

• Conclusions
• The findings of this initial round of Stakeholder
  modeling are preliminary because
• Potentially significant errors may exist in the
  baseline modeling inventories (e.g., motor
  vehicle emissions in the northeast U.S
  discrepancies in the ammonia source inventory)
  and,
• The analysis has focused largely on only three
  days (25 June 01, 21 July 01, and 8 January
  00) out of two long summer and winter episodes
  in order compare and contrast the many control-
  and model-configuration options currently
  available.
• Additional analysis of the existing CAMx/M4
  modeling results should be guided by a prefatory
  identification of specific hypotheses to be
  tested and specific questions to be answered so
  that the data analysis and reporting efforts are
  properly focused.
• The PM2.5 and 8-hr ozone results summarized
  herein confirm the fact that substantial
  non-linearities exist in the relationships
  between ozone and fine particulate precursor
  emissions and resultant gas-phase (ozone) and
  secondary aerosol (PM2.5) concentrations in the
  Midwest and eastern U.S. These non-linearities
  also exist with respect to choices among model
  configurations, particularly the selection of
  horizontal grid resolution.
• These preliminary results appear to confirm the
  need to utilize regional photochemical models at
  high resolution (e.g., 4 km) in order to
  elucidate more clearly the efficacy of VOC
  and/or NOx controls on 1-hr and 8-hr ozone. The
  need for finer grid resolution for PM2.5
  concentrations is less clearly demonstrated
  however, given the inextricable coupling of the
  gas-phase and particulate aerosol chemistries, it
  is likely that high resolution bill prove
  essential for PM2.5 as well.

20
10. Conclusions and Recommendations (concluded).

• Recommendations
• Modeling
• Continue using CAMx/M4 as the primary model
  corroborate with CMAQ later.
• Continue diagnostic model performance evaluation,
  focusing on other modeling days, and additional
  species (e.g., particulate nitrates, secondary
  organic aerosols).
• Strengthen understanding of conditions
  necessitating fine grid (4 km) modeling.
• Continue to encourage and support cooperative
  modeling among agencies and stakeholders
• Emissions
• Implement concerted effort to refine and QA the
  ammonia inventory (given strong response of PM2.5
  concentrations to NH3 reductions).
• Correct known deficiencies in existing base year
  (00 and/or 01 inventories.
• Develop future 2010 baseline inventory with best
  available growth/control forecasts
• Implement vigorous, cooperative QA and inventory
  improvement activity
• Analysis
• Develop guidance concerning the selection of
  subsets of modeling days to be used in
  synthesizing overall simulation results.
• Identify and evaluate candidate control measures

21
11. MRPO Schedule