WRAP Experience: Investigation of Model Biases

1
WRAP ExperienceInvestigation of Model Biases
Uma Shankar, Rohit Mathur and Francis
Binkowski MCNCEnvironmental Modeling
Center Research Triangle Park, NC 27709
2
Acknowledgements

• Studies performed under contract with the Western
  Regional Air Partnership
• Model results provided by the WRAP Regional
  Modeling Center (Gail Tonnesen, Chao-Jung Chien,
  Mohammed Omary)

3
Outline

• Overview of Simulations
• Analysis of Modeling Results
• January nitrate overprediction
• Planetary Boundary Layer (PBL) heights and
  nitrate bias
• Role of ammonia emissions reduction nitrate bias
  in different chemical regimes
• Coarse mass (CM) underprediction
• Comparison of CM emission and deposition fluxes
• Summary
• Recommendations

4
CMAQ Configuration

• Advection Piecewise-Parabolic Method (PPM)
• Diffusion K-theory
• Gas-phase Chemistry Carbon Bond Mechanism 4
• extensions include SO2 oxidation to particulate
  SO4, secondary organic aerosol formation by
  oxidation of 6 VOC groups including monoterpenes
• Gas-phase Solver Modified Euler Backwards
  Integration
• Particulate dynamics using the modal approach
• Kuo-Anthes cloud scheme for deep convection
• Shallow convection scheme and aqueous chemistry
  in clouds as in the Regional Acid Deposition
  Model (RADM )
• Size-dependent dry and wet removal algorithms

5
Overview of the Simulations

• Analysis Period
• 62 days of CMAQ simulations (January and July,
  1996)
• Compared model predictions for all PM species and
  visibility metrics with IMPROVE network
  measurements to evaluate model performance
• on days for which measurements are reported
  (January and July 10, 13, 17, 20, 24, and 27,
  1996)
• on an event average basis
• excluded 31st due to lack of 24-hr output (output
  time-shifted to PST)

6
Overview of the Simulations (contd)

• Boundary Conditions (BCs)
• default BCs from the REgulatory Modeling System
  for Aerosols and Deposition (REMSAD)
• choice of BCs based on earlier sensitivity tests
  for better inter-model comparison between REMSAD
  and CMAQ
• Time-independent
• SO42- reduced from 1.2 mg/m3 to 0.3 mg/m3 based
  on CARB measurements of background aerosol in
  coastal areas, and NH3 reduced from 0.3 ppb to
  0.1 ppb
• Emissions
• Wildfires included
• NH3 reduced by 50 over the whole domain for the
  winter months based on reported uncertainties
  from prior studies by the EPA ORD

7
Surface Level CMAQ NH3 Emissions January Average
1996 - Base
8
Sulfate Response to NH3 and BC Changes
Base NH3 Emissions, BCs
50 Base NH3 Emissions, New BCs
9
Aerosol NO3 to Total NO3 Ratio in January
Base NH3 Emissions, BCs
50 Base NH3 Emissions, New BCs
10
Bias vs. IMPROVE SO4 and NO3 January 1996
11
Daily Average Nitrate January 1996
January 13
January 17
January 24
January 27
12
PBL Heights and Total Nitrate January 13 1996
Columbia River Gorge
Yellowstone
Bridger W
PBL Height (m) Nitrate x 100 (mg/m3)

13
PBL Heights and Total Nitrate January 13 1996
(contd)
Upper Buffalo
Lone Peak
Pinnacles NM
PBL Height (m) Nitrate x 100 (mg/m3)
14
PBL Height vs. Nitrate Bias January 1996
January 17
3000
Nighttime avg.
Daytime avg.
2500
2
y 1e03 - 5.4e02x R
0.28
2
y 1.2e03 - 5e02x R
0.17
2000
1500
1000
500
0
-2
-1.5
-1
-0.5
0
0.5
1
1.5
3
D
NO
(CMAQ - Obs) (
m
g/m
)
3
15
PBL Height vs. Nitrate Bias January 1996 (contd)
16
MM5 Wintertime PBL Height Predictions

• Wintertime PBL heights not well-examined against
  obs data in previous analyses
• MM5 simulations performed in 5-day chunks
• Snow cover fields have crude spatial resolution,
  are updated only once a week, and remain in
  effect through each five-day period
• Could contribute to varying degrees of
  underestimation in PBL heights at different
  periods most significant on the worst days of
  overprediction
• Simulations used MRF improved land-surface
  models available in MM5 and could provide better
  surface temperature and PBL predictions over
  water bodies and snow cover

17
January NO3 Bias in Different Chemical Regimes
Free NHx / Total Nitrate (NH3 NH4
2SO42-) / (HNO3 NO3-) Ratio gt 1.0
NO3 formation limited by HNO3 lt
1.0 NO3 formation limited by NH3
18
Surface Level NHx/Total Nitrate in January
Base NH3 Emissions, BCs
50 Base NH3 Emissions, New BCs
19
SO4 Response to Change in Emissions, BCs
January Avg DSO4
January Avg Cloud Fraction
20
Event-Average NO3 and Bias January 1996
21
Understanding the NO3 Bias

• NHx/total nitrate ratio best applies to closed
  systems
• Biases highest for high values of the ratio,
  i.e., HNO3-limited regime HNO3 too high in
  some locations
• Some NH3 source regions become more HNO3-limited
  possible offsetting role of SO4 reductions
• Need observations of NH4, NH3 and HNO3 to help
  further evaluation (compute observed ratio)
• Need to isolate effects of BC changes from the
  effects of NH3 emissions reductions
• Aerosol nitrate to total nitrate ratio should be
  compared with observations (e.g., CASTNet)

22
Who are the Bad Guys?
23
Comparison with IMPROVE PM2.5 and PM-Coarse
24
Comparison of Area PM10 Emissions from WRAP and
NEI Inventories
25
PM-Coarse Deposition and Emission Fluxes (Domain
Average)
January 13
July 13
Deposition Flux (gm/s)
Emission Flux (gm/s)
26
PM-Coarse Deposition and Emission Fluxes(Domain
Average)
January 27
July 27
27
Summary

• Biases in nitrate tend to be anti-correlated with
  PBL height for large biases less of a trend for
  smaller biases
• PBL height and ground temperature show anomalous
  behavior at one location nitrate bias
  correspondingly very high
• Ammonia emission reductions have a strong impact
  on both the SO4 and NO3 concentrations, and on
  the chemical regime
• Ammonia reductions have less of an impact on the
  nitrate bias if the regime is severely
  HNO3-limited
• Positive nitrate bias is not systematic, and may
  be due to transport or overestimates of NOx
  emissions at such locations

28
Summary (contd)

• Coarse mode deposition and emission fluxes are
  consistent with predicted concentrations on a
  domain-average basis
• Little or no day-to-day variability in emission
  fluxes, probably due to exclusion of wind-blown
  dust
• More variability in deposition fluxes during the
  daytime in January, and between January and July

29
Recommendations

• Future MM5 simulations should use a land surface
  model option to better predict ground
  temperature and PBL heights over water and snow
  cover
• NOx emission sensitivity studies, along with
  comparisons of total nitrate and NHx against
  measurements would help characterize the source
  of the most severe overpredictions in nitrate
• Additional sensitivities could examine the effect
  of NH3 emissions reductions without the
  confounding influences of BC changes on the
  nitrate bias

30
Recommendations (contd)

• Coarse mass dry deposition measurements should be
  compared with model predictions to determine the
  source of the coarse mass underprediction
• The effect of including wind-blown dust emissions
  on the model predictions should be evaluated