Source-Specific Forecasting of Air Quality
Impacts with Dynamic Emissions Updating
&
Source Impact Reanalysis
Yongtao Hu1, Cesunica Ivey1, M. Talat Odman1, Peter Vasilakos,
Michael E. Chang2 and Armistead G. Russell1
1School
of Civil & Environmental Engineering,
2Brook Byers Institute of Sustainable Systems
Georgia Institute of Technology
With thanks to Pius Lee and the NOAA ARL Forecasting Team
AQAST 9th Semi-Annual Meeting, June 3rd, 2015
Georgia Institute of Technology
Motivation
• Air Quality Forecasting
–
–
Source specific air quality impacts used in air quality management
•
Both for “static” and, in the future, targeted “dynamic” air quality
management, and for health related public and individual exposure
management.
Current source specific air quality impact forecasting/hindcasting
accuracy call for improvement.
•
Prediction with 3-D models relies on accuracy of emissions
–
–
Uncertain (particularly for smaller sources)
Change with time
• Source impact assessment with chemical reanalysis
–
Desire to integrate advanced, hybrid source impact assessment as
part of chemical re-analysis and emissions updates
• Improving model performance
–
–
Most CTMs currently are biased low when simulating organic
aerosol (OA) during the summer
Linked to secondary (SOA) formation
•
Experiments suggest biogenic SOA formation underestimated
Georgia Institute of Technology
Objective
• Provide information that can assist air quality
management and exposure assessment
–
–
–
Source impact forecasting in addition to air quality forecasting
•
Critical for dynamic air quality management
•
Utilize measurements (gases, PM, AOD and PM composition)
Improve air quality and source impact forecasting accuracy using
near real time measurements through dynamic adjustment of
emissions
Improve source impact hindcasting
• Work with Georgia EPD, Forest Service, Georgia Forestry
Commission, Atlanta Regional Commission, and EPA on
use of the products and information
• Provide source-specific impacts as part of Tiger Team
reanalysis
Georgia Institute of Technology
Source-Specific Air Quality Impact
Forecasting
Georgia Institute of Technology
Hi-Res2 Forecasting System
Forecasting Air Quality
for CONUS
(https://forecast.ce.gatech.edu open
since November 28th,2014)
• Updated base emissions to
2011NEI
• WRF3.6.1 and CMAQv5.02 used
• 72-hour forecasts at 4-km
resolution for Georgia and
surrounding states, 12-km for
most of Eastern states and 36-km
for the rest of CONUS
Georgia Institute of Technology
Forecasting Source Impacts at 4-km for Georgia
(https://forecast.ce.gatech.edu open since November 28th,2014)
Georgia Institute of Technology
Hi-Res2: Online Auto-Adjustment Inverse
Modeling Approach for Adjusting Emissions
An emissions and air quality auto-correction system
utilizing near real-time satellite and surface observations
Currently working with
measurements at ~20 sites in
Georgia and Soon with
MODIS C6 AOD
• Minimizes the differences
between forecasted and
observed concentrations
• Minimal adjustment to source
emissions
• Uses impacts of emission
sources calculated by CMAQDDM-3D
– Source impacts can be used for
dynamic air quality
management.(e.g., traffic and fires)
Georgia Institute of Technology
Inverse Model Formulation
• Solve for the Adjustment Factors, Rj, that minimize 2
DDM-3D calculated sensitivity of
concentration i to source j emissions
emission adjustment ratio
2


J

  ciobs  cisim   Si , j ( R j  1)  


N
J (ln R ) 2


j 1
j

2
  


  2
2
 C obs
i 1
j 1
ln R j


i

 weight


 Ci
 Rj 2
2
Remaining Error
Amount of Change
in Source Strengths
L-BFGS algorithm is used for the optimization (R package nloptr)
Georgia Institute of Technology
Hindcasting Air Quality Impacts from
Specific Sources: 2010 Nationwide Source
Apportionment Effort within the Tiger
Team Project of Operational Chemical
Reanalysis
Georgia Institute of Technology
Spatial Hybrid Approach (Ivey et al, GMDD, 2015)
1. CMAQ-DDM Source Impacts (~30
sources, daily)
2. Hybrid Analysis at Monitors to find
Adjustment Factors (Rj’s)
Hybrid Adjustment Factors
(Atlanta, GA 1/24/2004)
Rjs
2
0
Rjs
2
0
Rjs
2
0
January 4, 2004
4. Temporal Interpolation of Adjustment
Factors
Monitor
Rj
3. Spatial Interpolation of Adjustment
Factors (Kriging)
1.2
1
0.8
0.6
0.4
0.2
0
Adjustment Factors (Example)
1/4 1/6 1/8 1/10 1/12 1/14 1/16 1/18 1/20 1/22 1/24 1/26 1/28
5. Adjust CMAQ-DDM Spatial Fields (Daily, Spatially Dense)
Original Woodstove Impact
Interpolated
Woodstove Adjustment Factors
Adjusted Woodstove Impact
2010 Annual PM2.5 Source Apportionment
WRF-CMAQ model physics and chemistry options
WRF-ARW
Both North America (12 km) & CONUS (4
km)
Map projection &
grid
Lambert Conformal & Arakawa C
staggering
Vert. co-ordinate
42 σ-p unevenly spaced levels
advection
RK3 (Skamarock and Weisman (2008))
SW & LW radiation
PBL Physics
RRTMG (Iacono et al. 2008))
Mellor-Yamada-Janjic (MYJ) level 2.5
closure
Surface layer scheme
Monin-Obukhov Similarity with viscous
sub-layer
Land Surface Model
NCEP NOAH
Cloud Microphysics
Thompson et al. (2008)
Cloud convective
mixing
CMAQ4.7.1
Map projection & grid
Vert. co-ordinate
Gas chemistry
Aerosol chemistry
Anthropogenic
emission
12 km nested to 4 km
Betts-Miller-Janjic Mass adjustment
Both CONUS(12 km) & SENEX (4 km)
Lambert Conformal & Arakawa C staggering
42 σ-p unevenly spaced levels
Cb05 with 156 reactions
Aero5 with updated evaporation enthalpy
2005NEI as base year, mobile projected using AQS*, area
and off-road used CSPR^, point source uses 2012 CEM data
WRAP oil and gas emissions data
Biogenic emission
Lateral BC
BEIS-3.14
RAQM (B. Pierce)
42 vertical layers
Biomass Impact: Before/After
Assimilation of Observations
• Many sources are
highly variable leading
to significant
differences between
observations and
simulations
– Biomass
– Dust
– Agriculture
• Seasonal results in
better agreement
– Suggests inventories
reasonable “on average”
CMAQ-DDM PM2.5 Biomass Burning Impact
10
8
04/01/04
6
4
2
0
Adjusted PM2.5 Biomass Burning Impact
10
8
6
4
2
0
2006 Seasonally-Averaged PM2.5 Impacts (ug/m3):
Biomass Burning (2006)
SUMMER
WINTER
CMAQ-DDM
Spatial Hybrid
CMAQ - Spatial Hybrid
Updated isoprene OA Mechanism
• Comparison of
CMAQ-simulated
SOA from isoprene to
observations from
AMS factors led to
updates
• Assimilated PBL
measurements
• Improved IEPOX physics
and chemistry:
•
•
•
Dry deposition resistance reduced
Updated reaction rate constants
Modified Henry’s law for IEPOX
• Not yet in forecasting system
Summary
• PM and ozone forecasting system (Hi-Res2) operational with source
impact forecasting and dynamic emission adjustments
– Supports dynamic air quality management through providing source specific
information
– Currently for traffic, power plant and prescribed-burn (Talat’s talk) emissions
– Expansion to include other species measurements underway
– Improved approach to assimilating AOD and PM measurements underway
(Utilizing data-fused fields)
• Hybrid approach for source apportionment
– Utilizes gases concentration and PM composition measurements
– Application nationwide for 2010: Tiger Team Chemical Reanalysis project
– Results for air quality management (EPD, ARC) and health (CDC)
• Extension and update of biogenic SOA formation improves SOA results
– Isoprene oxidation and NO3+terpene reactions
Acknowledgements
•
•
•
•
NASA
Georgia EPD
Georgia Forestry Commission
US Forest Service
– Scott Goodrick, Yongqiang Liu, Gary Achtemeier
• Environmental Protection Agency (EPA)
RD83521701
• Atlanta Regional Commission (ARC)
Georgia Institute of Technology