Adaptation and Application of the CMAQ Modeling System for Real-time Air Quality Forecasting During the Summer of 2004

R. Mathur, J. Pleim, T. Otte, K. Schere, J. Young, G. Pouliot, B. EderAtmospheric Sciences Modeling Division, ARL/NOAA, NERL/U.S. EPA

D. Kang, S. Yu, H.-M. LinScience and Technology Corporation

J. McQueenNational Centers for Environmental Prediction

P. Lee, M. TsidulkoScience Applications International Corporation

D. WongLockheed Martin Information Technology

Although this work was reviewed by EPA and approved for publication, it may not necessarily reflect official Agency policy

Eta-CMAQ AQF System

Eta-12Eta-12

CMAQCMAQ

Eta PostEta Post

PRDGENPRDGEN

PREMAQPREMAQ

AQF PostAQF Post

Verification ToolsVerification Tools

Vertical interpolation from eta to sigma

Horizontal interpolation to Lambert grid

CMAQ-ready meteorology and emissions

Gridded ozone files for users

Chemistry model

Meteorology model

Performance feedback for users and developers

CMAQ Configuration

• Structural– Netcdf replaced with binary IOAPI

• Advection– Horizontal: Piecewise Parabolic Method– Vertical: Upstream with rediagnosed vertical velocity

to satisfy mass conservation

• Turbulent Mixing– K-theory; PBL height from Eta

– New scheme for specification of minimum Kz

CMAQ Configuration (contd.)

• Gas phase chemistry– CB4 mechanism with EBI solver

• Cloud Processes– Mixing and aqueous chemistry: following the scheme in RADM

• Deposition– Dry : M3dry modified to use Eta land surface parameters– Wet

• Aerosols– 2004 release version

166

142

268 grid cells

259gridcells

Northeast“1x” Domain

East “3x” Domain

CMAQ Modeling Domains

Ozone forecasts on 3x and 1x Experimental PM forecasts on 3x

Lateral Boundary Condition Specification

A key uncertainty in long term modeling over limited area domains– Determines “model background”

• Default profiles– “Clean” tropospheric background values– Used in 1x

• Seasonal Profiles– Derived from continental CMAQ simulations for 2001– Used in 3x

Lateral Boundary Conditions (contd.)

• Ozone profiles from NCEP’s Global Forecast System (GFS)– O3 is a 3-d prognostic variable

– Initialized with Solar Backscatter Ultra-Violet (SBUV-2) satellite observations

– Motivation• Simulating varying dynamical conditions

• Improve model representation of O3 in the free troposphere

– Effects associated with intrusions

– Study FT-BL exchange mechanisms

Specification of Minimum Kz

• Minimum value of Kz allowed to vary spatially depending on urban fraction (furban)

• Kz = 0.1 m2/s, furban = 0• Kz = 2.0 m2/s, furban = 1

– allows min. Kz in rural areas to fall off to lower values than urban regions during night-time; mimics urban heat island effects

– prevents precursor concentrations (e.g., CO, NOx) in urban areas from becoming too large at night

– lower Kz (and reduced mixing intensity) in non-urban areas results in increased night-time O3 titration

• Helps reduce night time over predictions of ozone “regionally”

Summer 2004: Atypical Ozone Season

July 21

Aug. 12Source: EPA AIRNOW

Model Performance Characteristics: Summer 2004

Bias

Effects of GFS Ozone and Cloud Mixing

O3

Vertical profile Cloud top

GFS Sensitivity Simulations

• Limit GFS use to above a specified altitude– 6 km– 10 km

• Motivation: to limit the use of GFS derived O3 profiles to the upper levels of the model, where there is greatest confidence in GFS predictions and to avoid abnormally high O3 within the boundary layer (noticed in early parts of May)

• Default BC (without GFS)

Target Day Stats: May 18, 2004

Impact of GFS 6km and min. Kz change1x domain

Solid lines-with changes Dashed lines-without changes

Implemented in 1x domain on 7/20/04

Hourly 1 Hr. Max 8 Hr. Max

Comparison of 3x and 1x PerformanceAt sites within the 1x domain

Comparison of 3x and 1x Lateral Boundary Conditions

Solid lines (3x); filled circles and dash line (1x)

CMAQ 2001 Performanceat 3x Western Boundary

Solid lines: seasonal BCsDash/dots: Default profile

“Switch-Off” top-down cloud mixing

Tropopause limiton cloud top

Diagnostic Tests: Cloud Mixing

Diagnostic Tests: Cloud Mixing Target Day Stats: May 18, 2004

FSIDE: switch-off top-down mixingTROPLIM: tropopause limit on cloud top

Diagnostic Tests: Cloud effects on Photolysis

O3 (ppb)

Cloud Fraction: Current (average)

Cloud Fraction: Modified (max)

Scale by Radiation reaching the surface 1-(J/Jclear)

Photolysis Attenuation August 12, 2004

Effects of Cloud Process ModificationsMaximum Reductions in O3

August 12, 2004

Below cloud attenuation based onradiation

Below cloud attenuation based onRadiation + switch-off top-down mixing+New CFRAC

Effects of Cloud Process ModificationsAugust 12, 2004

rad_atten: cloud mixing + photolysis attenuation modifications

Comparison with Previous Day PersistenceMax. 1 Hr. Ozone

Persistence Eta-CMAQ

Correlation coefficient plots from S. McKeen

PM2.5 Forecast Comparisons with AIRNOW : Preliminary Daily Average

August 15 August 16 August 17

Summary

• Lateral boundary conditions play a dominant role in regulating modeled O3 background levels– Critical when ozone levels are relatively low as in the past two

summers– Higher O3 BC led to a systematic higher bias in the 3x simulation

• Careful consideration needs to be given in deriving LBCs from larger scale models– Are conditions representative?

• Bias propagation

– Consistent coupling

• Over-predictions at low O3 range related to representation of cloud processes– Top-down mixing– Photolysis attenuation

Looking ahead ….

• Methods to improve coupling between models– Boundary conditions

• GFS, Eta, CMAQ• Layer structure and model top to improve representation of

tropopause dynamics• Comparison of model (GFS, CMAQ) and observed free-

tropospheric O3 values

– Radiation• Photolysis attenuation

– Boundary layer mixing• Revisit Eta-Kh

– Transition to WRF• Advection on E-grid

– Minimize interpolations

Looking ahead ….

• Continue testing alternate formulations over a wider range of conditions – Mixing

– Below cloud washout (low observed O3 during precipitation events)

• Assessment of experimental PM forecasts results– Surface data: AIRNOW, IMPROVE, CASTNET– Satellites

• AOD