COMPARATIVE MODEL PERFORMANCE EVALUATION OF CMAQ-VISTAS, CMAQ-MADRID, AND CMAQ-MADRID-APT FOR A NITROGEN DEPOSITION ASSESSMENT OF THE ESCAMBIA BAY, FLORIDA.

COMPARATIVE MODEL PERFORMANCE EVALUATION OF CMAQ-VISTAS, CMAQ-MADRID, AND CMAQ-

MADRID-APT FOR A NITROGEN DEPOSITION ASSESSMENT OF THE

ESCAMBIA BAY, FLORIDA WATERSHED

6th Annual CMAS ConferenceChapel Hill, NC

1-3 October 2007

Presented by Jay Haney ICF International, San Rafael, CA

Co-Authors: Sharon Douglas Tom Myers

Justin Walters John Jansen

Krish Vijayaraghavan AER

Project sponsored by Southern Co.

ICF

Southern Company

Background/Objectives

Atmospheric deposition of nitrogen is a source of contamination in Escambia Watershed

Air quality modeling performed to estimate change in nitrogen deposition in watershed due to controls at a local EGU as part of larger combined air/water quality modeling analysis

Objective for this part of study: Assess the ability of air quality models to replicate observed gaseous and particulate concentrations and wet and dry deposition

Air Quality Models Used:Based on CMAQ 4.5.1

CMAQ-VISTAS: CB-IV, AERO4, modified SOA by VISTAS

CMAQ-MADRID: Sectional representation of particle size distribution as opposed to modal for CMAQ

CMAQ-MADRID-APT: “Advanced plume treatment” based on SCIPUFF with CHEMistry – SCICHEM

Air Quality Modeling Databases

Meteorological inputs: VISTAS 2002 inputs from RPO modeling analysis

Emissions: CMAQ-VISTAS: Base_G1 MADRID & APT: Base_F

Domain: 12-km ALGA, subset of VISTAS domain centered on Alabama & Georgia

Annual simulations for 2002

CMAQ ALGA Subdomain/Escambia Watershed

Oak GroveOLF

Gulfport Pensacola

Centreville

N. BirminghamJefferson St.

Yorkville

Oak GroveOLF

Gulfport Pensacola

Centreville


Yorkville

CMAQ ALGA Subdomain

Escambia Watershed

Plant Crist

Air Quality Data Used in Evaluation

SEARCH: Hourly gaseous and 3-day speciated PM2.5 concentrations

IMPROVE: 3-day speciated PM2.5 concentrations

CASTNET: Weekly particulate concentrations and derived dry deposition based on concentration/ambient conditions

NADP: Weekly particulate concentrations and wet deposition

Model Performance Measures

Mean bias, normalized bias, fractional bias, mean error, normalized gross error, and fractional gross error

Paired for appropriate time interval

Statistics calculated using daily averages, except for CASTNET and NADP weekly measurements

Statistics calculated for all sites/species in ALGA domain with focus on sites near Escambia watershed

Location of SEARCH and CASTNET Sites in CMAQ

Subdomain

SEARCH Sites CASTNET Sites

Coffeeville

Sand Mountain

Georgia Station

Sumatra

Coffeeville

Sand Mountain

Georgia Station

Sumatra

Oak GroveOLF

Gulfport Pensacola

Centreville


Yorkville

Oak GroveOLF

Gulfport Pensacola

Centreville


Yorkville

Location of IMPROVE and NADP Sites in CMAQ Subdomain

IMPROVE Sites NADP Sites

St. Marks

Sipsey

Chassahowitzka

Cohutta

Okefenokee

CapeRomain

Linville Gorge

Shining Rock

Great Smoky MtnCadiz

Mammoth Cave

Breton

St. Marks

Sipsey

Chassahowitzka

Cohutta

Okefenokee

CapeRomain

Linville Gorge

Shining Rock

Great Smoky MtnCadiz

Mammoth Cave

Breton

Mobile Co.

Baldwin Co.

Quincy

SumatraMobile Co.

Baldwin Co.

Quincy

Sumatra

Results for Gaseous Species: SO2 for SEARCH Sites

Mean Observed & Simulated SO2

0

2

4

6

8

PNS OLF GFP OAK CTR BHM YRK JST

ppb

OBS

CMAQ

MADRID

APT

Results for Gaseous Species: SO2 for SEARCH Sites

Fractional Bias & Error: SO2

-100-80-60-40-20

020406080

100


Perc

ent (

%)

MFB-CMAQ

MFE-CMAQ

MFB-MADRID

MFE-MADRID

MFB-APT

MFE-APT

Results for Gaseous Species: HNO3 for SEARCH Sites

Mean Observed & Simulated HNO3

0.0

0.5

1.0

1.5

2.0


ppb

OBS

CMAQ

MADRID

APT

Results for Gaseous Species: HNO3 for SEARCH Sites

Fractional Bias & Error: HNO3

-200-160-120

-80-40

0

4080

120160200


Perc

ent (

%)

MFB-CMAQ

MFE-CMAQ

MFB-MADRID

MFE-MADRID

MFB-APT

MFE-APT

Gaseous Species Summary

For SO2, all models slightly underestimate concs nearby (evidence of differences between MADRID and APT in Atlanta area)

For HNO3, all models consistently overestimate at nearby sites

For NO2, all models do well and for NO, all models underestimate, but these are typically not major contributors to nitrogen deposition

Results for Particulate Species: NO3 for SEARCH Sites

Mean Observed & Simulated NO3

0.0

0.4

0.8

1.2

1.6

2.0


ugm

-3

OBS

CMAQ

MADRID

APT

Results for Particulate Species: NO3 for SEARCH Sites

Fractional Bias & Error: NO3

-200-160-120

-80-40

0

4080

120160200


Perc

ent (

%)

MFB-CMAQ

MFE-CMAQ

MFB-MADRID

MFE-MADRID

MFB-APT

MFE-APT

Results for Particulate Species: NH4 for SEARCH Sites

Mean Observed & Simulated NH4

0

1

2

3

4


ugm

-3

OBS

CMAQ

MADRID

APT

Results for Particulate Species: NH4 for SEARCH Sites

Fractional Bias & Error: NH4

-100-80-60-40-20

0

20406080

100


Perc

ent (

%)

MFB-CMAQ

MFE-CMAQ

MFB-MADRID

MFE-MADRID

MFB-APT

MFE-APT

Particulate Species Summary

For nitrate, CMAQ better simulates mean conc. but fractional bias and error are lower for MADRID and APT at nearby sites

For ammonium, all models show consistent underestimation at nearby sites, and overestimation at BHM and ATL

Results for Dry Deposition: NO3 for CASTNET Sites

Fractional Bias & Error: NO3 Dry Dep.

-200-160-120

-80-40

04080

120160200

Coffeeville, MS Sand Mtn, AL Georgia Station, GA Sumatra, FL

Perc

ent (

%)

MFB-CMAQ

MFE-CMAQ

MFB-MADRID

MFE-MADRID

MFB-APT

MFE-APT

Results for Dry Deposition: NH4 for CASTNET Sites

Fractional Bias & Error: NH4 Dry Dep.

-200-160

-120-80-40

0

4080

120

160200

Coffeeville, MS Sand Mtn, AL Georgia Station, GA Sumatra, FL

Perc

ent (

%)

MFB-CMAQ

MFE-CMAQ

MFB-MADRID

MFE-MADRID

MFB-APT

MFE-APT

Results for Dry Deposition: HNO3 for CASTNET Sites

Fractional Bias & Error: HNO3 Dry Dep.

-200-160-120-80-40

04080

120160200

Coff eeville, MS Sand Mtn, AL Georgia

Station, GA

Sumatra, FL

MFB-CMAQ

MFE-CMAQ

MFB-MADRID

MFE-MADRID

MFB-APT

MFE-APT

Results for Wet Deposition: NO3 for NADP Sites

Fractional Bias & Error: NO3 Wet Deposition

-100-80-60

-40-20

02040

6080

100

Baldwin Co., AL Mobile Co., AL Quincy, FL Sumatra, FL

Perc

ent (

%)

MFB-CMAQ

MFE-CMAQ

MFB-MADRID

MFE-MADRID

MFB-APT

MFE-APT

Results for Wet Deposition: NH4 for NADP Sites

Fractional Bias & Error: NH4 Wet Deposition

-100-80

-60-40-20

0

204060

80100

Baldwin Co., AL Mobile Co., AL Quincy, FL Sumatra, FL

Perc

ent (

%)

MFB-CMAQ

MFE-CMAQ

MFB-MADRID

MFE-MADRID

MFB-APT

MFE-APT

Dry Deposition Summary

For nitrate and ammonium dry deposition, all models show consistent gross underestimation

For HNO3 dry deposition, all models show consistent overestimation, with MADRID and APT showing more overestimation than CMAQ

With HNO3 higher than NO3 (simulated and observed), net result is that all models overestimate dry deposition of nitrates

Dry deposition estimates complicated by potential differences in meteorology used for data vs. model

Wet Deposition Summary

Models do better in simulating wet deposition and are consistent in underestimating wet deposition at nearby sites

Larger differences seen between models: effects of plume-in-grid treatment for APT?

Summary and Key Findings

Results are mixed: none of the models stand out as better performing

Greatest contributor to nitrogen deposition is dry deposition of HNO3, followed by wet deposition of nitrate (all forms)

Simulated net wet deposition of nitrogen is lower than observed while simulated net dry deposition is higher, so total loading of nitrogen in domain may be adequately simulated

Summary and Key Findings

Dry deposition monitoring not available in Escambia watershed, so performance may not be representative

Deposition output from all three models was used in water quality modeling assessment

COMPARATIVE MODEL PERFORMANCE EVALUATION OF CMAQ-VISTAS, CMAQ-MADRID, AND CMAQ-MADRID-APT FOR A NITROGEN DEPOSITION ASSESSMENT OF THE ESCAMBIA BAY, FLORIDA.

Documents

search sitesresults

sitesnadp sitesresults

fractional bias

ability of air quality

alga domain

normalized bias

fractional gross error

mean error