PBL Modeling for Air Quality Models

Jonathan PleimU.S. EPA/ORD/NERL

NOAA/OAR/ARLAtmospheric Modeling Division

Research Triangle Park, NC

`

Outline• Overview of PBL modeling

Interdependencies with land surface model (LSM), and radiation and clouds

• Our approach in WRF-CMAQ systemNew PBL and LSM in WRFConsistent PBL and surface flux in CMAQ

• Description and testing of the Asymmetric Convective Model v2 (ACM2)

• Model results in WRF and CMAQ• PBL modeling issues

Goal: To model effects of PBL processes on air quality

1. AQM PBL should be consistent with PBL model in Met model

Same subgrid transport schemeSame PBL heightSame grid structure (horizontal and vertical)

2. Met model PBL model should realistically represent fluxes and profiles of heat, momentum, and all tracers

3. Land surface model should produce realistic representation of surface fluxes of heat, momentum, and moisture

4. Chemical surface fluxes should be consistent with LSM (e.g. Rst, Ra)

WRF PBL/Surface Fluxes

• PBL/Surface model components are in three separate, but interdependent modules in WRF

1. Land surface model – computes soil moisture and temperature and moisture fluxes from ground, wet leaves, evapotranspiration. Added PX LSM

2. Surface layer – Solves flux-profile relationships. Outputs: L, u*, Ra. Added Pleim (2006) surface layer scheme

3. PBL – Computes subgrid turbulent vertical transport. Added ACM2

PBL Modeling for Met and AQ

• Need simple computationally efficient schemes (generally 1st order)

• Stable:Local diffusion that’s responsive to wind shear, stability, and eddy length scale

• Unstable:Local and non-local transport scaled by surface heat flux and depth of PBL

Convective conditions• Local flux-gradient proportionality (i.e. Eddy diffusion)

is not appropriate for Convective Boundary LayersUpward heat flux penetrates to ~80% of h while potential temperature gradients are very small through most of the PBLEddies in CBL are larger that vertical grid spacing (local closure is not appropriate)

• Two common alternative approaches:1. Gradient adjustment term:

⎟⎠

⎞⎜⎝

⎛⎟⎠⎞

⎜⎝⎛ −∂∂

∂∂

=∂∂

hh zK

ztγθθ Deardorff 1966,Troen and Mahrt 1986,

Holstlag and Boville 1993, Noh et al. 2003

2. Transilent or non-local closure:

jiji M

tθθ =

∂∂ Stull 1984, Blackadar 1976,

Pleim and Chang 1992

Asymmetric Convective Model version 2 (ACM2)

• Combined local and nonlocal closure for convective conditions

• Simple Transilient model (asymmetric)Rapid upward transport by convectively buoyant plumes Gradual downward transport by compensatory subsidence

• Combined with local eddy diffusionAllows local mixing at all levelsMore realistic (continuous) profiles in lower layersSmooth transition from stable to unstable

• Mass flux scheme is equally applicable to heat, momentum, and any tracer

ACM ACM2

Partitioning local and non-localMixing rate, defined by Kz, is partitioned into local andnon-local components

( ) MfzhzzKf

Mu convzconv =−∆

=+

+

21

21

11

1 )(

( )convzi fzKK −= 1)(

The Question is: How to define fconv? Clearly it should increase with increasing instability, but what is its upper limit for free convection?

First a sensitivity study for convective conditions

309.5 310 310.5 311 311.5 312 312.5 313

0

500

1000

1500

2000

Eddy25%50%75%ACM1

Theta (K)

1-D experiments –

Variations in partitioning of local and non-local transport

-0.2 0 0.2 0.4 0.6 0.8 1 1.2

0

0.2

0.4

0.6

0.8

1

1.2

H/Hsfc

θ-θmin

H/Hscf

, potential temperature (K)

Normalized heat flux and potential temperature profiles

LES (Stevens 2000) ACM2

Non-local partitioning

• These tests suggest that the upper limit of fconv should be about 50%

• An expression for fconv can be derived from gradient adjustment models (e.g. Holstlag and Boville 1993) at top of surface layer:

13

13

2

1.01

−−−

⎟⎟⎠

⎞⎜⎜⎝

⎛⎟⎠⎞

⎜⎝⎛−+=

Lh

akfconv

Non-local fraction (fconv) as function of stability

0

0.1

0.2

0.3

0.4

0.5

0.6

0 5 10 15 20

-h/L

Testing and evaluation of ACM2

• LES experiments• Participation in the GABLS2 experiment• Evaluation of MM5 w/ ACM2 and PX LSM• Evaluation of WRF w/ ACM2 and PX LSM• Testing in CMAQ

LES experiment low heat flux (Q* = 0.05 K m s-1), weak cap

-0.05 0 0.05 0.1 0.15 0.2 0.250

200

400

600

800

1000

1200

1400Kinematic heat flux - 24SC

LESACM2-localACM2-nonlocalACM2-total

kinematic heat flux (K-m/s)

Profiles of sensible heat flux. Local, non-local, and total sensible heat flux compared to LES

Comparisons of ACM2 PBL Heights to Observations

• MM5 and WRF simulations using ACM2 and the PX LSM

• PBL heights derived for radar wind profilers by Jim Wilczak and Laura Bianco NOAA/ESRL

• 2 profiler site from ICARTT 2004• 10 radar sites in Texas area - TexAQS

II, 2006• Observations and models averaged by

hour of the day

PBL Height averaged from July 13 to August 18, 2004

Concord, NH

0

500

1000

1500

2000

12 14 16 18 20 22

ETAMM5ACM2OBS

Hour (UT)

0

500

1000

1500

2000

12 14 16 18 20 22

ETAMM5ACM2OBS

Hour (UT)

Pittsburgh, PA

PBL height averaged for August 2006

0

500

1000

1500

2000

10 12 14 16 18 20

aclSTATS

OBSNOAHYSUPXACM2

Hour (LT)

0

500

1000

1500

2000

8 10 12 14 16 18 20 22

bhmSTATS

OBSNOAHYSUPXACM2

Hour (LT)

0

500

1000

1500

2000

2500

3000

8 10 12 14 16 18 20

mdySTATS

OBSNOAHYSUPXACM2

Hour (LT)

0

500

1000

1500

2000

2500

8 10 12 14 16 18 20 22

lvwSTATS

OBSNOAHYSUPXACM2

Hour (LT)

0

200

400

600

800

1000

1200

1400

8 10 12 14 16 18 20

lptSTATS

OBSNOAHYSUPXACM2

Hour (LT)

0

500

1000

1500

2000

8 10 12 14 16 18 20 22

hveSTATS

OBSNOAHYSUPXACM2

Hour (LT)

0

500

1000

1500

2000

2500

3000

8 10 12 14 16 18 20

cleSTATS

OBSNOAHYSUPXACM2

Hour (LT)

0

200

400

600

800

1000

1200

1400

1600

8 10 12 14 16 18 20 22

bpaSTATS

OBSNOAHYSUPXACM2

Hour (LT)

Further WRF Development

• Implement new, high resolution, more accurate 2001 National Land Cover Data

Based on 30 m Landsat-7 ETM• WRF-CMAQ 2-way “On-line” system

under development Beta release September 2008

WRF12km Grids and NLCD DataResearch Triangle Park, NC

CMAQ

• PBL options:Eddy diffusion

• Boundary layer scaling approach • PBL height read from Met model• Default option up-through version 4.5 (2005)

ACM2• Local and nonlocal transport identical to ACM2 in

Met model • Default option starting for version 4.6 (2006)

• Surface fluxesDry deposition and bidirectional fluxes (ammonia and mercury) use Ra and Rstdirfectly from LSM in Met model

Eddy vs ACM2

Hourly COMaximum 1-hr Ozone

O3 Vertical Profiles for urban andrural grid cells

40 50 60 70 80 90 1000

500

1000

1500

2000

2500

3000

Aug 1, 16-22Z - R97C48

O3 ACM2O3

O3

45 50 55 60 650

500

1000

1500

2000

2500

3000

Aug 1, 16-22Z - R97C88

O3 ACM2O3

O3

NOx

0 5 10 15 200

500

1000

1500

2000

2500

3000

Aug 1, 16-22Z - R97C48

NOX ACM2NOX

NOX

0 0.2 0.4 0.6 0.8 1 1.2 1.40

500

1000

1500

2000

2500

3000

Aug 1, 16-22Z - R97C88

NOX ACM2NOX

NOX

CO

50 100 150 200 250 300 3500

500

1000

1500

2000

2500

3000

Aug 1, 16-22Z - R97C48

CO ACM2CO

CO

65 70 75 80 85 900

500

1000

1500

2000

2500

3000

Aug 1, 16-22Z - R97C88

CO ACM2CO

CO

Ozonesonde at BeltsvilleJuly 19, 16 Z

295 300 305 310 315 3200

500

1000

1500

2000

2500

3000

3500

4000

Obs719beltsville 16Z?

OTHTMTHT

OTHT

0 2 4 6 80

500

1000

1500

2000

2500

3000

3500

4000

Obs719beltsville

OQVMQV

OQV

MM5 w/ ACM2

10 12 14

CMAQ ozone Eddy vs ACM2

45 50 55 60 65 70 75 800

500

1000

1500

2000

2500

3000

3500

4000

Obs719beltsville

O3ppbMO3acm2MO3eddy

O3ppb

Pellston, MI July 25, 18Z

290 295 300 305 3100

500

1000

1500

2000

2500

3000

Obs725pellston

OTHTMTHT

OTHT

0 2 4 6 8 100

500

1000

1500

2000

2500

3000

Obs725pellston

OQVMQV

OTHT

30 40 50 60 70 800

500

1000

1500

2000

2500

3000

Obs725pellston

O3ppbMO3acm2MO3eddy

O3ppb

ACM2 Summary• ACM2 is a combination of local and non-local closure

techniquesSimilar capabilities to eddy diffusion w/ counter-gradient adjustment but more readily applicable to any quantity (e.g chemistry)

• LES and 1-D tests show accurate simulation of vertical profiles and PBL heights

• MM5 and WRF tests show good ground level performance and accurate PBL heights

• CMAQ testing shows comparable ground level performance as the eddy diffusion model

• Vertical profiles show larger differences from eddy diffusion with shallower and more well mixed CBL

• Ozonesonde comparisons often show good agreement for θ and qv but not as often for O3

• Recent publications: Pleim (2007) Part 1 and 2 JAMC

Conclusions

• PBL treatment in Air Quality models should be identical to PBL model in the Meteorology model

• PBL scheme used should be equally valid for Heat, momentum, met scalars, and chem tracers.

• PBL models should be evaluated in terms of vertical profiles of θ,q,u,v,χ

• PBL performance depends on other physics, especially LSM but also radiation and clouds

Issues• Stable Boundary layers becoming more

important in AQ modeling (beyond O3)Daily average standardsNocturnal cold pools

• Nocturnal SBL mixing is often determined by minimum eddy diffusivity

Min Kz is usually larger in AQ model than in Met modelMet model issue is run-away coolingAQ model issue is extreme surface concs

• Great need to develop more realistic SBL models that address both met and AQ needs

Acknowledgements

• Rob Gilliam for help with WRF development and evaluation

• Jim Wilczak and Laura Bianco for the PBL height data

• Keith Ayotte for the LES dataDisclaimerDisclaimer: The research presented here was performed under the : The research presented here was performed under the

Memorandum of Understanding between the U.S. Memorandum of Understanding between the U.S. Environmental Protection Agency (EPA) and the U.S. Environmental Protection Agency (EPA) and the U.S. Department of Commerce’s National Oceanic and Atmospheric Department of Commerce’s National Oceanic and Atmospheric Administration (NOAA) and under agreement number Administration (NOAA) and under agreement number DW13921548. This work constitutes a contribution to the DW13921548. This work constitutes a contribution to the NOAA Air Quality Program. Although it has been reviewed by NOAA Air Quality Program. Although it has been reviewed by EPA and NOAA and approved for publication, it does not EPA and NOAA and approved for publication, it does not necessarily reflect their views or policies.necessarily reflect their views or policies.

Land-Atmosphere Interactions

Extra

Met Modeling Summary• Both MM5-PX-ACM2 and WRF-PX-

ACM2 show good performance for surface statistics (at least as good as other PBL/LSM combinations)

• Preliminary PBL height analysis shows that PX-ACM2 generally agrees well with observations and is generally better than other options

Model equations

i

i

zz∆∆

∂∂ +1

1+i1+iii1i CMd + CMd - CMu =

tC

⎟⎟⎠

⎞⎜⎜⎝

⎛

∆

−+

∆

−

∆ −

−−

+

++

21

21

21

21 )()(1 11+

i

iii

i

iii

i zCCK

zCCK

z

( ) iii zzhMuMd ∆−= − /21

Where Ci is mass mixing ratio at center of Layer iKi+1/2 is eddy diffusivity at top of layer i [m2s-1]Mu is upward convective mixing rate [s-1]Mdi is downward mixing rate from layer i to layer i-1 [s-1]

h is top of PBL

Non-local Mixing ratesConvective mixing rate derived by conservation of buoyancy flux:

( )( )[ ]2111 2121 / vvzhBM θθ −−= ++

( )2

12

12

1

1

1211 )(

+++ ∆

−=

zzKB vvz

θθBuoyancy flux at top of first layer defined by eddy diffusion:

( ) ( )hzzhzzukK sLzz s

1.0,min;/1)(

2* =−=φ

Where:

Thus, Convective mixing rate is a function of Kz but not a function of potential temperature gradient:

( )2

1

21

11

1 )(

+

+

−∆=

zhzzK

M z

vegetation

PBL model in Met andAQM

Surface layer

Root zonesurface

L, u*, Ra

Stomatal conductance

q-flux

Tg

H-flux

T2Soil Moisture

T-2m, WRF PX – Analysis, August 2006

T-2m, WRF NOAH – Analysis, August 2006

WRF Analysis 2-m Temperature Statistics

2006 WRF 12km T-2m Stats

Temperature Error

1

1.2

1.4

1.6

1.8

2

2.2

2.4

1-Aug

2-Aug

3-Aug

4-Aug

5-Aug

6-Aug

7-Aug

8-Aug

9-Aug

10-A

ug11

-Aug

12-A

ug13

-Aug

14-A

ug15

-Aug

16-A

ug17

-Aug

18-A

ug19

-Aug

20-A

ug21

-Aug

22-A

ug23

-Aug

24-A

ug25

-Aug

26-A

ug27

-Aug

28-A

ug29

-Aug

30-A

ug

Day

RM

SE/M

AE

(K)

Analysis (RMSE)Analysis (MAE)WRF PX ACMWRF PX ACMWRF NOAH YSUWRF NOAH YSU

Temperature BIAS

-1

-0.75

-0.5

-0.25

0

0.25

0.5

0.75

1

1-Aug

2-Aug

3-Aug

4-Aug

5-Aug

6-Aug

7-Aug

8-Aug

9-Aug

10-A

ug11

-Aug

12-A

ug13

-Aug

14-A

ug15

-Aug

16-A

ug17

-Aug

18-A

ug19

-Aug

20-A

ug21

-Aug

22-A

ug23

-Aug

24-A

ug25

-Aug

26-A

ug27

-Aug

28-A

ug29

-Aug

30-A

ug

Day

BIA

S (K

)

Analysis (RMSE)WRF PX ACMWRF NOAH YSU

WRF PX 2-m Temperature Statistics

WRF NOAH 2-m Temperature Statistics

Mixing Ratio Error

0.5

1

1.5

2

2.5

1-Aug

2-Aug

3-Aug

4-Aug

5-Aug

6-Aug

7-Aug

8-Aug

9-Aug

10-A

ug11

-Aug

12-A

ug13

-Aug

14-A

ug15

-Aug

16-A

ug17

-Aug

18-A

ug19

-Aug

20-A

ug21

-Aug

22-A

ug23

-Aug

24-A

ug25

-Aug

26-A

ug27

-Aug

28-A

ug29

-Aug

30-A

ug

Day

RM

SE/M

AE

(g/k

g)

Analysis (RMSE)Analysis (MAE)WRF PX ACMWRF PX ACMWRF NOAH YSUWRF NOAH YSU

Mixing Ratio BIAS

-1.5

-1

-0.5

0

0.5

1

1.5

1-Aug

2-Aug

3-Aug

4-Aug

5-Aug

6-Aug

7-Aug

8-Aug

9-Aug

10-A

ug11

-Aug

12-A

ug13

-Aug

14-A

ug15

-Aug

16-A

ug17

-Aug

18-A

ug19

-Aug

20-A

ug21

-Aug

22-A

ug23

-Aug

24-A

ug25

-Aug

26-A

ug27

-Aug

28-A

ug29

-Aug

30-A

ug

Day

BIA

S (g

/kg)

Analysis (RMSE)WRF PX ACMWRF NOAH YSU

WRF PX 2-m MixR Statistics

WRF NOAH 2-m MixR Statistics

Wind Speed Error

0.75

1

1.25

1.5

1.75

1-Aug

2-Aug

3-Aug

4-Aug

5-Aug

6-Aug

7-Aug

8-Aug

9-Aug

10-A

ug11

-Aug

12-A

ug13

-Aug

14-A

ug15

-Aug

16-A

ug17

-Aug

18-A

ug19

-Aug

20-A

ug21

-Aug

22-A

ug23

-Aug

24-A

ug25

-Aug

26-A

ug27

-Aug

28-A

ug29

-Aug

30-A

ug

Day

RM

SE/M

AE

(m/s

)

WRF PX ACMWRF PX ACMWRF NOAH YSUWRF NOAH YSU

Revised PBL Height• Original Ri based scheme (e.g. Troen

and Mahrt,1986):

• Revised scheme for CBL:

))(()( 2

sv

vcrit hg

hURihθθ

θ−

=

mixsv

mixvcrit zhg

zUhURih +−

−=

))(())()(( 2

θθθ

zmix is where θv(zmix) = θs

Wind Speed BIAS

-0.75

-0.5

-0.25

0

0.25

0.5

0.75

1-Aug

2-Aug

3-Aug

4-Aug

5-Aug

6-Aug

7-Aug

8-Aug

9-Aug

10-A

ug11

-Aug

12-A

ug13

-Aug

14-A

ug15

-Aug

16-A

ug17

-Aug

18-A

ug19

-Aug

20-A

ug21

-Aug

22-A

ug23

-Aug

24-A

ug25

-Aug

26-A

ug27

-Aug

28-A

ug29

-Aug

30-A

ug

Day

BIA

S (m

/s)

WRF PX ACMWRF NOAH YSU

290

292

294

296

298

300

4 8 12 16 20 24

Mean ObsMean Model

Hour of day (UT)

-0.5

0

0.5

1

1.5

2

4 8 12 16 20 24

Mean Abs ErrorMean Bias

Hour of day (UT)

2 m TemperatureAveraged over all NWS/FAA sites 7/15 – 8/18 2004

Mean Absolute ErrorMean Bias

2.5

3

3.5

4

4.5

0 5 10 15 20 25

ObsModel

Hour (UT)

-1

-0.5

0

0.5

1

1.5

0 5 10 15 20 25

MAEMB

Hour (UT)

10 m Wind speedAveraged over all NWS/FAA sites 7/15 – 8/18 2004

Mean Absolute ErrorMean Bias

The second GABLS model intercomparison

• Multi-day Intercomparison of 23 PBL models for CASES99 field study

Given initial profilesTg time series Constant geostrophic windLarge scale subsidence2.5% of potential evaporationP, zo

GABLS T-2m Intercomparison

270

275

280

285

290

295

24 32 40 48 56 64 72

1.5-m Air Temperaturemain towerstn 1stn 2stn 3stn 4stn 5stn 6H

Hour (LT)

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

0.3

24 32 40 48 56 64 72

Sensible Heat Flux

main towerACM2

Hour (LT)

GABLS Experiment –Simulation of CASES99October 22-24, 1999

CASES99 profiles

286 288 290 292 294 296 298 3000

500

1000

1500

2000Leon, KS October 23, 1999 - 19z

RadiosondeACM2

Theta (K)

0 5 10 15 20 250

500

1000

1500

2000Leon, KS October 23, 1999 - 19z

RadiosondeACM2

wind speed (m/s)

Near surface Temperature profile from CASES99 main tower

283.5 284 284.5 285 285.5 2860

10

20

30

40

50

60

70

80Temperature profile at 19Z (hr 38) October 23, 1999

TowerACM2

T (K)

MM5 Evaluations• Domain: 202 x 208 x 34 @ 12 km res• Physics:

ACM2 PX LSMKF2Reisner 2RRTM w/ Dudhia SW

• Data Assimilation:Winds at all levels, T and qv above PBLIndirect soil moisture nudging

• July 13 – August 18, 2004

Evaluation of WRF-PX-ACM2 and WRF-NOAH-YSU

• Grid configuration:Horizontal grid resolution = 12 km34 vertical layer

• Physics:RRTM long wave radiationDudhia SWWSM6 microphysicsKF2 convective cloud scheme

• FDDA3-D analysis nudging for winds (all levels), T, qv(above PBL only)Indirect soil moisture nudging using NCEP surface analysis of T and RH for PX LSM

PBL Modeling for Air Quality ModelsOutlineGoal: To model effects of PBL processes on air qualityWRF PBL/Surface FluxesPBL Modeling for Met and AQConvective conditionsAsymmetric Convective Model version 2 (ACM2)Partitioning local and non-localNormalized heat flux and potential temperature profilesNon-local partitioningNon-local fraction (fconv) as function of stabilityTesting and evaluation of ACM2LES experiment low heat flux (Q* = 0.05 K m s-1), weak capProfiles of sensible heat flux. Local, non-local, and total sensible heat flux compared to LESComparisons of ACM2 PBL Heights to ObservationsFurther WRF DevelopmentWRF12km Grids and NLCD Data Research Triangle Park, NCCMAQEddy vs ACM2Ozonesonde at BeltsvilleJuly 19, 16 ZPellston, MI July 25, 18ZACM2 SummaryConclusionsIssuesAcknowledgementsExtraMet Modeling SummaryModel equationsNon-local Mixing ratesT-2m, WRF PX – Analysis, August 2006T-2m, WRF NOAH – Analysis, August 20062006 WRF 12km T-2m StatsRevised PBL HeightThe second GABLS model intercomparisonGABLS T-2m IntercomparisonGABLS Experiment – Simulation of CASES99October 22-24, 1999CASES99 profilesNear surface Temperature profile from CASES99 main towerMM5 EvaluationsEvaluation of WRF-PX-ACM2 and WRF-NOAH-YSU