Suitability of WRF for air quality simulations: comparison of two dynamical cores Oriol Jorba 1,...

Suitability of WRF for air quality Suitability of WRF for air quality simulations: comparison of two simulations: comparison of two

dynamical coresdynamical cores

Oriol Jorba1, Thomas Loridan2, Pedro Jiménez-Guerrero1, José M. Baldasano1,3

Corresponding author: [email protected]

1Barcelona Supercomputing Center, Earth Sciences Department, Barcelona2King’s College London, Department of Geography, London3Environmental Modelling Laboratory, Technical University of Catalonia,

Barcelona

EGU 2008, Vienna, April 16th, 2008

Evaluation of the two dynamical cores of the WRF modelling system: WRF-NMM WRF-ARW

12 km horizontal resolution for a European domain

Annual simulation for 2004 – surface and boundary layer

Temperature Moisture Wind speed Wind direction

ObjectiveObjective

To analyse the skills of meteorological mesoscale models as drivers of air quality modelling systems

Sistema Sistema CALIOPECALIOPE

http://www.bsc.es/http://www.bsc.es/caliopecaliope

WRF-ARW current driver of

CALIOPE air quality modelling

system

Common air pollution episodesCommon air pollution episodes

Strong anthropogenic PM episode over western Europe - wintertime

Strong anthropogenic O3 episode over Europe - summertime

Common characteristics: high pressure, low winds

High stability, cold temperature High mixing, warm temperature

•2 important episodes of air quality pollution during 2004:• February 2004: PM and primary pollutants episode over Europe• July-August 2004: O3 season

WRF v2.2.1 WRF-NMM

Time step: 30 s NX,NY,NZ: 360, 564, 38 Dx, Dy: 0.078º, 0.07248º (~12km) PTOP: 50 hPa Hydrostatic

WRF-ARW Time step: 72 s NX, NY, NZ: 481, 401, 38 Dx, Dy: 12 km PTOP: 50 hPa Hydrostatic

IC-BC: 6h 1-degree FNL analysis 366*36 h simulation -12 hours of cold-start

2004 annual simulation: Model configuration2004 annual simulation: Model configuration

Microphysics: Ferrier LW, SW radiation: GFDL Surface-layer: Janjic scheme Land-surface: NMM LSM Boundary layer: Mellor-Yamada-Janjic Cumulus scheme: Betts-Miller-Janjic

Microphysics: WSM 3-class LW, SW radiation: RRTM, Dudhia scheme Surface-layer: Monin-Obukhov scheme Land-surface: Noah LSM Boundary layer: YSU scheme Cumulus scheme: Kain-Fritsch

Evaluation against 2 datasetsEvaluation against 2 datasets

931 surface meteorological stations from the NCEP ADP Global Surface

Observations archive (includes METAR, SYNOP, AWOS and ASOS)

61 meteorological soundings from European network of

radiosoundings

Classical StatisticsClassical Statistics

Surface evaluation:

931 surface meteorological stations from METAR and SYNOP network

Performance of the models: general overviewPerformance of the models: general overview

ARW Temp MAE (C)

NMM Temp MAE (C)

ARW SPD MAE (m/s)

NMM SPD MAE (m/s)

We have identified problems with NMM simulation for May month related to technical aspects. For the following evaluation, the

period 29/4/2004 to 26/5/2004 is not taken into account.

Wind direction

0

5

10

15

20

25

30

35

40

45

50

1 2 3 4 5 6 7 8 9 10 11 12

MAE-ARW MAE-NMM ME-ARW ME-NMM

Monthly evolution of mean surface errors Monthly evolution of mean surface errors (MAE , MBE) Temperature

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

1 2 3 4 5 6 7 8 9 10 11 12


Dewpoint temperature

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

1 2 3 4 5 6 7 8 9 10 11 12


Wind speed

0

0.5

1

1.5

2

2.5

3

1 2 3 4 5 6 7 8 9 10 11 12


• Temperature and moisture: underprediction of ARW/ overprediction of NMM. Similar absolute errors.• Increased underestimation of ARW during spring• Wind speed overestimation – similar errors.• Lower wind speed error during summertime

di diobs

+-

(ºC)

(deg)(ºC)

(m/s)

2m Temperature Monthly errors – January to March2m Temperature Monthly errors – January to March

ARW

NMM

• Good performance over continental flat terrain• Major problems - complex terrain and coastal areas with complex topography

(Mediterranean sea).

Jan 2 m Temp MAE (ºC) Mar 2 m Temp MAE (ºC)Feb 2 m Temp MAE (ºC)

Lower MAE – NMM

2m Temperature Monthly errors – July to September2m Temperature Monthly errors – July to September

ARW

NMM

• Stagnant situation over the Mediterranean and low baric gradients over Europe• Summertime - increased errors, but generally below 2.5ºC• Lower errors with NMM in western and central Europe

Jul 2 m Temp MAE (ºC) Sep 2 m Temp MAE (ºC)Aug 2 m Temp MAE (ºC)

2m Temperature2m Temperature Monthly errors – October to Monthly errors – October to DecemberDecember

ARW

NMM

• Coastal areas:• Good performance during the whole year in northwestern European

coasts.• Major problems in Mediterranean coasts – systematic error.

• Complex terrain:• Higher errors during the whole year• Need high resolution for a better representation of orography

(mountains-valleys)• Continental flat terrain:

• Overall best performance – larger errors related to meteorological conditions

Oct 2 m Temp MAE (ºC) Dec 2 m Temp MAE (ºC)Nov 2 m Temp MAE (ºC)

Systematic errors of the models – annual mean RMSE, Systematic errors of the models – annual mean RMSE, RMSEsRMSEs2m Temperature2m Temperature

ARW

NMM

•Lower errors over northern Europe – larger errors over southern Europe.

•Complex topographic areas and Mediterranean coasts – large errors.

•From the generally 1.5 to 2.5ºC total RMSE, 0.5 to 1.5ºC corresponds to systematic error.

•Higher systematic RMSE of ARW in western and central Europe.

Systematic errors of the models – annual mean RMSE, Systematic errors of the models – annual mean RMSE, RMSEsRMSEs2m Dewpoint temperature2m Dewpoint temperature

ARW

NMM

•Similar errors of both models ranging between 1.5-2ºC

•Lower errors – northern European coast

•Higher errors – Alpine region.

•Homogeneous error distribution over flat continental terrain areas.

•Lower systematic RMSE in NMM (0 to 1ºC)

Systematic errors of the models – annual mean RMSE, Systematic errors of the models – annual mean RMSE, RMSEsRMSEs10m Wind speed10m Wind speed

ARW

NMM

•No clear spatial pattern of the annual error for wind speed.

•RMSE ranges from 1 to 3 m/s

•Similar absolute and systematic errors between ARW and NMM.

•Systematic error ranges from 0.5 to 1.5 m/s, accounting for the half part of total error.

Monthly MAE – Wind speed (10 m)Monthly MAE – Wind speed (10 m)

Major problems in complex terrain regions, and coastal regions with near mountain ranges (Mediterranean, Scandinavia, north UK)

ARW

NMM

Feb July

Different performance in northern and southern Europe

Major skills over the wetter continental Europe

Vertical structure evaluation:

61 meteorological soundings from European network of radiosoundings

Vertical temperature monthly errors 1000-2000m

-3

-2

-1

0

1

2

3

1 2 3 4 5 6 7 8 9 10 11 12

ARW-MAE-annual NMM-MAE-annual ARW-ME-annual NMM-ME-annual

•MAE errors ranging around 2.5ºC for ARW and 1.5ºC for NMM during the whole 2004.

•Larger underestimation of ARW.

•Larger diurnal errors in ARW but similar for NMM.

•Both models presents an underestimation of temperature in lower layers that is increased in upper levels of the ABL.

•NMM presents a regular MAE from 0 to 2000 m, while ARW tends to have higher errors in upper levels.


-3

-2

-1

0

1

2

3

1 2 3 4 5 6 7 8 9 10 11 12



-4

-3

-2

-1

0

1

2

3

4

1 2 3 4 5 6 7 8 9 10 11 12


TemperatureTemperature profile 0-2000 m evaluation (MAE, MBE, RMSEs) profile 0-2000 m evaluation (MAE, MBE, RMSEs)

Vertical temperature monthly errors 0-2000m Daytime

-4

-3

-2

-1

0

1

2

3

4

1 2 3 4 5 6 7 8 9 10 11 12

ARW-MAE-annual NMM-MAE-annual ARW-ME-annual NMM-ME-annualVertical temperature monthly errors 0-2000m Nighttime

-4

-3

-2

-1

0

1

2

3

4

1 2 3 4 5 6 7 8 9 10 11 12



-3

-2

-1

0

1

2

3

1 2 3 4 5 6 7 8 9 10 11 12



0

0.5

1

1.5

2

2.5

3

1 2 3 4 5 6 7 8 9 10 11 12

ARW-RMSEs-annual NMM-RMSEs-annual

ARW-RMSEu-annual NMM-RMSEu-annual

Vertical mixing ratio monthly errors 0-2000m

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1 2 3 4 5 6 7 8 9 10 11 12

ARW-MAE-annual NMM-MAE-annual ARW-ME-annual NMM-ME-annualVertical mixing ratio monthly errors 0-2000m Daytime

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1 2 3 4 5 6 7 8 9 10 11 12



-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1 2 3 4 5 6 7 8 9 10 11 12

ARW-MAE-annual NMM-MAE-annual ARW-ME-annual NMM-ME-annualVertical mixing ratio monthly errors 500-1000m

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1 2 3 4 5 6 7 8 9 10 11 12

ARW-MAE-annual NMM-MAE-annual ARW-ME-annual NMM-ME-annualVertical mixing ratio monthly errors 0-500m

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1 2 3 4 5 6 7 8 9 10 11 12


Mixing ratioMixing ratio vertical structure evaluation vertical structure evaluation (MAE, MBE, RMSEs)(MAE, MBE, RMSEs)

•Moisture errors higher under dryer conditions.

•NMM has slightly larger errors during summertime

•Both models presents an homogeneous overestimation of moisture for winter and spring and a slight underestimation in summer.

•Day-night conditions over-underestimation.

•Overestimation to underestimation from 0 to 2000 m for NMM

Vertical temperature monthly errors 0-2000m Nighttime

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1 2 3 4 5 6 7 8 9 10 11 12


Major differences during summertime


0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

1 2 3 4 5 6 7 8 9 10 11 12



Vertical wind speed monthly errors 0-2000m

0

0.5

1

1.5

2

2.5

3

3.5

1 2 3 4 5 6 7 8 9 10 11 12

ARW-MAE-annual NMM-MAE-annual ARW-ME-annual NMM-ME-annualVertical wind speed monthly errors 0-2000m Daytime

0

0.5

1

1.5

2

2.5

3

3.5

1 2 3 4 5 6 7 8 9 10 11 12



-0.5

0

0.5

1

1.5

2

2.5

3

3.5

1 2 3 4 5 6 7 8 9 10 11 12

ARW-MAE-annual NMM-MAE-annual ARW-ME-annual NMM-ME-annualVertical wind speed monthly errors 500-1000m

-0.5

0

0.5

1

1.5

2

2.5

3

3.5

1 2 3 4 5 6 7 8 9 10 11 12

ARW-MAE-annual NMM-MAE-annual ARW-ME-annual NMM-ME-annualVertical wind speed monthly errors 0-500m

0

0.5

1

1.5

2

2.5

3

3.5

1 2 3 4 5 6 7 8 9 10 11 12


Wind speedWind speed Vertical structure evaluation Vertical structure evaluation (MAE, MBE, (MAE, MBE, RMSEs)RMSEs)

•Similar pattern of both models – general overestimation of wind speed.

•Daytime larger errors under wintertime conditions of ARW.

•Nighttime same performance.

•Larger overestimation at surface levels (0-500 m).

Vertical wind speed monthly errors 0-2000m Nighttime

0

0.5

1

1.5

2

2.5

3

3.5

1 2 3 4 5 6 7 8 9 10 11 12



0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

1 2 3 4 5 6 7 8 9 10 11 12



Summary of results and impacts over an air quality Summary of results and impacts over an air quality modelling systemmodelling system

2004 Annual evaluation of the two dynamical cores of WRF modelling system over Europe: WRF-NMM WRF-ARW

Results have highlighted: Coastal areas:

Good performance during the whole year in northwestern European coasts. Major problems in Mediterranean coasts – systematic error.

Complex terrain: Higher errors during the whole year for temperature, but reasonable good wind results. Need high resolution for a better representation of topography (mountains-valleys)

Continental flat terrain: Overall best performance – larger errors related to meteorological conditions

Larger errors over southern Europe and under high pressure or stagnant conditions situation Systematic errors higher with ARW, more work for improvement than NMM

Summary of results and impacts over an air quality Summary of results and impacts over an air quality modelling system (cont.)modelling system (cont.)

Normal range of mean absolute errors: Temperature: 1º to 3 ºC – ARW underestimation : NMM overestimation Dewpoint: 1º to 2.5ºC – ARW underestimation : NMM overestimation Wind speed: 1 to 5 m/s – ARW and NMM overestimation Wind direction: 30 to 70 degrees

Impacts of accurate meteorological fields for air quality modelling: From Pérez et al. (2006): increase in 3ºC of temperature leads to an increase of 7.2 - 21.1 g m-3 of

ozone concentration in background areas. Moisture errors impact in the formation of secondary organic aerosols and in the aqueous chemistry by

modifying the OH radical concentrations. Overestimation of wind speed will produce an increase mixing of pollutants within the boundary layer

and a major advective effect, the result will be a simulation of lower background pollutant concentrations over the affected areas.

The direction will strongly impact with the geographical accuracy and the levels of the polluted air mass.

[email protected]

THANK YOU FOR YOUR THANK YOU FOR YOUR ATTENTIONATTENTION

Suitability of WRF model for air quality simulations: Suitability of WRF model for air quality simulations: comparison of two dynamical cores on a yearly basiscomparison of two dynamical cores on a yearly basis

AcknowledgmentsThis work was funded by the project CALIOPE project 441/2006/3-12.1 and A357/200/2-12.1 of the Spanish

Ministry of the Environment. The FNL analysis and verification data are provided by the NCAR’s Data

Support Section. WRF-ARW and WRF-NMM simulations were performed with the MareNostrum supercomputer

held by the Barcelona Supercomputing Center.

Suitability of WRF for air quality simulations: comparison of two dynamical cores Oriol Jorba 1,...

Documents

Transcript of Suitability of WRF for air quality simulations: comparison of two dynamical cores Oriol Jorba 1,...