Forecasting Air quality in China Using

CAMS Boundary Conditions:

the PANDA Project

Guy P. Brasseur

and Idir Bouarar

June 206

The PANDA ProjectCoordinator: Guy Brasseur

Deputy Coordinator: Prof. Xuemei Wang

Period: Jan 2014 - Dec. 2016

Budget: 2 Millions Euros

The Different Elements of Prediction

System

Some Elements of the AQ

Downscaling System

• Boundary and Initial Conditions (CAMS C-IFS))

• Emissions (natural, anthropogenic)

• Weather forecasts (IFS, NCEP, etc.)

• Representation of PBL processes

• Learning from daily diurnal predictions

Assimilation

in CAMS

WRF-Chem prediction

20x20km

MACC forecast/reanalysis as IC & BC

WRF-Chem 60x60km

20x20km

Downscaling to Regional Scale in Asia

7 x 7 km

Satellite data

Air Quality Index (AQI)

WRF-Chem prediction

AQI

PANDA Methodology

Global Model

Ensemble of Regional

60 km

Ensemble of Regional

Models

60 km

Local AQ Model Local AQ Model Local AQ Model

Ensemble of Sub-

20 km

Ensemble of Sub-

regional Models

20 km

Satellite

Observations

AssimilationCAMS

operational

validation

Dissemination of AQ predictions

validationvalidation

validation

Do the Initial and Boundary

Conditions Matter?

EUROPEAN CENTRE FOR MEDIUM-RANGE WEATHER FORECASTS

Do initial and boundary conditions matter?

Monthly mean surface O3 concentrations for January 2010 simulated by

WRF-Chem using MOZART (left) and MACC (right) initial and boundary

conditions.

MOZART ECMWF

Idir Bouarar

Do the Emissions Matter?

Ratio between REAS and MACCITY

NOx Emissions in Asia

Analysis from the

CAMS Group on

Emissions

Do emissions matter?

CO

NOx

WRF-Chem simulations

(Jan. 2010) with:

-HTAPv2 emissions (HTP)

-REASv2 emissions (RAS)

-MACCity emissions

(MCT)

With HTAPv2

Absolute differences with

RAS and MCT simulations

O3

PM2.5

Do emissions matter?

Does the formulation of the Boundary

Layer (PBL) Matter ?

Formulation of the PBL

• Surface concentrations depend critically on adopted PBL parameters.

• Height of the PBL (which decreases abruptly in early evening)

• Vertical mixing, specifically in the nighttime PBL.

• Needs adjustment in urban areas to account for heat island effects and mechanical turbulence generated by the buildings

• How well are the models doing?

PBL Height

Diffusion Coef. At 18:00 Diffusion Coef. At 19:00

Below:

Nightime Vertical

exchange coefficient in

rural areas

In urban areas, this

coefficient is set equal to

2 m2 s-1

Rural nighttime PBL

Daytime PBL

Effect of increasing diffusion in the urban area of Beijing

CNTRL: Control run (2 m2 s-1)

TED: run with increased vertical diffusion from 2 to 10 m2 s-2

Increased diffusion in urban areas leads to

decrease in NOx

increase in O3

NOx

O3

What Do we Learn from Daily

Forecasts?

Ensemble Predictions

for 37 Chinese Cities

Using 7 Models

Predictions of NO2 and O3

Guangzhou

3 June 2016

CHIMERE

C-IFS

NO2 O3

Predictions of NO2/O3

Guangzhou

3 June 2016

WRF-

Chem

/HH

SILAM

CHIMERE

C-IFS

SILAM

WRF-Chem/HH

EMEP

LOTOS-EUROS

PM2.5

Xi’An

3 June 2016

Conclusions

• CAMS predictions provide unique data used as boundary conditions for regional predictions in Asia.

• CAMS assimilates data and accounts for highly variable sources such as fire emissions and global meteorology.

• This allows regional predictions to be “relaxed” to space observations and to account for long-range influences.

Towards a Constellation of Similar Initiatives?

• North America

• South America

• Africa

• Europe

• Asia

• Russia

• Driven by the same global models and space observations, using similar methodologies, data bases, models, and providing the same type of products and services.

• MAP-AQ Initiative in support of WMO/GAW project

MAP-AQ

Modeling, Analysis and Prediction of Air Quality

To develop and implement a global air pollution monitoring, analysis and prediction system with downscaling capability in

regions of the world (e.g., Asia, Latin America, Africa) that are affected by high levels of atmospheric pollutants.

Thank you!

Questions?