Forecasting the atmospheric composition with COSMO-ARTin southern West AfricaKonrad Deetz and Bernhard VogelInstitute of Meteorology and Climate Research, Karlsruhe Institute of Technology (KIT), Karlsruhe, Germanykonrad.deetz@kit.edu

Pantone 362 Pantone 122

ObjectivesThe Dynamics-aerosol-chemistry-cloud interactions in West Africa (DACCIWA) project (Knippertz et al., 2015) investigates the influenceof anthropogenic and natural emissions on the atmospheric composition over southern West Africa (SWA). Between 1 June and 31 July2016 the DACCIWA measurement campaign took place in SWA, including ground based and airborne observations. Within this work,the model system COSMO-ART was applied for the first time for operational numerical forecasts with full chemistry and aerosols. Dailyforecasts were realized during the complete measurement campaign to support the decision making of the research aircraft flights and toderive model climatologies of atmospheric aerosol and chemical composition.

Model frameworkThe regional-scale model frameworkCOSMO-ART combines the operationalweather forecast model COSMO of theGerman Weather Service (DWD) withthe ART modules (Aerosols and Reacti-ve Trace gases, Vogel et al., 2009). Gasphase chemisty and aerosol dynamicsare online coupled with the meteorology.Aerosol particles are treated via a modalapproach, considering mineral dust, seasalt and anthropogenic aerosols.

The research domain is affected by a variety of pollution sources (e.g. emis-sions from coastal megacities and biomass burning emissions from centralAfrica). For this study a new parameterization for mineral dust and gas flaringemissions have been developed and applied with COSMO-ART. The forecastswere realized for the domain 25◦W-40◦E, 20◦S-35◦N on 50 vertical levels witha grid mesh size of 28 km. Interactions with radiation and cloud formation arenot considered in this study.

Methods

12 0 12 0 12 0F1F2F3

UTC

ICON forecast(DWD)

GFASv2.1 NRT

(MACC)

MOZART forecast(NCAR)

dacciwa.sedoo.fr dacciwa.sedoo.fr

Morning Weather Briefing

Daily Planning Meeting

Mission selection team Meeting

Morning Weather Briefing

Daily Planning Meeting

Mission selection team Meeting

DDDDDDDDD D + 1D + 1D + 1D + 1D + 1D + 1D + 1D + 1D + 1 D + 2D + 2D + 2D + 2D + 2D + 2D + 2D + 2D + 2

Forecast

Computing

Postprocessing

The forecasts were initialized daily at 12 UTC with a lead time of 57h. Fig. 1shows the structure for three subsequent forecasts F1-F3. As meteorologicaland aerosol/chemistry boundary ICON and MOZART were used, respectively.Fire emissions were taken from GFAS near real-time. The visualizationsof the forecast results were uploaded to the dacciwa.sedoo.fr databaseto be available for the decision makers. The model data was archived forsubsequent analysis.

Fig.1. COSMO-ART forecasts - structure and embedding in DACCIWA

Result 1 - StratusThe low-level stratus (Fig.2) frequently developed in SWA during the cam-paign. Good agreement can be seen for simulated cloud water and the obser-

0 3 6 9 12 15 18 210

1

2

3

4

Time (UTC)

Hei

ght (

km)

0.001

0.005

0.01

0.05

0.1

−−−−−−−−−−−−−−−−−−−−−−−−

−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−

−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−

−−−−−−−−−−−−−−−−−−−−−−−−−−

ved cloud base location of the stratus. It develops in the early morning, rea-ches maximum intensity around 9 UTC and is lifted and transformed to cumu-lus, subsequently. During night, Savé frequently observed clear sky whereasCOSMO-ART shows midlevel clouds. Realizations with smaller mesh size andwithout convection parameterization show better results in this regard.

Fig.2. Diurnal cycle of simulated cloud water (shading, g kg−1) and ceilometer cloud basebelow 2000m (bars, black and grey denote average and minimum, respectively), averagedover the period 13 June to 31 July 2016 at Savé.

Result 2 - Ozone

12 18 00 06 12 18 00 06 12 18

0 6 12 18 24 30 36 42 48 54

0

10

20

30

40

50

60

70

Time (UTC)

O3

volu

me

mix

ing

ratio

(pp

bv)

Lead time (h)

Savé supersite (obs)COSMO−ART (sim)

COSMO-ART captures the diurnal cycle of O3 concentration with maximum at15 UTC but generally overestimates the concentrations especially in the after-noon peak. It is assumed that the overestimation in CO (see Fig.5) leads to anincrease in the O3 formation. A sensitivity study with reduced anthropogenicCO emissions has shown improvement.

Fig.3. Near surface O3 volume mixing ratio (ppbv) temporal evolution for the 57h forecastleadtime, averaged over the period 13 June to 30 July 2016 at Savé. Shaded areas denotethe standard deviation.

Result 3 - AOD

0

0.1

0.2

0.3 Mineral dust COSMO−ART (sim)CAMS (sim)

0

0.01

0.02

0.03Sea salt

0

0.5

1

1.5 Anthropogenic

13 17 21 25 29 3 7 11 15 19 23 27 310

0.5

1

1.5

Day (June−July 2016)

Total

●●●●

●

●●●

●

●●

●

●●●

●●

●

●

●

●

●

●

●

●

●●●●

●●●●

●

●●● ● ●

●

●●●

●

●●●●

●●●●●●

●●●

●

●●●●● ●●●●

●

●●●● ●●●●●●●

●

●

●●●

●

●●●●

●

●●

●

●

●

●●

● ●●●●●●

●●

●

● AERONET (obs)Total

Fig.4. Time series of ae-rosol optical depth (AOD,550nm) between 13 Ju-ne and 31 July 2016 atSavé, simulated by COS-MO-ART (blue) and Co-pernicus Atmosphere Mo-nitoring Service (CAMS,green) and observed byphotometer at Savé su-persite (AERONET, red).

Although the mineral dust influence on SWA is reduced during the monsoontime, it significantly contributes to the total AOD. COSMO-ART performs wellcompared to CAMS in terms of absolute values and temporal evolution. Seasalt influence is one order of magnitute smaller. Highest uncertainty can beseen in terms of anthropogenic aerosol. COSMO-ART seems to overestimatethe anthropogenic aerosol which is also reflected in the differences towardsAERONET.

Result 4 - Aircraft

0.1

0.512

510

0 1E 2E 3E5N

6N

7N

8N

●●LoméLoméLoméLoméLoméLoméLoméLoméLomé

●●SavéSavéSavéSavéSavéSavéSavéSavéSavé

●●CotonouCotonouCotonouCotonouCotonouCotonouCotonouCotonouCotonou

●●AccraAccraAccraAccraAccraAccraAccraAccraAccra

NOx,surf (ppbv)NOx,surf (ppbv)NOx,surf (ppbv)NOx,surf (ppbv)NOx,surf (ppbv)NOx,surf (ppbv)NOx,surf (ppbv)NOx,surf (ppbv)NOx,surf (ppbv)

0.1

0.512

510

0 m

500

1000

1500

2000

6.9N/1.2E 6.7N/1.4E 6.5N/1.6E 6.3N/1.8E

NOx (ppbv)NOx (ppbv)NOx (ppbv)NOx (ppbv)NOx (ppbv)NOx (ppbv)NOx (ppbv)NOx (ppbv)NOx (ppbv)

0

500

1000

1500

2000Alt (m)

0

0.5

1

1.5

2

NO (ppbv)

0

2

4

6

8 NO2 (ppbv)

0

150

300

450

600

CO (ppbv)

0845 0900 UTC 09150

20

40

60

80

100

O3 (ppbv)

0845 0900 UTC 09150

1

2

3

SO2 (ppbv)

Fig.5 shows the intercomparison of COSMO-ART with the Twin Otter aircraftchemistry measurements during a NW-SE leg focussing on urban pollutionplumes in the boundary layer downwind of Lomé and Accra. Two distinct NOxpeaks have been observed. The first is likely related to Lomé but is not visiblein the model results. The second peak around 9 UTC at 500m altitude is cap-tured reasonable in location and amplitude and relates to Accra. O3 is slightlyoverestimated, confirming the findings of Fig.3.

Fig.5. Trace gas con-centrations (ppbv)along a 30 minu-te flight track of theBAS Twin Otter Air-craft at 14 July 2016(8:45-9:15 UTC,black) and simu-lated by COSMO-ART (9 UTC, blue).The upper figuresshow the locationof and the transectalong the flight track.The colored lineson top of the pa-nels denote the flightposition accordingto the altitude plot.

Result 5 - Flaring4W 0 4E 8E 12E

5S

0

5N

5S

0

5N

CO2

5S

0

5N

CO 5S

0

5N

SO2

4W 0 4E 8E 12E

5S

0

5N

NO4W 0 4E 8E 12E

5S

0

5N

NO2

10−5 10−4 10−3 0.01 0.1 1 10

Fig.6. Flaring emission inventory (t h−1)for June-July 2015.

A new flaring emission dataset has be-en created, combining remote sensingobservations and physically-based com-bustion equations (Deetz and Vogel,2017). It considers the trace gases CO,CO2, NO, NO2 and SO2. We have ap-plied our method to derive the flaringemissions for SWA (Fig.6) and to in-clude the dataset into COSMO-ART.Within a case study the fraction of airpollution from flaring to the total air pol-lution has been analyzed (Fig.7). O3concentrations from flaring of about 15ppbv can be observed in flaring areas.O3 contributions from flaring are up to 35%. Fig.7 also denotes a long-range transport offlaring pollution (westward O3 transport from the Niger Delta, affecting Ivory Coast andGhana). Although the flaring dataset does not include aerosol particles, aerosol nucleationis considered in COSMO-ART. For the offshore area of the Niger Delta, as a flaring hotspot,COSMO-ART results show enhanced aerosol formation.The flaring parameterization and a flaring emission dataset for SWA covering 2012-2016on a daily basis is made available for the research community.

10W 0 10E

5S

0

5N

O3 contribution (%)O3 contribution (%)O3 contribution (%)O3 contribution (%)O3 contribution (%)O3 contribution (%)O3 contribution (%)O3 contribution (%)O3 contribution (%)

10W 0 10E

NOx contribution (%)NOx contribution (%)NOx contribution (%)NOx contribution (%)NOx contribution (%)NOx contribution (%)NOx contribution (%)NOx contribution (%)NOx contribution (%)

6E 7E 8E 9E

4N

5N

●

● ● ●●●

●

● ●

●

●

●

●

●

●●

●

●

●

●●

●

●●●●

●

●

●

●

●●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●●

●

●

●

●

●

●●

●

●●●●

●

●

●

●

●●

●●

●

●

●

●

●

●

●

●●●

●

●●●

●

●

●●

●

●

●

●●●●●●●●●●

●

●

●

●

Submicron particlesSubmicron particlesSubmicron particlesSubmicron particlesSubmicron particlesSubmicron particlesSubmicron particlesSubmicron particlesSubmicron particlescontribution (%)contribution (%)contribution (%)contribution (%)contribution (%)contribution (%)contribution (%)contribution (%)contribution (%) 0.5

123510203050100

Fig.7. Contribution of pollution from flaring to the total air pollution (%, vertically integrated) at 3July 2016 12 UTC after a 36h simulation with and without flaring consideration. Focus of rightpanel is on Niger Delta (referring to the black rectangle in the middle panel) including the flarelocations as circles.

Forecastanimations

ConclusionsDuring the DACCIWA campaign in June-July 2016, operational forecasts have been conducted with the regional-scale model COSMO-ART. The results were made available via dacciwa.sedoo.fr to the research community. Model results resonably reproduce the observati-ons with a tendency to overestimate anthropogenic emissions. Further added value was contributed by the development and applicationof a new flaring emission parameterization. The results emphasize the need to consider flaring emissions in air pollution modelling.

AcknowledgmentsThe research leading to these results has received funding from the European Union 7th Framework Programme (FP7/ 2007-2013) under Grant Agreement no. 603502 (EUproject DACCIWA: Dynamics-aerosol-chemistry-cloud interactions in West Africa). We thank the Steinbuch Centre for Computing (SCC) for providing us with the computationalresources to realize the COSMO-ART forecasts and the University of York, University of Manchester and UPS for providing the observational data.

ReferencesDeetz, K. and Vogel, B., 2017: Development of a new gas-flaring emission dataset for southern West Africa, Geosci. Model Dev., 10, 1607-1620Knippertz, P., et al., 2015: The DACCIWA Project: Dynamics-Aerosol-Chemistry-Cloud Interactions in West Africa, BAMSVogel, B., et al., 2009: The comprehensive model system COSMO-ART - Radiative impact of aerosol on the state of the atmosphere on the regional scale, Atmos. Chem. Phys., 9

KIT – The Research University in the Helmholtz Association www.kit.edu