Introduction
The United Nations Department of Economic and Social Affairs, Population
Division reported 31 megacities globally (urban agglomerations with more than
10 million inhabitants in 2016) and that their number is projected to rise up
to 41 by 2030. In southern West Africa (SWA), Lagos is considered a
megacity (with more than 13 million inhabitants) and is expected to reach
24 million in 2030. The urban agglomeration extends along the coast to
Cotonou (Benin) and even to Lomé (Togo). Moreover Accra in Ghana (with a
population predicted to increase from 2.3 in 2016 to 3.3 million in 2030),
Kumasi in Ghana (with a population predicted to increase from 2.7 in 2016 to
4.2 million in 2030) and Abidjan in Côte d'Ivoire (with a population predicted
to increase from 5.0 in 2016 to 7.8 million in 2030) will all contribute to
form a more or less continuously urbanized strip at some point during the
twenty-first century. This growth is associated with enhanced pollutant
emissions and low air quality, which leads to chronic health problems
and contributes to anthropogenically forced climate
change.
The vertical structure of air pollution is complex along the Guinea coast
during the period when the West African monsoon (WAM) is established in the
boreal summer. From the surface to the top of the planetary boundary layer
(PBL), marine air transported by the northeastward monsoon flow gets enriched
with anthropogenic pollution emitted at the coast before moving further
inland . Above the marine PBL, biomass burning (BB) aerosol
layers, resulting from incomplete combustion of fires in central Africa
, can on occasion be observed reaching the
Guinea coast after being transported over thousands of kilometers
. In higher layers, at altitudes from 3 to 5 km, the
Saharan air layer (SAL) is generally observed to be advected from the north
depending on the meridional disturbances of the African easterly jet (AEJ),
carrying desert dust . This
general picture is often perturbed by the presence of organized convective
systems, which propagate along the Guinea coast from Nigeria to Liberia
. The latter authors also note the presence of land–sea
breeze convective systems in the immediate coastal strip.
The EU-funded project Dynamics–Aerosol–Chemistry–Cloud Interactions in West
Africa (DACCIWA) was designed to focus specifically on the Guinea coastal
atmospheric dynamics and the interactions among aerosols, chemistry, and
clouds . An intensive measurement campaign took place in
Nigeria, Benin, Togo, Ghana, and Côte d'Ivoire during June–July 2016, which
corresponds to the climatological onset period of the WAM
. Three research aircraft flew over the Guinea coastal
region with different scientific objectives, notably with flight plans
designed to map out city, shipping, and flaring emissions, or they focused on
sampling BB aerosol layers . The DACCIWA
field campaign took place in so-called post-WAM post-onset conditions,
i.e., after deep convection (and related precipitation) had migrated
from the coast inland over the Sahel .
During the WAM, the atmospheric composition over the Gulf of Guinea coastal
region is the result of a complex mix of natural and anthropogenic sources,
which include urban, BB, biogenic, desert dust, and oceanic
compounds. Using numerical tracer experiments, have
highlighted that fire emissions in central Africa impacting the surface
aerosol and gaseous species concentrations over the Gulf of Guinea are mostly
transported over the southeast Atlantic above the marine PBL. Using
WRF–CHIMERE numerical simulations of the WAM during the African Monsoon
Multidisciplinary Analyses (AMMA) campaign period from May to July 2006,
quantified the relative contributions of anthropogenic
and BB sources to carbon monoxide (CO) concentrations over SWA,
which in July 2006 were about 25 % local anthropogenic and 50 %
BB from central Africa. The remaining 25 % are the
background corresponding to long-range transport from outside of Africa. In
order to better distinguish the contributions from different sources to
background concentrations, additional studies are needed focusing on Africa
as has been carried out for other regions
e.g.,. However the high BB contribution is partly due to the significant
underestimation of
anthropogenic emissions for the Gulf of Guinea region
.
Characteristics of the studied cities with country, latitude,
longitude, elevation above mean sea level (a.m.s.l.), and number of
inhabitants of urban agglomerations. The population in bold is given for the
year 2015 according to the World Urbanization Prospects report
. National general population and habitat census is used
to estimate the population of Cotonou and Savè .
City
Country
Latitude
Longitude
Elevation
Number of inhabitants
Abidjan
Côte d'Ivoire
5.36∘ N
4.00∘ W
50 m a.m.s.l.
4 923 000
Accra
Ghana
5.60∘ N
0.19∘ W
30 m a.m.s.l.
3 013 000
Lomé
Togo
6.17∘ N
1.23∘ E
10 m a.m.s.l.
1 830 000
Cotonou
Benin
6.36∘ N
2.38∘ E
10 m a.m.s.l.
2 194 000
Savè
Benin
8.03∘ N
2.49∘ E
130 m a.m.s.l.
87 000
Lagos
Nigeria
6.49∘ N
3.36∘ E
10 m a.m.s.l.
13 121 000
AMMA was focused on the Sahelian region combining a multi-scale approach to
better characterize the interactions among atmosphere, land, and ocean
during the monsoon , whereas the DACCIWA
project is dedicated to the interactions among aerosols, clouds, and
radiation along the highly urbanized coastline of the Gulf of Guinea
. Over SWA, and have
documented a regular occurrence of a coastal front, which is located where
the strongest horizontal gradients of wind speed and potential temperature
occur. It develops during the daytime and propagates inland in the evening. After
the frontal passage, the wind in the lowermost troposphere brings air masses, and probably also anthropogenic pollutants
emitted from coastal urban areas e.g., and BB pollutants imported by monsoon flow e.g.,,
from the coast northward, especially at night with the nocturnal low-level
jet (NLLJ) .
The main objective of this article is to understand the diurnal cycle of
anthropogenic pollutant transport from the coast to the continental SWA. We
present numerical tracer experiments made with high-resolution CHIMERE
simulations set up in order to separate the contribution of each important
urban agglomeration, namely Abidjan, Accra, Lomé, Cotonou, and Lagos. We
take advantage of the DACCIWA measurements made by the three research
aircraft, by an enhanced radio-sounding network, and at the super-site of
Savè in central Benin . The super-site of Savè
is a representative location to assess the impact of pollution transport from
the coast on the air quality of remote inland cities characterized by low
local emissions. We focus on the period of 1–7 July 2016, during which a case
of long-range transport of BB aerosol from central Africa was
observed . We aim at answering the following questions:
What is the relative contribution of each coastal urban area to the air
pollution at Savè? How does it evolve during the day?
How are BB and anthropogenic pollutants mixed along the
coast and inland? Is it usually a mixture of the two pollution types that is
transported inland in the PBL?
This study is focused on one specific period and location. However, the
conclusions are representative of a longer time period as meteorological
conditions at the coast during the so-called monsoon post-onset period were
found to be quite stable for several weeks . Spatially,
the results are directly representative of the studied region only, the main
goal being to estimate the influence of the emissions from four coastal
cities on the atmospheric composition in the lower troposphere over inland
Benin. Given the broad southwesterly monsoon flow in the region, a similar
transport from coastal pollution inland will likely be found along most
of the Guinea coast. We shall answer these questions using a synergistic
combination of observations and numerical modeling experiments, described in
Sect. 2. Section 3 analyzes the temporal evolution of meteorology and air
pollution over a portion of SWA including Côte d'Ivoire, Ghana, Togo, Benin, and
Nigeria. Then, we focus on urban anthropogenic (URB) and long-range BB pollutant transport in Sect. 4. Conclusions are given in
Sect. 5.
DACCIWA project: observations and modeling
In the DACCIWA project, there are strong components on both in situ
observations and modeling. Here, we present all studied sites
(Sect. ), observational datasets used (Sect. ),
and numerical simulations performed to analyze the pollution transport
pathways (Sect. ).
Studied sites
We focus on six locations, five major urban agglomerations of the Guinean
coastal region, and one small town, Savè, which is 185 km north of
Cotonou (Benin). Table shows the coordinates and the
population of the urban agglomerations studied. For Abidjan, Accra, Lomé,
and Lagos, we present estimations for the year 2015 of the department of
social and economic affairs of United Nations . In this
report, these cities are associated with large administrative areas in
contrast to Cotonou.
Comparing Lomé and Cotonou, Lomé has a large administrative area
while Cotonou is a city with a very high population density over a small
area. The population of Lomé is about 839 000 inhabitants in the city
according to the general population and habitat census of Togo
and about 1 830 000 inhabitants in the administrative
state . Cotonou city is estimated by the World
Urbanization Prospects report to have about 1 086 000 inhabitants
.
According to the general population and habitat census of Benin
, the population of Cotonou only slightly increased by
2.09 % over the period of 2002–2013 because of the limited possibility of
expansion. They note that along the shores of Lake Nokoué the population
has increased rapidly, thus forming an agglomeration of 2 194 000
inhabitants, calculated as the sum of the Cotonou district and several cities
of the Atlantique district (Abomey-Calavi and Sô-Ava) and of the Ouémé
district (Sèmè Kpodji, Porto Novo, Avrankou, and Akpro-Missérété).
Observational datasets
During the DACCIWA field campaign, several observational platforms were
deployed to perform in situ and remote-sensing measurements
. In this study, we use datasets acquired by
ground-based stations, aircraft, radiosondes, and satellites. Table
gives the main information on each dataset. Figure presents the
location of the aircraft flight tracks and of the stations. Our studied
domain is located in the Greenwich mean time (GMT). Hence local time is the
same as UTC. During the aircraft campaign period, sunrise occurred around
06:00 UTC and sunset around 18:00 UTC.
Datasets used in this study with acquisition platform, variables, and
sampling frequency.
Datasets
Platform
Variables
Frequency
Ground-based station
Savè super-site (8.03∘ N, 2.49∘ E)
NO2, NO,
Raw data: 1 hz
operated by KIT–UPS universities
and CO concentrations
Presented: hourly averages
Aircraft
ATR 42, Twin Otter, Falcon
Relative humidity
Raw data: 1 hz
operated by SAFIRE, BAS, and DLR teams
Wind direction and speed
Presented: 3 min averages
Radiosonde
Launch sites:
Wind direction and speed
High resolution 1 hz
Abidjan, Accra, Cotonou, Savè
Relative humidity
Presented: 100 m averages
Satellite
MODIS on Terra
AOD (550 nm)
Daily
and Aqua
level 3 (1∘×1∘)
Map of the modeling domain (red rectangles) with location of the
major cities (red dots), of the Lomé airport, and of the Savè super-site
(white dots). Superimposed are the flight tracks of the three research
aircraft during the 1–7 July 2016 period (gray lines). The aircraft flight
tracks on 5 July are colored for the German Falcon (blue line), the French
ATR 42 (violet line), and the British Twin Otter (yellow line).
Three super-sites have been implemented in the framework of the DACCIWA
project in Kumasi (Ghana), Ile-Ife (Nigeria), and Savè (Benin). Unlike the
two others, Savè is representative of transport-related air quality
issues affecting small cities, characterized by low local emissions,
downstream of large coastal cities. It is ideal in that the terrain is very
flat with no orographically induced circulation impacting the monsoonal flow.
Thus to study NOx and CO from coastal urbanized areas, this
rural environment is well suited.
At Savè, the Karlsruhe Institute of Technology (KIT) and the Paul
Sabatier University (UPS) have set up meteorological and atmospheric
composition measurements. UPS installed a chemical analyzer
(ThermoEnvironment instrument), which measured NO2, NO, and CO
surface concentrations . Raw observations
acquired at 10 s are averaged hourly. The detection limit of the instrument
is 0.05 ppb for NO2 and NO and 12 ppb for CO
. The measurement site is upwind of Savè city
when the wind corresponds to the monsoon flow (SW sector). On 3 July from
18:00 to 21:00 UTC, the wind direction was shifted, which corresponds to
local pollution. This period has been removed from the analysis.
The DACCIWA aircraft campaign took place during the period of 25 June–14 July
2016 and was based at the Lomé (Togo) airport . Three
research aircraft were involved: a Twin Otter operated by the British
Antarctic Survey (BAS), an ATR 42 operated by the French Service des Avions
Français Instrumentés pour la Recherche en Environment (SAFIRE), and
a Falcon operated by the German Deutsches Zentrum für Luft- und
Raumfahrt (DLR). We base our study on three variables, namely relative
humidity (RH), wind direction, and wind speed, which are measured by core
meteorological instrumentation. The flight trajectories used are depicted in
Fig. . For the three aircraft, raw observations acquired at
1 Hz are averaged over 3 min time steps.
The DACCIWA project included a large radiosonde component with locations
carefully chosen building on the AMMA radiosonde campaign experiences
. We use radiosondes
launched from four locations: Abidjan, Accra, Cotonou, and Savè (see
Table ). There were four releases per day at around 00:00,
06:00, 12:00, and 18:00 UTC. In Savè, more radiosondes were launched
every 1.5 to 3 h at the super-site during the intensive observation
period of 1–7 July 2016 .
We analyze the horizontal spatial extent of the main aerosol plumes from
satellite observations of aerosol optical depth (AOD) at 550 nm made by
MODIS (Moderate Resolution Imaging Spectroradiometer) on both the Aqua
platform (MYD08-D3-6 dataset DOI: 10.5067/MODIS/MYD08_D3.061, passing
over the studied region at 13:30 UTC) and the Terra platform (MOD08-D3-6
dataset DOI: 10.5067/MODIS/MOD08_D3.061, passing over the studied
region at 10:30 UTC). Daily MODIS AOD averages at 1∘ resolution have
been calculated from the Collection 6 combined product of the Dark Target
retrieval available over oceans or non-bright continental surface and the
Deep Blue retrieval available over deserts
.
In order to identify the altitude of aerosols together with their speciation,
we use the spaceborne Cloud-Aerosol Lidar with Orthogonal Polarization
(CALIOP) aerosol type classification . This classification
is suited and accurate to distinguish homogenous aerosol plumes with
different optical properties such as sea salt, dust, and BB
e.g.,. The CALIOP cross sections are very useful since
this is a realistic way to have an instantaneous evaluation of the aerosol
layer altitudes, with their type depending on backscattered optical
measurements, e.g.,. Data are available on
https://www-calipso.larc.nasa.gov/ (last
access: 10 January 2019).
Numerical modeling by WRF–CHIMERE models
The WRF–CHIMERE simulations presented in this study have a setup similar to
those used by . Both models are run offline in a nested
configuration on the same grids with two domains: a regional domain
(10 km × 10 km, extending from 1∘ S to 14∘ N and
from 11∘ W to 11∘ E) and a high-resolution coastal domain
(2 km × 2 km). The simulations over the regional domain are
started on 1 June 2016. In the following, we present only results modeled
over the high-resolution domain (Fig. ). The simulated period
over the high-resolution domain (1 to 7 July) is entirely included in the
2016 WAM post-onset phase, which has been defined from 22 June to
20 July 2016 by .
Meteorological fields from the WRF model
Meteorological variables are modeled with the regional non-hydrostatic WRF
model version 3.7.1, . The domains have a constant
horizontal resolution with 32 vertical levels from the surface to 50 hPa,
including about 10 vertical levels below 1 km a.m.s.l. We use a two-way
nesting for the communication between different domains.
Global meteorological fields are taken from the US Global Forecast System
(operational final analyses) produced by the National Center for
Environmental Prediction (ds083.3 dataset DOI: 10.5065/D65Q4T4Z). These
fields are used to provide meteorological initial and boundary conditions
and to nudge hourly fields of pressure, temperature, humidity, and wind in the
WRF simulations, with spectral nudging, which has been evaluated for regional
models by . In order to enable the PBL variability to be
resolved by WRF, low-frequency spectral nudging is used only above 850 hPa.
The WRF model setup is as follows: the microphysics scheme is the WRF Single-Moment 6-class Microphysics Scheme (WSM6), the radiation scheme is the Rapid
Radiative Transfer Model for General Circulation Models (RRTMG) with the
Monte Carlo independent column approximation (McICA) method of random cloud
overlap from , the PBL physics are computed using the
Yonsei University scheme , the cumulus parametrization is
the ensemble Grell–Dévényi scheme (for the high-resolution domain,
convective precipitation is explicitly calculated and not parametrized), the
surface layer scheme is the Carlson–Boland viscous sublayer, and the surface
physics is calculated with the Noah land surface model scheme with four
soil temperatures and moisture layers . This setup has already
been used by because it allows the reproduction of a
satisfactory diurnal cycle of wind speed over SWA according to
.
MODIS AOD 1-day moving average of two products acquired by Aqua and
Terra (the combined Dark Target and Deep Blue MYD08-D3 and MOD08-D3 products,
respectively) on 3 July 2016 (a) and 5 July 2016 (b). Data
excluded by the cloud screening process are in gray. The modeling domain is
presented by the black square.
Gaseous tracers transport from the CHIMERE model
CHIMERE is a regional chemistry-transport model (version 2017), fully
described in and . The 32 vertical
levels of the WRF model are projected onto the 20 levels for CHIMERE from the
surface and up to 200 hPa.
In this study, the model is used in its tracer version and there is no
atmospheric chemistry. We choose to release passive gaseous tracers in the
simulation because we want to analyze only their transport (no chemistry, no
deposition) caused by the monsoonal flow. Since we want to distinguish the
relative contribution of several coastal cities from the pollution further
inland, we designed a first experiment for which we impose the tracer
emissions at specific urbanized locations: Abidjan (Côte d'Ivoire), Accra
(Ghana), Lomé (Togo), Cotonou (Benin), and Lagos (Nigeria) (see
Table for coordinates). Specific tracers are emitted for a
given city in order to distinguish their relative contributions at inland
locations. Thus the tracer emissions occur in a single grid cell
corresponding to the center of each city.
The tracer emissions are constant and continuous during the modeled period
(1–7 July). This allows the quantification of the variability due to the meteorology
only. Emissions are released at the lowest level of the model (below 10 m
in altitude) and are proportional to the population of each city; this approach
has also been used by . We defined an arbitrary emission
for 1 million inhabitants. Then we multiply this emission by a factor
depending on the population (see Table ): 1.8 for Lomé,
2.2 for Cotonou, 3 for Accra, 5 for Abidjan, and 13 for Lagos.
A second objective is to understand the interactions between urban (URB) and
BB pollutants resulting from long-rang transport on 5 July.
For this, we design a second numerical tracer experiment in which BB tracers
are added to the URB tracers and are tagged to be different from the URB
tracers. BB tracers are released to reproduce the BB layer observed with
MODIS on 5 July at 00:00 UTC to 6 July at 23:00 UTC with a spatial
horizontal extent going from 1∘ W to 2∘ E at
4.5∘ N and at an altitude of ≈1.5 km (see
Fig. and Sects. and for
justifications).
The two tracer experiments are summarized in Table . They
are accessible on the DACCIWA database (10.6096/BAOBAB-DACCIWA.1760,
).
Main characteristics of the two numerical tracer experiments using
high-resolution modeling at 2 km grid spacing with tracer emissions relevant
for biomass burning (BB) and urban pollutants (URB).
Tracer Experiment 1
Tracer Experiment 2
Tracer type
URB tracers only
BB and URB tracers
Release duration
1–7 July
5–7 July (BB) and 1–7 July (URB)
Release altitude
lowest level
at 1500 m (BB) and lowest level (URB)
Release location
five cities
from 1∘ W to 2∘ E at 4.5∘ N (BB) and five cities (URB)
Number of tracer
five (each city)
two (BB and URB)
The tracers are transported using the van Leer scheme .
There is no sink for tracers (no deposition and no chemical reaction). The
tracers are chosen to be gaseous and are representative of the gaseous part
of the URB emissions and the BB plume. This choice of gaseous tracers was
made to be consistent with the gaseous concentrations measured by the
aircraft and the surface gaseous concentrations measured at the Savè
super-site. The period is not associated with widespread rain. At Savè,
there were only some small precipitation events . We
focus our analysis on gaseous species but we suppose similar transport
patterns for aerosol and gaseous pollutants because of the constant monsoon
flow blowing over SWA during our studied period. The only difference from
aerosol is the absence of settling. But this long-term impact could be
considered negligible in this study focused on a few days and a spatially
restricted region: it is assumed that gaseous and aerosol species are
transported in the same way by the meteorological flow.
Large-scale atmospheric patterns over the Gulf of Guinea
This section is dedicated to analysis of the atmospheric dynamics,
thermodynamics, and composition across SWA using AOD from satellites in
Sect. , together with RH and wind from
radiosondes in Sect. and from aircraft in Sect. .
The prerequisite to realistic numerical tracer experiments is the accuracy of
the meteorological simulation. The WRF meteorological simulation is therefore
extensively compared to in situ observations made by both radiosondes and
aircrafts.
Regional-scale aerosol distribution
This section investigates the daily MODIS AOD observations for the period of 1
to 7 July 2016. Two important types of aerosols can be advected towards SWA:
dust from the north and BB from the south. We focus on two different days, 3
and 5 July 2016 (Figs. and ). Note that we
present 1-day moving averages (Figs. ) because we analyze
the long-range transport of aerosols (using the MODIS level 3 product with a
coarse resolution of 1∘).
During the studied period, high AOD values are found north of the domain over
the Sahel (north of 14∘ N) and south of the domain over the Gulf of
Guinea (on average over the period of 1–7 July, Fig. ). The
origin of the high AOD over the Gulf of Guinea is well known. This is the BB
layer coming from central Africa, where intense vegetation fires occur during
this season with increasing trends over the period of
2001–2012 . Part of this pollution is transported over the
Gulf of Guinea and the BB plume reaches the Guinea coast, as seen during the
AMMA campaign . The BB pollutant concentrations
observed along the Guinea coast depend on the synoptic wind patterns
. The presence of this layer is confirmed over the Gulf of
Guinea using CALIPSO data acquired on 5 July (Fig. ), which
gives a layer altitude between 1 and 3 km above mean sea level (a.m.s.l.).
It is worth noting that during this period, there is no evidence of mineral
dust transport over the studied cities. The dates 3 and 5 July 2016 are two contrasting
days in terms of AOD values over the Gulf of Guinea (Fig. ).
On 2 July, AOD values are low over the continent and moderate over the ocean
(Fig. ), which is in agreement with , who
have shown that the BB layer is present close to the Guinea coast but it does
not reach the coast. On 3 July, AOD values are low to moderate over our
domain (AOD < 0.5, Fig. ). On 4 July, there is a pattern
of high AOD (AOD > 0.5) 100 km south of the coast
(Fig. ). On 5 July, the BB layer reaches the coastline
(Fig. ), then on 6 July it seems to penetrate inland but
clouds prevent AOD retrievals over Togo and Benin (Fig. ). On
7 July, this layer is no longer visible close to the Guinea coast
(Fig. ).
For 6 July, have shown a clear large-scale BB signature
between Abidjan and Accra with in situ measurements made onboard the research
aircraft. Moreover have analyzed atmospheric chemistry and
demonstrated mixing of urban pollutants with advected BB into the region.
This interpretation has been supported by backward trajectories locating the
origin of the BB plume in central Africa.
Vertical layers in the lowermost troposphere
In this section, we combine observations from the high-resolution radiosondes
(Sect. ) and the three aircraft (Sect. )
over the period of 1–7 July. For radiosondes, we analyze 32 vertical profiles
in Abidjan, 32 in Accra, 26 in Cotonou, and 51 in Savè
(Fig. ). For aircraft data, we analyzed 11 flights
including six of the ATR 42, four of the Falcon, and one of the Twin Otter
(Table ). Aircraft observations are acquired only during the daytime.
The modeled and observed dynamical and thermodynamical variables are compared
in order to identify the different layers. The modeled variables have been
interpolated along the balloon trajectories and aircraft flight tracks using
a spatial bilinear interpolation and then temporal and vertical linear
interpolations with a 3 min time step.
Observed and modeled mean vertical profiles of wind speed (m s-1) and direction (360∘ circle with 0 and
360∘ is the north and 90∘ is the east) and relative humidity
(RH in %) averaged for all profiles over the period of 1–7 July 2016 at
Abidjan in Côte d'Ivoire (green line), Accra in Ghana (blue line), Cotonou in
Benin (purple line), and Savè in Benin (orange line). The mean and
standard deviation at the four locations are represented by the black line
and the gray shading. WRF-derived variables are interpolated to the
radiosonde positions. The right panel presents the (mod–obs) mean vertical
bias of each location and of the average of the four locations.
Identification from radiosondes
For wind speed, the vertical profiles observed at the four locations have a
similar shape (Fig. ). The mean wind speed increases from
the surface to 300 m a.m.s.l., then decreases to 3 m s-1 at
1.5 km a.m.s.l. and finally increases to a maximum of about 8 m s-1
between 3 and 5 km a.m.s.l. The model predicts a vertical wind profile in
good agreement with observations from the surface to 300 m a.m.s.l. but
there is an increasing positive bias from 300 m to 1 km a.m.s.l., reaching
+2 m s-1 at 1 km a.m.s.l. At 300 m a.m.s.l., observed wind speed
reaches 6 m s-1 on average, which shows the NLLJ signature
. The model reproduces this signature with wind speed
reaching 7 m s-1 but over-estimates its altitude (at 400 m a.m.s.l.)
at the three coastal sites (Abidjan, Accra, Cotonou).
Observed and modeled distribution (first quartile, median, third
quartile), mean and bias (absolute and relative) of relative humidity (%),
and
wind direction (∘) and speed (m s-1) measured by the three
aircraft over the period of 1–7 July 2016 separated into three altitude ranges:
surface to 1 km, 1 to 2 km, and 2 to 4 km a.m.s.l.
Var
N
Q1
Median (Q2)
Q3
Mean
Bias
Obs
Mod
Obs
Mod
Obs
Mod
Obs
Mod
Absolute
Relative (%)
0 to 1 km
RH
349
90.75
85.24
94.80
88.65
98.61
93.50
94.19
88.12
-6.07
-6 %
W. speed
349
3.80
3.98
5.25
5.55
6.47
6.80
5.31
5.51
0.19
4 %
W. dir.
349
220.08
214.17
229.43
228.47
244.41
237.67
231.78
228.19
-3.59
-2 %
1 to 2 km
RH
116
91.41
88.10
95.58
90.85
98.77
94.79
94.33
90.86
-3.47
-4 %
W. speed
116
2.10
1.88
2.90
2.80
4.04
3.94
3.30
3.04
-0.26
-8 %
W. dir.
116
191.17
207.29
244.71
246.59
285.29
272.28
228.56
228.41
-0.15
>1 %
2 to 4 km
RH
62
76.34
73.99
84.43
76.86
90.44
85.18
76.12
79.63
3.51
5 %
W. speed
62
3.34
2.23
4.71
5.69
9.08
9.75
6.37
6.67
0.30
5 %
W. dir.
62
55.73
45.76
72.12
69.04
137.67
145.84
107.36
105.24
-2.12
-2 %
The vertical profile observed at Savè stands out from the three other
cities with lower wind speed below 1 km and higher wind speed above
2 km a.m.s.l. The lower wind speed near the surface may be related to the greater
distance from the coast, resulting in a stronger deceleration by friction,
which is not reproduced by the model. When looking at 3 km a.m.s.l., this
is close to the altitudinal maximum of the AEJ. Savè is located at a
latitude closer to the AEJ core, which is seen at about 10∘ N
. The jet is clearly observed only at Savè, with
wind speeds of up to 10 m s-1 at 3 km a.m.s.l., which is modeled in
good agreement.
For wind direction, the four cities have again similar profiles. The mean
observed and modeled vertical profiles are composed of three distinct layers.
From the surface to 1 km a.m.s.l., the monsoon layer corresponds to wind
coming from the sector between 210 and 240∘. From 2 to
5 km a.m.s.l., wind direction is also almost constant between 80 and
120∘. In between these two layers, which are well defined in terms of
direction, there is a layer characterized by a quick change of direction from
240 to 120∘. This layer associated with weak wind speed is a
directional shear layer. On average, the monsoon layer depth seems to be
overestimated by about 200 m because the modeled wind direction is biased
by about +20∘ between 1.2 and 2 km a.m.s.l.
For the three variables, the profiles of their standard deviation present the
same modeled and observed characteristics (gray shading in
Fig. ). For wind speed, the standard deviation is about
2 m s-1 from the surface to 2 km a.m.s.l., and it increases up to
4 m s-1 from 2 to 4 km a.m.s.l. For wind direction, the standard
deviation is low (about 45∘) from the surface to 1 km a.m.s.l.; it
increases from 1 to 2 km in the directional shear layer and it decreases
from 2 to 4 km a.m.s.l. For RH, the standard deviation is about 10 %
from the surface to 2 km a.m.s.l., and it increases in the AEJ layer but
the model does not reproduce the low observed RH values in this layer.
This analysis shows that the modeled monsoon layer is too deep when it
arrives at the coast. Further inland the monsoon flow is too fast when it
reaches Savè. The comparison between observed and modeled meteorology
also reveals that the model reproduces the several vertical
layers in terms of wind direction and speed and thus most likely the
transport that we want to characterize using the tracer experiments well enough.
Identification from research aircraft
The atmospheric vertical structure can be separated into three different
layers (see Sect. ): (i) from the surface to
1 km a.m.s.l., there is the monsoon flow with RH > 90 %, wind speed
>4 m s-1 and direction from the southwesterly sector; (ii) between
1 and 2 km a.m.s.l., there is a directional shear layer associated with high
RH > 90 % and with low wind speed and changing wind direction;
(iii) between 2 and 4 km a.m.s.l., this is the AEJ layer with
RH < 80 %, reversed wind direction coming from the northeast, and wind
speed up to 8 m s-1. In order to evaluate the model, aircraft
measurements during the daytime are separated into three corresponding altitude
ranges (Table ).
From the surface to 1 km a.m.s.l., the modeled wind speed and direction
match well with the observations (absolute bias lower than 0.2 m s-1
and 4∘, respectively). The observed distribution of the monsoon wind
is captured by the model up to 1 km a.m.s.l. (interquartile range
3.80–6.47 m s-1 for the observations and 3.98–6.80 m s-1 in
the simulation). The modeled distribution of RH shows a dry bias in the
monsoonal flow (of -6 %).
In the directional shear layer, from 1 to 2 km a.m.s.l., observed and
modeled wind speed distributions are narrower than in the monsoon layer
(interquartile ranges being 2.10–4.04 m s-1 and
1.88–3.94 m s-1, respectively), showing that this layer is well
defined over the domain. The modeled wind direction is in good agreement with
observations (relative bias lower than 1 %), although with a wider
distribution (observed interquartile range 191.17–285.29∘ and
modeled 214.17–237.67∘) than in the monsoon layer (observed
interquartile range 220.08–244.41∘ and modeled
207.29–272.28∘).
In the main easterly layer, from 2 to 4 km a.m.s.l., the observed
distribution of wind speed is wider at lower altitudes, which is in good
agreement with the modeled distribution. Observed and modeled RH and wind
direction distributions are also consistent.
In conclusion, during the daytime, the monsoon layer is reproduced with a dry
bias. The low (relative) biases of wind speed (+4 %) and direction
(-2 %) in this layer are of prime importance to accurately model the
URB transport.
From the coast to the north: 5 July Lomé–Savè flights
As we want to understand northward pollution transport, we need to focus on
the wind direction and speed from the coast to the north. In this section, we
analyze the spatial variability in the wind over the Lomé–Savè
transect. We compare aircraft measurements of wind to modeled values using
data acquired during three specific flights conducted on 5 July at different
times of day with similar flight plans (see Fig. ) and similar
altitude ranges (i.e., flying mostly below 2 km a.m.s.l.). The French ATR 42
flight took place between 08:00 and 11:00 UTC, that of the German Falcon
between 11:20 and 15:00 UTC, and that of the British Twin Otter between
16:00 and 17:50 UTC.
Using ceilometer measurements, have described the cloud
base height evolution on 5 July at Savè, which was between 200 and
1000 m during the ATR 42 fight, between 400 and 1800 m during the Falcon
flight, and between 1000 and 3800 m during the Twin Otter flight. This was a
cloudy day, which allowed the operational center to plan for characterizing
the diurnal cycle of low level clouds .
Time series on 5 July 2016 of (a) the French ATR 42,
(b) German Falcon, and (c) British Twin Otter aircraft data,
composed of three panels for each aircraft: (top) altitude (m) with the
latitude (∘N), (middle) wind speed (m s-1), and (bottom) wind
direction (degrees). Modeled values with the WRF model are interpolated
along the flight positions (red line). Observed value averages are the black
dots with the hourly standard deviation (error bars).
During the ATR 42 flight in the morning, measurements of wind speed range
from 2 to 10 m s-1 (Fig. ). The highest values are
observed close to the coast (greater than 8 m s-1) during the two
times the aircraft passes there. The model reproduces the spread of the
observed values below 2 km a.m.s.l. When the altitude reaches
2 km a.m.s.l., observed wind speed decreases below 4 m s-1. Wind
direction ranges from 240 to 300∘, even at about 2 km a.m.s.l. The
model predicts a constant direction at 250∘, except when flying above
2 km a.m.s.l. because the modeled direction changes, revealing an
underestimation of the modeled PBL depth.
In the morning, the monsoon layer is modeled with an overestimation of the
wind speed and with a sharp directional shear at 2 km a.m.s.l., whereas we
observe an important variability in wind speed without reaching the
directional shear layer up to 2 km a.m.s.l. This behavior of the model
suggests that low-level clouds are not well represented, leading to modeled
PBL depth underestimation.
During the Falcon flight around midday, wind speed also decreases from the
beginning of the flight at the coast (up to 10 m s-1) to 100 km
further north (less than 4 m s-1). Wind direction varies smoothly from
250∘ at the coast to 300∘ close to Savè. The model is
able to reproduce the weakening of the monsoon layer linked to daytime dry
convection , and the variability in observed and
modeled wind direction and speed is in better agreement.
In contrast to the ATR 42 and Falcon flights, the Twin Otter flew only one
time over Savè and made three vertical profiles up to 3 km a.m.s.l.
During the Twin Otter flight in the afternoon, the range of wind speed
increases compared to the two previous flights, reaching between 1 and
12 m s-1. The model reproduces the wind in the monsoon layer well but
does not capture the wind direction changes between 1 and 3 km a.m.s.l.
There is no clear change of the direction during the first sounding, but
during the latter two soundings, wind direction changes from southwesterly
winds at 1 km a.m.s.l. (about 225∘) to northeasterly winds at
3 km a.m.s.l. (about 45∘). The model predicts a too sharp wind
direction change from 1 to 3 km a.m.s.l., which shows that the directional
shear layer depth is underestimated.
Overall below 1 km a.m.s.l., wind speed ranges from 4 to 10 m s-1
with a direction from about 250∘, which is in good agreement with the
model. have shown that 5 July was during a period when
the AEJ weakens and becomes more fragmented, which has led to relatively
patchy signals in wind and vorticity. This results in observed wind direction
mostly greater than the third quartile of the distribution measured over the
period of 1–7 July 2016 (Q3 = 244∘ below 1 km a.m.s.l.; see
Table ).
The three aircraft cover the same region from 08:00 to 17:00 UTC. It is thus
possible to quantify the evolution of dynamical variables during the daytime. We
have selected a box crossed many times by the aircraft in order to compare
hourly averages of in situ wind speed and direction observations to the
modeled values. The box is delimited in latitude from 6.6 to 7.8∘ N,
in longitude from 1.5 to 2.2∘ E, and in altitude from 300 to 1000 m a.m.s.l. (Fig. ). When the three flights cross this box, the average
and standard deviation are calculated from observed values. For the model, we
present the average and standard deviation of each hour calculated from all
grid cells included in the box. Wind speed observations decrease from 08:00
to 13:00 UTC (Fig. ), then increase again in the afternoon (but
there are only a few measurements made by the Twin Otter aircraft in the two
boxes at 17:00 UTC). The model does not capture the morning evolution,
when the NLLJ is eroded, well. We note that the minimum of wind speed is modeled
and observed in the early afternoon, when vertical mixing is strongest.
Observed wind directions are almost constant in the box at about
225∘. There is a change of the direction between 16:00 and 18:00 UTC
for both the model and the observations, which shows the establishment of the
NLLJ.
On the one hand, these comparisons with aircraft measurements reinforce our
confidence in the model to adequately reproduce the wind speed and direction
and
thus the main characteristics of pollution transport between Lomé and
Savè. On the other hand, this analysis confirms that the PBL depth is not
accurately modeled, especially in the morning, which could in turn impact the
pollution mixing and dilution. During the day, the surface concentration
could be overestimated and the concentration at the PBL top height
underestimated, especially going further away from the sources.
Inland pollution transport from the coast
Firstly, this section investigates CO and NOx concentrations
at the Savè super-site (Sect. ). Secondly, we analyze
the contribution of the major cities along the coastline to the pollution
budget in Savè (Sect. ). Thirdly, we study how URB plumes
and the BB layer observed on 5 July interact at the coast and are transported
inland.
Surface pollutant concentrations at Savè
We analyze the temporal evolution of CO, NO2, and NO concentrations.
Trace gas concentrations were measured at the ground level at the Savè
DACCIWA super-site. We study the hourly temporal variability in observed
concentrations over the studied period (1 to 7 July).
Time series of (a) carbon monoxide (CO), nitrogen dioxide
(NO2), and nitrogen monoxide (NO) hourly concentration averages (ppb) observed at Savè (Benin) for the period of 1–7 July 2016 and
(b) of the ratios CO/NOx (in black) and
NO2/NO (in red). The vertical dashed lines indicate periods of 6 h
starting at 00:00 UTC.
During this period, the hourly CO concentration varies between 140 and
250 ppb (Fig. a). Moreover, looking at the entire
campaign period (25 June to 15 July), hourly CO concentration ranges from 110
to 250 ppb (Fig. a). Comparing these two periods,
we note that our studied period seems representative for the diurnal cycle
over the campaign period. There is a clear diurnal cycle with the maximum
occurring every day at the beginning of the night (between 220 and 250 ppb)
and the minimum at the beginning of the day.
Hourly NO2 concentration ranges from 0.2 to 3.5 ppb. We note also
that every day there are periods of high NO2 concentrations in the
evening and low NO2 concentrations from the morning to the
afternoon. It is worth noting that high CO values noticed in the evening are
associated with high NO2 but not with high NO concentrations.
Given the short lifetime of NO (less than 1 h) in the PBL
, the analysis of NO concentration gives some clues to
understand NO2 variability because NO is mostly linked to local
sources (i.e., not transported). The baseline of NO concentration is 0.09 ppb
(median). High NO concentrations (>0.5 ppb) are measured on 1 and 7 July
in the evening. Moreover, there is an increase every evening, which shows
that there are local sources close to the instrument location, probably
associated with charcoal stove cooking or traffic time.
In order to identify periods of high NO2 associated with low NO
concentrations, we have computed the NO2/NO ratio
(Fig. b). This ratio is expected to increase at night by
the ozone titration (O3+NO reaction). During the daytime, it depicts
local or transported pollution (respectively a low or high NO2/NO
ratio). Every day we note a sharp increase from 18:00 to 00:00 UTC
(NO2/NO > 15) associated with low NO concentrations
(NO < 0.2 ppb), which suggests transported pollutants.
In order to identify periods of high CO associated with low
NOx concentrations, we have computed the
CO/NOx ratio (Fig. b). When a BB layer
reaches the Guinea coast, gaseous nitrogen oxide concentrations are lower
than 0.1 ppb because gaseous nitrogen oxides
have been converted into the particulate phase during the transport over the
southeast Atlantic. We therefore expect an increase in CO and constant
NOx, when the BB layer reaches Savè without being mixed
with URB (containing NOx in the gaseous phase). At Savè,
the CO/NOx ratio is not higher after the arrival of the BB
layer on 5 July (see Sect. ). There is an increase in CO
together with a NOx increase; thus when the BB layer reaches
Savè, it is mixed with urban pollution or transported above the PBL.
Hourly diurnal cycles of (a) carbon monoxide (CO in black), (b) nitrogen dioxide (NO2 in blue), and nitrogen
monoxide (NO in red) concentrations (ppb) observed at Savè (Benin).
Means of each hour are presented by the lines over the period of 1–7 July 2016
and the upper and lower shading limits correspond to the hourly ranges
(maximum and minimum of each hour over the period).
In order to determine the diurnal cycle of the three pollutants, we present
observed hourly averages over 1–7 July together with the maximum and minimum
of each hour measured (Fig. ). There is a clear
diurnal cycle of hourly CO concentration averages with the minimum occurring
at 08:00 UTC (about 160 ppb) and with the maximum occurring between 18:00
and 22:00 UTC (greater than 200 ppb) over the period of 1–7 July
(Fig. ) and also over the entire campaign period
(Fig. ). This time is in agreement with
and , who have found using the super-site
instrumentation that the coastal front starts moving northward after
16:00 UTC, reaching Savè in the evening. It also corresponds to the
highest hourly minimum (190 ppb) and maximum (250 ppb). It is worth noting
that CO concentration remains greater than 180 ppb from 22:00 to 04:00 UTC.
For NO2, we also note a clear diurnal cycle with low hourly
concentration averages between 08:00 and 15:00 UTC and with high hourly
concentration averages (greater than 1 ppb) from 18:00 to 21:00 UTC over
the studied period and also over the entire campaign period
(Fig. ). The two small NO increases (from 0.2 to
0.6 ppb) at 12:00 and 19:00 UTC are consistent with the usual time of local
activities such as traffic and charcoal cooking. At 21:00 UTC, there is
a high NO2 concentration, as well as a high CO concentration, which
is not associated with a high NO concentration, suggesting pollution
transport because it is the time of the coastal front passage
.
The NO2 diurnal cycle is similar to the one of CO with a minimum at
08:00 UTC (about 0.6 ppb) and a maximum between 18:00 and 21:00 UTC. The
main difference of the NO2 and CO diurnal cycles occurs at night
between 21:00 and 02:00 UTC because CO remains high (≈200 ppb),
whereas NO2 decreases from 1.3 to 0.7 ppb. This result could be
linked to a higher ratio of BB compared to URB.
In conclusion, at Savè, there are similar diurnal cycles of CO and
NO2 with maxima between 18:00 and 21:00 UTC. Moreover NO
concentration is very low at 21:00 UTC, indicating pollution transport from
the coastal urban agglomerations and not local production. The BB layer could
interact with the URB plumes in the PBL, thus increasing the CO
concentration. We need to understand how the BB layer is mixed with URB at
the coast and how it is transported further inland.
Contribution of major coastal cities to the pollution budget at Savè
This section aims at identifying which major cities have a significant
contribution to inland pollution at Savè. For this, we analyze Tracer
Experiment 1 described in the model
section (see Table in
Sect. 2.3.2).
In order to present this experiment, the synoptic wind patterns and the
pollution plumes of the coastal cities with the URB tracers are displayed in
a single figure (Fig. ). This figure represents an average of
the modeled plumes over the period of 1–7 July in the monsoon layer (from
surface to 1 km). Results are presented in an arbitrary unit with the same
isocontour value (Iso1) of tracer concentration for each city (color shadings
in Fig. ).
Maps of URB tracer concentration (in arbitrary units) averaged over
the period of 1 to 7 July 2016 in the monsoon layer (from the surface to
1 km a.m.s.l.). Tracers are released from Abidjan (Côte d'Ivoire) in green,
from Accra (Ghana) in blue, from Lomé (Togo) in orange, from Cotonou
(Benin) in violet, and from Lagos (Nigeria) in red. The same threshold of tracer
concentration is used for all city plumes for the color shading (Iso1). Wind
vectors at 10 m modeled by WRF are presented by the light blue arrows. The
location of the DACCIWA super-site (Savè in Benin) is presented by the
light green dot.
Over the Gulf of Guinea, we can see markedly higher wind speed than over the
continent. This figure shows that the pollution plumes of Accra, Lomé, and
Cotonou could reach Savè, while the direction of the Lagos and Abidjan
plumes is not oriented towards Savè.
We now focus on the temporal variability reproduced by the tracer experiment
at Savè (light green dot in Fig. ). The tracer
concentration of the five cities has been interpolated to Savè
coordinates (Fig. a). The tracer emissions started on
1 July at 00:00 UTC. The first tracer plume that reaches Savè is that
from Cotonou in the evening. We can see that tracers emitted from Cotonou
reach Savè every day in the evening, typically at around 19:00 UTC. In
the morning, the Lomé pollution plume reaches Savè, while the Accra
pollution plume reaches Savè in the afternoon. There is a short period
when hourly concentrations are at a maximum every day, and this peak is
associated with the arrival of the Cotonou plume in the evening. This pattern
is seen repeatedly over the entire 1–7 July period. The model clearly
predicts identified periods when Savè is under the influence of different
cities, which implies that these periods correspond to pollution plumes
characterized by different chemical ages.
From 5 to 6 July, the contribution of Cotonou decreases and the Accra and
Lomé contributions increase, which suggests a modification of wind
patterns. From midnight to the end of the night, there is no city plume
reaching Savè. However, we have seen in Sect. that a
high CO concentration persists during the night.
The average diurnal cycle of tracers is presented with the contribution of
each city (Fig. b). It confirms that there are
distinct periods when Savè is under the successive influences of Lomé
in the morning (06:00 to 12:00 UTC), of Accra in the afternoon (12:00 to
18:00 UTC), and of Cotonou in the evening (18:00 to 01:00 UTC). Lagos
tracers do not reach Savè because of the southwesterly monsoon flow.
The observed increases in surface concentration may be explained by a couple
processes: long-range transport of the URB plume and/or a local collapse of
the nocturnal boundary layer, quickly concentrating locally emitted
pollutants. In our case, the dominant effect is the long-range transport,
which is mainly associated with the transport of the Cotonou plume. Moreover,
during the entire campaign, we did not note any increase in NO at night
(Figs. and ). These
results suggest that the Cotonou plume affects Savè during a short
period with a maximum between 21:00 and 22:00 UTC (about 2 times greater
than the peak magnitude due to Lomé). This is in agreement with
observations of CO and NOx concentrations (see Sect. ).
We now need to investigate the diurnal cycle of pollutant transport from
coastal cities.
Mixing and transport of urban anthropogenic and biomass burning
In this section, results of Tracer Experiment 2 (see Table ) are discussed. We present the
spatial patterns of URB and BB tracer concentrations averaged over the three
layers described in Sect. and then we analyze the vertical
structure of these two types of pollution. We focus on 5 July when the BB
layer reaches the Guinea coast (see Sect. ). The first layer
height is 300 m, which is roughly the minimum PBL top height at night.
Time series of hourly URB tracer concentrations (in arbitrary units)
modeled by the CHIMERE model at the DACCIWA field campaign ground station in
Savè (Benin) for the period of 1–7 July 2016. Urban tracers are released
from five coastal cities: Abidjan (Côte d'Ivoire) in green, Accra (Ghana) in
blue, Lomé (Togo) in orange, Cotonou (Benin) in violet, and Lagos (Nigeria)
in red.
Maps of concentrations of URB (in gray) and BB (in brown) numerical
tracers on 5 July 2016 at 13:00 UTC (column a, c, e) and at
21:00 UTC (column b, d, f). Concentrations are averaged over three
layers (a, b) from the surface to 300 m a.m.s.l., (c, d) from 0.3 to 2 km a.m.s.l., and (e, f) from 2 to
4 km a.m.s.l. Wind vectors at 10 m from the WRF model are presented by the
light blue arrows. The meridional–vertical transect shown in
Fig. corresponds to the zonally averaged area of the
red rectangle.
In order to analyze the interactions between URB emissions and the BB layer,
we use the information on the spatial characteristics and vertical extent of
the BB layer derived from MODIS and CALIPSO observations
(Sect. ) by releasing passive gaseous tracers with a spatial
horizontal extent from 1∘ W to 2∘ E at 4.5∘ N and
with an altitude of ≈1.5 km a.m.s.l. from 5 July at 00:00 UTC to
6 July at 23:00 UTC (see Table ). Although a non-negligible
part of BB is transported in the marine PBL, measurements performed during
the DACCIWA campaign have confirmed that the BB layer altitude over the ocean
is mostly between 1 and 3 km a.m.s.l. . URB tracers are
not separated by city in this experiment. The threshold of URB tracer
concentration for the isocontour presented in the maps is
Iso2 = Iso1 × 5 (Fig. ).
On 5 July at 13:00 UTC when shallow dry convection is well developed and
looking at the URB in the surface layer, the Accra and Abidjan plumes are
transported towards the north, whereas the Lomé, Cotonou, and Lagos plumes
have a strong eastward component (Fig. a). This
matches the difference in wind direction along the coastline. From 1 to
2 km a.m.s.l., the URB tracer distribution is almost the same (with lower
wind speed), which shows that the PBL reaches 1 km a.m.s.l. at
13:00 UTC. The BB layer emitted at 4.5∘ N is transported northward
between 1 and 2 km a.m.s.l. over the Gulf of Guinea. BB tracers are not
mixed with the marine PBL but with URB tracers from the coast to ≈7∘ N.
On 5 July at 21:00 UTC, when the dry convection has stopped, the model
predicts consistent wind speed between the ocean and the continent with a
stronger northward component than at 13:00 UTC, especially over the
continent. The shape of URB plumes in the first two layers presents two
distinct parts, which seems to follow the change of the wind patterns. The
area where BB and URB tracers are mixed extends from the coast to ≈8∘ N.
Meridional–vertical transect of concentrations of URB (in gray) and
BB (in brown) numerical tracers on 5 July 2016 at 13:00 UTC (a) and
at 21:00 UTC (b). PBL top height from the WRF model is the green
line. The vertical dashed gray lines show the latitude of Accra
(5.6∘ N), Lomé (6.2∘ N), and Cotonou (6.4∘ N).
The black area is the topography. Vectors (light blue arrows) represent wind
in the plan of the transect (with an aspect ratio of 500 between the
meridional and vertical lengths).
The vertical structure of the wind is now analyzed using cross sections along
a meridional transect from 1 to 3∘ E (the red square in
Fig. ). We present the same hours of the simulation
(13:00 and 21:00 UTC on 5 July), with isocontours Iso2 and other
isocontours (Iso3) for both tracers in order to see the core of the plumes:
Iso3 = Iso2 × 5.
The vertical structure changes markedly from 13:00 and 21:00 UTC
(Fig. ). At 13:00 UTC, shallow dry convection occurs
between the coast and ≈7∘ N, which leads to a vertical
mixing of BB and URB tracers (Fig. ). The rising motion
at the coast is linked to the coastal front that occurs during the day
. The sinking motion over the ocean is linked to
the land–sea breeze circulation . BB tracers are
transported over the marine PBL without mixing. They reach the surface
between the coast and ≈7∘ N. URB and BB tracers accumulate
along the coastline until the coastal front begins moving northward, which is
in agreement with and .
At 21:00 UTC, the NLLJ is established at about 400 m a.m.s.l. from the
coast up to 9∘ N where the front is located. At the coast itself,
the front is not present; thus URB and BB tracers are not mixed anymore. The
mixture of URB and BB occurring during the daytime is transported northward
up to 8∘ N. The Iso3 isocontour of BB does not reach the surface. It
shows that at night the BB layer penetrates further inland. BB and URB plumes
are mostly transported above (between 0.5 and 2.5 km a.m.s.l.) and within
the PBL, respectively.
The discussion in this section is supported by two video supplements (click
on the links below):
three layer maps
10.5446/37228 ()
vertical–meridional transects
https://doi.org/10.5446/ 39159 ().
The two videos further illustrate the analysis made at 13:00 and 21:00 UTC
on 5 July. They also provide useful additional information to analyze the day-to-night transition leading to the evening maximum at Savè
(Sect. ). Our simulation reproduces the main features of the
diurnal cycle, which are that vertical mixing occurs during the daytime, while
meridional advection of pollutants is most efficient at night
.
From the wind patterns, we note that the coastal front is present from 09:00
to 15:00 UTC from the coast up to ≈7∘ N. It leads to the
accumulation of URB pollutants. This period is referred as “daytime drying”
by . From 16:00 to 02:00 UTC (from
03:00 to 08:00 UTC), the meridional wind increases (decreases) in the
PBL and URB pollutants are mostly transported northward within the PBL. This
period is referred to “Atlantic inflow” (“moist morning”)
by .
During the daytime drying period, we notice that URB and BB tracers
accumulate along the coastline in the PBL (on 5 and 6 July from 11:00 to
15:00 UTC). When dry convection stops (at 16:00 UTC), wind speed quickly
increases with a stronger northward component. BB and URB tracers are
simultaneously transported northward from the coast. From 16:00 to
02:00 UTC, the front moves towards the north, and the mixture of BB and URB
tracers is advected accordingly. A similar diurnal evolution of BB and URB
transport is simulated on 5 and 6 July.
The timing of the coastal front propagation in our simulation is in agreement
with and , who have shown the same regular
occurrence of a coastal front that develops during the daytime and propagates
inland in the evening. After the frontal passage, there is the establishment
of the NLLJ (with a jet axis around 250 m a.m.s.l.), which is also
reproduced in our simulation (with an overestimation of jet axis altitude of
about 150 m).
Conclusions
In this study, several observational datasets together with high-resolution
model simulation are used to analyze the diurnal cycle of atmospheric
pollution transport over SWA. We focus on two distinct pollution sources,
URB and BB pollutants,
in order to understand their mixing and their advection inland. We first
studied the dynamics and thermodynamics in the lowermost troposphere over SWA
using aircraft, radio-sounding, and ground-based measurements made during part
of the DACCIWA field campaign (from 1 to 7 July 2016). The second part of the
study uses high-resolution numerical tracer experiments. We analyzed
pollution transport from the main urban emission centers of the Guinea coast
(Abidjan, Accra, Lomé, Cotonou, and Lagos) at the super-site of Savè
in order to assess the impact on the air quality of remote inland cities
characterized by low local emissions.
Observations at Savè (185 km to the north of Cotonou) show that there is
a clear diurnal cycle of NO2 and CO, with a maximum occurring every
day between 18:00 and 22:00 UTC, suggesting URB transport from remote
emission sites. From the tracer experiments, we demonstrated that there are
clear and successive periods of the day when air quality in Savè is
affected by different city plumes. More precisely, the contribution of tracers
released from Lomé is greater than 50 % (of the total amount of
tracers) between 02:00 and 12:00 UTC, while from 12:00 to 18:00 UTC tracers
released from Accra constitute the main contributor. Then, during 3 h (from
20:00 to 22:00 UTC), tracers released from Cotonou reach Savè, leading to
a contribution of greater than 80 %, while it is lower than 10 % during
the other 15 h (from 03:00 to 18:00 UTC). Over this period, tracers released
from Cotonou represent a contribution of 40 %, from Lomé of 36 %,
and from Accra of 23 %. Our results suggest that the successive periods
affected by different city plumes are characterized by different chemical
ages.
Previous studies have already highlighted the importance of the diurnal cycle
of the wind along the coastline in the lowermost troposphere over the Gulf of
Guinea during the monsoon
and suggested an influence on pollution transport. The WRF simulation
reproduces a diurnal cycle of the wind over SWA in agreement with
and . Indeed, the structure of the wind
changes from the morning to the evening. When the shallow dry convection over
the land is well developed, wind speed in the PBL reaches a minimum. A
coastal front develops during the day (from 09:00 to 15:00 UTC) and when it
ceases, wind speed quickly increases with a stronger northward component.
There is a specific time of the day (16:00 to 02:00 UTC) for the transport
of pollutants from the coast toward the north, which affects inland air
quality over SWA, human health, and radiative transfer and the diurnal
cycle of low-level clouds.
Our results based on modeling experiments suggest that URB and BB
(transported from central Africa) are accumulated and mixed along the
coastline during the day from 09:00 to 16:00 UTC, whereas at night, URB
plumes are transported within the shallow PBL (below about 300 m a.m.s.l.)
and the BB layer is mostly transported between the PBL top and
2 km a.m.s.l. The mixture of both URB and BB accumulated over coastal areas
is transported northward in the surface layer from 16:00 UTC onward and
reaches Savè at 21:00 UTC.
Over SWA, both wind and URB emissions have a diurnal cycle. The strength of
numerical tracer experiments is to enable the dichotomy between the
variability linked to the meteorology and the emissions by imposing a
constant emission (i.e., we do not account for any diurnal cycle). In this
article, only the observed variability linked to the meteorology is analyzed;
we demonstrated that there are clear periods of the day when Savè is
impacted by pollution plumes from different cities. In future research,
integrated analyses should be conducted to characterize both the URB plumes
and the BB layer in terms of composition, gaseous and particulate phase,
oxidation of the organic components, and spatiotemporal variability. The
DACCIWA campaign provides unique and valuable observations that will allow
the investigation of the perspectives in this article based on tracer
experiments.