WRF 1960-2014 Winter Season Simulations of Particulate Matter in the Sahel: Implications for Air Quality and Respiratory Health.

We use the Weather Research and Forecast model using the Goddard Global Ozone Chemistry Aerosol Radiation and Transport (GOCART) dust module (WRF-CHEM) to simulate the particulate matter (PM) variations in the Sahel during the winter seasons (January-March) of 1960-2014. Two simulations are undertaken where the direct aerosol feedback is turned off, and only transport is considered and where the direct aerosol feedback is turned on. We find that simulated Sahelian PM10 and PM2.5 concentrations were lower in the 1960s and after 2003 and higher during the period between 1988 and 2002. Higher Sahelian PM10 concentrations are due to stronger winds between the surface and 925 hPa over the Sahara, which transport dust into the Sahel. Negative PM10 concentration anomalies are found over the Bodele Depression and associated with weaker 925 wind anomalies after 1997 through 2014. Further west, positive PM10 concentration anomalies are found across the Adrar Plateau in the Sahara and responsible for dust transport to the Western Sahel. The North Atlantic Oscillation (NAO) is positively correlated to Sahelian dust concentrations especially during the periods of 1960-1970 and 1988-2002. The temporal/spatial patterns of PM10 concentrations have significant respiratory health implications for inhabitants of the Sahel. ©2018. The Authors.

Introduction

The Sahara Desert is the most significant dust emission source in the world with long‐range transport to the Caribbean, the southeastern United States, Europe, and into West Africa (Engelstaedter et al., 2006; Weinzierl et al., 2017). Higher amounts of dust can be transported into West Africa and the Sahel (10–18°N, 17.5°W–15°E) in particular with hazardous dust concentrations located at the surface during the winter and spring seasons (Diokhane et al., 2016; Marticorena et al., 2010). Land‐use change can also contribute to dust emissions, potentially driven by increasing population across West Africa (340 million in 2015), which can create new or activate existing Sahelian dust sources (Mbourou et al., 1997). Dust transport is a concern because Sahelian countries are all low‐income countries, with hazardous dust concentrations creating a financial burden on the public health sector. Also, communicable diseases such as meningitis have occurred primarily in the Sahel (Molesworth et al., 2003) where a meteorological connection has been observed (Sultan et al., 2005).

Mineral dust acts as a negative radiative forcing on the climate system, offsetting positive infrared forcing linked to accumulating anthropogenic greenhouse gases (Stocker et al., 2013). Consequently, it can cool sea surface temperatures on a regional basis over parts of the Atlantic Ocean (Lau & Kim, 2007) and drive tropical variability (Evan et al., 2011). During Northern Hemisphere summer, dust is found within the Saharan air layer at altitudes of 1.5–5 km (Carlson & Prospero, 1972). The Saharan air layer can inhibit tropical cyclones (Evan et al., 2006) although there is some evidence that dust can lead to convective invigoration (Jenkins et al., 2008). Evidence also supports the possibility the Saharan dust may have played a role in the late 20th century summer season below average rainfall in the Sahel (Konare et al., 2008; Prospero & Lamb, 2003).

While there has been an emphasis on long‐term precipitation trends in West Africa, much less attention has been given to long‐term trends in dry season (December–March) dust emissions. Currently, long‐term surface measurements of dust concentrations are sparse with limited surface observations of particulate matter (PM) 10 in five West African countries in 2012 and only one long‐term station in the Sahel at Dakar Senegal. There are, however, long‐term records of dust concentrations at Barbados, and long‐term records of dust variability have been constructed in Cape Verde (CV) using proxy data and satellite‐based observations (Evan & Mukhopadhyay, 2010; Prospero et al., 2014).

In West Africa, surface visibilities also provide evidence of dust variability with higher dust loading associated with lower visibility observations during the 1980s (Mbourou et al., 1997; Silue et al., 2013). Recently, Evan et al. (2016) show that there is a multidecadal trend in dust emissions associated with 10‐m wind speeds and lower dust amounts during the 1950s and after 2000, but higher dust amounts 1980s. While satellite measurements of dust loading using scattering/absorption principles extend back to 1978, they do not give a direct measure of how much dust is at the surface. For example, after the months of May–July, the highest aerosol optical depth is often found when the dust is elevated above the monsoon layer (Drame et al., 2011). In addition, clouds and nighttime conditions limit satellite measurements of dust loading.

During the Northern Hemisphere winter months, high dust concentrations (PM10 and PM2.5) are found near the surface creating a health threat to the public in West Africa because of observed hazardous levels (De Longueville et al., 2013; Diokhane et al., 2016; Marticorena et al., 2010). Pollution is a global risk factor impacting an estimated 9 million premature deaths worldwide with ambient air pollution contributing (Landrigan et al., 2018). A number of studies have linked winter season dust conditions to meningitis, which is a dry season disease (January–April). Hyper‐endemic meningitis cases and occasionally epidemics occur during this period (Agier et al., 2013; Martigny & Chiapello, 2013; Mueller & Gessner, 2010). For example, Diokhane et al. (2016) using observed and simulated and PM10 concentrations over Senegal find higher suspected meningitis cases during 2012 relative to 2013, which had higher observed and simulated PM10 concentrations during January through March. High dust concentrations can also drive lower respiratory infection and nontransmissible respiratory/cardiovascular (asthma, chronic obstructive pulmonary disease, and bronchitis) and most likely other serious diseases in West Africa (Landrigan et al., 2018). Roy (2016) estimates approximately 90,137 premature deaths from ambient PM pollution in West Africa.

The objective of this work is to examine long‐term January through March trends (1960–2014) in dust concentrations with a focus on the Sahel region of West Africa using a regional model with dust emissions. The advantage of this approach is the consistent framework between PM10 concentrations and other meteorological variables over the period. There are other emission sources of PM that can impact human health that we do not consider: biomass burning, automobile traffic, industry, and indoor cooking.

Model Description and Experiment

We design a long‐term simulation beginning 1 December 1959 through 31 December 2014, which only examines the transport of dust into the Sahel, which we have defined between 10–18°N, 17.5°W–15° E. We use the Weather Research and Forecasting (WRF) model coupled with chemistry (Grell et al., 2005) using the Goddard Global Ozone Chemistry Aerosol Radiation and Transport (GOCART) model (Ginoux et al., 2001). We have undertaken other summer and winter season simulations involving Saharan dust with the WRF‐CHEM model (Diokhane et al., 2016; Drame et al., 2011; Jenkins & Diokhane, 2017). The WRF‐CHEM simulations use a horizontal grid spacing of 50 km, 31 vertical levels, with 90 east‐west points and 72 north‐south grid points covering all of West Africa and the Eastern Atlantic. Greenhouse gas concentrations are fixed in the simulations. The aerosol radiative feedback is turned off, and we denote the simulation as NOFEED.

In a second simulation, we turn on the direct aerosol feedback, which will reduce the amount of solar radiation reaching the surface by scattering and denote this simulation as FEED. The FEED simulation is not run continuously but rather begin on 1 January and end on 31 March from 1960 to 2014. The WRF simulations are forced at the lateral boundaries every 6 hr using the National Centers for Environmental Prediction (NCEP) Reanalysis (Kalnay & Coauthors, 1996), which has a horizontal resolution of 2.5° × 2.5°. GOCART is widely used in global and regional models and has a source function for potential dust sources across North Africa, and dust emissions occur when a threshold in 10‐m winds is exceeded (Ginoux et al., 2001). GOCART aerosol scheme simulates PM2.5 and PM10 dust concentrations and uses five sectional bins, with effective radius centered on 0.5, 1.4, 2.4, 4.5, and 8 μm, respectively.

We also use the station‐based North Atlantic Oscillation (NAO) index, which uses the difference of normalized sea level pressure between Lisbon Portugal and Stykkisholmur/Lisbon Iceland (Hurrell, 1995).

The focus of the analysis is during the period of 1 January through 31 March (JFM) when the highest surface PM10 and PM2.5 dust concentrations are observed in Senegal and most likely the rest of the Sahel (Diokhane et al., 2016). As shown in Jenkins and Diokhane (2017), the WRF‐CHEM model compares well to observations in Senegal although it has a negative bias for PM10 concentrations. We examine the Sahara region, the west, central, and eastern sectors of the Sahel. We also examine desert source regions: the Bodele depression (15–20°N, 15–20°E) and the plateau region to the southeast of the Hoggar mountains in Algeria, which we denote as the Adrar Plateau (15–20°N, 3°W–2°E). Finally, we undertake a regional analysis for population centers across the Sahel and CV Islands. This includes CV (14–17°N, 21–24°W); Dakar, Senegal (13–16°N, 17.5–14.5°W); Bamako, Mali (11–14°N, 9.5–6.5°W); Ouagadougou, Burkina Faso (11–14°N, 3°W–0°); Niamey, Niger (12–15°N, 0.5–3.5°E); and Kano, Nigeria (10.5–13.5°N, 7–10°E). Figure S1 shows the regions used for analysis.

Results

Figure 1a show the mean simulated JFM 1961–1990 PM10 concentrations and the associated wind streamlines for the FEED simulation across North Africa. The highest dust concentrations (>255 μg/m3), which are associated with northeasterly winds, are found over Eastern Niger, Central Chad, Northeast Mali, and Southwestern Mauritania and Northern Senegal. Figure 1b shows that the most substantial JFM variations for 1961–1990 mean in PM10 concentrations are found over Eastern Niger, and Central Chad followed by smaller variations over Western Sahara, Northern Mali, Southern Mauritania, and Northern Senegal. Figures 2a–2d show the time‐averaged JFM monthly standardized anomalies for FEED and NOFEED simulations based on the mean period of 1961–1990 for PM10 and PM2.5, and 10‐m and 925 hPa wind speed magnitude differences from 1961 to 1990 mean values across the Sahel (10–18°N, 17.5°W–15°E). Negative PM10 and PM2.5 anomalies are found in both simulations early and late during the simulated period, specifically in the 1960s, the late 1970s through mid‐1980s, and after 2002 (Figures 2a and 2b), while positive PM anomalies occur during the mid‐1970s, and especially between 1988 and 2002. During the latter period, PM10 and PM2.5 are 2–3 standard deviations above the mean. Corresponding to periods of higher and lower PM concentrations are stronger and weaker winds at 10 m and at 925 hPa in the FEED simulation over the Sahel region. In contrast, weaker 10 meter and 925 hPa winds occur in the NOFEED simulation during the period of high dust concentrations of 1988–2002, but similar to the FEED simulation, a trend toward weaker wind speeds occur with lower PM concentrations after 2002.

Figure 1

Simulated FEED JFM: (a) mean 1961–1990 PM10 concentrations; (b) standard deviation of PM10 concentrations. Units are in μg/m3.

Figure 2

The 1960–2014 JFM Sahelian of (a) standardize anomalies of PM10 concentrations, (b) standardize anomalies of PM2.5 concentrations, (c) 10‐m wind magnitude differences from 1961 to 1990 means, and (d) 925‐hPa wind magnitude from 1961 to 1990 means for NOFEED and FEED simulations.

Figures 3a and 3b show PM10 and PM2.5 standardized anomalies across the Sahara (16–23°N, 16°W‐19°E), which serves as the source region for the Sahel. Overall, the anomalous patterns match those of the Sahel, with lower PM concentrations between 1960 and 1970, parts of the 1980s and after 2002, while higher PM concentrations are found between 1998 and 2002. While there are some year‐to‐year differences between the NOFEED and FEED simulations, the anomaly patterns are very similar. Figures 3c and 3d show the 1960–2014 JFM 10 m and 925 hPa wind speed differences subtracted from 1961 to 1990 mean values for the Sahara. The Sahara region shows trends of interannual and interdecadal variability, with reduced winds at the beginning and end of the time series. Higher wind speed magnitudes are shown between 1988 and 2002. The wind speed differences show good agreement between the FEED and NOFEED simulations.

Figure 3

Same as Figure 2. but for the Sahara region (17°W–19°E, 16–23°N).

In the NOFEED simulation, the lack of solar attenuation allows for more solar radiation to reach the surface and can lead to additional changes in regional circulation. Figure 4 shows JFM differences in PM10 concentrations, 2‐m temperatures, specific humidity, and 925 wind speeds between the NOFEED and FEED simulation. Relative to the FEED simulation higher PM10 concentrations are found over much of central and western parts of the Sahel and Sahara with slightly lower PM10 concentrations over Chad and Nigeria (Figure 4a). As expected, 2‐m air temperatures are up to 4° warmer over the land areas, with the warmest values found over the Central Sahel and Northeastern Nigeria in the NOFEED simulation (Figure 4b). The NOFEED simulation also has higher 2‐m absolute humidity especially over the Guinea region, where there is up to a 4 g/kg difference from Nigeria westward to the coast in the (Figure 4c). A stronger southwesterly monsoon circulation transports moisture from the Gulf of Guinea into the Guinea region (Figure 4d). The warming in Figure 4a from the NOFEED simulation creates a thermal low type circulation with a counterclockwise flow relative to the FEED simulation (Figure 4d). With these differences identified, the results shown below will only use the FEED simulation.

Figure 4

JFM differences between the NOFEED and FEED simulations. (a) PM10 concentrations, (b) 2‐m air temperatures, (c) 2‐m specific humidity, and (d) 925‐hPa wind magnitude and direction.

Regional Analysis of the FEED Simulation

Figures 5a–5c show the variations in JFM PM10 concentration anomalies across the Western Sahel (10–18°N, 17.5–7.5°W), Central Sahel (10–18°N, 7.5°W–2.5°E), and Eastern Sahel (10–18°N, 2.5–15°E). The Western Sahel shows the overall pattern of improved air quality prior (negative PM10 anomalies) to 1970, and after 2002, however, there are additional years of poor air quality during the 1980s (Figure 5a). A similar pattern is also found for the Central Sahel with additional years of reduced air quality (positive PM10 anomalies) after 2002 and 6 out of 10 years during the 1980s (Figure 5b). The Eastern Sahel, on the other hand, shows improved air quality during the 1960s, most of the 1980s, and after 2000, with reduced air quality for several years during the 1970s and then between 1990 and 2000 (Figure 5c).

Figure 5

The 1960–2014 JFM standardize anomalies of PM10 concentrations for (a) West Sahel (10–18°N, 17.5–7.5°W), (b) Central Sahel (10–18°N, 7.5°W‐2.5°E), and (c) East Sahel (10–18°N, 2.5° E‐15°E.

To gain insight on the Sahelian regional differences occur, we examine two significant dust source regions in the Sahara Desert near the Bodele Depression and the Adrar Plateau, which is located just downstream of the Hoggar mountains. Dust is transported from the Bodele Depression into the Eastern parts of the Sahel, while dust is transported from the Adrar Plateau into the Central and Western Sahel. Figure 6. shows the PM10 anomalies and 925 hPa wind magnitude differences relative to 1961–1990 for both regions. Positive PM10 anomalies across the Bodele depression are found in the early 1960s, the mid‐1970s, and then from approximately 1986–1996 in the FEED simulation. Negative PM10 anomalies are found from the mid‐1960s through early 1970s and then from 1997 to 2013 (Figure 6a). The Adrar Plateau region shows quite a different pattern of PM10 concentration anomalies, with the most extended period of negative anomalies between 1960 and 1972. Positive PM10 anomalies are found during each decade after 1970 and continue through 2012.

Figure 6

The 1960–2014 JFM: (a) Bodele Depression standardize anomalies of PM10 concentrations, (b) Adrar Desert standardize anomalies of PM2.5 concentrations, (c) 10‐m wind magnitude differences from 1961 to 1990 means, and (d) 925‐hPa wind magnitude from 1961 to 1990 means for NOFEED and FEED simulations.

Figure 6c shows that 925 hPa wind speed differences relative to 1961–1990 and shows a general pattern of weaker winds during the period of negative PM10 anomalies and positive wind speed differences during periods of positive PM10 anomalies for the Bodele Depression. The weaker wind speeds are associated with a reduction in 925 hPa zonal winds, even though there is a slight strengthening of meridional winds (not shown) between 1997 and 2014. A positive correlation is found between PM10 and 925 hPa wind speeds (0.90) for the Bodele Depression. Figure 6d shows the wind speed differences relative to 1961–1990 mean for the Adrar Plateau, with weaker winds being simulated after 2002 but stronger winds between 1960 and 2002. However, we find a smaller positive correlation of 0.58 between PM10 and 925 hPa wind magnitude anomalies for the Adrar Plateau.

Next, we examine smaller regions near metropolitan centers across the Sahel with the analysis occurring over a 3° × 3° box located near an urban center. Urbanization has been a significant contributor to rapid population growth in West Africa and the Sahel (Ollson et al. 2005). From east to west we examine the regions near Kano Nigeria Niamey, Niger, Ouagadougou, Burkina Faso); Bamako, Mali; Dakar, Senegal; and the Cabo Verde Islands (Figure S1). CV has a population of approximately 500,000, while all of the other cities have a population of at least 1 million; Dakar, Senegal; Bamako, Mali; and Kano, Nigeria, have populations exceeding 2 million inhabitants. The JFM 1960–2014 PM10 anomalies for these locations are shown in Figure 7. where we have put the metropolitan regions into Western, Central, and Eastern Sahelian sectors. Figure 6a shows PM10 anomalies for CV and Dakar with negative anomalies during the 1960s and after 2002. Positive PM10 concentration anomalies are found between in varying periods between 1970 and 2002. There is a positive correlation of 0.71 between Dakar and CV PM10 anomalies during the period. The Central Sahelian sectors linked to Bamako, Mali, and Ouagadougou, Burkina Faso, show a pattern similar to CV and Dakar with positive PM10 anomalies found primarily between 1972–1978 and 1988–2002. A positive correlation of 0.93 between Bamako and Ouagadougou PM10 anomalies occurs from 1960 to 2014. The eastern sector regions linked to Niamey and Kano show the period of positive PM10 anomalies in the 1970s and 1988–2002 (Figure 7c). Negative PM10 anomalies are found during each decade outside of the 1990s for Niamey and Kano. A positive correlation of 0.89 is found for simulated PM10 anomalies between Kano and Niamey for the simulated period.

Figure 7

The 1960–2016 JFM standardized PM anomalies for (a) CV and Dakar, Senegal; (b) Bamako, Mali, and Ouagadougou, Burkina Faso; and (c) Niamey, Niger, and Kano, Nigeria.

Figure 8a shows the frequency of daily PM10 concentrations exceeding 255 μg/m3 and defined as the threshold for unhealthy air quality for the U.S. Environmental Protection Agency (De Longueville et al., 2013; Fitz‐Simons, 1999) between 1960 and 2014 for locations near Dakar, Senegal; Bamako, Mali; and Kano, Nigeria. Dakar has the highest number of unhealthy air quality events on an annual basis relative to Kano and Bamako. The mean number of days during the 1960–2014 period with PM10 exceeding 255 μg/m3 is 24 (Dakar), 2 (Bamako), and 14 (Kano). Between 1988 and 2002, the number of unhealthy air quality days increased by a factor of 2 for Dakar and 2–3 in Kano. While the frequency of unhealthy air quality days for PM10 is higher at both Dakar and Kano during the 1988–2002 period, there is a significant reduction in the rate of unhealthy days at Kano after 2002. Figure 8b shows the frequency of daily PM10 concentrations exceeding 355 μg/m3, which are defined as the threshold for very unhealthy air quality (Fitz‐Simons, 1999). The mean number of days during the 1960–2014 period with PM10 exceeding 355 μg/m3 is 7 (Dakar), 0 (Bamako), and 4 (Kano). However, during 1988–2002, the number of very unhealthy PM10 days increased by a factor of 2–3 for Dakar and Kano. After 2002, there is a reduction in the numbers of days when very unhealthy air quality conditions exist at Kano, while in Dakar the number of very unhealthy days exceeds the mean values by a factor of 1.5 to 2 as recently as 2012.

Figure 8

The 1960–2014 JFM frequency of(a) unhealthy air quality for Dakar, Senegal; Bamako, Mali; and Kano Nigeria, and (b) very unhealthy air quality for Dakar, Senegal; Bamako, Mali; and Kano, Nigeria.

Next, we compare daily values for at Dakar, Bamako, and Kano for the individual years of 1964, 1973, 1983, 1990, and 2012. First, we compare the simulated values (FEED and NOFEED) for 2012 to observed values in Dakar, which is similar to what has been done in Diokhane et al. (2016) at a finer grid spacing. Figure 9a shows that the significant dust events are present in the simulations, but PM10 concentrations are underestimated relative to observed values in Dakar. Mean JFM values during 2012 are 357 μg/m3 (observed), 265 μg/m3 (FEED), and 289 μg/m3 (NOFEED) with the mean values exceeding unhealthy air quality conditions for 2012. Figure 9b shows the JFM daily values for Dakar relative to earlier years with 1990 producing the highest mean values, with similar mean concentrations between 2012 and 1983. Relative to 2012, PM10 concentrations are much lower in 1964 with a mean value of 128 μg/m3.

Figure 9

JFM daily PM10 concentrations for (a) observed and simulated 2012 values at Dakar, Senegal; (b) 1964, 1973, 1983, 1990, and 2012 for Dakar, Senegal; (c) 1964, 1973, 1983, 1990, and 2012 for Bamako, Mali; and (d) 1964, 1973, 1983, 1990, and 2012 for Kano, Nigeria. Units are in μg/m3.

Figure 9c shows the daily PM10 concentrations near Bamako, Mali, for 1964, 1973, 1983, 1990, and 2012 with the largest mean values occurring during 1983; mean values of 159 μg/m3 are found in 1983 relative to 111 μg/m3 during 2012. Similar to Dakar, the lowest simulated mean PM10 concentrations of 84 μg/m3 occur in 1964. Figure 9d shows the daily PM10 concentrations for Kano for five years. The mean simulated 2012 PM10 concentrations are 165 μg/m3, which are slightly higher than mean values of 153 μg/m3 in 1964 because of the negative PM10 anomaly trends shown in Figure 7a. Table 1 shows the frequency of unhealthy (255–354 mg/m3) concentrations, at the Dakar, Senegal; Bamako, Mali; and Kano, Nigeria, for 1964, 1973, 1983, 1990, and 2012 for the three locations. For Kano and Dakar, 1990 has the highest number of unhealthy air quality days, followed by 1983. Bamako shows that in general there are many fewer days with unhealthy air quality, with 13 days found in 1983. Table 1 also shows that 39 unhealthy air quality days occurred in Dakar relative to 10 days in Kano in 2012.

Table 1

WRF Simulated Frequency of Unhealthy Air Quality Days at Dakar, Kano, and Bamako

	1964	1973	1983	1990	2012
Dakar,	7	21	41	52	39
Bamako	0	3	13	3	3
Kano	11	19	35	42	10

Linkages to the NAO

The phase of the NAO (Hurrell, 1995) can increase or decrease dust emissions over the Sahara, with more dust emission during the positive phase (Chiapello & Moulin, 2002; Ginoux et al., 2004; Moulin et al., 1997). Figure 10 shows the NAO index based on station data for JFM 1960–2014 with the positive phase of the NAO index being associated with potentially stronger winds over North Africa in association with a stronger Azores high‐pressure system. Table 2 shows a positive correlation of 0.61 for simulated PM10 anomalies from the Sahel with the Western Sahel having the highest correlation (0.66) and the Eastern Sahel having the lowest (0.49) value for the 1960–2014 period. The highest correlations are found for the period of 1960–1970 when the NAO was in a negative phase and there were negative PM10 anomalies and during 1988–2002 when the NAO was in a positive phase and positive PM10 anomalies are found. In more recent times, correlations are less than 0.53 for the Sahel and its subregions, suggesting that the relationship has weakened and other regional factors may be influencing PM10 anomalies.

Figure 10

JFM 1960–2014 station‐based NAO index.

Table 2

Correlations Between the Simulated PM10 for Various Locations and the NAO Index

Location	1960–2014	1960–1970	1971–1987	1988–2002	2003–2014
Sahel	0.61	0.86	0.45	0.73	0.49
Western Sahel	0.66	0.89	0.50	0.67	0.50
Central Sahel	0.62	0.91	0.45	0.56	0.39
Eastern Sahel	0.49	0.74	0.15	0.70	0.53
Sahara	0.63	0.87	0.41	0.77	0.49
Bodele	0.23	0.54	−0.03	0.58	0.44
Adrar	0.63	0.90	0.41	0.58	0.44

Over the Sahara Desert, simulated PM10 anomalies are positively correlated to the NAO with a value of 0.65 for the entire period. However, similar to the Sahel, the highest correlations between the NAO index and simulated PM10 anomalies are found between 1960–1970 and 1988–2002 with correlations less than 0.5 for the other two time periods in Table 2. For the two desert sources, the Bodele Depression PM10 anomalies have a much lower correlation to the NAO index than the region near Adrar Plateau for 1960–2014. While the Bodele Depression region has higher correlations during the periods of 1960–1970 and 1988–2002, these are much lower than the Sahara as a whole and the Adrar Plateau. Consequently, the Western and Central Sahara deserts may have a greater connection to subtropical and midlatitude circulation associated with the NAO.

The higher correlation between desert sources and the NAO index manifests itself when examining the correlation between the NAO to simulated PM10 anomalies near large urban centers (Table 3). In general, we find the highest correlations for urban centers that are in the West/Central Sahel as compared to the Eastern Sahel. Similar to the regional correlations, the period of 1960–70 and 1988–2002 has the highest PM10 anomalies/NAO index correlation, when negative and positive anomalies are found respectively. The simulated PM10 anomalies for Kano, Nigeria, have a lower correlation to the NAO index than other urban locations; however, the period after 2002 shows higher correlations and is similar to the other five sites.

Table 3

Correlations Between the Simulated PM10 for Large METRO Locations and the NAO Index

Location	1960–2014	1960–1970	1971–1987	1988–2002	2003–2014
CV	0.61	0.86	0.47	0.71	0.39
Dakar, Senegal	0.70	0.89	0.55	0.66	0.55
Bamako, Mali	0.57	0.86	0.47	0.62	0.35
Ouagadougou, Burkina Faso	0.63	0.89	0.47	0.62	0.45
Niamey, Niger	0.61	0.90	0.30	0.60	0.46
Kano, Nigeria	0.47	0.68	0.08	0.64	0.51

Conclusions

In this work, we examine the trends and variability from the transport of PM into the Sahelian region between 1960 and 2014 using the WRF GOCART module. Some of the primary findings include the following: For the simulated time period we find that higher dust concentrations (1988–2002) lagged the higher dust loading of the 1970s and 1980s reported by Evans et al. (2016). However, this could be due to the fact that we focus on JFM instead of annual timescales and only consider the Sahelian land areas although analysis over the CV Islands shows positive PM10 anomalies between 1998 and 2002.

There are both interannual variability and decadal trends in PM, with a trend toward improved air quality through reduced PM10 and PM2.5 concentrations in the Sahel after 2002.

Periods of high PM10 and PM2.5 are found in the Sahel during the period of 1988–2002 and driven by stronger winds and the higher PM10 transport from the Sahara region.

The simulation without direct radiative feedback (NOFEED) has a warm bias, higher PM10 concentrations, and a positive moisture bias across the Guinea region. An anomalous thermal low circulation is found relative to the simulation with direct radiative feedback (FEED).

The NAO index is positively correlated to simulated PM10 anomalies, with the highest correlations during the periods of 1960–1970 and 1988–2002; areas in the Western Sahel and Sahara also show higher correlations between PM10 anomalies and the NAO index than those in eastern sectors.

The results suggest that poor air quality, which is responsible for one in nine deaths globally, indirectly improved between 2003 and 2014 in the Sahel (World Health Organization, 2016). But, trends in respiratory health on a national/regional basis need to be established to determine if indeed this is the case. If indeed air quality is improving, it represented a reduced burden on public health infrastructures in Sahelian low‐income countries. However, significant increases in West African population over the last 50 years (75 million to 340 million) and anticipated changes over the next 30 years (700 million in West Africa) will increase the total number of persons with various types of respiratory disease, even if the rate per 100,000 remains constant. Improved health and environmental monitoring will help to understand the relationships between respiratory disease and the ambient PM concentrations.

Development and the increase in mega‐cities across West Africa have increased the emissions of anthropogenic PM from transportation and industrial sources (Landrigan et al., 2018). An additional source of PM, biomass burning, occurs during the dry season in West Africa creating poor air quality. These sources were not considered in this work and would contribute to the total PM but general fall under the category of fine mode PM (PM2.5). Future research related to air quality associated with PM in West Africa will have to consider these additional sources.

Two other factors may be relevant to future PM concentrations over West Africa: (a) land use change and (b) anthropogenic climate change and subsequent changes to the NAO. The growth of West African population will likely lead to land use change with the reduction of natural vegetation and may increase dust emissions in semiarid regions of the Sahel reducing air quality locally and regionally. Satellite‐based observations of vegetation can be used to monitor trends, while satellite‐based aerosol optical depth can be used for identifying new sources of dust emissions over time. Hanna et al. (2015) suggest that Representative Concentration Pathways (RCP) 8.5 forcing could lead to a more frequent positive NAO phase during the late 21st century and could have significant implications for respiratory health across West Africa where the overall population and urbanization will be significantly larger.

Conflict of Interest

The authors declare no conflicts of interest relevant to this study.

Supporting Information S1

Click here for additional data file.

Supporting Information S2

Click here for additional data file.