1. 

Introduction

The impact of climate change in Africa is projected to be adverse for ecosystems and water resources (GIEC, 2014) and could undermine socio-economic and health stability. Also, of the ten most threatened countries by climate change in the world, seven are African (i.e. Central African Republic, Eritrea, Ethiopia, South Sudan, Chad, Nigeria and Sierra Leone). Furthermore, the sub-Saharan West African region is subject to a fast population increase (>3% year−1) (UNDP, 2015; Ago et al., 2016) and land-use change with cropland expansion over natural ecosystems. In addition, the FAO report of 2011 (FAO, 2011) highlighted reducing areas of wetlands and forests, with a 1.1% year−1 average forest cover loss during the 2000-2010 period, with important losses of –5.1% year−1 for Togo, –3.7% year−1 for Nigeria and –2.1% year−1 for Ghana. In the same vein, the regional temperature increased by 0.2 to 0.8 °C in West Africa since the late 1970s (CEDEAO-CSAO/OCDE, 2008). Also, West Africa has experienced decrease by 40% to 60% in the annual average flow of the major rivers (Niasse, 2004). This region experienced extreme heat events and heat waves such as those observed in Niger in 2010 (Ringard et al., 2014; Karimou et al., 2015) causing deaths among vulnerable populations. Africa has been exposed to more than 136 drought episodes in the 1995–2015 period of which 77 (i.e. 56.61%) for the East African region alone (https://www.afdb.org/en/cop23/cop23-overview/). The region is also a hotspot of fire activity, accounting for more than 50% of global fire emissions (van der Werf et al., 2010; Scholes et al., 2011) although a reduction of burned areas (Andela and van der Werf, 2014) and emissions are observed from satellite data since the year 2000, related to rainfall increase and cropland area expansion.

Although it is widely recognized that African ecosystems play a major role in the carbon cycle, uncertainties remain very important in both the mean magnitude of carbon fluxes and their inter-annual variability (Bombelli et al., 2009). Few long-term carbon cycle observations are available in Africa, particularly in West Africa, with five eddy covariance towers (Merbold et al., 2009) and up to this study, no atmospheric station. As a result of atmospheric data scarcity, the results from atmospheric inversions used to infer the distribution of sources and sinks at the surface from atmospheric concentration observations and atmospheric transport models (Ciais et al., 2010) have large uncertainties across the region. Chevallier et al. (2014) attributed natural (fossil fuel removed) CO2 emissions to North Africa, ranging from 0.5 to more than 1 GtC yr−1 depending if they use surface in-situ or satellite data (GOSAT). More recently the major role of northern tropical Africa in the CO2 budget has been confirmed with an emission estimated to about 1.5 GtC yr−1 (Palmer et al., 2019) based on the inversion of CO2 total column retrieved from GOSAT and OCO-2 satellites. The scarcity of ground-based measurements constitutes a hindrance to provide robust estimates of CO2 sources and sinks in this region (Palmer et al., 2019). Moreover, the estimation based on satellite inversions are sensitive to the aerosol bias in the retrieval of column CO2 and to transport uncertainty.

In the present study, we present the longest in situ time series of CO2 and CH4 concentrations monitored in West Africa, from the Lamto station (LTO), located in the wet tropical region of West Africa in Côte d’Ivoire. The observatory of Lamto was setup in 2008 as part of the CARBOAFRICA European project (www.carboafrica.eu). The main use of this type of data is to feed global to regional inversion studies. This paper presents a synthesis of the seasonal and diurnal cycles of greenhouse gases recorded over the 10 years (2008–2018). Besides, we present a record of carbon monoxide from the station, CO being a key pollutant and a surrogate tracer that can detect combustion processes (Turnbull et al., 2006; Wang et al., 2010; Duren and Miller, 2012). The concentration of CO2, CH4, and CO continuously measured from the Lamto station are compared with flask air samples from three stations located in the Sahara Desert (Assekrem, ASK), in the Canarias Is. (Izana, IZO) and the tropical Atlantic Ocean (Ascension Island, ASC) as part of the National Oceanic and Atmospheric Administration (NOAA-ESRL) global air sampling network and to the continuous measurements of the Cape Verde Atmospheric Observatory (CVO) (Kozlova and Manning, 2009) to highlight significant trends of the greenhouse gases concentrations variations. This work is structured as follows; Section 2 describes the study area, the material, and the method used. Section 3 shows the results of the CO2, CH4 and CO atmospheric records. Conclusions and perspectives are given the last section.

2. 

Study area, material, and methods

2.1. 

Study area

The Lamto station (6°31 N and 5°02 W) is located within a 2700 ha nature reserve, about 160 km north of Abidjan, at an altitude of 155 m above the sea level (Fig. 1) in a flat region without striking relief. The surrounding vegetation is a wet tropical savanna, in the transition area between forest and savanna (Diawara et al., 2014). The Taabo hydroelectric dam, built on the Bandama River, is located 6 km east of the station (Kouassi et al., 2008). Long times series measurements of climate data (i.e. temperature, rainfall, relative humidity, wind speeds and direction, radiative heat, etc.) have been carried out at the Geophysical station since 1962 (Abbadie et al., 2006; Diawara et al., 2014). In addition, the mean annual rainfall and temperature records are at 1194 mm and 27.8 °C, respectively (Diawara et al., 2014; N’dri et al., 2018). The climate of this area is controlled by the West African Monsoon (WAM) with the main dry season (GSS) from December to February, the main rainy season (GSP) from March to July, a short dry season (PSS) in August and a short rainy season from September to November (Fig. 2). On the other hand, the anthropogenic activities around the station are mainly related to agriculture, fishing and livestock. Also, the A3 highway connecting Abidjan to Yamoussoukro, which records a very dense road-traffic during the day, is located 14 km west of the station. Seasonal bush fires (from December to February) affect more than 80% of the ground biomass (Nacro, 2003).

Furthermore, Fig. 3 shows the average influence functions from surface fluxes given short and medium range transport of air masses to LTO during the main dry (Fig. 3a) and wet (Fig. 3b) seasons. We observe that during the main dry season (GSS), Lamto region is permanently influenced by air masses originating from the northern and north-eastern sectors (Tiemoko et al., 2020b). The presence of these air masses from the northern and north-eastern sectors is related to the harmattan flow regime in the region. However, in the main wet season (i.e. from May to July), the air masses arriving at Lamto are preferentially observed from a tropical Atlantic Ocean and Guinean Gulf coast origin in the south and south-west sectors which are characteristic of the west African monsoon flow.

2.2. 

Atmospheric measurements

2.2.1. 

Continuous CO2, CH4, CO measurements at Lamto

The CO2 and CH4 mole fractions are continuously measured at Lamto station (LTO) since August 2008, and CO measurements started in March 2014. The measurements were obtained using two versions of cavity ring down spectroscopy instruments (CRDS_Picarro_Envirosens and CRDS_Picarro_2401). The Envirosens model (SN: CFADS-02) was used from August 2008 to late 2013 for CO2/CH4/H2O measurements while the G2401 model (SN: 1703-CFKADS-2124) is used since April 2014 for CO2/CH4/H2O/CO measurements. These instruments based on cavity ring-down spectroscopy have linear and stable responses, and are recognized for their high accuracy in CO2, CH4 and CO molar fractions measurements (Crosson, 2008; Chen et al., 2010; Yver Kwok et al., 2015). Lamto measurements data are calibrated to the World Meteorological Organization (WMO) reference scales using four calibration gases spanning the atmospheric ranges of concentrations (Table A1). Those tanks are themselves calibrated at LSCE using six calibration tanks from NOAA/ESRL. The air is sampled continuously at the top of a 50 m tower and passes through 2 µm filters to protect the pump and the analyzer from dusts and combustion aerosol (see Fig. 4 and Fig. A1). Since January 2010 a Nafion dryer is used to reduce the amount of water inside the instrument and to minimize the water vapor correction (Rella et al., 2013). In addition to the calibration gases, we measured 2 to 3 times per day a target gas used as a quality control of the measurements. Since 2014 a second target gas was measured at each calibration sequence once or twice a month. This additional target gas (long term target) allows to monitor the measurement reproducibility on longer term periods. This reproducibility over the 10 years period gives a mean bias of –0.02 ± 0.04 ppm for CO2 and –0.01 ± 0.17 ppb for CH4 for target gas values with no significant trend (Table 1), compliant with the recommendations of WMO (Conil et al., 2019), within ±0.1 ppm for CO2 and ±2 ppb for CH4. On the other hand, in the case of CO, we found a mean bias of –6.58 ppb, and a total drift of 2.00 ppb over 4 years, that can be attributed to a drift of the target gas cylinder, or in one of the cylinders used for regular calibration. In order to investigate this problem, new cylinders will be sent to the station to verify the consistency of the calibration scale. Until then, we did not apply any corrections to the CO measurements presented here. We consider that this problem has a low impact on diurnal, synoptic and seasonal variations. However, we did not analyse the long-term trend and the growth rates of CO in view of calibration cylinders drift.

Fig. A1

Schematic diagram of the air sampling analysis line for the measurement of CO2, CH4 and CO concentrations by "PICARRO" at Lamto.

Maintaining high precision continuous measurements of CO2 and CH4 in a remote site like Lamto is a technical and logistical challenge. With the first CRDS analyser we several problems (one per year on average) leading to data gaps larger than one month. Since the installation of the new analyser in April 2014, we had only one three (3) weeks data gap due to a lightning strike at the station. The regular maintenances of the analyser and different parts of the measurement system are performed by the technical staff of the Geophysical Station of Lamto, under the supervision of LSCE engineers who visit the station once a year on average. The major data gaps can be recalled as follows:

  • 04-10-2008 to 22-10-2008: unidentified problem with the analyzer
  • 12-05-2009 to 04-07-2009: unidentified problem with the analyzer
  • 17-01-2010 to 04-03-2010: crash of the computer (the timestamp no longer works), replaced in March 2010.
  • 12-11-2010 to 30-08-2011: breakdown of the analyser due to a problem with the regulation of the warm chamber. Analyzer ESP1000 sent back to the manufacturer in the USA.
  • 24-07-2012 to 04-09-2012: stopping the measurements following numerous interruptions of the Picarro analyzer (since the end of May 2012). A new computer was installed.
  • 01-09-2013 to 07-04-2014: breakdown of the analyser which cannot regulate the temperature of the cavity. New CRDS analyser set up on 07 April 2014. At the same time a weather station is installed in the vicinity of the air inlet.
  • 25-05-2017 to 15-06-2017: lightning of the station. The analyser is safe, but not the computer, time server, and power converter. A maintenance mission is organized in mid-June to replace the damaged parts.

2.2.2. 

Regional data from ASC, ASK, IZO and CVO stations

The closest measurements of CO2 and CH4 are performed on a weekly basis by NOAA/ESRL as part of their cooperative air sampling network (www.esrl.noaa.gov/gmd/ccgg/flask.php) at Assekrem, Algeria (ASK); Ascension Island (ASC) and Izana, Canary Islands (IZO). In situ measurements are also performed at the Cape Verde Atmospheric Observatory (CVO) and retrieved from the WDCGG database (Kozlova and Manning, 2009) (see Table 2). These stations are the closest to LTO in the global network (Fig. 1), at distances of 1900 to 2700 kilometers from LTO. The air flask samples are taken twice a week between 11 h and 16 h (sampling by glass flask) and then analysed at the NOAA/ESRL/GMD laboratory in Boulder, Colorado for the three NOAA stations. The data for CO2, CH4 and CO concentrations are available since 1996 (ASK), 1997 (ASC), 1992 (IZO), and 2012 (CVO), respectively.

2.3. 

Methods

To characterize the long-term trend, seasonal cycle and short-term variations of CO2 and CH4 over the 2008-2018 period, we proceeded by filtering the data measured at LTO, CVO, and at the three nearby NOAA stations (ASC, ASK and IZO). The HYSPLIT (Stein et al., 2015) and FLEXPART (Stohl et al., 2005) models were used to calculate air mass trajectories associated with significant peaks of CH4 and CO concentrations observed in December 2014 during the peak of the fire season.

2.3.1. 

Calculating the trend and seasonal cycles

The curve fitting and data filtering routine (CCGvu) developed by NOAA/CMDL (Thoning et al., 1989) was used to calculate short-term variations, growth rates, and the average seasonal cycle of CO2, CH4 and CO. For this purpose, the time series of mid-afternoon averages X(t) of each gas (i.e. CO2 (t), CH4 (t) and CO (t)) is decomposed into an adjustment function f(t) and into a residual function R(t) (Equation (1)). The function f(t) (Equation (2)) includes a 2-degree polynomial p(t) representing the long term trend and a series of four harmonics that fit the seasonal cycle (Levin et al., 2002; Ramonet et al., 2002; Fang et al., 2017). The residual function R(t) that represents the synoptic variations, obtained by the difference between the hourly mean measurements and the fitted curve f(t) (Lin et al., 2015).

((1))
X (t) = f(t) + R(t)
((2))
f (t)=p(t)+k=14[a2k+1sin(2kπt)+a2k+2cos(2kπt)] 
where, p(t) = a0+ a1t + a2t2, t is the time and the terms ai are coefficients determined by the least square’s method. A combination of high-pass and low-pass filters is applied to R(t) with two respective cut-off frequencies at 80 and 667 days to determine the smoothed time series S(t) (Equation (3)) and long-term trends T(t) (Equation (4)) (Bakwin et al., 1998) . The seasonal cycle Sc(t) (Equation (5)) is finally obtained by the difference between the smoothed time series S(t) and the long-term trends T(t).
((3))
S(t) = f(t) + R80(t)
((4))
T(t) = p(t) + R667(t)
((5))
Sc(t) = S(t)  T(t)

2.3.2. 

Calculating of trend by Liebmann's method

An analysis of long and short-term trends in CO2 and CH4 is provided using a statistical diagnosis based on linear regression by least square fit (Liebmann et al., 2010). Trends are calculated by multiplying the slope of linear trends by the length of the time-serie. The most significant trends are obtained with a 95% confidence level by the Student t-test. This method shows robust and specific performances to objectively determine one or more trend breaks in time-series when they occur. In addition, it has been used in many studies on climate variability in West Africa (Diawara et al., 2014; Ta et al., 2016; Tiemoko et al., 2020a).

2.3.3. 

Computing of species ratios

The computing of concentrations ratios allows diagnosing co-variations between the different gases (i.e. CO2, CH4 and CO). Hourly residual time series (i.e. ΔCO2, ΔCH4 and ΔCO) were calculated by the difference between the raw data and the smoothed time series described above. These residual time series represent synoptic-scale variations (Harris et al., 2000; Ramonet et al., 2002; Grant et al., 2010; Tohjima et al., 2014; Lin et al., 2015). Concentration ratios of residuals (ΔCO2/ΔCH4, ΔCO2/ΔCO and ΔCH4/ΔCO) are finally calculated from the regression slope between residual anomalies of different gases, using the Deming regression method (Guérette et al., 2018; Wu and Yu, 2018).

3. 

Results and discussions

3.1. 

Time series analysis and general statistics

Figure 5a–c shows the time series of the hourly CO2, CH4 and CO atmospheric concentrations. In blue are represented daytime observations from 11 h am to 17 h local time. The choice of separating night and day is based on the variations of the ABL (Atmospheric Boundary Layer) diurnal cycle. The ABL is practically in a stable state during the night (not very turbulent) while it is most often unstable (turbulent) during the day (Fernández-Duque et al., 2017; Mai et al., 2020) resulting in different vertical dilution of local fluxes and influence from more remote fluxes.

The diurnal variations of CO2 and CH4 are characterized by lower concentrations during the day (i.e. 11 h to 17 h) and by higher concentrations during the night (17 h of the day j to 11 h of the day j + 1), mainly due to ABL development, which dilutes CO2 and CH4 exchanged at the surface and decreases their values during the morning (Murayama et al., 2003; Marnas, 2009). The relative diurnal CO2 variations are more pronounced than those of CH4 due to the covariance of photosynthesis and respiration and ABL depth. The existence of a diurnal cycle of CH4 also suggests CH4 sources in the surroundings of the station, especially for the night time footprint. One possibility for this local CH4 emission corresponds to the Taabo dam take it off located 6 km from the station. The CO2 and CH4 time series also show clear upward trends over the studied period, which are discussed in the section about the growth rates. During the observation period, 29% of the CO2 values are less than 395 ppm, 64% between 395 and 420 ppm and 7% are higher than 420 ppm. For CH4, about 8% of the data below 1800 ppb, 74% are between 1800 and 1900 ppb and only 18% are greater than 1900 ppb. This statistic shows that more than 50% of CO2 and CH4 observations are higher than reported at several monitoring stations listed by the WDCGG (World Data Centre for Greenhouse Gases) Data Summary No. 41. (WMO, 2017), indicating the significant impact of anthropogenic activities and natural sources (e.g. biomass fires, wetlands) on the content of these compounds at LTO.

Figure 5c shows the time series of mean hourly CO concentrations from March 2014 to June 2018. Significant seasonal and interannual variabilities of CO concentrations are observed. The CO variations observed during the wet season from April to October are much smaller than the variations in the dry season from December-January-February (i.e. bushfire regime, agricultural slash-and-burn activities, etc.). The CO concentrations are in the range [83.69–5990.79] ppb, thus abnormally high values compared to those provided at clean air stations like Mt. Kenya-Kenya (MKT) or, Assekrem-Algeria (ASK) (WMO, 2017) mainly linked to biomass burning in the LTO region. About 1.5% of the hourly CO molar fractions values are lower than 150 ppb, 82% are between 150 and 350 ppb and 16% are higher than 350 ppb.

3.2. 

Diurnal cycle

The diurnal cycle of CO2 in Fig. 6a shows a maximum concentration at 7 hr (local time) and a minimum between 14 hr and 17 hr during all the seasons. A gradual increase in concentrations starts towards 18 h and peaks in the morning. This signal can be explained by the combination of the variability of CO2 fluxes exchanged between the vegetation, and the ABL (Putaud, 2019). Generally, ABL height increases after sunrise, reaches a maximum during noon, and decreases from sunset (Mahalakshmi et al., 2011; Mai et al., 2020). The accumulation of CO2 at the surface stops abruptly after 7 hr, which result on the one hand from the radiative soil heating effects, which break up the inversion layer established during the night to mix vertically the trapped gases, and on the other hand reversal from CO2 fluxes that become negative because of photosynthesis. CO2 uptake by plants induces a slight decrease in CO2 concentrations of ∼4 ppm between 12 hr and 18 hrs when the atmospheric boundary layer is developed to altitudes between 1 and 2 km (Goudie and Middleton, 2001; Aryee et al., 2020; Mai et al., 2020; Pedruzo-Bagazgoitia et al., 2020). The highest amplitudes of the CO2 diurnal cycle (peak to peak amplitude greater than 40 ppm) are observed during the rainy season GSP and PSP (Table 3).

The diurnal cycle of CH4 in Fig. 6b shows a similar phase compared to CO2, despite different surface fluxes, suggesting a dominant influence of the ABL. However, in contrast to CO2, there is an increased concentration of CH4 during the afternoon of about 8 ppb, after a minimum concentration reached around 12 hr, indicating daytime emissions of CH4 in the footprint of the station, e.g. from the nearby dam. Although the phase of the diurnal cycle’s changes very little from one season to another, their amplitudes show marked seasonal variations (Table 3). The largest diurnal amplitudes for CH4 are observed during the main wet (GSP) and dry (GSS) seasons (Table 3).

The CO diurnal cycle is characterized by a double peak with two maxima, in the morning (7–8 hr) and the evening (18–20 hr), respectively. The causes of these peaks could result from the impact of the traffic occurring not far from the area, but the morning peak is probably related to accumulation near the surface just before ABL mixing decreases CO. The evening peak of CO is greater than the morning peak, possibly because of less convective evening boundary layer with evening denser traffic. The largest diurnal amplitude of CO is observed during the long rainy season (GSP), and the smallest during the short rainy season (PSP).

3.3. 

Growth rate of CO2 and CH4

Long-term trends and interannual variability of CO2 and CH4 have been calculated from the CCGvu software (Thoning et al., 1989). The mean growth rate of CH4 over the 2008-2018 period is of 7 ppb year−1, which is equal to the global mean growth rate obtained from marine boundary layer stations (Dlugokencky and Tans, 2020; https://www.esrl.noaa.gov/gmd/ccgg/trends_ch4/). The mean growth rate of CO2 of 2.24 ppm year−1 is close to the estimated global trend of 2.3 ppm year−1 from marine sites (Dlugokencky and Tans, 2020; www.esrl.noaa.gov/gmd/ccgg/trends). In addition, further analysis of the CH4 and CO2 trends is illustrated in Fig. 7. This figure shows all trends for time segments of 2 to 11 years (total length of the time window), associated with a 95% confidence level by Student’s t-test. Thus, it is obvious that time windows of several years or less may have an excess or a deficit of CH4 or CO2 concentrations. There is an increase and/or a decrease of concentrations during time segments shorter than 3 years for both CH4 (Fig. 7a) and CO2 (Fig. 7b), while for segments longer than 3 years, there is always an increase in concentrations (>8 ppb for CH4 and >5 ppm for CO2). For example, the 8-year changes observed between 2009 and 2017 indicate an increase in CH4 (∼+50 ppb) and CO2 (∼+17 ppm) concentrations at Lamto region. Two variability modes (i.e. high and low frequency variability) are thus observed and characterise the trends of CH4 and CO2 concentrations. The low frequency variability (i.e. time segment ≥3 years) is associated only with an increase in CH4 and CO2 concentrations while the high frequency variability (i.e. time segment <3 years) is associated with relatively low positive and negative values. Furthermore, we note that the low frequency variability indicates significant positive trends for many time segments, while the high frequency variability shows two positive and significant trends for each compound, notably during 2015–2016 and 2017 for CH4, and during 2010–2012 and 2012–2014 for CO2.

Furthermore, Fig. 8 shows evolutions of the growth rates of CO2 (Fig. 8a) and CH4 (Fig. 8b) observed at the LTO, ASC, ASK, IZO and CVO stations during the 2008–2018 decade derived from daytime measurements. The annual growth rates of CO2 from these different stations fluctuated from one year to the next, between 0.7 ppm yr−1 and 4.20 ppm yr−1. The long-term mean growth rate is, however, similar to the global average.

Highest CO2 rates (>2.50 ppm yr−1) were observed in 2010 at IZO and ASK, in 2015 and 2016 at IZO, ASK, ASC, CVO and LTO when the most severe El Nino Southern Oscillation (ENSO) event in this decade occurred. A similar result was found in studies of Patra et al. (2005) using 87 stations from 1959 to 2004. These studies have shown that the magnitudes of the observed CO2 growth rate increase or decrease are mainly modulated by the ENSO variability. However, any analysis trying to characterize relationships between regional CO2 growth rate and ENSO requires a longer time interval, which is not possible for the region of Western Africa. We observed a decrease in growth rates at LTO after the peak in 2015. This change may result in part from a decrease in precipitation, with a recovery of tropical vegetation after the 2015 El Nino. The CO2 average annual growth rate at LTO was around 2.23 ppm. yr−1, 1.5%, 2.7%, 3% and 16% lower than those observed at ASC, ASK, IZO and CVO, respectively. For CH4, the growth rate shows a similar long-term mean across all stations with important interannual variability for LTO and CVO stations. This strong interannual variability observed at the CVO and LTO stations could be due to variable inter-annual air mass transport and flux exposure from the continent (Fomba et al., 2014; Tiemoko et al., 2020b). In addition, the average annual growth rate of CH4 at LTO presents values ∼6.65 ppb. yr−1 that is 9% higher than that observed at ASC and 19%, 14% and 44% lower than those observed at ASK, IZO and CVO, respectively.

3.4. 

Seasonal cycles of CO2, CH4 and CO

3.4.1. 

CO2

Figure 8 shows the smoothed curves derived from CO2 observations over the last 10 years, and the mean seasonal cycles at the LTO, ASC, ASK, IZO and CVO stations.

The highest amplitude of the seasonal cycle is observed at LTO (13.6 ppm on average); ASK, IZO and CVO stations have a seasonal amplitude equivalent to half (∼8 ppm peak-to-peak) of that of LTO, while AUC, which is located in the southern hemisphere and most often south of ITCZ, has an amplitude that is a quarter (∼2.5 ppm peak-to-peak). The seasonal cycle at Lamto is characterized by a gradual decrease in CO2 concentrations from January to July, followed by a broad minimum-till October and a rapid increase from November to January. On the other hand, the ASK, IZO and CVO stations show seasonal cycles which are more typical of the background stations in the Northern mid-latitudes, with decreasing CO2 concentrations from May to September followed by an increase in autumn (September to December) and a broad maximum in winter (December to March). This seasonality of the CO2 cycle reflects the seasonality of the terrestrial vegetation fluxes and atmospheric transport. Also, the studies of Ago (2016) and Tiemoko et al.(2020a) on West African ecosystems indicated that photosynthesis is dominant during the rainy season and respiration exceeds it during the dry season. Compared to ASK, IZO and CVO, the maximum concentration at LTO occurs about 4 months earlier, in January. The LTO station is influenced by the biomass burning in West and Central Africa from December to February (Lacaux et al., 1995), and by the long-range transport of air masses from the north-east of the continent (Harmattan or boreal trade winds) coinciding with the fire season (Jonquières et al., 1998). Touré et al. (2012) showed that air masses from the north of the continent cross warm (i.e. Sahara) regions to reach the coastal areas of the Gulf of Guinea. These air masses contain dust and elevated levels of CO2 from respiration and fire emissions. The decrease of CO2 after January is probably related to the end of fire emissions with a dominant northerly wind regime (Kocha, 2011). The small secondary CO2 peak observed in August could be due to the occurrence of the short dry season (Diawara et al., 2014) during which the respiration activity of plants is important.

3.4.2. 

CH4

Figure 8e,f shows the CH4 seasonal cycles at LTO, ASK, IZO and ASC stations. Mean annual molar fractions of CH4 (Table 4) at LTO increased from 1822 ppb in 2008 to 1898 ppb in 2018. At LTO, ASK, IZO stations, the observed CH4 concentrations show fairly similar seasonal cycles (Fig. 9e), with a larger amplitude in LTO. The CH4 seasonal cycle at LTO shows a peak-to-peak amplitude of about 75 ppb, five times higher than in ASK (∼13 ppb), 2.5 times higher than at IZO (∼27 ppb) stations and practically similar in the CVO station. At the CVO station, the CH4 seasonal cycle shows a peak in the month of April accompanied by a decrease until July followed by an increase starting in February. At LTO, there is a decrease of CH4 concentrations from January to August-September, and an increase until December, similarly to the CO2 seasonal cycle. This observed phase is in advance compared to the ASK and IZO stations, where the decrease in CH4 concentrations occurs from March to June, and from March to July respectively. The decoupling of the LTO cycle compared to ASK, IZO and CVO in December and January is probably explained by CH4 emissions from biomass burning, explaining the correlation with the CO2 and CO cycles at Lamto. The seasonal variations in CH4 concentrations are also related to changes in the OH concentration, which is the major CH4 sink (Dlugokencky et al., 1994; Henne et al., 2008; McNorton et al., 2016). Satar et al. (2016) and Xia et al. (2020) showed, for example, that in the northern hemisphere, the CH4 chemical elimination and atmospheric dilution peak in summer, which is the period when low CH4 concentrations are measured. At the LTO station, this period extends from May to September and has significant cloud cover whose persistence induces low temperatures followed shortly afterwards by a drop in rainfall (Fig. 2) not favouring CH4 emissions from biogenic sources.

3.4.3. 

CO

In the Lamto region, the most important sources of CO emissions are biofuels (i.e. firewood, charcoal, residues from agriculture), photodegradation of litter and wildfires. Seasonality and observed trends of CO at LTO, ASC, ASK, IZO and CVO are shown in Fig. 9c,d, respectively. In the case of the LTO station, there is a strong seasonal and interannual variability in CO. The highest concentrations are observed in January during the peak of the fire season (monthly average up to 450 ppb), while minimum concentrations are recorded between May and November (down to 130 ppb). At ASK, IZO, ASC and CVO, the mean monthly concentrations measured are systematically lower than 120 ppb and vary little from one month to another, compared to LTO. The mean annual molar fraction measured at LTO is 216 ± 97 ppb over the 2014–2018 period (Table 4 and Fig. 10d), and is above those observed at the other stations with differences of 140 ± 89 ppb, 116 ± 86 ppb, 118 ± 82 ppb and 117 ± 85 ppb ppb from ASC, ASK, IZO and CVO, respectively. The differences observed in CO concentrations between LTO and the other three stations are likely explained by sources of CO emissions (e.g. biofuels, biomass fires) in the Lamto region, particularly wildfires during the dry season from December to February. These sources of emission are both local and regional. Indeed, CO concentration levels at LTO are influenced by the transport of polluted air, especially during the great dry season when air masses from the north cross-areas where agricultural activities and fires are high. On the other hand, the ASC, ASK, IZO and CVO stations are far from sources of emission (biofuel, biomass combustion, etc.). Also, there are strong similarities between the CO and CO2 seasonal cycles at Lamto that reflect a strong contribution of combustion sources in the CO2 seasonal cycle at Lamto.

3.5. 

Synoptic variations

In this section, we analyse the synoptic-scale variations of CO2, CH4 and CO by looking more specifically to the correlation’s residual values (i.e. ΔCO2, ΔCH4 and ΔCO) from the smoothed seasonal cycle (see Section 2.3.3). In this analysis we have been using only the mid-afternoon measurements (blue points on Fig. 5).

3.5.1. 

ΔCO2/ΔCO

The monthly (ΔCO2/ΔCO) slopes are calculated using regression lines orthogonal distance (i.e. Deming regression) over the 2014–2018 period. The numbers in red (Fig. 10a) are the correlation coefficients (R) between CO2 and CO residues. These correlation coefficients are low, positive and significant (R < 0.50, p < 0.001) in all months except in July, which presents a high and significant correlation value (R = 0.56, p < 0.001). (R = 0.56, p < 0.001). The low correlation coefficients show that changes in CO anomalies cannot explain those of CO2. The atmospheric CO2 variations are significantly controlled by plant uptake and respiration at very short time scales (Murayama et al., 2003). As illustrated by the diel cycles, the magnitude of CO2 variance driven by surface exchange with soils and vegetation exceeds the variance from combustion sources and has little impact on CO, hence relatively low correlations are observed. We notice that respiratory CO2 from soils decreased with low temperatures and the unavailability of water in July, which could explain the higher correlation. The ΔCO2/ΔCO ratio is characterized by significant seasonal variations as shown in Fig. 10a, with high values outside the fire season between May (0.13 ppm ppb−1) and July (09 ppm ppb−1) on the one hand, and between September (0.08 ppm ppb−1) and October (0.09 ppm ppb−1) on the other hand. In August and the November-April season, ΔCO2/ΔCO ratios take lower values in the range of 0.02–0.05 ppm ppb−1. This seasonal distribution of the ΔCO2/ΔCO variations follows the one of the rainfall patterns in the Lamto region characterized by alternating dry and wet seasons (Diawara et al., 2014). The low values of the concentration ratios recorded in the dry season also correspond to the influence of fires which provide high CO to CO2 emission factors. Although CO is an important tracer of biomass combustion (Langenfelds et al., 2002), its short lifetime does not enable to fully explain (inconsistent coefficients of correlation) changes in recorded CO2 fluxes during this season. The different components of CO2 emissions from fires, but also from respiration after the fire period, causes uncertainties on the characterization of emission factors from ΔCO2/ΔCO concentration ratios. However, general estimates of fire emission rates are often used to characterize the nature of CO2 emissions assuming that the observed enhancements of CO are mainly due to biomass fires (Suntharalingam, 2004; Henne et al., 2008; Wada et al., 2011; Denjean et al., 2020) The emission ratios (i.e., ΔCO2/ΔCO) of combustion sources (Christian, 2003; Koppmann et al., 2005; Wada et al., 2011), for African savannas are estimated between 0.0031 and 0.032 ppm ppb−1. During the FOS/DEFACE experiment in Lamto (Côte d’Ivoire), Bonsang et al. (1995) reported molar ratios of ΔCO/ΔCO2 in the range of 0.004 to 0.012 ppm ppb−1. In the present study, the mean ΔCO2/ΔCO ratio estimated during the biomass burning period is 0.025 ppm ppb−1 (0.034 ppm ppb−1 in December, 0.016 ppm ppb−1in January and 0.027 ppm ppb−1in February), in agreement with Koppmann et al. (2005). The difference with the results of Bonsang et al. (1995) could be explained by the fact that these measurements and other such studies measured emission factors directly in fire plumes, whereas our data mix the signature of fires, respiration and regional background concentrations. In the LTO area, Diawara et al. (2014) showed that in particular years, the climate of Lamto can alternate between climate of arid region (i.e. dry savanna) and climate of sub-humid or humid region (i.e. humid savanna) with year to year variations of fire intensity, distribution and CO to CO2 emission ratios. However, additional differences between biofuels types, fire intensity, firewood, charcoal and agricultural residues CO to CO2 emission factors could also contribute to the variance of ΔCO2/ΔCO slopes at LTO.

3.5.2. 

ΔCH4/ΔCO

Figure 9b shows the mean seasonal variations of the ΔCH4/ΔCO concentration ratios at LTO over the 2014–2018 period. The maximum ratio is observed in May (2.4 ppb ppb−1), while the minimum values are in January (0.26 ppb ppb−1), February (0.30 ppb ppb−1) and August (0.14 ppb ppb−1) in agreement with the range found in the literature on forest and savanna fire emission ratios ranging from 1 to 17%, i.e. 0.02 to 0.30 ppb ppb−1 (Ward et al., 1982; Hurst et al., 1994; Koppmann et al., 2005). The seasonality in the ratios is likely dominated by the seasonality in CO emissions reaching its maximum in December-January-February (Fig. 9). The correlation coefficients between ΔCH4 and ΔCO are high and significant (R > 0.5, p < 0.001) in all months except in May (R = 0.35, p < 0.001), August (R = 0.32, p < 0.001) and September (R = 0.48, p < 0.001), unlike those obtained between ΔCO2 and ΔCO. This observation shows that CH4 and CO are closely related and moreover these positive coefficients suggest that the anthropogenic emissions and fires dominate the carbon cycle at the Lamto station. The periods corresponding to the high CO emissions also coincide with the periods of high values of the correlation coefficients (R > 0.6, p < 0.001) between ΔCH4 and ΔCO. High values of correlation coefficients (R > 0.5, p < 0.001) between ΔCH4 and ΔCO indicate similar sources of emission (i.e. fire emissions) for these two gases (Fang et al., 2015).

Significant variations in ΔCH4/ΔCO concentration ratios are also observed during the wet seasons in April-July (GSP) and September-November (PSP). The ratios observed during those periods reflect significant CH4 emissions compared to CO, while the smaller ratios observed in December-March (GSS) and August (PSS) indicate significant CO emissions compared to CH4. The predominance of CH4 or CO is established according to the rainfall regime at Lamto. The calculated concentration ratios vary between 0.26 ppb ppb−1 and 2.41 ppb ppb−1 and are for some months different to the provided regional (Bonsang et al., 1995; Pak, 2000; Koppmann et al., 2005; Lin et al., 2015) and global ratios (Xiao, 2004).This difference could be induced by the fact that the Lamto area is in the transition area between forest and savanna, seasonally influenced by large scale transport from the West African Monsoon (WAM) and Harmattan winds in the main dry season, but also by strong anthropogenic pressures (i.e. agricultural activities, burning, farming, etc.). As a result, seasonal cycles of CO and CH4 potential sources of emissions may differ from a region to another considering the environment. To go further in this analysis, and to have a better estimate of regional emissions of these gases from the values of the concentration ratios, it would be interesting to have specific emission inventory that we could use in parallel with a source-receptor relationship method. In addition, CH4 isotopic measurements could also be used to identify the influence of the dominant sources from biogenic vs. biomass burning origins (Bergamaschi et al., 1998; Mikaloff Fletcher et al., 2004; Barker et al., 2020).

3.5.3. 

ΔCH4/ΔCO2

Figure 9c shows the seasonal variations of the ΔCH4/ΔCO2 ratio and correlation coefficients between ΔCH4 and ΔCO2 over the 2014–2018 period. The slopes between ΔCH4 and ΔCO2 show significant seasonal variations ranging from 4.2 to 13 ppb ppm−1, with an average of 7.8 ppb ppm−1. This mean value is low compared to those calculated for anthropogenic emissions in North Africa (39.1 to 46.2 ppb ppm−1), but relatively similar to that reported by Lacaux et al. (1995) (0.31% or 8.5 ppb ppm−1) in biomass burning plumes in the tropical savannas of Côte d'Ivoire. The molar ratios of fire emissions found by Bonsang et al. (1995), during the FOS/DEFACE experiment at Lamto from December to February, range 0.32 to 0.46% (11 to 21.5 ppb ppm−1) for fires with high combustion efficiency, is in accordance with the range (11 to 13 ppb ppm−1) that we obtain during the fire season (i.e. from December to February) (Fig. 10c).We note that the average value obtained in Lamto (7.8 ppb ppm−1) is relatively close to the concentrations found in suburban areas (7.1 ppb ppm−1) in India by Sreenivas et al. (2016) and in the Los Angeles megacity (6.4 ppb ppm−1) by Wong et al. (2014). These authors underlined that the ΔCH4/ΔCO2 low values indicate dominance of CO2 anthropogenic emissions in the studied area. On the other hand, October (12.1 ppb ppm−1), November (9.59 ppb ppm−1), December (12.92 ppb ppm−1), January (12.95 ppb ppm−1) and February (11.0 ppb ppm−1) record higher ΔCH4/ΔCO2 concentration ratios than the average rate, indicating the dominance of CH4 during these months. This dominance shows that local to regional emissions influencing the LTO station are probably controlled by fire emission in these months. The molar emission rates overall observed for different biomass burning events in various ecosystems range from lowest values of about 0.1% to highest values of about 2% (i.e. 2.75 to 55 ppb ppm−1)depending on the biofuel and the phase of the fire (Greenberg et al., 1984; Crutzen et al., 1985; Cofer et al., 1990; Bonsang et al., 1991; Kaufman et al., 1992; Hurst et al., 1994; Lacaux et al., 1995; Yokelson et al., 1997; Pak, 2000; Koppmann et al., 2005). The molar fractions calculated in our study are within the large range provided by these authors. Moreover, the calculated correlations show positive and significant values (p < 0.001) between 0.24 and 0.59. The correlations lower than 0.5 are obtained from February to November and those higher than 0.5 in December and January (12.9 ppb ppm−1). Strong correlations (>0.5) indicate that the factors which control the emissions and variability of CH4 and CO2 in December and in January, could be similar, especially as their molar fractions vary in the same way. On the other hand, the very tight correlations suggest that biospheric CO2 fluxes play only a minor role and that the ratios are dominated by collocated anthropogenic emissions and fires of CH4 and CO2. Fang et al. (2015) suggested similar patterns of CO2 and CH4 sources for correlation coefficients greater than 0.5. Hence, the observed correlation between ΔCH4 et ΔCO2 in this study during December-January is not only due to spatially and temporally correlated sources but is caused to a large extent by meteorological variability associated with more or less accumulation of trace gases in the boundary layer irrespective of their sources. In conclusion the processes controlling both CH4 and CO2 emissions during February-November are heterogeneous.

3.6. 

Case study from December 10 to 19, 2014

Figure 11 shows the CO2, CH4 and CO concentrations recorded at Lamto from December 10th to 21st 2014. The observed signals of these three species show typical variations at the scale of a few hours and a few days. For CO2 we observe systematically night-time increase with a large amplitude (40 to 120 ppm). CO2 peaks are generally associated with CO peaks, indicating a large contribution of emissions by combustion processes. CH4 has less pronounced diurnal cycles, and sometimes in phase opposition with CO2 and CO (e.g. Dec. 16–17). Superimposed on the diurnal cycles, we observe a significant increase around 13 December, framed by two periods of lesser variability and lower concentrations. The co-occurrence of this peak of CO2, CH4 and CO at Lamto suggests influences of polluted air masses with common sources and origins. The determination of the contribution of regional and/or distant sources on the levels of CO2, CH4 and CO concentrations observed passes necessarily by identification of the origin of the air masses arriving at the station. For this purpose, we calculated three-day back trajectories of air masses above Lamto (Figs. 12 and 13) coinciding with observed peaks (December 13, 2014) and the period of low concentrations (December 16-17, 2014) using the FLEXPART and HYSPLIT models (Draxler and Rolph, 2003). The back trajectories highlight two distinct origins: air masses coming from the East during the peak, indicating advection of polluted air coming from the continent, and coming from the South bringing cleaner air to the station on 17 December. It should be noted that even when the air mass is originating from the South, with the Atlantic Ocean located 165 km away, we still observe a significant CO2 diurnal cycle generated by the biospheric activity and the correlation with the diurnal cycle of the boundary layer development. The eastern air masses from the continent may be influenced by emissions from cities in Ghana, but also by biomass fires occurring during the dry season (van der Werf et al., 2010).

4. 

Conclusion and perspectives

The measured atmospheric CO2, CH4 and CO concentrations at LTO station were analyzed at different scales, from hours to the decade. The times series from August 2008 to June 2018 (for CH4 and CO2) and from 2014 to June 2018 (for CO) show a wide range of hourly mean concentrations ranging from 368 and 510 ppm for CO2, from 1771 to 2350 ppb for CH4 and from 84 to 5990 ppb for CO. Strong contrasts are generally observed between day time and night time values for all gases. We observe a sharp decrease in the concentrations around 7 hr in the morning, probably due to the effects of the soil radiative heating, which break the inversion layer formed during the night and releases the entrapped gases. The lowest variations in concentrations are observed during the wet season from April to October, whereas the highest variations appear during the bushfire regime (December-January-February).

Seasonal cycles compared to those of NOAA’s stations (ASC, ASK, IZO) show stronger seasonal and interannual variability associated with the biomass fire regime from December to February, but also with long-distance air mass transport from the southern Atlantic (monsoon) and from the north-east of the continent (Harmattan). Besides, the ΔCO2/ΔCO concentration ratio variations are significant and show a seasonal coherence with the rainfall regime of the Lamto region but these values ​​differ from those provided by over studies for African savannas. On the other hand, the ΔCH4/ΔCO ratios indicate significant CH4 emissions with respect to CO during April-July and September-November, and significant CO emissions compared with CH4 during December-March and August. The predominance of CH4 or CO is thus established according to the rainfall regime at Lamto. The correlations between CH4 and CO are positive and significant, indicating that the variations of these two gases are closely related. Also, the ΔCH4/ΔCO2 ratios have low values compared to those calculated in anthropogenic emissions in North Africa, Central Asia, and the Middle East region. The associated correlation coefficients indicate values ​​greater than 0.5 in December and January with concentration ratios of 12.9 ppb ppm−1. This means that during these two months the factors controlling the CH4 and CO2 emissions are similar.

This paper demonstrates the possibility of carrying out precise and long-term GHG measurements in Lamto, in a region of the world that previously had no continuous observation site. There is therefore no doubt that this station will provide valuable information on understanding GHG fluxes in West Africa using atmospheric inversions. Because ground level GHG data from this site may not capture the full transport of GHG and pollutants produced by fires, being uplifted in the troposphere e.g. by pyro-convection. For this reason, an analyzer (EM27/Sun) of total GHG columns will be installed at Lamto in 2021, which will allow a more complete view of the dispersion of biomass burning plumes during fire periods, and validation of satellite observations.