Carbon Tracker provides an analysis of atmospheric carbon dioxide distributions and their surface fluxes . It is a data assimilation system that combines observed in situ carbon dioxide concentrations from 81 sites around the world with model predictions of what concentrations would be based on a preliminary set of assumptions (“the first guess”) about sources and sinks for carbon dioxide. Carbon Tracker compares the model predictions with reality and then systematically tweaks and evaluates the preliminary assumptions until it finds the combination that best matches the real-world data. It has modules for atmospheric transport of carbon dioxide by weather systems, for photosynthesis and respiration, air–sea exchange, fossil fuel combustion, and fires. Transport of atmospheric CO2 is simulated by using the global two-way nested transport model (TM5). TM5 is an offline atmospheric tracer transport model driven by meteorology from the European Centre for Medium-Range Weather Forecasts (ECMWF) operational forecast model and from the ERA-Interim reanalysis to propagate surface emissions. TM5 is based on a global 30×20 and at a 10×10 spatial grid over North America. The model can be used in a wide range of applications, which includes aerosol modelling, stratospheric chemistry simulations, and hydroxyl-radical trend estimates. A detailed description of the TM5 model can be found in the works of and .

CT data from the CT2015 release and onwards use aircraft profiles from the stratosphere to the top of the atmosphere , and co-location errors are also quantified . The older data versions have been used and also compared with different datasets over other parts of the globe in previous studies . Most of the studies confirm that CT XCO2 captures observations reasonably well. In this study, we use Carbon Tracker release version CT2016 , hereafter CT2016, and a near-real-time version (CT-NRT.v2017). Both versions of NOAA CT provide 3-hourly CO2 mole-fraction data for the global atmosphere at 25 pressure levels at a 30×20 spatial resolution for a period covering 2000 to 2016. The data can be accessed freely in the public domain (ftp://aftp.cmdl.noaa.gov/products/carbontracker, last access: 27 February 2018).

GOSAT is the world's first spacecraft particularly designed to measure the concentrations of carbon dioxide and methane, the two major greenhouse gases, from space. The spacecraft was launched successfully on 23 January 2009 and has been operating properly since then. GOSAT records reflected sunlight using three near-infrared band sensors. The field of view at nadir allows a circular footprint of about 10.5 km in diameter . GOSAT consists of two instruments. The sensors for the two instruments can be broadly labelled as thermal, near infrared and imager. The first two sensors are used as part of a Fourier transform spectrometer for carbon monitoring which is referred to as TANSO-FTS, while the imager for cloud and aerosol observations is referred to as TANSO-CAI. The details on spectral coverage, resolution, field of view, and different products of TANSO-FTS in the three SWIR bands can be found in a number of previous studies and references therein. In this study ACOS B3.5 Lite XCO2 from GOSAT Level 2 (L2) retrieval based on the SWIR spectra of FTS observations and made available by Atmospheric CO2 Observations from Space (ACOS) of NASA is used. ACOS B3.5 Lite XCO2 has lower bias and better consistency than NIES GOSAT SWIR L2 CO2 globally . However, this version of ACOS XCO2 was found to suffer systematic retrieval error over the dark surfaces of high-latitude lands and over African savanna . shows systematic error in the African savanna associated with underestimating the intensity of fire during March at the end of the savanna burning season. Therefore, our choice of the ACOS B3.5 Lite, hereafter GOSAT XCO2, is motivated by these differences.

OCO-2 is the world's second full-time dedicated CO2 measurement satellite. It was successfully launched by the National Aeronautics and Space Administration (NASA) on 2 July 2014 . OCO-2 measures atmospheric carbon dioxide with the accuracy, resolution, and coverage required to detect CO2 source and sink on a global and regional scale. OCO-2 has a three-band spectrometer, which measures reflected sunlight in three separate bands. The O2 A-band measures molecular absorption of oxygen from reflected sunlight near 0.76 µm, while the CO2 bands are located near 1.61 and 2.06 µm . In this study, both the nadir and glint-mode measurements of OCO-2 XCO2 V7 lite level 2 covering the period from January 2015 to December 2016, hereafter referred to as OCO-2 XCO2, are used. Due to the scarcity of data, CT values from the two releases CT2016 for the year 2015 and CT-NRT.v2017 for the year 2016, hereafter CT16NRT17, are employed in this study. The OCO-2 project team at the Jet Propulsion Laboratory, California Institute of Technology, produced the OCO-2 XCO2 data used in this study. The data can be accessed from NASA Goddard Earth Science Data and Information Service Center.

Measurements of CO2 from nine ground-based flask observations near and within Africa's land mass were accessed from the NOAA/ESRL/GMD CCGG cooperative air sampling network https://www.esrl.noaa.gov/gmd/ccgg/flask.php (last access: 1 May 2019). Site description is given in Table .

The GOSAT and CT model XCO2 time series used in this investigation span 5 years, ranging from May 2009 to April 2014. Atmospheric CO2 concentrations of NOAA Carbon Tracker have global coverage with a 30×20 longitude–latitude resolution which covers 426 grid boxes in our study area. Satellite observations, however, are different from model assimilation and have gaps for various reasons (e.g. cloud and the observational mode of the satellite). As a result, there is no one-to-one spatiotemporal match between the two datasets. For example, CO2 products from the two datasets are not directly comparable since CT is a 3-hourly smooth and regular grid dataset, whereas GOSAT XCO2 is irregularly distributed in space and time. Thus, the CT CO2 is extracted on the time and location of GOSAT-XCO2 data. Using the grid point of CT as a reference bin, the corresponding GOSAT XCO2 found within a rectangle of 30×30 with centre at the reference bin and with a temporal mismatch of a maximum of 3 h is extracted. Moreover, CT has higher vertical resolutions than GOSAT. As a result, the two can not be directly compared. It is customary to smooth the high-resolution data (in this case CT) with averaging kernels and a priori profiles of the low-resolution satellite measurements (in this case GOSAT). Besides, due to a difference between CT and GOSAT on the number of vertical levels, CT CO2 is interpolated to vertical levels of GOSAT. The CT XCO2 (XCO2model) used in the comparison is computed from the interpolated CT CO2 (CO2interp), pressure weighting function (w), XCO2 a priori (XCO2a), column averaging kernel of the satellite retrievals (A) and a priori profile (CO2a) of the retrievals as per the procedure discussed by , , , , and and given as 1XCO2model=XCO2a+∑iwiTAi×(CO2interp-CO2a)i, where i is the index of the satellite retrieval vertical level and T is the matrix transpose. To compare the CT simulations and the satellite observations with the flask observations, the vertical profiles of the satellite and CT were extracted at the corresponding pressure level and location within a box of 1.50.

Correlation coefficients (R), bias and root mean square deviation (RMSD) are used to assess the level of agreement between the two datasets. The mean bias determines the average deviations in XCO2 between Carbon Tracker simulation and satellite observations. In this work the bias at the jth grid point is computed as 2Biasj=1n∑i=1n(Si-Oi), where Si and Oi are CT and GOSAT XCO2 values over the jth pixel at the ith time respectively. To quantify the extent to which XCO2 of CT and GOSAT agree, the pattern correlations at the jth grid point are computed as 3Rj=1n∑i=1n(Si-S‾)(Oi-O‾)1n∑i=1n(Si-S‾)21n∑i=1n(Oi-O‾)2, where S‾ and O‾ are the mean values of Si and Oi over the jth pixel. The RMSD which shows the standard error of the model with respect to the observation at the jth grid point is computed as 4RMSDj=1n∑i=1n((Si-S‾)-(Oi-O‾))2; this is the centered pattern root mean squared (rms) difference which is obtained from the rms error after the difference in the mean has been removed .

Comparison with in situ flask observation is achieved in a way that the Carbon Tracker and satellite observations are taken at a corresponding pressure level of the in situ flask observation (as mentioned in Table ) in order to correspond to flux towers' surface observation. Furthermore, the datasets are resampled to fit the flask observations in a 30X30 window centered on the flux towers, and the available months were averaged.

The column-averaged mole fraction of CO2 obtained from the NOAA Carbon Tracker model and GOSAT observation was compared. The results are based on 426 grid boxes uniformly distributed to cover the whole of Africa's land region. The analysis was based on 5 years of daily data starting from May 2009 to April 2014.

Figure  shows the temporal average of CT2016 (Fig. a) and GOSAT (Fig. b) XCO2 distribution. The major common spatial feature in the mean map of XCO2 from GOSAT and CT2016 reanalysis is dipole structure characterized by high XCO2 northward of the Equator and low XCO2 southward of the Equator, with the exception of some part of Equatorial Guinea and the Republic of Congo for CT (Fig. a) and part of the Democratic Republic of Congo for GOSAT (Fig. b); these are characterized by spatially anomalous high XCO2. The Southern Africa region is characterized by weaker anthropogenic CO2 emission and higher CO2 uptake by the vegetation than North Africa . This contributed to the observed dipole distribution. Another important pattern is the anomalous peak over the annual average location of the ITCZ (Fig. b) which appears to fade over eastern Africa. This is in agreement with the fact that carbon stocks and net primary production per unit land area are high over Equatorial Africa and decrease northward and southward of the Equator over arid environments . However, Fig. b shows that GOSAT observations have some limitations in simulating this spatial pattern in comparison to CT.

Figure c shows the mean difference (CT2016-GOSAT) XCO2 which ranges from -4 to 2 ppm. The highest difference between the CT2016 and GOSAT XCO2 (as high as -4 ppm) is observed over the northern part of Equatorial Africa (e.g. southern Guinea, southern Ghana, southern Nigeria, south-east of central Africa, western Ethiopia and South Sudan.), which is also known for near-year-round rainfall and relatively dense vegetation. The regions are known for their rainforest . The likely explanation could be that the XCO2 mean (over 5 years) may be slightly positively biased due to fewer GOSAT observations as shown in Fig. d. The satellite retrievals have noise which can be smoothed out when a large number of datasets is averaged. The strategy and methods for cloud screening in GOSAT retrievals could lead to a smaller number of observations in the equatorial region . The number of datasets used for comparison range from 14 to 4288 from grid box to grid box, with a spatial mean of 1109 data over the continent. Figure c also shows CT2016 simulations are overall lower than the values of GOSAT observation over most regions, with exceptions in Gabon, Congo, southern Kenya and southern Tanzania, where CT2016 simulations are higher than GOSAT observations by more than 1 ppm. The spatial distribution of global atmospheric CO2 is not uniform because of the irregularly distributed sources of CO2 emissions, such as large power plant and forest fire and biospheric assimilation as clearly noted above.

Figure a shows differences between CT2016 and GOSAT XCO2, which ranges from -4 to 3 ppm. Out of 100 % occurrence, more than 90 % of observed differences are within ±2 ppm. The mean difference between CT2016 and GOSAT means is about -0.27 ppm, with the standard deviation of 0.98 ppm indicating better regional consistency and low potential outliers. Moreover, a negative mean of the difference implies that XCO2 simulated from CT2016 is lower than that of GOSAT retrievals over Africa's land mass.

Because of selection criteria which permit a difference of 3∘ long and wide, the two datasets are not exactly at the same point. The impact of the relative distance between them should be assessed before performing any statistical comparison. Figure b depicted the colour-coded scatter plot of CT2016 model simulation versus GOSAT to determine whether the discrepancy between the datasets arises from the spatial mismatch. The colour code indicates the relative distance between the model and observation datasets. For these datasets the 50th percentile has a relative distance of 1.190, which means 50 % of the data have a relative distance of shorter than 1.190. The maximum relative distance between them is 2.120. However, there is no indication that this has been the case since the scatter is not a function of the relative distance between the datasets. For example, data points with blue colour with the lowest location difference are scattered everywhere instead of along the 1:1 line. Furthermore, we found the bias of -0.26 ppm, correlation coefficient of 0.86 and RMSD of 2.19 ppm for datasets which have a relative distance shorter than 1.190. On the other hand, the bias, correlation coefficient, and RMSD are -0.33, 0.86 and 2.22 ppm for those which are above 1.190. These statistics confirm that there is no strong discrepancy due to our selection criteria.

Figure  shows a statistical comparison of XCO2 from the CT2016 and GOSAT over Africa. The number of data used in this comparison are shown in Fig. d. As is depicted in Fig. a, the bias ranges from -4 to 2 ppm with a mean bias of -0.28±1.05 ppm (see Table ). A larger negative bias of about -2 ppm was found along the annual mean position of the ITCZ, the main climatic mechanisms controlling rainfall in Africa. Systematic errors due to the ITCZ and the East African Monsoon need to be addressed well in satellite retrievals and modelling works. The correlation varies from 0.4 over some isolated pockets in Congo, Tanzania, Mozambique, Uganda, and western Ethiopia to 0.9 over the northern part of Africa above 13∘ N, eastern Ethiopia and the Kalahari. Figure b depicts the correlation coefficient between GOSAT and Carbon Tracker XCO2. The region with poor correlation also exhibits high RMSD as shown in Fig. c. To understand whether this discrepancy originates from model weakness alone or terrible satellite visibility when the ITCZ is present and clouds are extremely thick and widely present, we have looked at the GOSAT posterior estimates of XCO2 error (Fig. d), which are high over regions where the bias and RMSD between GOSAT and Carbon Tracker XCO2 is high. GOSAT's posterior estimate of XCO2 error is a combination of instrument noise, smoothing error and interference error . This posterior estimate of XCO2 error does not include forward model error, which may lead to underestimation of the true error of satellite XCO2 by a factor of 2 . Therefore, part of the discrepancy is clearly linked to satellite retrieval uncertainty, which might have been amplified due to the small number of data points used to calculate the mean error of GOSAT XCO2 measurements (see Fig. d). In general, the two datasets are characterized by a high spatial mean correlation of 0.83±1.20, a global offset of -0.28±1.05 ppm, which is the average bias, a regional precision of 2.30±1.46 ppm, which is average RMSD, and a relative accuracy of 1.05 ppm, which is the standard deviation in the bias as depicted in Table .

Figures – show monthly mean XCO2 from CT2016 and GOSAT averaged over North Africa, Equatorial Africa, and Southern Africa respectively. Figures a–a depict the existence of an overall very good agreement for the monthly averages with respect to amplitudes and phases of XCO2. However, XCO2 from the two datasets slightly disagree in capturing the seasonal cycle over Southern Africa.

Figure a shows that XCO2 concentration reaches maximum in April and minimum in September over North Africa. Consistent with this evidence, other authors e.g. have indicated the presence of strong absorption of CO2 by vegetation during August in the Northern Hemisphere. This is the most likely the cause of the minimum concentration observed during September over North Africa. Both datasets show a concentration of XCO2 increases from October to April and decreases from May to September (see also Table ). Moreover, the two datasets show a monthly mean regional mean bias of -0.36 ppm with a correlation of 1.0 and a small root mean square deviation of 0.36 ppm (see Table ).

Figure a shows that XCO2 concentration reaches maxima (392.99 ppm) for CT2016 in March and (393.53 ppm) for GOSAT in January and minima (389.56 ppm for CT2016 and 389.32 ppm for GOSAT) in October over Equatorial Africa. The largest monthly mean difference of -1.34 ppm and the smallest of -0.05 ppm between the two datasets were observed in December and in April respectively (Table ). Moreover, both datasets show that concentration of CO2 increases from October to March, while it decreases from June to October. This similarity in the seasonal variability of the two datasets shows that they are in good agreement in terms of amplitude and phase. In addition, the two datasets show a monthly average regional average bias of -0.17 ppm, correlation of 0.98 and a small root mean square deviation of 0.71 ppm over Equatorial Africa (see Table ). Figure a shows maximum XCO2 concentration in April (391.04 ppm) for CT2016 and in October (391.28 ppm) for GOSAT and minimum in May (389.30 ppm) for CT2016 and (388.46 ppm) for GOSAT over Southern Africa. The largest monthly mean difference of 1.53 and 0.03 ppm between the two datasets is observed in April and in July (Table ) respectively. Both datasets show a concentration of CO2 increases from May to July, while it decreases from October to November. However, the XCO2 from CT2016 shows a gradually increasing trend from January to April. Conversely, GOSAT XCO2 shows decreasing values. This is most likely the result of the fact that CT2016 simulation is more sensitive to the growing size of the sink following the rainy season. Moreover, the two datasets show a monthly mean regional mean bias of 0.07 ppm, correlation of 0.97 and RMSD of 0.87 ppm over Southern Africa (see Table ).

Figures b–b show regional averaged bias in the monthly mean XCO2 from CT2016 and GOSAT. Figure b shows the presence of seasonally varying negative bias over North Africa. A high (<-0.5 ppm) negative bias in dry seasons (April to June) and low (≧-0.1 ppm) negative bias in wet seasons (August to September) are observed. Moreover, the strength of the bias increases from February to June. Conversely, the bias decreases from June to September. Similarly, Figs. b and b show seasonally fluctuating bias. For example, Fig. b shows a positive bias from February to July and negative bias from August to December over Southern Africa.

Figures c–c show the histogram of difference. The mean difference between CT2016 simulation and GOSAT observation of XCO2 is -0.36 ppm with a standard deviation of 0.35 ppm over North Africa (see Fig. c); Fig. c presents a mean difference of -0.17 ppm with a standard deviation of 0.71 ppm over Equatorial Africa and Fig. c reveals a mean difference of 0.01 ppm and a standard deviation of 0.85 ppm, which indicates that XCO2 from CT2016 was slightly higher than that of GOSAT over Southern Africa on average. In addition, the low standard deviation of monthly mean difference over North Africa typically indicates good regional consistency between CT2016 and GOSAT. This is mainly because North Africa is dominated by the Sahara, which is a vegetation-free area, and the systematic bias due to the local atmosphere–biosphere interaction is minimum. However, the spatial mean of monthly mean bias is slightly higher (-0.36 ppm) over North Africa than over Equatorial Africa (-0.17 ppm) and Southern Africa (0.01 ppm). This is possibly due to the presence of strong local emissions from Egypt, Algeria and Libya as well as due to long-range transport from the Northern Hemisphere as reported in other studies .

Figures d–d display the annual growth rate of XCO2, which ranges from 1.5 to 2.7 ppm yr-1. Moreover, the two datasets are consistent in determining the annual growth rate. The results are found to be in good agreement with the observed variability in the global annual growth rate from surface measurements (http://www.esrl.noaa.gov/ gmd/ccgg/trends/global.html, last access: 20 March 2018) which is 1.67, 2.39, 1.70, 2.40, and 2.51 ppm yr-1 globally during 2009–2013 respectively and 1.89, 2.42, 1.86,2.63, and 2.06 ppm yr-1 for Mauna Loa during 2009–2013 respectively, with error bars of 0.05–0.09 ppm yr-1 for global datasets and 0.11 ppm yr-1 for Mauna Loa datasets . The growth rate may not be conclusive due to the short length of the datasets used. However, it reflects how the CT and GOSAT observations perform with respect to each other.

The seasonal cycle has important implications for flux estimates . It is important to analyse whether there are seasonally dependent biases that are affecting the seasonal cycle and whether the datasets are capturing the same seasonal cycle. The four seasons considered here are December/January/February (DJF), March/April/May (MAM), June/July/August (JJA), and September/October/November (SON). DJF corresponds to northern winter/southern summer, MAM to northern spring/southern autumn, JJA to northern summer/southern winter, and SON to northern autumn/southern spring respectively. Figure  shows the seasonal distributions of CT2016 (left panels) and GOSAT (middle panels) XCO2 and their difference (CT2016-GOSAT, right panels). The distribution clearly shows that XCO2 concentration is maximum during MAM and minimum during SON over the North Africa. On the other hand, maxima are found during SON and minima during DJF over Southern Africa. These features are in good agreement with the rainfall climatology of the Northern Hemisphere and Southern Hemisphere. Moreover, Table  shows seasonally varying biases. Seasonal biases affect the seasonal cycle and amplitudes, which are important for biospheric flux attribution .

The right panels in Fig.  show that the seasonal mean difference (CT2016-GOSAT) ranges from -4 to 6 ppm, with a maximum difference of 6 ppm over the Gulf of Guinea and Congo during JJA. However, such a maximum difference was also observed over Southern Africa during DJF. A minimum of -4 ppm over the annual mean ITCZ region was observed during DJF and MAM. Moreover, the difference is above 1 ppm over Southern Africa during DJF and MAM (wet season of the region). This implies high spatial variability of the seasonal mean difference during different seasons (see also Table ). It also suggests that the discrepancy between the CT2016 and GOSAT becomes significant when vegetation cover is weak during DJF and MAM (dry seasons) over North Africa.

During SON the seasonal difference in most of Africa's land region ranges from -2 to 1 ppm. The result implies that CT2016 simulates lower values of XCO2 than that of GOSAT observation, indicating that there is a better spatial consistency during this season. Furthermore, during these seasons both North and Southern Africa have a moderate vegetation cover following their respective summer seasons. The two datasets show lower regional variation (i.e. only from -2 to 2 ppm) over most of Africa's land mass. However, Equatorial Africa exhibits a mean difference lower than -2 ppm during DJF and MAM. This indicates that the model tends to simulate lower than GOSAT XCO2 over the region. Figure  (right panels) reveals XCO2 from CT2016 is lower than GOSAT XCO2 over North Africa. The underestimation of observed XCO2 by the NOAA CT2016 model is likely related to the skill of driving ERA-Interim data as noted from previous studies. For example, has shown that the ERA-Interim data have a wet bias over Ethiopian highlands. have also shown that ERA-Interim precipitable water is higher than measurements from radio-sonde, FTIR and GPS observations. Therefore, such wet bias in the driving ERA-Interim global circulation model (GCM) might have forced NOAA CT2016 to generate dense vegetation which serves as a CO2 sink.

Figure  shows the mean difference between CT2016 and GOSAT XCO2 seasonal means which ranges from -0.37 to 0.04 ppm with a standard deviation within a range of 1.00 to 1.91 ppm over the continent. The highest mean difference of XCO2 (-0.37 ppm) occurs during SON and the lowest (0.04 ppm) occurs during MAM. Table  presents the summary of statistical values for the spatial mean of each season means. The comparison between the two datasets also shows there is a strong correlation (>0.5) during each season over the continent. However, there are moderate correlations (0.3 to 0.5) during DJF and MAM over North Africa and during DJF over Southern Africa. The low correlation over North Africa may be linked to a weak absorption by vegetation and a strong emission from human activities during winter as reported elsewhere . Moreover, Table  shows that the seasonal biases are negative over North Africa, while they are mostly positive over Equatorial and Southern Africa. Negative biases are observed during DJF and SON over Equatorial and Southern Africa respectively, implying that XCO2 from CT2016 are lower than from GOSAT during dry seasons.

Comparison of GOSAT and CT2016 with flask observation is carried out over six available ground-based flask observations. For the comparison, the volume mixing ratio of CO2 from GOSAT and CT2016 at the pressure level that corresponds to surface flask observations (see Table ) was considered.

Monthly mean CO2 from flask observations at IZO and ASK in North Africa shows an excellent agreement with both CT2016 and GOSAT CO2. Moreover, CT2016 has a better sensitivity in capturing the amplitudes than GOSAT, where observations from GOSAT mostly underestimate higher values of flask CO2 (Fig. ). However, this agreement has deteriorated over sites in Equatorial Africa (ASC and MKN) and Southern Africa (MNB). Over MKN, CT2016 shows better correlation (0.43) than GOSAT observation (0.08). In addition, monthly amplitudes from CT2016 were closer to the flask observations, suggesting that satellite retrievals need much attention over the region. On the other hand, GOSAT observations were found to be in better agreement with flask observations over ASC. also show that GOSAT data were correlated well with ground observation and found to be more centralized, having high system stability, especially over the ocean.

CT2016 has a better sensitivity over IZO, ASK and NMB. Moreover, CT2016 compared better with flask observations than GOSAT over these sites; almost all flask observations are within the standard deviations of the monthly mean of CT2016. However, GOSAT observations were found to be in better agreement with flask observations than CT2016 was over WIS and ASC. On the other hand, both CT2016 and GOSAT have low sensitivity to flask observation over MKN (see Fig. ). Similar to our previous discussion on sites in North Africa (IZO, ASK and WIS), CT2016 underestimates XCO2 during August, September, and October (wet season) compered to GOSAT observation and overestimates XCO2 during January to June. However, the CT2016 and the flask observations exhibit better agreement, indicating a bias in GOSAT observation during the wet season.

The strong El Niño event that occurred during 2015–2016 provides an opportunity to compare the performance of CT16NRT17 during strong El Niño events. Because of the decline in terrestrial productivity and enhancement of soil respiration, the concentration of CO2 increases during El Niño events . In this section we compare mean XCO2 of NOAA CT16NRT17 and NASA's OCO-2 covering the period from January 2015 to December 2016.

The comparison was made based on the selection criteria discussed in Sect. . Figure  shows the mean distribution of XCO2 from CT16NRT17 (Fig. a) and OCO-2 (Fig. b) over Africa's land mass. CT16NRT17 shows high (>400 ppm) XCO2 values over North Africa, while these high XCO2 values are observed over Equatorial Africa in the case of OCO-2 observation. The two datasets show a discrepancy over Equatorial Africa, where CT16NRT17 simulates low XCO2 values (<401 ppm), while OCO-2 observes high values of XCO2 (>401 ppm). Both datasets show moderate XCO2 values which range from 397 to 400 ppm over Southern Africa. The XCO2 distribution from OCO-2 is consistent with the maximum CO2 concentration reported in a past study by , implying that the CT16NRT17 likely underestimates XCO2 values over Equatorial Africa. It is also possible that the discrepancy is a compounded effect of OCO-2 XCO2 positive bias over the region . Figure c shows the mean difference between the 2-year mean of XCO2 from CT16NRT17 and OCO-2, which is in the range from -2 to 2 ppm. However, high (<-2 ppm) negative mean difference between the two datasets over rainforest regions (Gulf of Guinea and Congo basin) and the ITCZ over eastern Africa (South Sudan and south-eastern Sudan) is observed, implying that CT16NRT17 simulates lower XCO2 values than that of OCO-2 observation over regions where vegetation uptake is strong. Conversely, high (>1) positive mean difference over the Sahara, Somalia and Tanzania implies CT16NRT17 simulates higher XCO2 values than OCO-2 observation where the vegetation uptake is weak. Moreover, a positive (>2) mean difference over Egypt, Libya, Sudan, Chad, Niger, Mali and Mauritania is likely due to overestimates of XCO2 emission from local sources by CT16NRT17. Overall, the two datasets show a fairly reasonable agreement with a correlation of 0.60 and an offset of 0.36 ppm, a regional precision of 2.51 ppm and a regional accuracy of 1.21 ppm.

Figure a shows the histogram of 2-year mean difference, which is characterized by a positive mean of 0.34 ppm and a standard deviation of 1.21 ppm. This suggests that CT16NRT17 simulates high XCO2 as compared to observations from OCO-2 over Africa’s land mass.

Because of the presence of spatial and temporal mismatch of some level between CT16NRT17 and OCO-2 datasets, it is important to assess the effect of relative distance between the datasets. Figure b shows a colour-coded distribution of the two datasets. In the figure colour codes indicate the relative distance. The random scatter of blue dots implies that the statistical discrepancies do not arise from the relative distance between the two datasets. More specifically, a statistical comparison of datasets lower and higher than the 50th percentile (1.20) shows bias of 0.58 and 0.57 ppm, correlation of 0.57 and 0.57 and RMSD of 2.65 and 2.67 ppm respectively.

Figure  shows the comparison of mean XCO2 from CT16NRT17 and OCO-2 covering the period from January 2015 to December 2016. The number of data used are displayed in Fig. d. Figure a depicts the bias which ranges from -2 to 2 ppm with a mean bias of 0.34 ppm. However, higher biases (<-2 ppm) are observed over Equatorial Africa along the annual average location of the ITCZ. Figure b shows the correlation map with values from 0.2 to 0.8 over Africa's land mass. Good correlations of above 0.6 are seen over many regions of the continent, while weak correlation of less than 0.2 and higher root mean square error (>3 ppm) are observed over small pockets of the Equatorial and eastern Africa regions (see Fig. c). These regions also show a higher (>0.65 ppm) error in satellite retrieval (see Fig. d). In addition, Fig. d shows the number of observations are small (<1000) over these regions. This may contribute to the observed discrepancy over these regions. However, weak correlations are also observed over a wider area in North Africa such as Mauritania, Mali, Algeria and some regions of Niger, where satellite errors are low and sufficient data are obtained. Poor correlation and higher RMSD values are observed over south-western Ethiopia.

Figures– show a 2-year monthly average time series comparison of XCO2 from CT16NRT17 and OCO-2 over North Africa, Equatorial Africa and Southern Africa respectively. Figure a shows the existence of good agreement between the two datasets in describing patterns over North Africa. Moreover, both datasets show a decreasing trend of XCO2 from May to September but an increasing trend from October to April. On the other hand, consistent with the climate condition and associated CO2 exchange, the monthly mean XCO2 shows a maximum value of 403.37 ppm for CT16NRT17 and 402.06 ppm for OCO-2 during May. Conversely, minimum concentrations of 398.77 ppm from CT16NRT17 simulation and 398.27 ppm from OCO-2 observation are found in September. In addition, both CT16NRT17 and OCO-2 show maximum XCO2 values (402.15 ppm for CT16NRT17 and 402.03 ppm for OCO-2) in December. These peak values in December are not surprising, because the 2015–2016 El Niño started in March 2015 and reached its peak in December 2015, which added extra CO2 to the atmosphere . Figure a also shows that XCO2 from CT16NRT17 simulation is higher than OCO-2 observation over North Africa.

Figure b shows the monthly mean difference between CT16NRT17 and OCO-2 which ranges from -0.5 to 2 ppm. OCO-2 XCO2 observations are lower than CT16NRT17 by 2 ppm during March and April 2015. Starting from August 2015, the difference between the two datasets is minimum; On the other hand, a maximum difference exceeding 1.5 ppm was observed during MAM which can be mentioned as a burning season of North Africa, as the area north of the Equator was burned mostly from March to June . The observed lower XCO2 values from OCO-2 observations than that of CT16NRT17 simulation will be a consequence of much respiration which exceeded photosynthesis when vegetation uptake is weak following the strong El Niño and dry season over North Africa. Furthermore, intense burning of the forest during this season which will further be intensified by the strong El Niño may cause unpredicted aerosol loading, and thereby this inaccurate estimation of aerosol loading could be suggested as the most likely source for the observed discrepancy. Moreover, Fig. c displays a monthly mean regional mean bias of 0.87 ppm, correlation of 0.95 and root mean square deviation of 0.72 ppm between CT16NRT17 and OCO-2 XCO2. This implies that CT16NRT17 is in good agreement with OCO-2. However, small discrepancies arose, most likely due to a strong anthropogenic emission from Nigeria, Egypt and Algeria.

Figures a–a show monthly mean time series of XCO2 from the model and OCO-2 instrument over Equatorial Africa and Southern Africa, which are also in good agreement in terms of pattern. However, the figures show that CT16NRT17 simulations are lower than those of OCO-2 during October, November and December, whereas it is opposite during April, May and June over Equatorial Africa and Southern Africa. Figures b and b depict a seasonal bias in the monthly time series over Equatorial Africa and Southern Africa respectively. Positive biases are observed during dry seasons, while negative biases are during wet seasons. Moreover, the datasets have monthly averaged regional mean biases of 0.13 and 0.11 ppm, correlation of 0.90 and 0.94, and RMSD of 0.84 and 0.73 ppm over Equatorial Africa and Southern Africa respectively. This shows the existence of better agreement between CT16NRT17 and OCO-2 over these regions in terms of monthly average regional mean values. Figures d–d show both CT16NRT17 and OCO-2 are in good agreement in estimating the annual growth rate. found a global mean of more than 3 Gt of CO2 added to the atmosphere due to the strong El Niño event that occurred during 2015–2016. In agreement with this, both CT16NRT17 and OCO-2 show an annual growth rate that ranges from 3.10 to 3.42 ppm yr-1 of XCO2 over Africa's land mass (see also Table ). However, over all regions of Africa's land mass CT16NRT17 shows a lower XCO2 annual growth rate than those of OCO-2.

Figure  depicts seasonal means of XCO2 over Africa's land mass from CT16NRT17 (left panels), OCO-2 (middle panels) and their difference (right panels) covering the period of January 2015 to December 2016. The white space seen over some regions (e.g. Mali during JJA) is due to insufficient coincident satellite data according to the selection criteria during these seasons. XCO2 increases from winter to spring and then decreases from spring peak to summer minimum over the whole continent. The decrease from spring maximum to summer continued into autumn over the northern half of Africa in contrast to the southern half of Africa, which exhibits an increase in XCO2. The decrease from spring to autumn (northward of the Equator) and until summer (southward of the Equator) is likely to be a consequence of the land vegetation awakening from dormancy of winter and partly spring. Conversely, the decomposition of died and decayed vegetation which began in autumn and continued throughout winter adds extra CO2, leading to a maximum concentration during spring . In agreement with this, both CT16NRT17 and OCO-2 show maximum XCO2 during MAM over North Africa and during SON over Southern Africa. Conversely, minimum concentrations are observed during SON over North Africa and during DJF over Southern Africa.

Figure  (right panels) shows the seasonal mean difference of CT16NRT17 and OCO-2. A higher mean difference greater than 1 ppm is observed over North Africa during DJF and MAM, when the vegetation cover over the region decreases and also in the presence of an intensive burning of the northern savanna during this season . This indicates that XCO2 values from CT16NRT17 are higher than that of OCO-2 when vegetation uptake is weak and there is more fire. On the other hand, higher negative mean differences of less than -2 ppm are observed over Equatorial Africa during DJF and SON over Southern Africa. This difference between the CT and OCO-2 arises likely due to grass fires from the dry savanna. Consistent with the report by , low seasonal variability is observed between CT16NRT17 and OCO-2 in the range from -4 to 4 ppm, with greater amplitude over North and Equatorial Africa than over Southern Africa (see Fig. , right panels). During dry seasons OCO-2 overestimates values over the North Africa, but it underestimates them for Southern Africa.

Figure  shows the histogram of seasonal mean difference of CT16NRT17 and OCO-2. The smaller standard deviations of 1.49 and 1.07 are observed during JJA and SON. On the other hand, higher standard deviations of 1.69 and 1.75 ppm are observed during DJF and MAM respectively. These results indicate that CT16NRT17 and OCO-2 show a better consistency during wet seasons, and this consistency decreases as the vegetation cover decreases over most regions of Africa's land mass during dry seasons.

Monthly CT16NRT17 XCO2 has a better sensitivity over IZO and ASK both in terms of temporal pattern (phase) and amplitude than OCO-2 (see Fig. ), where observations from OCO-2 mostly underestimate XCO2 at the two flask sites. Over LMP and WIS, both CT16NRT17 and OCO-2 have moderate sensitivity in capturing the seasonal cycle. On the other hand, OCO-2 has a better sensitivity over ASC and SEY. In addition, XCO2 from both CT16NRT17 and OCO-2 is found to have poor correlations with flask observations over NMB and CPT. However, OCO-2 has closer sensitivity in capturing amplitudes than CT16NRT, where CT16NRT17 overestimates XCO2 at these flask sites. In general, CT has a better performance over sites located at high altitude (IZO, ASK) where satellite observations underestimate XCO2. Conversely, satellite observations have better performance over low-altitude island sites (ASC and SEY) as revealed by better agreement with flask XCO2 observations.