Reviews and syntheses : Greenhouse gas emissions in natural and agricultural lands in sub-Saharan Africa : synthesis of available data and suggestions for further studies

Introduction Conclusions References

Greenhouse gas fluxes in Africa play an important role in the global GHG budget (Thompson et al., 2014;Hickman et al., 2014;Valentini et al., 2014;Ciais et al., 2011;Bombelli et al., 2009).In recent years, conversion rates of African natural lands, including forest, grassland and wetland to agricultural lands have increased (Gibbs et al., 2010;FAO, 2010).The dominant type of land use change has been the conversion of forest to agriculture with average deforestation rates of 3.4 million ha per year (FAOSTAT, 2014) (Fig. 1).This land-use conversion results in an estimated release of 0.32 ± 0.05 Pg C yr −1 (Valentini et al., 2014) or 157.9 ± 23.9 Gt CO 2 eq in 1765 to 2005 (Kim and Kirschbaum, 2015), higher than fossil fuel emissions for the continent (Valentini et al., 2014).Figures

Back Close
Full Soil emissions of all the major GHGs from Africa can be potentially significant at global scales.For example, CO 2 eq emissions from 12 river channels in SSA and wetlands of the Congo River were about 0.9 Pg C per year, equivalent to about 25 % of the global terrestrial and ocean carbon sink (Borges et al., 2015).Nitrous oxide emissions in Africa contribute between 6-19 % of the global total, and changes in soil N 2 O fluxes in Africa drive large inter-annual variations in tropical and subtropical N 2 O sources (Thompson et al., 2014;Hickman et al., 2011).Nitrous oxide emissions from biogenic sources and fires in natural lands were estimated to contribute to 34 % of total N 2 O emissions in the region (Valentini et al., 2014).Even with the low fertilizer rates used across the continent, agricultural GHG emissions in Africa are substantial; amounting to 26 % of the continent's total GHG emissions (Valentini et al., 2014).According to Lassaletta et al. (2014), mean N application rates in Africa were 34 kg N ha −1 in 2009 and only 16 kg N ha −1 in sub-Saharan African countries.Only Mauritius, Botswana and South Africa had average N application rates exceeding100 kg N ha −1 .However, use of synthetic fertilizers such as urea has increased in the last four decades as well as the number of livestock (and their manure and urine products) (Bouwman et al., 2009 and2013) (Figs. 2 and 3).The increasing trend in N application rates is expected to cause a twofold increase in agricultural N 2 O emissions in Africa by 2050 (from 2000) (Hickman et al., 2011).In the case of CH 4 emissions, there are important differences between ecosystems.Tropical humid forest, wetlands and termite mounds are likely sources of CH 4 , while seasonally dry forests and savannahs are typically CH 4 sinks (Valentini et al., 2014).
Our current understanding of GHG emissions in Africa is particularly limited when compared to the potential the continent has as both a GHG sink and source.This lack of data on GHG emissions from African natural and agricultural lands and the lack of a comprehensive analysis of existing data hinder the progress of our understanding of GHG emissions on the continent (Hickman et al., 2014;Valentini et al., 2014;Ciais et al., 2011;Bombelli et al., 2009).In order to identify mitigation measures and other climate smart interventions for the region it is important to quantify baseline GHG emis-Figures

Back Close
Full sions, as well as understand the impacts of different land-use management strategies on GHG emissions (e.g., Palm et al., 2010).In this study our objectives are to synthesize currently available data on GHG emissions from African AFOLU; create an inventory of information from studies on emissions; and select priority topics for future GHG emission studies in natural and agricultural lands in SSA.

Data collection
Data were acquired by searching existing peer-reviewed literature using the names of the sub-Saharan countries and the GHGs (i.e.CO 2 , CH 4 and N 2 O) as search terms (using Web of Science and Google Scholar;1960-2015).We selected studies that reported in situ annual GHG emissions or those that provided enough information to estimate annual GHG emissions through unit conversion and/or extrapolation of given data.Data from 73 studies, conducted in 22 countries (n = 244) in SSA were used and were further categorized as GHG emission in natural lands (n = 117; Table S1 in the Supplement) and agricultural lands (n = 127; Table S2) (Fig. 4).The category of GHG emissions in natural lands were further divided into emissions from forest/plantation/woodland (n = 55), savannah/grassland (n = 31), streams/rivers (n = 14), wetlands/floodplains/lagoons/reservoirs (n = 11), termite mounds (n = 5), and salt pans (n = 1) (Table 1).The category of GHG emission in agricultural lands, were subdivided into emissions from cropland (n = 105), rice paddy (n = 1), vegetable garden (n = 5), and agroforestry (n = 16) (Table 1).Across all categories there were 174 follows: emission control is control treatments with no N fertilizer additions, and N input is the amount of added N.
It should be noted that our data compilation includes a wide variety of studies that were conducted under diverse biophysical conditions using a range of methodologies for quantifying GHG emissions (e.g., different sampling protocols, chamber design, and emission rate calculations), soil properties, and climatic factors.Therefore, the overall figures on GHG emissions shown are based on results achieved by different measurement techniques with inherent and contrasting sources of error.

Statistical analyses
The compiled datasets were used to examine the best model fit and to derive the corresponding model parameters for N 2 O emissions and yield-scaled N 2 O emissions as a function of the respective N input levels.Different data fitting models (linear, nonlinear, natural log, logarithm and sigmoidal) were tested for each dataset.The regression models were checked for violation of assumptions of normal distribution (Shapiro-Wilk test), homoscedasticity (Breusch-Pagan test), and constant variance (Durbin-Watson statistic) (Motulsky and Christopoulos, 2004).Separate t tests were used to assess significance of regression coefficients and intercepts in the fitted parametric models and adjusted coefficients of determination (adjusted R 2 ) of fitted parametric models were used as criteria for model selection: the model with the higher adjusted R 2 was selected.Statistical significance was considered at the critical level of 5 %.These statistical analyses were conducted using SAS ® ver.9.2 (SAS Institute, Cary, NC, USA) and SigmaPlot ® ver.11.0 (Systat Software Inc., San Jose, CA, USA).Figures

Back Close
Full 3 Results and discussion

Terrestrial systems
Soil GHG emissions from African natural terrestrial systems such as natural forest, plantation, woodland, savannah, grassland, termite mounds and salt pans ranged from 3.3 to 130.9 Mg CO 2 ha −1 yr −1 , −4.8 to 3.5 kg CH 4 ha −1 yr −1 and −0.1 to 13.7 kg N 2 O ha −1 yr −1 (Tables 1 and S1).The high variability in emission rates was likely related to differences in soil temperature, moisture content and physical-chemical properties as well as the type of natural vegetation present.Within forest systems, CO 2 emissions were strongly related to both soil moisture and temperature.For example, soil moisture explained about 50 % of the seasonal variability in soil CO 2 efflux in a Croton macrostachys, Podocarpus falcatus and Prunus africana forest in Ethiopia (Yohannes et al., 2011), as well as much of the seasonal variation in soil CO 2 efflux in a 3-year-old Eucalyptus plantation in Republic of Congo (Epron et al., 2004).Thomas et al. (2011) found that the Q 10 of soil CO 2 efflux (a measure of the temperature sensitivity of efflux, where a Q 10 of 2 represents a doubling of efflux given a 10 • C increase in temperature) was dependent on soil moisture at sites across the Kalahari in Botswana, ranging from 1.1 in dry soils, to 1.5 after a 2 mm rainfall event and 1.95 after a 50 mm event.Similarly, in a Zambian woodland, the main driving factor controlling CO 2 emissions at a seasonal time scale was a combination of soil water content and temperature (Merbold et al., 2011).Soil physical (e.g., bulk density, porosity and soil texture) and chemical properties (e.g., pH, C and N) also affected soil GHG emissions (e.g., Saggar et al., 2013;Smith, 2010;Snyder et al., 2009).Soil CO 2 efflux was positively related to total soil C content in undisturbed miombo woodland in Zambia, although not in an adjacent disturbed woodland (Merbold et al., 2011).In a Kenyan rainforest, CO 2 emissions were negatively correlated with subsoil C and positively correlated with subsoil N concentrations, Figures

Back Close
Full while N 2 O emissions were negatively correlated with clay content and topsoil C : N ratios (Werner et al., 2007).However, soil bulk density and pH were the most influential factors driving spatial variation of in situ N 2 O emissions in a tropical highland rainforest in Rwanda (Gharahi Ghehi et al., 2014).Similarly, a laboratory-based experiment using soils from 31 locations in a tropical mountain forest in Rwanda showed that N 2 O emissions were negatively correlated with soil pH, and positively correlated with soil moisture, soil C and soil N (Gharahi Ghehi et al., 2012).In many temperate systems, vegetation type also affects soil GHG emissions, likely because of differences in litter quality and production rate, amount of below-ground biomass, the structure of root systems as well as plant-mediated effects on soil microclimate (e.g., Díaz-Pinés et al., 2014;Masaka et al., 2014;Kim et al., 2010).This is consistent with findings from African systems where annual soil CO 2 efflux also varied with vegetation types.For example, annual soil CO 2 emissions were significantly lower in N-fixing acacia monocultures than in eucalypt monocultures and mixed-species stands in Republic of Congo (Epron et al., 2013).The differences were attributed to leaf area index in another study from the Republic of Congo where they found 71 % of seasonal soil CO 2 efflux variability was explained by the quantity of photosynthetically active radiation absorbed by the grass canopy (Caquet et al., 2012).Also in the Republic of Congo, it was found that litterfall accounted for most of the age-related trends after the first year of growth, and litter decomposition produced 44 % of soil CO 2 flux in the oldest stand (Nouvellon et al., 2012), strongly suggesting that the amount and quality of litter plays a major role in determining soil CO 2 flux.However, the effect of vegetation type can also interact with soil physical-chemical properties.For example in Benin, root respiration contributed to 30 % of total soil CO 2 efflux in oil palms when the soil was at field capacity and 80 % when soil was dry (Lamade et al., 1996).
Forest soils predominantly act as sinks for CH 4 (Werner et al., 2007).In Cameroon, the largest CH 4 oxidation rates were observed from relatively undisturbed nearprimary forest sites (-14.7 to -15.2 ng m −2 s −1 ) compared to disturbed forests (-10.5 to 0.6 ng m −2 s −1 ) (Macdonald et al., 1998).be both a sink and source of CH 4 .In Mali, CH 4 uptake was observed in dry sandy savannah (Delmas et al., 1991), while a savannah in Burkina Faso was found to be both a CH 4 sink and source during the rainy season, although overall it was a net CH 4 source (Brümmer et al., 2009).
Soil rewetting typically has a large impact on GHG emissions.Two broad mechanisms responsible for changed soil GHG flux following rewetting have been hypothesized: (1) enhanced microbial metabolism by an increase in available substrate due to microbial death and/or destruction of soil aggregates (i.e.commonly known as the Birch effect, Birch, 1964), and (2) physical mechanisms that can influence gas flux, including infiltration, reduced diffusivity, and gas displacement in the soil (e.g., Kim et al., 2012).Consistent with this mechanisms of re-wetting effects in soils of other continents (e.g., Kim et al., 2012), soil CO 2 efflux increased immediately after rainfall in a sub-tropical palm woodland in northern Botswana, however the increase was short-lived (Thomas et al., 2014), while large pulses of CO 2 and N 2 O, followed by a steady decline were also observed after the first rainfall event of the wet season in a Kenyan rainforest (Werner et al., 2007).Soil CO 2 efflux in a South African savannah was strongly stimulated by addition of rainfall (Fan et al., 2015;Zepp et al., 1996) and soil N 2 O concentrations increased markedly 30 min after wetting and peaked between 2 and 5 h after rainfall in a semi-arid savannah (Scholes et al., 1997).In Zimbabwe, the release of N 2 O from dryland savannahs was shown to constitute an important pathway of release for N, and emissions were strongly linked to patterns of rainfall (Rees et al., 2006).In Botswana, Thomas and Hoon (2010) reported large and short-lived pulses of soil CO 2 efflux after artificial wetting of dry soils: soil CO 2 efflux on dry soils was between 2.8-14.8mg C m −2 h −1 but increased to 65.6 mg C m −2 h −1 in the hour after light wetting and 339.2 mg C m −2 h −1 in the hour after heavy wetting.
Forest management such as burning, which is a common practice in SSA, and thinning, affects GHG emissions (Table 2).The IPCC Tier 1 methodology only calculates the amount of GHG emissions as a percentage of the carbon that is released through the burning; however it may also increase forest soil GHG emissions once the fire has Figures passed.For example, soil CO 2 efflux immediately increased after burning of woodland in Ethiopia (Andersson et al., 2004); also, five days after burning rainfall resulted in a 2-fold increase in soil CO 2 efflux from the burned plots compared to the unburned plots.In contrast, 12 days after burning soil CO 2 efflux was 21 % lower in the burned plots (Andersson et al., 2004).However, contrasting impacts of fire on soil GHG emission were observed in a savannah/grassland in the Republic of Congo where fire did not change soil CO 2 , CH 4 and N 2 O fluxes (Castaldi et al., 2010;Delmas et al., 1991).
Similarly, in South Africa, soil CH 4 efflux was not significantly affected by burning (Zepp et al., 1996).In contrast, annual fires decreased soil CH 4 oxidation rates in a Ghanaian savannah (Prieme and Christensen, 1999).These case studies demonstrate that fire impacts are not always consistent and this is likely the result of different fire characteristics (e.g., intensity or frequency), soil type (e.g., Kulmala et al., 2014;Kim et al., 2011) and post-fire weather conditions.Thinning forest cover also can increase soil CO 2 efflux.Yohannes et al. (2013) reported 24 and 14 % increases in soil CO 2 efflux in the first and second years following thinning of a 6 year old Cupressus lusitanica plantation in Ethiopia.
Termite mounds are known sources of CH 4 and CO 2 , and a study in a Burkina Faso savannah found that CH 4 and CO 2 released by termites (Cubitermes fungifaber) contributed 8.8 and 0.4 % of total soil CH 4 and CO 2 emissions, respectively (Brümmer et al., 2009).In Cameroon, the mounds of soil-feeding termites (Thoracotermes macrothorax and Cubitermes fungifaber) were point sources of CH 4 ranging 53.4 to 636 ng s −1 mound −1 , which at the landscape scale may exceed the general sink capacity of the soil (Macdonald et al., 1998).In Zimbabwe, it was found that Odontotermes transvaalensis termite mounds located in dambos (seasonal wetlands) were an important source of GHGs, and emissions varied with catena position for CO Makgadikgadi salt pan in Botswana.Annual CO 2 emissions in salt pan were estimated as 0.7 Mg CO 2 ha −1 yr −1 (Thomas et al., 2014).
In the Nyong River (Cameroon), CO 2 emissions (5.5 kg CO 2 m −2 yr −1 ) were four times greater than the flux of dissolved inorganic carbon (Brunet et al., 2009).In Ivory Coast, three out of five lagoons were oversaturated in CO 2 during all seasons and all were CO 2 sources (3.1-16.2g CO 2 m −2 d −1 ) due to net ecosystem heterotrophy and inputs of riverine CO 2 rich waters (Koné et al., 2009).In the flooded forest zone of the Congo River basin (Republic of Congo) and the Niger River floodplain (Mali), high CH 4 emissions (5.16 × 10 20 -6.35 × 10 22 g CH 4 m −2 d −1 ) were recorded on flooded soils (Tathy et al., 1992;Delmas et al., 1991).In Zimbabwe, dambos can be major or minor sources of GHGs depending on catena position.Upland dambos were important sources of N 2 O and CO 2 , and a sink for CH 4 ; while those in a mid-slope position were a major source of CH 4 , but a weak source of CO 2 and N 2 O; and those at the bottom were a weak source for all GHGs (Nyamadzawo et al., 2014a).In the Congo Basin (Republic of Congo), streams and rivers in savannah regions had higher CO 2 emissions (46.8-56.4g CO 2 m −2 d −1 ) than swamps (13.7-16.3g CO 2 m −2 d −1 ) and tropical forest catchments (37.9-62.9g CO 2 m −2 d −1 ) (Mann et al., 2014).In the Okavango Delta (Botswana), the average CH 4 flux in river channels (0.75 g CH 4 m −2 d −1 ) was higher than that in floodplains and lagoons (0.41-0.49g CH 4 m −2 d −1 ) (Gondwe and Masamba, 2014).In the Zambezi River (Zambia), while CO 2 and CH 4 concentrations in the main channel were highest downstream of the floodplains, N 2 O concentrations were lowest downstream of the floodplains (Teodoru et al., 2015).Overall, 38 % of the total C in the Zambezi River is emitted into the atmosphere, mostly as CO 2 (98 %) Figures

Back Close
Full ( Teodoru et al., 2015).A recent study of 10 river systems in SSA estimated water-air CO 2 , CH 4 and N 2 O fluxes to be 8.2 to 66.9 g CO 2 m −2 d −1 , 0.008 to 0.46 g CH 4 m −2 d −1 , and 0.09 to 1.23 mg N 2 O m −2 d −1 , respectively (Borges et al., 2015).The authors suggested that lateral inputs of CO 2 from soils, groundwater and wetlands were the largest contributors of the CO 2 emitted from the river systems (Borges et al., 2015).
The concentration and flux of GHGs are strongly linked to stream discharge but clear patterns have not yet been identified.In the Congo River, surface CO 2 flux was positively correlated with discharge (Wang et al., 2013), while in Ivory Coast, rivers were often oversaturated with CO 2 and the seasonal variability of partial pressure of CO 2 (pCO 2 ) was due to dilution during the flooding period (Koné et al., 2009).Similarly, CO 2 fluxes show a very pronounced seasonal pattern strongly linked to hydrological conditions in the Oubangui River in the Central African Republic (Bouillon et al., 2012).Although higher CH 4 concentrations were found during low-discharge conditions, N 2 O concentrations were lowest during low-discharge conditions (Bouillon et al., 2012).In the Zambezi River (Zambia), inter-annual variability was relatively large for CO 2 and CH 4 and significantly higher concentrations were measured during wet seasons (Teodoru et al., 2015).However, inter-annual variability of N 2 O was less pronounced and generally higher values were found during the dry season (Teodoru et al., 2015).
The relationship between GHG fluxes from aquatic systems and water temperature is not clear.In the Okavango Delta (Botswana), CH 4 emissions were highest during the warmer, summerrainy season and lowest during cooler winter season suggesting the emissions were probably regulated by water temperature (Gondwe and Masamba, 2014).However, Borges et al. (2015) found no significant correlation between water temperature and pCO 2 and dissolved CH 4 and N 2 O in 11 SSA river systems, although there was a positive relationship between pCO 2 and dissolved organic C in six of the rivers.They also found the lowest N 2 O values were observed at the highest pCO 2 and lowest %O 2 levels, suggesting the removal of N 2 O by denitrification (Borges et al., 2015).Introduction

Conclusions References
Tables Figures

Back Close
Full 3.2 Greenhouse gas emissions from agricultural lands
The effects of the amount and type of N input on N 2 O emissions in croplands have been studied in several locations (Table 2).In western Kenya, the rate of N fertilizer application (0 to 200 kg N ha −1 ) had no significant effect on N 2 O emissions (620 to 710 g N 2 O-N ha −1 for 99 days) (Hickman et al., 2014), however another study from western Kenya, found a relationship between N input and N 2 O emissions that was best described by an exponential model with the largest impact on N 2 O emissions occurring when N inputs increased from 100 to 150 kg N ha −1 (Hickman et al., 2015).
An incubation study in Madagascar demonstrated that application of mixed urea and diammonium-phosphate resulted in lower N 2 O emissions (28 vs. 55 ng N 2 O-N g −1 h −1 for 28 days, respectively) than a mixed application of urea and NPK fertilizer (Rabenarivo et al., 2014).Incorporation of crop residues to the soil has frequently been proposed to increase soil fertility (Malhi et al., 2011), however incorporation of crop residues also affects CO 2 and N 2 O emissions (Table 2).In Tanzania, incorporation of plant residue into soil increased annual CO 2 fluxes substantially (emissions rose from 2.5 to 4.0 and 2.4 to 3.4 Mg C ha −1 yr −1 for clay and sand soils, respectively) (Sugihara et al., 2012), although a study in Madagascar showed that rice-straw residue application resulted in larger fluxes of CO 2 but reduced N 2 O emissions due to N immobilization (Rabenarivo et al., 2014).In contrast, application of Tithonia diversifolia (tithonia) leaves led to greater N 2 O emissions compared to urea application in maize fields in Kenya (Sommer et al., 2015;Kimetu et al., 2007).The higher N 2 O emissions after application of Introduction

Conclusions References
Tables Figures

Back Close
Full Tithonia diversifolia were attributed to high levels of nitrate and available carbon in the soil caused by the application that subsequently enhanced denitrification rates.In incubation studies with cultivated soil from Ghana, N 2 O emissions were significantly higher from soils amended with low C : N ratio clover residues compared to high C : N ratio barley residues (Frimpong et al., 2012) and increasing the proportion of maize in a cowpea-maize residue significantly decreased N 2 O emissions compared to cowpea residue incorporation alone (Frimpong et al., 2011), again likely due to the higher C : N ratio of the maize residue compared with the cowpea.Another incubation study with cultivated soil from Ghana showed that N 2 O emissions increased after addition of residues of three tropical plant species (Vigna unguiculata, Mucuna pruriens and Leucaena leucocephala) and emissions were positively correlated with the residue C : N ratio, and negatively correlated with residue polyphenol content, polyphenol : N ratio and (lignin + polyphenol) : N ratio (Frimpong and Baggs, 2010).It is rare for N 2 O emissions to be positively correlated to C : N ratio and the authors of the study suggest that it was either because soil C was limiting denitrification rates or that release of N from the residues was slow (Frimpong and Baggs, 2010).The results demonstrate that the quality of residues (e.g., C : N ratio, N, lignin and soluble polyphenol contents) affect GHG emissions and further studies are needed to clearly identify the relationship between them (Snyder et al., 2009;Mafongoya et al., 1997).
Adding an additional source of N (mineral or organic) when crop residues are incorporated into the soil could stimulate mineralization of crop residues, increase N-use efficiency and produce higher yields (e.g., Garcia-Ruiz and Baggs, 2007) (Table 2).It was found that application of mixed crop residue or manure and inorganic fertilizers resulted in different response of CO 2 and N 2 O emissions.In maize (Zea mays L.) and winter wheat (Triticum aestivum L.) fields in Zimbabwe, application of inorganic fertilizer (ammonium nitrate, NH 4 NO 3 -N) with manure increased CO 2 emissions (26 to 73 %), compared to sole application of manure (Nyamadzawo et al., 2014a).However, the mixed application resulted in lower N 2 O emissions per yield (1.6-4.6 g N 2 O kg −1 yield), compared to sole application of inorganic fertilizer (6-14 g N 2 O kg −1 yield) (Nya-Introduction

Conclusions References
Tables Figures

Back Close
Full  , 2014a).Similarly, in a maize field in Zimbabwe, N 2 O emissions were lower after the application of composted manure and inorganic fertilizer (NH 4 NO 3 -N) compared to sole application of inorganic fertilizer.The same treatments, however, led to the opposite results for CO 2 emissions (Mapanda et al., 2011).In Mali, pearl millet (Pennisetum glaucum) fields treated with both manure and inorganic fertilizer urea emitted significantly less N 2 O than plots receiving only urea fertilizer (Dick et al., 2008).The lower N 2 O emissions in soils amended with manure were attributed to the initial slow release and immobilisation of mineral N and the consequently diminished pool of N available to be lost as N 2 O (Nyamadzawo et al., 2014a, b;Mapanda et al., 2011;Dick et al., 2008).In an incubation study with cultivated soils from Zimbabwe, Ghana and Kenya, combining organic residue (maize, calliandra, and tithonia) and urea fertilizers decreased N 2 O emissions in coarse-textured soils but it increased N 2 O emissions in fine-textured soils due to the higher level of available N (Gentile et al., 2008).
The effects of crop type and management on GHG emissions have also been studied by several groups (Table 2).In Uganda, there were no significant differences in soil CO 2 effluxes from different crops (lettuces, cabbages, beans) (Koerber et al., 2009).However, in Zimbabwe, rape production resulted in greater N 2 O emissions (0.64-0.93 % of applied N was lost as N 2 O) than tomatoes (0.40-0.51 % of applied N was lost as N 2 O) (Masaka et al., 2014).The results suggest that the effect of crop type on GHG emissions is difficult to predict and more research is needed to elucidate the relationship between crops, crop management and GHG emissions.
In Mali, growing N-fixing haricot beans in rotation did not significantly increase N 2 O emissions (Dick et al., 2008).In Madagascar, N 2 O emissions were not significantly affected by management practices such as direct seeding mulch-based cropping and traditional hand-ploughing after harvesting (Chapuis-Lardy et al., 2009).However, the authors admitted the lack of difference between treatments may be partially due to the short duration of the experiment and suggested more complete monitoring to validate the observation.In highland Tanzanian maize fields, GHG fluxes were similar from soils under conventional and various conservation agriculture practices (Kimaro et al., Figures Back Close Full 2015).However, when fluxes were yield-scaled the global warming potential (Mg CO 2 eq Mg grain −1 ) was lower from fields with reduced tillage plus mulch and leguminous trees (2.1-3.1) and from fields with reduced tillage plus mulch and nitrogen fertilizer (1.9-2.3)compared to fields under conventional agriculture (1.9-8.3)(Kimaro et al., 2015).
Croplands were found to be both a sink and a source of CH 4 .In Burkina Faso, CH 4 flux rates from croplands ranged from −0.67 to 0.70 kg CH 4 -C ha −1 yr −1 (Brümmer et al., 2009), while in Republic of Congo, CH 4 uptake was observed in cassava and peanut fields and a recently ploughed field (Delmas et al., 1991).However, cropped and fertilised dambos in Zimbabwe were consistently sources of CH 4 (13.4 to 66.7 kg CH 4 ha −1 yr −1 ) (Nyamadzawo et al., 2014b).

Grazing grassland
Only one study measured GHG emissions in grazing grasslands.Thomas (2012) found that soil CO 2 efflux from a Botswana grazing land was significantly higher in sandy soils where the biological soil crust (BSC) was removed and on calcrete where the BSC was buried under sand.The results indicated the importance of BSCs for C cycling in drylands and indicate that intensive grazing, which destroys BSCs through trampling and burial, will adversely affect C sequestration and storage (Thomas, 2012).

Rice paddies
Rice paddies are known to be sources of CH 4 (e.g., Linquist et al., 2012).Experiments measuring GHG emissions in rice paddies were conducted in Kenya (Tyler et al., 1988) and Zimbabwe (Nyamadzawo et al., 2013).In Kenya, the range of δ 13 C in CH 4 for rice paddies was from −57 to −63 ‰ and δ 13 CH 4 did not show any seasonal trend and did not indicate appreciable variability among two different strains of rice (Tyler et al., 1988).In Zimbabwe, intermittently saturated dambo rice paddies were a source of GHG

Vegetable gardens
Greenhouse gas emissions from soils in vegetable gardens in peri-urban areas of Burkina Faso (Lompo et al., 2012) and Niger (Predotova et al., 2010) were much higher than all other land uses, ranging from 73.3 to 132.0 Mg CO 2 ha −1 yr −1 and 53.4 to 177.6 kg N 2 O ha −1 yr −1 (Tables 1 and S1).In Burkina Faso annual CO 2 and N 2 O emissions from the garden soils were 68 to 85 % and 3 to 4 % of total C and N input, respectively (Lompo et al., 2012).The N 2 O EFs (3 to 4 %) were higher than the IPCC default value of 1.0 % for all cropping systems (IPCC, 2006) and the global N 2 O EF of vegetable fields (0.94 %) (Rezaei Rashti et al., 2015).The high N 2 O EFs may be attributed to the large amount of applied N in vegetable gardens (2700-2800 kg N ha −1 yr −1 ) since surplus N will stimulate N 2 O production and also indirectly promote N 2 O production by inhibiting biochemical N 2 O reduction (e.g., Shcherbak et al., 2014;Kim et al., 2013).
In vegetable gardens of Niger, a simple plastic sheet roofing and addition of ground rock phosphate to stored ruminant manure decreased N 2 O gaseous losses by 50 % in comparison to dung directly exposed to the sun (Predotova et al., 2010).The authors argued that a decreased evaporation rate was behind this abating effect.Figures

Back Close
Full

Agroforestry
Soil CO 2 and N 2 O emissions from African agroforestry were 38.6 Mg CO 2 ha −1 yr −1 and 0.2 to 26.7 kg N 2 O ha −1 yr −1 , respectively (Tables 1 and S1).In agroforestry homegardens in Sudan, CO 2 (16.6 Mg CO 2 ha −1 from June to December) and N 2 O emissions (17.3 kg N 2 O ha −1 from June to December) accounted for two-thirds of total C output and one-third of total N output, respectively and the CO 2 and N 2 O fluxes were positively correlated with soil moisture (Goenster et al., 2015).Improving fallow with N-fixing trees is a common practice in several areas of Africa since it provides additional N to the soil that can be utilised by the subsequent cash crop (e.g., Makumba et al., 2007;Chikowo et al., 2004;Dick et al., 2001).However, the practice is also thought to increase CO 2 and N 2 O emissions compared to conventional croplands (Table 2).In an intercropping system with a N-fixing tree (Gliricidia) and maize in southern Malawi, soil C was depleted as a result of enhanced CO 2 emissions, with over 67 % of soil C lost over the first 7 years of intercropping (Kim, 2012).In Zimbabwe, N 2 O emissions in improved-fallow agroforestry systems were 7 times higher than emissions in maize monoculture (Chikowo et al., 2004).In Senegal, soil collected under the Nfixing tree (Acacia raddiana) emitted significantly more N 2 O than soil collected under the N-fixing crop (Arachis hypogaea) and non-N fixing tree (Eucalyptus camaldulensis) (Dick et al., 2006).In western Kenya, N 2 O emissions increased after incorporation of fallow residues and emissions were higher after incorporation of improved-fallow legume residues than natural-fallow residues (Baggs et al., 2006;Millar and Baggs, 2004;Millar et al., 2004).It was found that N 2 O emissions were positively correlated with residue N content (Baggs et al., 2006;Millar et al., 2004) and negatively correlated with polyphenol content and their protein binding capacity (Millar and Baggs, 2004), soluble C-to-N ratio (Millar and Baggs, 2005) and lignin content (Baggs et al., 2006).
While high residue N content likely leads to more available soil N and consequently increased N 2 O production (Baggs et al., 2006;Millar and Baggs, 2005;Millar et al., 2004), polyphenols and lignins are both resistant to decomposition and could result in N immobilization resulting in less labile soil N and less N 2 O production (Baggs et al., 2006;Millar and Baggs, 2004).The type and quality of plant residue is likely to be an important control factor affecting N 2 O emissions.
As in natural systems, improved fallow with N-fixing trees also results in increased N 2 O emissions following rainfall events.In an incubation experiment in Uganda, N 2 O emissions following simulated rainfall were a least 4 times larger for soils from under N-fixing trees (Calliandra calothyrsus) compared to soils with non-N fixing trees (Grevillea robusta) (Dick et al., 2001).Similarly, in Mali, N 2 O emissions were around six times higher from improved fallow with N-fixing trees (Gliricidia sepium and Acacia colei) following a simulated rainfall event, compared with the emissions from soil under traditional fallow and continuous cultivation (Hall et al., 2006).Replacing traditional natural fallow with improved-fallow systems in the humid tropics of Kenya also increased N 2 O emissions by up to 3.9 kg N 2 O-N ha −1 over a 122 day maize cropping season (Millar et al., 2004).

Greenhouse gas emissions from land use change
Land-use change affects soil GHG emissions due to changes in vegetation, soil, hydrology and nutrient management (e.g., Kim and Kirschbaum, 2015) and the effects of land-use change on soil GHG emissions have been observed in woodlands and savannah.In Zimbabwe, clearing and converting woodlands to croplands increased soil emissions of CO 2 , CH 4 and N 2 O (Mapanda et al., 2012) and soil CO 2 emissions from the converted croplands were higher than Eucalyptus plantations established in former natural woodlands (Mapanda et al., 2010).In Republic of Congo, early-rotation changes in soil CO 2 efflux after afforestation of a tropical savannah with Eucalyptus were mostly driven by the rapid decomposition of savannah residues and the increase in Eucalyptus rhizospheric respiration (Nouvellon et al., 2012).Introduction

Conclusions References
Tables Figures

Back Close
Full 3.4 Summary of greenhouse gas emissions in natural and agricultural lands in Africa

CO 2 emissions
Carbon dioxide emissions ranged from 3.3 to 130.9 Mg CO 2 ha −1 yr −1 in natural terrestrial systems and from −11.9 to 232.0 Mg CO 2 ha −1 yr −1 in aquatic systems.The area weighted average was 27.6±17.2Mg CO 2 ha −1 yr −1 (Tables 1 and S1).Aquatic systems such as water bodies or water submerged lands were the largest source of CO 2 followed by forest, savannah, termite mounds and salt pans (Table 1).Soil CO 2 emissions in agricultural lands were similar to emissions from natural lands and ranged from 6.5 to 141.2 Mg CO 2 ha −1 yr −1 with an area weighted average of 23.0 ± 8.5 Mg CO 2 ha −1 yr −1 (Tables 1 and S2).Vegetable gardens were the largest sources of CO 2 emission largely due to the large C inputs, followed by agroforestry, cropland and rice fields (Tables 1 S2).

N 2 O emissions and emission factor (EF)
Nitrous oxide emissions in natural lands ranged from −0.1 to 13.7 kg N 2 O ha −1 yr −1 and the area weighted average was 2.5 ± 0.8 kg N 2 O ha −1 yr −1 (Tables 1 and S1).Our study Introduction

Conclusions References
Tables Figures

Back Close
Full reveals that forest, plantation and woodland were the largest source of N 2 O followed by rivers and wetlands, savannah and termite mounds (Table 1).Soil N 2 O emissions in agricultural lands ranged from 0.051 to 177.6 kg N 2 O ha −1 yr −1 and the area weighted average was 4.5 ± 2.2 kg N 2 O ha −1 yr −1 (Tables 1 and S2).The largest N 2 O source in agricultural lands was vegetable gardens followed by agroforestry, cropland and rice fields (Table 1).The N 2 O EF was 0.5±0.2% and 3.5±0.5 % for cropland and vegetable gardens, respectively (Tables 1 and S1).The results indicate that the N 2 O EF of African cropland is lower and the N 2 O EF of African vegetable gardens is higher than IPCC default N 2 O EF (1 %, IPCC, 2006).
The relationship between N input and N 2 O emissions varied depending on N input level (Fig. 4).N 2 O emissions increase slowly up to 150 kg N ha −1 yr −1 , after which emissions increase exponentially up to 300 kg N ha −1 yr −1 (Fig. 5a).Consistent with van Groenigen (2010) N inputs of over 300 kg N ha −1 yr −1 resulted in an exponential increase in emission (Fig. 5b), slowing to a steady state with N inputs of 3000 kg N ha −1 yr −1 .Overall, the relationship between N input and N 2 O emissions shows a sigmoidal pattern (Fig. 5c).The observed relationship is consistent with the proposed hypothetical conceptualization of N 2 O emission by Kim et al. (2013) showing a sigmoidal response of N 2 O emissions to N input increases.The results suggest that N inputs over 150 kg N ha −1 yr −1 may cause an abnormal increase of N 2 O emissions in Africa.The relationship between N input and N 2 O emissions show that the lowest yield-scaled N 2 O emissions were reported for N application rates ranging from 100 to 150 kg N ha −1 (Fig. 6).The results are in line with the global meta-analysis of Philiber et al. ( 2012) who showed that from an N application rate ∼ 150 kg N ha −1 the increase in N 2 O emissions is not linear but exponential.

CO 2 eq emission
Carbon dioxide eq emission (including CO 2 , CH 4 and N 2 O) in natural lands ranged from 11.7 to 121.3 Mg CO 2 eq ha −1 yr −1 and the area weighted average of CO 2 eq emissions Introduction

Conclusions References
Tables Figures

Back Close
Full (excluding salt pans) was 29.9 ± 22.5 Mg CO 2 eq ha −1 yr −1 (Table 1).Water bodies or water submerged lands such as rivers and wetlands were the largest source of CO 2 eq emissions followed by forest/plantation/woodland, savannah/grassland and termite mounds (Table 1).Carbon dioxide eq emissions in agricultural lands ranged from 7.3 to 26.1 Mg CO 2 eq ha −1 yr −1 and had an area weighted average of CO 2 eq emissions (excluding vegetable gardens and agroforestry) of 25.6 ± 12.4 Mg CO 2 eq ha −1 yr −1 (Table 1).
Total CO 2 eq emissions in natural lands (excluding salt pans) were 43.4 ± 9.3 Pg CO 2 eq yr −1 with forest/plantation/woodland the largest source followed by savannah/grassland, stream/river, wetlands/floodplains/lagoons/reservoir, and termite mounds (Table 1).Total CO 2 eq emissions in agricultural lands (excluding vegetable gardens and agroforestry) were 13.5 ± 3.4 Pg CO 2 eq yr −1 with crop land the largest source followed by rice fields (Table 1).Overall, total CO 2 eq emissions in natural and agricultural lands were 56.9 ± 12.7 Pg CO 2 eq yr −1 with natural and agricultural lands contributing 76.3 and 23.7 %, respectively.

Suggested future studies
Despite an increasing number of published estimates of GHG emissions in the last decade, there remains a high degree of uncertainty about the contribution of AFOLU to emissions in SSA.To address this and reduce the uncertainty surrounding the estimates, additional GHG emission measurements across agricultural and natural lands throughout Africa are urgently required.Identifying controlling factors and their effects on GHG fluxes is a pre-requisite to enhancing our understanding of efflux mechanisms and a necessary step towards scaling up the field-scale data to landscape, national and continental scales.It is important to know how GHG fluxes can be affected by management practices and natural events such as logging (e.g., Yashiro et al., 2008), thinning (e.g., Yohannes et al., 2013), storms (e.g., Vargas, 2012), pest outbreaks (e.g., Reed et al., 2014), fires (e.g., Andersson et al., 2004), and wood encroachment (e.g., Smith Introduction

Conclusions References
Tables Figures

Back Close
Full and Johnson, 2004) in natural terrestrial systems and changing discharge (e.g., Wang et al., 2013) and water table (e.g., Yang et al., 2013) in aquatic systems.It is also important in agricultural lands to know how GHG fluxes are affected by management factors such as soil compaction (e.g., Ball et al., 1999), tillage (e.g., Sheehy et al., 2013), removal of crop residues (Jin et al., 2014), incorporation of crop residues and synthetic fertilizer (e.g., Nyamadzawo et al., 2014a), N input (e.g., Hickman et al., 2015) and crop type (e.g., Masaka et al., 2014).However, because management and soil physical/chemical interactions cause different responses in soil GHG emissions (e.g.Pelster et al., 2012), it is critical to measure these interaction effects in the African context.The effect of predicted climatic change in Africa such as increased temperature (e.g., Dijkstra et al., 2012), changing rainfall patterns (e.g., Hall et al., 2006), increase in droughts incidence (e.g., Berger et al., 2013), rewetting effects (e.g., Kim et al., 2012) and increased atmospheric CO 2 concentration (e.g., Lane et al., 2013) also require further testing using laboratory and field experiments.
Where possible studies should seek to identify and separate driving processes contributing to efflux of soil CO 2 (e.g., autotrophic and heterotrophic sources), CH 4 (e.g., methanogenesis and methanotrophy) and N 2 O (e.g., nitrification, denitrification, nitrifier denitrification).This is important because the consequences of increasing GHG emissions depend on the mechanism responsible.For example, if greater soil CO 2 efflux is primarily due to autotrophic respiration from plant roots, then it simply reflects greater plant growth.If however, it is due to heterotrophic microbial respiration of soil organic carbon then it represents a depletion of soil organic matter and a net transfer of C from soil to the atmosphere.Currently there are very few studies that differentiate these sources making it impossible to truly determine the consequences and implications on changes in soil GHG efflux.
Land-use change has been recognized as the largest source of GHG emission in Africa (Valentini et al., 2014).Hence, various types of conversion from natural lands to different land-use types should be assessed to know how these changes may affect the GHG budget (e.g., Kim and Kirschbaum, 2015).The focus of the assessment should be Figures on deforestation and wetland drainage, followed by a conversion to agricultural lands, since they are dominant types of land-use change in Africa (Valentini et al., 2014).

Strategic approaches for data acquisition
A strategic plan for acquisition of soil GHG emission data in sub-Saharan Africa is required.The success of any plan is dependent on long-term investment, stakeholder involvement, technical skill and supporting industries, which have not always been available in the region (Olander et al., 2013;Franks et al., 2012).A major challenge is to address the lack of consistency in the various methodologies used to quantify GHG emissions (Rosenstock et al., 2013).Relatively low cost and simple techniques can be used to determine GHG emission estimates in the first instance.Soil CO 2 fluxes can be quantified with a soda lime method (Tufekcioglu et al., 2001;Cropper et al., 1985;Edwards, 1982) or an infra-red gas analyzer (Bastviken et al., 2015;Verchot et al., 2008;Lee and Jose, 2003) and these do not require advanced technology or high levels of resource to undertake.Later, other GHG such as N 2 O and CH 4 fluxes in addition to CO 2 flux can be measured with more advanced technology (e.g., gas chromatography, photo-acoustic spectroscopy, or laser gas analyzers).Initially, the measurement can be conducted using manual gas chambers with periodical sampling frequencies.The sampling interval can be designed so that it is appropriate to the particular type of landuse or ecosystem, management practices and/or for capturing the effects of episodic events (e.g., Parkin, 2008).For example, GHG measurement should be more during potentially high GHG emission periods following tillage and fertilizer applications and rewetting by natural rainfalls or irrigation.With more advance technology and utilisation of automatic chamber systems measurements can be conducted at a much high frequency with relative ease.In order for the challenges associated with improving our understanding of GHG emissions from African soils it is critical to establish networks of scientists and scientific bodies both within Africa and across the world.Good communication and collaboration between field researchers and the modelling community should also be established 16503 Figures

Back Close
Full during the initial stages of research, so results obtained from field scientists can be effectively used for model development and to generate hypotheses to be tested in the field and laboratory (de Bruijn et al., 2009).Furthermore, lessons learned from scientific experiments can only really be successfully implemented by farmers if local stakeholders are involved from the start and throughout (see for example Stringer et al., 2012).Interviews, focus-groups, on-site or farm demonstrations, local capacity building training, local farmers and extension staff can all improve dialogue and understanding between local communities and scientists, ultimately improving the likelihood of successful GHG emission and mitigation strategies.These will equip local researchers and stakeholders (including farmers and extension staff) with state of art methodologies and help motivate them to develop their GHG mitigation measures and assist them in understand their roles and contributions to global environmental issues.

Conclusions
This paper synthesizes the available data on GHG emissions from African agricultural and natural lands.Emissions of CO 2 , CH 4 and N 2 O in a variety of environments (forests, savannahs, termite mounds, salt pans, agricultural areas and water bodies) were considered.Two broad conclusions can be drawn from the work.The first one is that African natural and agricultural lands may be a significant source of GHG and that the emissions may increase through land-use change and management strategies.
Secondly, there are huge research gaps.Africa is a vast continent, with a multitude of land uses, climates, soils and ecosystems.Field-based data on soil GHG emissions from many areas, soil types and environments are extremely sparse and as a result our understanding of Africa's contribution to global GHG emissions remains incomplete and highly uncertain.There is an urgent need to develop and agree on a strategy for addressing this data gap.The strategy may involve identifying priorities for data Figures

Back Close
Full  Full  Full Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | CO 2 , 201 CH 4 and 184 N 2 O emissions measurements.Where N 2 O emission studies included experimental data from control plots with no N fertilizer additions (i.e. for background N 2 O emissions) and from plots with different levels of applied N, a N 2 O emission factor (EF) was calculated following the IPCC (2006) Tier I methodology as Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Savannah and grassland were found to Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | madzawo et al.
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Zepp, R. G., Miller, W. L., Burke, R. A., Parsons, D. A. B., and Scholes, M. C.: Effects of moisture and burning on soil-atmosphere exchange of trace carbon gases in a southern African savanna, J. Geophys.Res., 101, 23699-23706, 1996Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

blog for open discussion and web based open databases
We have created a blog entitled "Greenhouse gas emissions in Africa: study summary and database" (http://ghginafrica.blogspot.com/)and an open-access database, which can be modified by the users, entitled "Soil greenhouse gas emissions in Africa database" (linked in the blog) based on this review.In the blog, we have posted a technical summary of each section of this review, where comments can be left under the posts.The database contains detailed information on the studies reported on GHG emissions, such as ecosystem and land use types, location, climate, vegetation type, crop type, fertilizer type, N input rate, soil properties, GHGs emission measurement periods, N 2 O EF, and corresponding reference.The database is hosted in web based spreadsheets and is easily accessible and modified.The authors do not have any relationship with the companies currently being used to host the blog and databases.

Table 1 .
Summary of greenhouse gas carbon dioxide (CO 2 ), methane (CH 4 ), nitrous oxide (N 2 O) emissions and CO 2 equivalents (CO 2 eq) in natural and agricultural lands in sub-Saharan African countries.Mean ± standard error (number of data) are shown.Except salt pan, vegetable gardens and agroforestry.