Volume 38, Issue 4 e2023GB008016
Research Article
Open Access

The African Regional Greenhouse Gases Budget (2010–2019)

Yolandi Ernst

Corresponding Author

Yolandi Ernst

Global Change Institute, University of the Witwatersrand, Johannesburg, South Africa

Correspondence to:

Y. Ernst and S. Archibald,

[email protected];

[email protected]

Contribution: Conceptualization, Methodology, Formal analysis, ​Investigation, Writing - original draft, Writing - review & editing, Project administration

Search for more papers by this author
Sally Archibald

Corresponding Author

Sally Archibald

School of Animal, Plant and Environmental Sciences, University of the Witwatersrand, Johannesburg, South Africa

Correspondence to:

Y. Ernst and S. Archibald,

[email protected];

[email protected]

Contribution: Methodology, Formal analysis, ​Investigation, Writing - original draft, Writing - review & editing, Project administration

Search for more papers by this author
Heiko Balzter

Heiko Balzter

Institute for Environmental Futures, School of Geography, Geology and the Environment, University of Leicester, Space Park Leicester, Leicester, UK

National Centre for Earth Observation, University of Leicester, Space Park Leicester, Leicester, UK

Contribution: Methodology, Formal analysis, Resources, Writing - review & editing

Search for more papers by this author
Frederic Chevallier

Frederic Chevallier

Laboratoire des Sciences du Climat et de l’Environnement, LSCE/IPSL, CEA-CNRS-UVSQ, Université Paris-Saclay, Gif-sur-Yvette, France

Contribution: Resources, Writing - original draft

Search for more papers by this author
Philippe Ciais

Philippe Ciais

Laboratoire des Sciences du Climat et de l’Environnement, LSCE/IPSL, CEA-CNRS-UVSQ, Université Paris-Saclay, Gif-sur-Yvette, France

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Carlos Gonzalez Fischer

Carlos Gonzalez Fischer

Department of Global Development, College of Agriculture and Life Sciences, Cornell University, Ithaca, NY, USA

Contribution: ​Investigation, Resources

Search for more papers by this author
Benjamin Gaubert

Benjamin Gaubert

Atmospheric Chemistry Observations & Modeling Laboratory (ACOM), NSF National Center for Atmospheric Research (NSF NCAR), Boulder, CO, USA

Contribution: Formal analysis, Writing - original draft

Search for more papers by this author
Thomas Higginbottom

Thomas Higginbottom

School of GeoSciences, University of Edinburgh, Edinburgh, UK

Contribution: Resources, Writing - original draft

Search for more papers by this author
Steven Higgins

Steven Higgins

Plant Ecology, University of Bayreuth, Bayreuth, Germany

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Shakirudeen Lawal

Shakirudeen Lawal

Department of Forestry and Environmental Resources, College of Natural Resources, North Carolina State University, Raleigh, NC, USA

Contribution: Resources, Writing - original draft

Search for more papers by this author
Fabrice Lacroix

Fabrice Lacroix

Climate and Environmental Physics, University of Bern, Bern, Switzerland

Oeschger Centre for Climate Change Research (OCCR), University of Bern, Bern, Switzerland

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Ronny Lauerwald

Ronny Lauerwald

INRAE, AgroParisTech, UMR ECOSYS, Université Paris-Saclay, Palaiseau, France

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Mauro Lourenco

Mauro Lourenco

School of Animal, Plant and Environmental Sciences, University of the Witwatersrand, Johannesburg, South Africa

National Geographic Okavango Wilderness Project, Wild Bird Trust, Johannesburg, South Africa

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Carola Martens

Carola Martens

Senckenberg Biodiversity and Climate Research Centre (SBiK-F), Frankfurt am Main, Germany

Institute of Physical Geography, Goethe University Frankfurt am Main, Frankfurt am Main, Germany

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Anteneh G. Mengistu

Anteneh G. Mengistu

Finnish Meteorological Institute, Helsinki, Finland

Contribution: Formal analysis, Writing - original draft

Search for more papers by this author
Lutz Merbold

Lutz Merbold

Integrative Agroecology Group, Strategic Research Division Agroecology and Environment, Agroscope, Zurich, Switzerland

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Edward Mitchard

Edward Mitchard

School of GeoSciences, King's Buildings, University of Edinburgh, Edinburgh, UK

Contribution: Resources, Writing - original draft

Search for more papers by this author
Mthokozisi Moyo

Mthokozisi Moyo

School of Animal, Plant and Environmental Sciences, University of the Witwatersrand, Johannesburg, South Africa

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Hannah Nguyen

Hannah Nguyen

Department of Geography, King's College London Strand, London, UK

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Michael O’Sullivan

Michael O’Sullivan

Faculty of Environment, Science and Economy, University of Exeter, Exeter, UK

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Pedro Rodríguez-Veiga

Pedro Rodríguez-Veiga

Sylvera Ltd, London, UK

Centre for Landscape and Climate Research, School of Geography, Geology and the Environment, University of Leicester, Leicester, UK

Contribution: Methodology, Resources, Writing - review & editing

Search for more papers by this author
Thais Rosan

Thais Rosan

College of Life and Environmental Sciences, University of Exeter, Exeter, UK

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Judith Rosentreter

Judith Rosentreter

Faculty of Science and Engineering, Southern Cross University, Lismore, NSW, Australia

Contribution: Resources, Writing - original draft

Search for more papers by this author
Casey Ryan

Casey Ryan

School of GeoScience, University of Edinburgh, Edinburgh, UK

Contribution: Resources, Writing - original draft

Search for more papers by this author
Simon Scheiter

Simon Scheiter

Senckenberg Biodiversity and Climate Research Centre (SBiK-F), Frankfurt am Main, Germany

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Stephen Sitch

Stephen Sitch

College of Life and Environmental Sciences, University of Exeter, Exeter, UK

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Nicola Stevens

Nicola Stevens

School of Animal, Plant and Environmental Sciences, University of the Witwatersrand, Johannesburg, South Africa

Environmental Change Institute, School of Geography and the Environment, University of Oxford, Oxford, UK

Contribution: Resources, Writing - original draft

Search for more papers by this author
Torbern Tagesson

Torbern Tagesson

Department of Physical Geography and Ecosystem Science, Lund University, Lund, Sweden

Department of Geosciences and Natural Resource Management, University of Copenhagen, Copenhagen, Denmark

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Hanqin Tian

Hanqin Tian

Center for Earth System Science and Global Sustainability, Schiller Institute for Integrated Science and Society, Department of Earth and Environmental Sciences, Boston College, Chestnut Hill, MA, USA

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Mengjia Wang

Mengjia Wang

School of Geoscience and Technology, Zhengzhou University, Zhengzhou, China

INRAE, UMR1391 ISPA, Université de Bordeaux, Villenave d'Ornon, France

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Joel S. Woon

Joel S. Woon

School of Environmental Sciences, University of Liverpool, Liverpool, UK

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Bo Zheng

Bo Zheng

Department of Earth System Science, Tsinghua University, Beijing, China

State Key Joint Laboratory of Environment Simulation and Pollution Control, School of Environment, Tsinghua University, Beijing, China

Contribution: Resources, Writing - original draft

Search for more papers by this author
Yong Zhou

Yong Zhou

Department of Wildland Resources, Utah State University, Logan, UT, USA

Ecology Center, Utah State University, Logan, UT, USA

Contribution: Formal analysis, Resources, Writing - original draft

Search for more papers by this author
Robert J. Scholes

Robert J. Scholes

Global Change Institute, University of the Witwatersrand, Johannesburg, South Africa

Contribution: Conceptualization, Formal analysis, Writing - original draft, Project administration

Search for more papers by this author
First published: 03 April 2024

Abstract

As part of the REgional Carbon Cycle Assessment and Processes Phase 2 (RECCAP2) project, we developed a comprehensive African Greenhouse gases (GHG) budget covering 2000 to 2019 (RECCAP1 and RECCAP2 time periods), and assessed uncertainties and trends over time. We compared bottom-up process-based models, data-driven remotely sensed products, and national GHG inventories with top-down atmospheric inversions, accounting also for lateral fluxes. We incorporated emission estimates derived from novel methodologies for termites, herbivores, and fire, which are particularly important in Africa. We further constrained global woody biomass change products with high-quality regional observations. During the RECCAP2 period, Africa's carbon sink capacity is decreasing, with net ecosystem exchange switching from a small sink of −0.61 ± 0.58 PgC yr−1 in RECCAP1 to a small source in RECCAP2 at 0.16 (−0.52/1.36) PgC yr−1. Net CO2 emissions estimated from bottom-up approaches were 1.6 (−0.9/5.8) PgCO2 yr−1, net CH4 were 77 (56.4/93.9) TgCH4 yr−1 and net N2O were 2.9 (1.4/4.9) TgN2O yr−1. Top-down atmospheric inversions showed similar trends. Land Use Change emissions increased, representing one of the largest contributions at 1.7 (0.8/2.7) PgCO2eq yr−1 to the African GHG budget and almost similar to emissions from fossil fuels at 1.74 (1.53/1.96) PgCO2eq yr−1, which also increased from RECCAP1. Additionally, wildfire emissions decreased, while fuelwood burning increased. For most component fluxes, uncertainty is large, highlighting the need for increased efforts to address Africa-specific data gaps. However, for RECCAP2, we improved our overall understanding of many of the important components of the African GHG budget that will assist to inform climate policy and action.

Key Points

  • Estimates of termite, herbivore, and fire emissions from novel methods

  • Global woody biomass products constrained with high quality local data

  • Africa a net source (approximately carbon neutral) between 2010 and 2019, sink capacity decreasing

Plain Language Summary

We developed a comprehensive greenhouse gases (GHG) budget for Africa as part of the REgional Carbon Cycle Assessment and Processes Phase 2 (RECCAP2) project over the 2010–2019 period. We used global and local data sets and innovative methods to estimate the different components of the budget. Our estimates show that wildfire emissions decreased; termite emissions may be less than previously expected and emissions from large mammals are increasing. We also used data from new satellite technology to estimate carbon that is stored in above-ground biomass in Africa. With increasing land use change and fossil fuel usage in Africa, the net bottom-up GHG estimate shows that Africa is a source at 4.5 (−3.3/14.1) PgCO2eq yr−1, with the top-down atmospheric inversion estimate smaller at 3.98 (3.13/4.85) PgCO2eq yr−1. However, our estimates continue to have large uncertainties owing to the differences between data sets and methods. It is therefore essential to increase efforts to expand the availability of high quality local data. Nevertheless, our work improved our understanding of all the components of the African GHG budget and will help to inform climate policy and action.

1 Introduction

Africa's role in the global greenhouse gases (GHG) cycles is of great interest due both to the large landmass covered by the continent, and the potential for rapid change in coming decades as the human population increases and land use patterns continue to evolve. Africa contains some of the largest tracts of untransformed land in the world, although it is often heavily utilized for grazing, fuelwood and other natural resources. With a current population of about 1.4 billion, set to increase to over 2 billion by 2040 (United Nations Urban Settlement Programme, 2019), it is expected that large areas of land will be converted for agricultural production to feed this increasingly urbanized community and to increase country-level GDP. Concurrently, there is massive interest in using African landscapes to store carbon and offset global carbon emissions (Armani, 2022). It is therefore imperative to develop reliable data on key carbon-cycle processes and GHG emissions to quantify the net effect of these competing trends.

Previous accounting efforts of the African GHG budget estimated the continent as a net biospheric sink but highlighted the large uncertainty associated with an inadequate observation network (Bombelli et al., 2009; Ciais et al., 2011; Valentini et al., 2014; Williams et al., 2007). Moreover, African savannas and woodlands, with seasonal rainfall, frequent fire and large populations of native and introduced herbivores, play a unique and significant role in the inter-annual variability of the continent's GHG fluxes that further contribute to uncertainty in estimates (Bombelli et al., 2009; Valentini et al., 2014).

Modeling studies indicate the risk for rapid and irreversible changes in vegetation cover in response to changing climates and CO2 fertilization (e.g., greening in northern ecosystems and browning in tropical biomes) (Winkler et al., 2021). Field observations further demonstrate both extensive woody thickening as well as areas of reduced productivity in recent years (Stevens et al., 2016). Since the last continental-scale GHG budget for the 1985–2009 period (Valentini et al., 2014), we have seen improved estimations of fire (Andela et al., 2017; Hantson et al., 2016; Lasslop et al., 2020) and herbivore emissions (Hempson et al., 2017; Pachzelt et al., 2015) and better representation of African landscapes and functional types in Dynamic Global Vegetation Models (DGVMs) (e.g., aDGVM—Scheiter & Higgins, 2009). Estimates for other GHG budget components such as inland waters (Borges et al., 2015; Borges, Deirmendjian, Bouillon, Okello, et al., 2022; Lauerwald et al., 2023a) and geological fluxes (Etiope et al., 2019; Lacroix et al., 2020) are also better represented.

The current synthesis of the GHG budget of Africa aims to integrate the most contemporary modeling and observational data sets to present a comprehensive and up to date summary of the key sources and sinks of carbon, CO2, CH4, and N2O greenhouse gases and their associated uncertainties from 2010 to 2019. Where possible, analyses that include the 1985–2009 period are presented for comparison. Due to the limitations imposed by the availability of some data sets, some estimates may represent alternative dates for the RECCAP1 (1985–2009) and RECCAP2 period (2010–2019) but reference periods are defined where necessary.

As part of the Regional Carbon Cycle Assessment and Processes Phase 2 (RECCAP2, https://www.globalcarbonproject.org/reccap/) initiative of the Global Carbon Project (https://www.globalcarbonproject.org/index.htm), this paper addresses the policy-relevant objectives of RECCAP2 through a comprehensive overview of improved estimates of CO2, CH4, and N2O fluxes and variability. In the following sections, we report the methodology and results for various component fluxes and uncertainties for Africa as a whole and for five ecoregions, delineated for interpretive purposes (Figure 1). The structure of the paper includes a section on carbon stocks represented by aboveground (Section 2.1.1) and below-ground (Section 2.1.2) biomass estimates, after which we report on the component fluxes estimated from various bottom-up methods. These broadly include gross and net primary production estimates (Section 2.2); fire, large mammals and termites as fluxes of special importance to Africa (Section 2.3); fluxes from geological, aquatic and coastal systems (Section 2.4); trade fluxes (Section 2.5); and anthropogenic emissions with special focus on fossil fuel emissions (Section 2.6). In Section 2.7, we present the top-down atmospheric inversion model estimates for CO2, CH4, and N2O, followed by a synthesis (Section 3) of all the estimates provided in the preceding sections. Our approach follows the guidelines by Ciais et al. (2022).

Details are in the caption following the image

The Scholes African Ecoregions Map (Ernst & Scholes, 2023) was delineated by regrouping and smoothing the vegetation classification of the UNESCO/AETFAT/UNSO (White's) Vegetation Map of Africa (White, 1983) in accordance with the delineations of the distributions of Mean Annual Precipitation-determined (“stable”) and Disturbance-determined (“unstable”) savannas in Africa by Sankaran et al. (2005).

1.1 Drivers of Change in the African Carbon Cycle

Together with increasing atmospheric CO2, changing climates and land use all impact carbon-cycle processes. African climates have warmed significantly over the last several decades (Engelbrecht et al., 2015), more so in the arid and semi-arid regions and particularly in East Africa. Rainfall has increased on average across all regions (Alahacoon et al., 2022) and variability between years is high and probably increasing. Consequently, aridity trends (as indexed by P/PET) are not uniform, with aridity increasing in East and Southern Africa, and decreasing in West Africa (Lickley & Solomon, 2018). Cropland area has increased, and over the two RECCAP periods Africa gained 7.15 ± 3.39 × 105 km2 new cropland area, and lost 1.83 ± 1.94 × 105 km2, resulting in a net increase of 5.32 ± 3.94 × 105 km2 from 2000 to 2019 (Potapov et al., 2022). Currently 20.83 ± 4.74 × 105 km2 (or ∼17%) of the global cropland area occurs in Africa, but mapping products disagree on whether cropland expansion has slowed in the last decade (see Text S1, Figures S1 and S2, Tables S1 and S2 in Supporting Information S1 for changes estimated by different products). Land use trends are discussed further in Section 2.2.2 on the TRENDY results. We summarize information on changing livestock numbers in Section 2.3.2 and above-ground biomass in Section 2.1.1.

2 African GHG Component Estimates

2.1 Biomass

2.1.1 Aboveground Biomass Change

Since the RECCAP1 period, novel L-VOD passive microwave data (Diouf et al., 2015) and LiDAR-based biomass data (Potapov et al., 2021) have become available. These data have the potential to provide more comprehensive information on AGB changes than estimates derived from changes in land cover as they measure AGB change within the land cover classes. They therefore account both for losses due to degradation and natural disturbance as well as gains from regrowing vegetation and environmental drivers such as CO2-fertilization. These within-land cover changes are important for Africa as land cover conversion is estimated to account for only about 25% of the AGB change on the continent (X. Feng et al., 2021; McNicol et al., 2018). However, although many papers reporting changes in AGB in Africa have been published within the 1985–2019 period, there is no agreement on the regional trends or magnitude of the changes (Text S2 in Supporting Information S1). These disagreements in AGB change estimates are largely due to the varied methods used, which include bookkeeping models, LiDAR-derived products, and various microwave-derived products. However, differences in the observation time periods might also add to the uncertainty due to large inter-annual variation in AGB.

For RECCAP2, we compared five microwave- and VOD-derived AGB change estimates from 2010 to 2017, three of which have been developed and calibrated specifically for Africa. The L-VOD product (Brandt et al., 2018) was calibrated against the Baccini et al. (2012) LIDAR-derived AGB. The X-VOD product (M. Wang et al., 2021) was retrieved from the AMSR2/AMSR-E brightness temperature observations at the X-band, with Saatchi et al. (2011) AGB (LiDAR-derived), (Bouvet et al., 2018) AGB (SAR-derived), GlobBiomass (SAR-derived AGB) and ESA-CCI AGB (SAR-derived AGB) as the calibration references. The National Center for Earth Observation (NCEO) product (Rodríguez-Veiga & Balzter, 2021; Rodríguez-Veiga et al., 2017) uses GEDI canopy-height data and L-band SAR to produce a canopy-height model calibrated against LiDAR-derived biomass data. The global ESA-CCI Biomass product (Santoro et al., 2021) uses both C- and L-band RADAR to estimate growing stock volume, and converts this to AGB using allometric equations from published wood density and biomass expansion data. The updated McNicol et al. (2018) product for southern Africa is focused on accurately estimating changes in non-forest African ecosystems (i.e., in contrast to L-VOD which is also sensitive to high-biomass regions), and trains its product with in situ biomass measurements. All products have potential artifacts from soil moisture and range in spatial resolution from 25 km (Brandt LVOD) to 25 m (McNicol product). More details on the products are available in Table S3 in Supporting Information S1.

For each product, we calculated the annual change as (AGB2017 − AGB2010)/7. As 2017 was the end of a severe multi-year drought in southern Africa (Blamey et al., 2018), the trends might not be reliable, but it is the first time that so many different products have been compared over the same period and regions.

All the products estimate net AGB losses at the scale of Africa, ranging from −71.9 to −309.9 Tg Cyr−1, but there was no consistency in predicted trends across biome classes or regions (Table 1, Figure 2). For example, the ESA-CCI biomass product predicted biomass gains of 44 Tg Cyr−1 in forest but losses of −118 TCyr−1 in sub-humid savannas, and the Brandt L-VOD product showed the opposite trend (forest loss: −20.8 Tg Cyr−1, sub-humid savanna gains: 36.6 Tg Cyr−1). Generally, these estimates are within the range reported by Valentini et al. (2014), but the uncertainty remains high for RECCAP2. Global RADAR and VOD products are currently unlikely to represent the dynamics of African woodlands accurately because they often lack African calibration data, and potentially require locally defined algorithms to represent the lower-biomass dynamics of African woodlands.

Table 1. Estimated Net Aboveground Biomass Annual Change 2010–2017 (in Tg Cyr−1) for Africa and Its Ecoregions
Region 1985–2009 2010–2017
Valentini et al. (2014) CCI NCEO L-VOD (Brandt et al., 2018) X-VOD (M. Wang et al., 2021) McNicol et al. (2018)
NH Desert 0.1 −1.4 −5.9 −3.0
Forest 44.8 −80.2 −20.8 −147.4
Sub-humid savanna −118.6 −63.0 36.2 −92.1
Semi-arid savanna −17.9 −7.5 −71.3 −62.6
Desert/shrubland −0.3 −0.2 −10.0 −4.8
Miombo Ecoregion −98.0 −22.0 −1.0 17.0 3.6
Africa −234 to −72 −92.0 −152.3 −71.9 −309.9
  • Note. Positive values are fluxes into the land-surface (sink); negative values represent loss from the living biomass pool (predominantly into the atmosphere as a source, rather than into the soil). Products ordered from global (left) to regional (right) calibrations. The Miombo Ecoregion was added to include the locally calibrated and developed (McNicol et al., 2018) product and because it is a region of rapid change.
Details are in the caption following the image

Change in aboveground biomass across seven countries in southern Africa for the period 2010–2017 as reported by five different RADAR-derived data products. Positive values are fluxes into the land-surface (sink); negative values represent loss from the living biomass pool (predominantly into the atmosphere as a source, rather than into the soil). There is no clarity on the trends between or within countries, but regionally and locally calibrated products report more sink capacity than globally calibrated products overall.

2.1.2 Belowground Carbon and Biomass

Since the previous synthesis of the African GHG budget, soil organic carbon (SOC) estimates (Table 2) have improved with the ISRIC (International Soil Reference and Information Center) producing soil property maps for the continent at 250 m resolution (Hengl et al., 2015, 2017a). These SoilGrids data (Hengl et al., 2017b) are interpolated from a network of several thousand soil cores and several hundred thousand surface samples, and estimate the SOC of Africa to be 87.7 PgC. Below-ground biomass carbon is poorly constrained and predicted from published root:shoot estimates. Recent quantification of biomass carbon in African grasslands (Gomes et al., 2021) indicates substantial below-ground stocks that are not accurately represented in existing continental-scale studies and are therefore likely to be under-estimates. These maps also still do not accurately map or account for peatlands, which are estimated to contain significant stores of carbon (Joosten, 2009). Currently, peat stocks are estimated at 36.9 PgC (UNEP, 2022), which is ∼3 times higher than previous estimates of ∼11 PgC due to new reserves found in the Congo basin (Dargie et al., 2017), and novel peat mapping methods (Lourenco et al., 2022). Peat loss, largely to the atmosphere, is estimated to be ∼0.013 PgC yr−1 (Joosten, 2009) and is increasing. Below-ground stocks modeled from DGVMs varied from 87.5 to 259.5 PgC in the previous RECCAP period (Valentini et al., 2014). For the RECCAP2 period, aDGVM, a dynamic vegetation model developed for African ecosystems (Scheiter & Higgins, 2009, see also Section 2.2.3), estimates total stocks to be 98.9 PgC, of which SOC is 76.8 Pg and belowground biomass carbon 22.1 Pg. The TRENDY models show a mean SOC of 148 ± 60 Pg and all but three show an increasing trend.

Table 2. Soil Organic Carbon, Peat Carbon Stocks, and Estimated Peat Loss Rates for Africa Per Ecoregion
Ecoregion SOC (Pg) from SoilGrids Peat carbon (Pg) Valentini et al. (2014)a aDGVMb 2009–2019
Total Joosten (2009) UNEP (2022) Loss rate (PgC yr−1) Total below-ground C SOC Biomass C Total belowground C
NA Desert 3.7 2.1 4.33 0.67 5
Forest 15.7 3.6 13.29 3.92 17.21
Desert/shrubland 1.0 0.0 1.03 0.15 1.18
Sub-humid savanna 46.9 4.0 40.98 12.91 53.89
Semi-arid savanna 20.3 1.1 17.15 4.42 21.57
Total 87.7 10.8 36.9 0.013 167 (87–259) 76.77 22.08 98.85
  • a Valentini et al. (2014) model average—including biomass carbon.
  • b aDGVM is a dynamic vegetation model developed for African ecosystems, see Section 2.3.3.

2.2 Gross and Net Primary Production Estimates

2.2.1 Satellite Observation Constrained Gross Primary Productivity Models

We used seven Earth observation based global scale vegetation gross primary productivity (GPP) data sets collected by Tagesson et al. (2021) for estimating Africa's GPP budgets 1985–2015. The contribution of Africa to the mean, trend, and inter-annual variability in the global scale GPP was estimated following Ahlström et al. (2015). The products with their spatial and temporal resolutions and estimates are listed in Table S4 in Supporting Information S1 and described in Tagesson et al. (2017). The average GPP budget for Africa over 1985–2015 was 23.50 ± 0.41 (± one standard deviation of inter-annual variability) ± 2.48 PgC yr−1 (± one standard deviation of model variability) (Table 3), which represents about 20% of the annual global GPP. This is relatively close to the 22.3% share Africa has of the global terrestrial surface area. Satellite observations indicate that the GPP is increasing by 28.60 ± 6.47 ± 33.69 TgC yr−1, over the 1985–2015 period (about 18.2% of the global trend), but the share of Africa in the inter-annual variability in the global GPP budgets was relatively low (6.77 ± 1.13 ± 3.74%).

Table 3. The Gross Primary Productivity Mean, Trend, and Inter-Annual Variability (± One Standard Deviation of Inter-Annual Variability ± Model Variability) From Seven Global Earth Observation Products for Africa and Its Ecoregions for the 1985–2015 Periods
Region Mean GPP (Pg Cyr−1) Trend GPP (TgCyr−2) Contributions (%) of Africa to global GPP budget/Ecoregions to Africa GPP budget
Mean IAV* Trend*
Africa (22.3% of global surface) 23.50 ± 0.41 ± 2.48 28.6 ± 6.47 ± 33.69 20.2 ± 0.4 ± 1.8 6.7 ± 1.1 ± 3.7 15.9 ± 3.6 ± 14.3
NA Desert (34.7% of Africa) 0.31 ± 0.02 ± 0.14 0.79 ± 0.41 ± 1.01 1.29 ± 0.1 ± 0.6 6.9 ± 0.4 ± 3.7 2.7 ± 1.4 ± 4.0
Forests (8.2% of Africa) 5.98 ± 0.06 ± 0.49 2.33 ± 1.05 ± 5.57 24.7 ± 0.2 ± 4.0 36.4 ± 1.2 ± 7.9 8.1 ± 3.7 ± 17.5
Desert/Shrubland (2.4% of Africa) 0.13 ± 0.01 ± 0.06 0.66 ± 0.15 ± 0.37 0.5 ± 0.0± 0.28 4.6 ± 0.1 ± 1.4 2.3 ± 0.5 ± 0.7
Sub-humid savanna (34.0% of Africa) 13.16 ± 0.23 ± 2.38 14.48 ± 3.73 ± 20.52 54.2 ± 0.9 ± 4.5 48.6 ± 4.7 ± 7.9 50.2 ± 12.9 ± 25.3
Semi-arid savanna (20.7% of Africa) 3.21 ± 0.15 ± 0.31 10.59 ± 2.47 ± 7.00 13.2 ± 0.6 ± 0.2 3.4 ± 2.3 ± 7.7 36.7 ± 8.6 ± 12.9

Sub-humid savannas and forests were the main contributors to African GPP, contributing more than 50% and ∼25%, respectively (Table 3). Sub-humid savannas drove both the increasing trends and the inter-annual variability in GPP, with forest GPP being more stable with less strong trends. Semi-arid savannas, which contributed relatively little (3.21 ± 0.15 ± 0.31 PgC yr−1) to the mean African GPP budgets, contributed substantially to the GPP trends (about a quarter of the GPP increases occurred in semi-arid savannas). Semi-arid regions in Africa are steadily becoming encroached with woody vegetation (Venter et al., 2018) and are important in terms of their inter-annual variability (Ahlström et al., 2015). The NH Desert and Desert/Shrubland regions have a very low share (about 1%) of the African GPP budget (Table 3). However, significant NA Desert trends and inter-annual variability (Table 3) indicate considerable changes in the vegetation cover during recent decades likely driven by CO2 fertilization (Song et al., 2018).

The GPP of Africa increased over the period 1985–2015, but the increase slowed down in the last decade (Table 3). This could be caused by the strong drought in southern Africa at the end of the study period in 2015 (Blamey et al., 2018). Other reasons for a slowing down of the GPP trends could be a decrease in the degree to which CO2 is upregulating photosynthesis (fertilization effect) (S. Wang et al., 2020), enhanced constraints from water supply, nutrient limitation, and land cover change (X. Feng et al., 2021; Peñuelas et al., 2013; Piao et al., 2020; Yuan et al., 2019). Still, Africa's contribution to the global GPP budgets are similar for both the RECCAP study periods: forest GPP contribution decreased slightly between RECCAP1 and RECCAP2, with increases in semi-arid savanna compensating for this. The semi-arid savanna also has an increasing GPP trend over 1985–2015 compared to forests, explaining their larger share during the RECCAP2 period.

2.2.2 Ecosystem Model Ensembles Including LUC: Trends in the Land Carbon Fluxes (TRENDY)

Outputs from an ensemble of 17 DGVMs from the TRENDY v.9 model suite were forced with observed changes in climate, CO2 and nitrogen deposition, and Land Use Change (LUC) (Land Use Land Cover Change HYDE3.2 within LUH2-GCB) over the period 1985 to 2019 (Friedlingstein et al., 2020a) (Table 4).

Table 4. Regional Carbon Fluxes (Pg Cyr−1) Decomposed Into the Three Main Drivers; Climate Change (CLIM), CO2 Fertilization (CO2), and Land Use Change (LUC) Over the Last Four Decades
Region Forcing Net ecosystem exchange (NEE PgC yr−1)
1980s 1990s 2000s 2010s
Africa CLIM 0.33 ± 0.21 0.16 ± 0.12 0.21 ± 0.13 0.00 ± 0.15
CO2 −0.41 ± 0.17 −0.39 ± 0.18 −0.56 ± 0.21 −0.55 ± 0.24
LUC 0.18 ± 0.12 0.22 ± 0.13 0.28 ± 0.1 0.46 ± 0.15
NET 0.10 ± 0.19 −0.01 ± 0.20 −0.07 ± 0.21 −0.09 ± 0.24
North Africa Desert CLIM 0.01 ± 0.02 −0.00 ± 0.01 0.01 ± 0.00 −0.00 ± 0.02
CO2 −0.01 ± 0.01 −0.01 ± 0.00 −0.01 ± 0.01 −0.01 ± 0.01
LUC −0.00 ± 0.01 −0.00 ± 0.01 −0.00 ± 0.01 −0.00 ± 0.01
NET 0.01 ± 0.01 −0.00 ± 0.01 −0.01 ± 0.01 −0.01 ± 0.02
Forest CLIM 0.03 ± 0.03 0.02 ± 0.03 0.03 ± 0.03 0.02 ± 0.02
CO2 −0.11 ± 0.04 −0.13 ± 0.05 −0.15 ± 0.05 −0.17 ± 0.07
LUC 0.04 ± 0.02 0.05 ± 0.03 0.05 ± 0.03 0.07 ± 0.04
NET −0.04 ± 0.04 −0.06 ± 0.05 −0.07 ± 0.04 −0.08 ± 0.06
Sub-humid savanna CLIM 0.18 ± 0.14 0.11 ± 0.09 0.13 ± 0.09 0.01 ± 0.08
CO2 −0.22 ± 0.13 −0.21 ± 0.13 −0.30 ± 0.17 −0.30 ± 0.17
LUC 0.12 ± 0.08 0.15 ± 0.08 0.20 ± 0.07 0.33 ± 0.12
NET 0.09 ± 0.13 0.05 ± 0.14 0.03 ± 0.14 0.04 ± 0.17
Semi-arid savanna CLIM 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.01 ± 0.00
CO2 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00
LUC 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00
NET 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00
Desert/Shrubland CLIM 0.10 ± 0.08 0.03 ± 0.04 0.04 ± 0.02 0.03 ± 0.06
CO2 −0.07 ± 0.03 −0.04 ± 0.03 −0.10 ± 0.03 −0.07 ± 0.04
LUC 0.02 ± 0.02 0.02 ± 0.03 0.03 ± 0.02 0.05 ± 0.04
NET 0.04 ± 0.06 0.01 ± 0.03 −0.02 ± 0.05 −0.04 ± 0.05
  • Note. Positive values represent fluxes out (source) of the biosphere and negative values, fluxes in (sinks).

We estimated changes in the African regional carbon fluxes and sinks and calculated their attribution to the underlying environmental drivers and the different ecoregions (Figure 3). Between 2000 and 2019, there were widespread but subtle losses due to climate change and variability (Figure 3c). The models also show a strong tropical forest uptake response driven by enhanced atmospheric CO2 concentrations (Figure 3b), while LUC losses were concentrated in East and West Africa (Figure 3d). These large opposing fluxes result in Africa acting as a net sink between 2000 and 2019 (Figure 3a), but there are still large uncertainties around the magnitude of the estimates.

Details are in the caption following the image

Spatial pattern of trends in annual mean NBP (gC m−2 yr−1) across Africa over 2000 to 2019 based on an ensemble of 17 Dynamic Global Vegetation Models from TRENDY v9. Large opposing fluxes result in a net sink of carbon (a), while (b) shows the attribution of CO2 fertilization and N deposits, (c) the attribution of climate change and variability and (d) the attribution of Land Use Change. Black isolines represent the boundaries of the ecoregions as depicted in Figure 1.

The model ensemble shows that losses due to LUC in Africa have increased over time (from 0.18 to 0.46 PgC yr−1) at a similar rate but in the opposite direction than the CO2 fertilization sink increase (from −0.41 to −0.55 PgC yr−1, Table 4). This estimate for the RECCAP2 period is within the range of LUC emission estimates for Africa reported from bookkeeping models: BLUE (Hansis et al., 2015): 0.57 ± 0.06 PgC yr−1 and HN2017 (Houghton & Nassikas, 2017): 0.43 ± 0.02 PgC yr−1. Climate-induced losses have decreased to almost zero (Table 4) likely due to the breaking of the decades-long drought in the Sahel, which compensated for increased aridity in East Africa over the same time period. Consequently, the biospheric sink capacity in Africa has increased to −0.09 ± 0.24 PgC yr−1 in the last decade. The LUC fluxes are spatially concentrated in the sub-humid savanna (a net source of 0.04 ± 0.17 PgC yr−1), while most of the sink capacity is concentrated in the tropical forests (−0.08 ± 0.06 PgC yr−1). This estimated sink capacity is an order of magnitude lower than that estimated from models that do not include land use and land cover: Africa NEE (including fire disturbances) estimated by TRENDY model ensembles (Section 2.2.2) was −0.09 ± 0.24 PgC yr−1 in 2010–2019 compared with −2.21 PgC yr−1 for aDGVM (Section 2.2.3).

We find large gross changes in the vegetation stocks but the net carbon stocks remain the same (Figure 4). Soil carbon pools are increasing: that is the DGVM models predict that the increase in CO2 uptake caused by CO2 fertilization continues to be larger than fluxes to the atmosphere due to increased microbial respiration rates, LUC and climate change.

Details are in the caption following the image

Change in carbon pools over the 1985 to 2019 period.

The TRENDY DGVM models vary in the processes simulated (see Table A1 in Friedlingstein et al. (2020a)). Most of them (11/17) simulate wildfires, and approximately half (8/17 include) nitrogen fertilization. Fuelwood harvest was commonly simulated (11/17 times), but tillage, irrigation, mowing, and other land use activities are included by very few models, and none include peatland drainage. The TRENDY protocol used the HYDE 3.2 land use product (Klein Goldewijk, 2017), but DGVM models varied in how they interpreted and used these data (Friedlingstein et al., 2020a). HYDE 3.2, unlike some land use data sets, does not show a leveling off of cropland expansion in Africa over the RECCAP2 period (see Text S1 in Supporting Information S1): all of the models used here are simulating increased cropland of approximately 50–100 km2 × 103 yr−1 whereas the HYDE 3.3 data set has cropland change of close to zero for most of the last decade (Figure 5). All of these factors might compound uncertainty in the TRENDY model estimates.

Details are in the caption following the image

Change in area for (a) land use, (b) cropland and (c) pasture estimated from HYDE 3.2 (orange line) and HYDE 3.3 (blue line). HYDE 3.2 indicates increases in cropland area over the RECCAP2 period, but HYDE 3.3 indicates no change. See Supporting Information S1 for further information on uncertainty in land use change trends.

2.2.3 Ecosystem Models Without Land Use (aDVGM)

The aDGVM is an individual-based model that has been developed specifically to simulate grass-tree dynamics in African ecosystems (Scheiter & Higgins, 2009). It has been shown to simulate the distribution of grasslands, savannas, and forests in Africa, but detailed assessments of carbon fluxes have not been conducted (Martens et al., 2021; Scheiter & Higgins, 2009). The aDGVM only represents potential natural vegetation without any land use driver (see Section 2.2.2 for results including land use). Here, aDGVM was forced with an ensemble of regionally-downscaled general circulation models over the 1985–2018 period.

In aDGVM simulated GPP, NPP, and NEE increased to 13.4, 7.4, and −3.0 PgC yr−1 for the 2009–2018 period (Table 5). These GPP values are lower than estimates from satellite observation (22.4–24.7 PgC yr−1 for different periods, Section 2.3.1, Table 3), and lower than values simulated by other DGVMs (GPP between 20.6 and 40.9 PgC yr−1, NPP between 9.2 and 20.5 PgC yr−1 for an ensemble of nine models, Valentini et al., 2014); NPP of 10.2 PgC yr−1 for the period 1980–2009 in simulations for Africa (Pan et al., 2015); NPP of 10.2 and 10.9 PgC yr−1 in the presence and absence of fire (Sato & Ise, 2012). However, the NEE of the forest region simulated by aDGVM (−0.51 PgC yr−1 for 1985–2008, increasing to −0.56 PgC yr−1 for 2009–2018) is slightly higher than the estimate of −0.34 PgC yr−1 (CI, −0.15 to −0.43) for observation data from sparse forest plots (Lewis et al., 2009). This supports results by Hubau et al. (2020) indicating that the forest carbon sink in intact African forests remained constant throughout the RECCAP2 period.

Table 5. Carbon Stocks and Fluxes Simulated by aDGVM
AboveGround (PgC) Belowground (PgC) Soil (PgC) Total (PgC) Trend (Pg Cyr−1)
Carbon stocks Region 1985–2008 2009–2018 1985–2008 2009–2018 1985–2008 2009–2018 1985–2008 2009–2018 1985–2008 2009–2018
Total carbon NH Desert 0.95 1.05 0.59 0.67 4.22 4.33 5.76 6.06 0.02 0.04
Forest 18.85 19.66 3.68 3.92 12.75 13.29 35.29 36.86 0.08 0.10
Desert/Shrubland 0.26 0.29 0.13 0.15 1.00 1.03 1.39 1.47 0.002 0.004
Sub-humid savanna 28.20 30.97 11.68 12.91 39.32 40.98 79.20 84.87 0.29 0.40
Semi-arid savanna 6.69 7.58 3.91 4.42 16.37 17.15 26.98 29.14 0.10 0.13
Africa 54.95 59.54 20.01 22.08 73.66 76.77 148.63 158.40 0.49 0.67
Total (Pg Cyr−1) Trend (PgC yr−1)
Carbon fluxes Region 1985–2008 2009–2018 1985–2008 2009–2018
NPP NH Desert 0.23 0.28 0.00 0.01
Forest 1.15 1.24 0.01 0.01
Desert/Shrubland 0.06 0.06 0.00 −0.00
Sub-humid savanna 3.82 4.14 0.02 0.04
Semi-arid savanna 1.50 1.68 0.01 0.01
Africa 6.75 7.40 0.04 0.06
GPP NH Desert 0.41 0.50 0.01 0.02
Forest 2.23 2.40 0.01 0.01
Desert/Shrubland 0.10 0.11 0.00 −0.00
Sub-humid savanna 6.86 7.45 0.03 0.07
Semi-arid savanna 2.63 2.94 0.02 0.02
Africa 12.22 13.41 0.07 0.11
NEE NH Desert −0.06 −0.09 −0.00 −0.01
Forest −0.51 −0.56 −0.00 −0.00
Desert/Shrubland 0.00 −0.01 0.00 0.00
Sub-humid savanna −1.62 −1.78 −0.01 −0.03
Semi-arid savanna −0.52 −0.60 −0.01 −0.01
Africa −2.72 −3.04 −0.02 −0.04
  • Note. Variables are averaged for whole Africa and ecoregions for the periods 1985–2008 and 2009–2018 and stocks include Aboveground, Belowground and Soil. Trends were derived by linear regression models using time series of monthly means of the respective variable. Detailed results in Supporting Information S1. Some values are zero due to rounding.

Both autotrophic and heterotrophic respiration increased in Africa according to aDGVM simulations (Table S5 in Supporting Information S1). Autotrophic respiration increased from 1.03 PgC yr−1 in the period 1985–2008 to 1.19 PgC yr−1 in the period 2009–2018, and heterotrophic respiration increased from 8.11 to 8.82 PgC yr−1 over the same periods. The highest respiration rates were simulated in the Sub-humid savanna region (0.65 and 4.72 PgC yr−1 for autotrophic and heterotrophic respiration in 2009–2018). Valentini et al. (2014) reported a multi-model mean heterotrophic respiration 11.8 PgC yr−1, which is higher than the aDGVM simulations.

In aDGVM simulations, carbon stored aboveground in Africa was 59.5 PgC in the period 2009–2018 (Table 5). This is lower than values by other models; 66.7 to 181.4 PgC for an ensemble of nine models (Valentini et al., 2014); 75.3 to 87.5 PgC with SEIB-DGVM (Sato & Ise, 2012); but falls within the range of estimates (48.3–64.5 PgC) by remote sensing AGB products (Avitabile et al., 2016; Baccini et al., 2012; Y. Y. Liu et al., 2015; Saatchi et al., 2011). Those remote sensing products do however represent slightly different periods within the RECCAP2 time period.

Aboveground carbon increased by 4.6 PgC between 2009 and 2018, with the highest increases in Sub-humid savannas. Belowground biomass increased by 2 PgC, and SOC increased by 3.1 PgC (Table 5), the overall rate of increase estimated without land use activities is 0.67 PgC yr−1 which is higher than for the 1985–2008 period.

2.3 Fluxes of Special Importance Within the African GHG Budget

2.3.1 Fires

Recent decades have seen reductions in the area burned per year in Africa from ∼3.1 × 106 km2 to ∼2.6 × 106 km2 (Andela et al., 2017; Zubkova et al., 2019) and consequently also a decline in total fire emissions (Figure 6) (Van Der Werf et al., 2017). Approximately 30% of this decline is attributed to land transformation and expansion of agricultural land (Zubkova et al., 2019); therefore, this does not necessarily imply increased C-sink potential. However the remaining ∼70% appears to be a result of higher effective rainfall and soil moisture, particularly in North Africa, producing less flammable vegetation (Zubkova et al., 2019).

Details are in the caption following the image

Spatial patterns of biomass burning emissions in Africa calculated from the FREMV2.1.

Emissions estimates from wildfire come from bottom up (based on burned area) and top-down (based on fire radiative power) methods (see Text S3 in Supporting Information S1). Several new data products have become available since the RECCAP1 period. Current bottom up burned area products omit small fires and analyses with higher resolution SENTINEL-2 data nearly double the estimated burned area (Roteta et al., 2019), possibly also doubling the estimated GFED fire emissions (Ramo et al., 2021). Here we present a new Africa-specific top-down fire emissions product (Nguyen & Wooster, 2020) and contrast it with estimates from other sources (Table 6).

Table 6. Comparing the Change in Mean Annual Emissions (Tg yr−1) for Different Chemical Species for Wildfires (Including Deforestation and Cropland Fires) and Fuelwood Burning Over the RECCAP1 and RECCAP2 Periods
Type Source Region RECCAP1a RECCAP2 2010-2019 Trend: Change/yr
Wildfire Valentini Africa 1031 (±87)
FREMv2.1 Africa 999 (±79) 953 (±113) −10.9
Northern Hemisphere 377
Southern Hemisphere 576
Forest 26
NH Desert 4
SH Desert 3
Sub-humid savanna 810
Semi-arid savanna 124
FuelWood Various (see SI) Africa 184 241 5.3
Total C wildfire + fuelwood 1,215 1,194 −9
Total CO2 FREM (range) 3,250 (2,225–5,475)
Total CH4 FREM (range) 6.8 (4.9–9.1)
Total CO FREM (range) 146 (142–224)
Total N2O FREM (range) 0.09 (0.09/0.42)
  • Note. Fuelwood burning was calculated from published sources (Amos, 1999; Bailis et al., 2015; Boden et al., 2013; Broadhead et al., 2001; FAO, 2010) integrated with the IEA World Energy Balances statistics (IEA, 2022).
  • Estimates come from FREMv2.1, a top-down regional product derived specifically for Africa (slightly modified from Nguyen & Wooster, 2020). Estimates for CO, CH4, and N2O emissions for RECCAP2 period are also provided, showing the FREM2.1 estimate and the range of other estimates for that time period. See Supporting Information S1 for more details of wildfire emissions data sources and the wood fuel burning estimates.
  • a Valentini et al. (2014) reported from 1997 to 2011, and FREMv2.1 was available from 2004 to 2009.

Existing estimates of total carbon emissions from wildfires in Africa range from 954 to 1,595 Tg Cyr−1, with CH4 ranging from 4.9 to 9.1 TgCH4 yr−1 and N2O from 0.8 to 0.4 TgN2O yr−1 (Table 6). Of these emissions, ∼85% come from sub-humid savannas which, due to their high productivity and long dry seasons, produce frequent fires that consume high amounts of biomass. Both top-down (calculated via energy released) and bottom-up approaches (calculated via burned area) show a clear decline over the last two decades (Table 6; Figure 7) in the order of ∼10 Tg Cyr−1. In contrast, total carbon emissions from wood fuel burning have increased steadily from 184 ± 24.6 Tg Cyr−1 for RECCAP1 to approximately 242 ± 36.1 Tg Cyr−1 for the RECCAP2 period (see Table S6 in Supporting Information S1 for more details). This represents an increase of approximately 5.3 Tg Cyr−1. Total fire emissions (wildfire and fuel wood burning) have therefore decreased slightly from 1,225 ± 99 to 1,197 ± 85 Tg Cyr−1. Of these fire emissions, approximately 134 TgC (or ∼12%) are considered a net source (Bailis et al., 2015; Scholes et al., 2011; van der Werf et al., 2017).

Details are in the caption following the image

Total carbon emissions from wildfires are decreasing while fuel wood emissions are increasing. Wildfire estimates are provided for a “bottom up” data product (GFED4.1s) (Randerson et al., 2017; Van Der Werf et al., 2017), a global “top-down” data product derived from an atmospheric inversion applied to MOPITT satellite CO data (Zheng et al., 2021), and a regional “top-down” data set for Africa derived from correlations between FRP and TPM and CO (FREMv2.1 slightly modified from Nguyen and Wooster (2020)). See Table 6 for the range of current estimates for all greenhouse gases.

2.3.2 Large Mammals

Herbivore CH4 emissions represent a small but increasing component of the African methane cycle, which is highly uncertain (Valentini et al., 2014). African livestock production systems differ from global averages in terms of diet, average body weights, herd structure, and body condition (Goopy et al., 2021; Ndung'u et al., 2022). The IPCC 2019 methodology estimates emission factors for free-ranging cattle in low productivity systems of Africa to be 48 kgCH4/head yr−1 (Table 10.11 in IPCC, 2019), but recent empirical papers from Africa report emissions factors closer to the IPCC 2006 estimate of 31 kgCH4/head yr−1 (Table S9 in Supporting Information S1).

Livestock represents 98% of the herbivore biomass in Africa (Hempson et al., 2017), and emissions from manure are small (<3%, Herrero et al., 2008); therefore, we focused here on enteric fermentation from livestock, whose numbers have increased by 30% in Africa in the last decade (Gilbert et al., 2018). The 11 African countries that regularly report livestock emissions to the UNFCCC showed livestock methane emissions increasing by ∼5% between the RECCAP1 and RECCAP2 periods, but the IPCC Tier 1 approach estimates increases closer to 30% for the same 11 countries. We produced a new African livestock emission factor (Africa_EF) calculated using the mean of a range of empirical data sources from African livestock production systems (see Table S9 in Supporting Information S1) of 35.6 kgCH4/head yr−1. When using Africa_EF instead of the IPCC value of 48 kgCH4/head yr−1 the overall methane emissions are reduced, but the increasing trend remains the same.

Models using metabolically based methane emissions model and different production systems (Herrero et al., 2008; Wolf et al., 2017) are less than half the IPCC 2019 Tier 1 approach (Table 7) and only show a 13% increase between the two periods caused both by increasing livestock numbers and a switch to more mixed production systems. The current best estimate of CH4 emissions from enteric fermentation of livestock in Africa for the RECCAP2 period is 17.6 (range 9.2–21.7) TgCH4 yr−1 which represents an annual increase of 2.9% (395 GgCH4 yr−1) from RECCAP1.

Table 7. Estimates of Annual Enteric Methane Emissions (TgCH4 yr−1) for Africa Calculated Using the IPCC Tier 1 Methodology (IPCC, 2019) and the Tier 1 Methodology With Africa-Specific Emissions Factors (IPCC2019_AfricaEF), Contrasted With Estimates From Published Sources, and From National UNFCCC Reporting
2000–2009 2010–2019 Trend: GgCH4 yr−1
UNFCCC (11 reporting countries)
UNFCCC 5.1 (±0.3) 5.3 (±0.1) 27
IPCC2019 5.2 6.8 161
IPCC2019_AfricaEF 4.1 5.4 131
Africa
Herrero et al. (2008) 8.1 9.2 109
Wolf et al. (2017) 12.7 ± 1.9
IPCC2019 16.8 21.7 482
IPCC2019_AfricaEF 13.7 17.6 395
  • Note. IPCC2019 uses emission factors from Table 10.11 which has a cattle emission factor of 48 for low-productivity systems. This is higher than all published emission factors for free-ranging cattle in Africa (See Table S9 in Supporting Information S1), so the IPCC2019_AfricaEF replaces this with the mean reported value of 35.6 kgCH4/head yr−1. Only 11 countries have UNFCCC data for both RECCAP periods so data are reported for these 11 countries, and for Africa as a whole.

2.3.3 Termites

Termites are an important source of methane due to the methanogenic degradation of lignocellulose in termite hindguts (Brune, 2014). The African continent hosts 39% of the total 2,600 species that have been described worldwide (Ahmed et al., 2011), contributing substantially to global termite CH4 emissions. Here, we provide new estimates of termite CH4 emissions across the African continent (Figure 8, Table 8) based on a new global termite biomass product predicted from 500 field transect measurements using a machine learning approach and the global mean and median of termite CH4 production rate from existing literature (mean = 3.74 μgCH4 g−1[termite] h−1, median = 2.88 μgCH4 g−1[termite] h−1, n = 251) (Zhou et al., 2023). Overall, termites across the African continent are predicted to emit 1.40 TgCH4 yr−1 (the 95% confidence intervals range: 1.31–1.49 TgCH4 yr−1) based on the mean termite CH4 production rate, with the largest emission from sub-humid savannas (0.63 TgCH4 yr−1) followed by semi-arid savanna (0.37 TgCH4 yr−1) and forests (0.19 TgCH4 yr−1) (also see Table 8 for the median estimate of termite CH4 production rate).

Details are in the caption following the image

Methane emission rates (mgCH4 m−2 d−1) from termites are estimated across the African continent.

Table 8. Predicted Termite Methane Emissions Across African Ecoregions
Ecoregion Termite methane emissions (TgCH4 yr−1)
Saunois et al. (2020) New estimate based on mean termite CH4 production rate New estimate based on median termite CH4 production rate
North Africa desert 0.067 0.134 (0.123–0.145) 0.103 (0.094–0.111)
Desert/shrubland 0.021 0.039 (0.036–0.042) 0.030 (0.028–0.032)
Semi-arid savanna 0.354 0.367 (0.342–0.392) 0.282 (0.263–0.301)
Sub-humid savanna 1.220 0.629 (0.589–0.670) 0.484 (0.452–0.516)
Forest 0.350 0.185 (0.175–0.195) 0.142 (0.134–0.150)
Africa (in total) 2.094 1.397 (1.305–1.489) 1.076 (1.004–1.247)
  • Note. Values in parentheses represent the 95% confidence intervals.

This new estimate is substantially lower than the estimate of 2.09 TgCH4 yr−1 from the global methane budget (Saunois et al., 2020) (Table 8) and other reported values (2.5–6.9 TgCH4 yr−1) from Valentini et al. (2014) for the African continent. Two prominent reasons for these inconsistencies are the lack of accurate data on termite biomass for upscaling, and the scarcity of empirical data on termite CH4 emission rates. Termite biomass is generally estimated by its dependence on GPP of ecosystems based on simple regression models (Kirschke et al., 2013; Saunois et al., 2020). Here, our global termite biomass estimate is based on available field measurements and predicted by a set of variables, including rainfall, soil pH, NPP, minimum/maximum temperature, SOC, and topography. Additionally, only a few studies measured CH4 emission rates at the individual species or mound scale across the African continent (Table S10 in Supporting Information S1) with CH4 emission rates varying significantly between species (0.68–17.4 μg CH4 g−1 hr−1), between mounds (81–5,478 ng CH4 s−1 mound−1) (Brauman et al., 2001; Macdonald et al., 1999; Rouland et al., 1993) and between seasons (Räsänen et al., 2023). However, more empirical measurements are still needed to improve the accuracy of termite biomass as well as termite methane emission rates across different ecosystems and regions.

2.4 Component Fluxes of NEE From Geological, Aquatic, and Coastal Systems

2.4.1 Geological Carbon Emissions

Africa's geogenic CO2 emissions are mostly due to volcanic and geothermal activity in the East African Rift (EAR), which is globally the largest active continental rift spanning a cumulative length of approximately 3,000 km (Lee et al., 2016). Extrapolation from first-order CO2 flux measurements of tectonic degassing in the Magadi-Natron basin amounts to a flux of 71 ± 33 TgCO2 yr−1 in the EAR (Lee et al., 2016). However, estimates based on extrapolation from surveys in the Main Ethiopian Rift (0.52–4.36 TgCO2 yr−1) give a flux range of 3.9–32.7 TgCO2 yr−1 (Hunt et al., 2017).

Geological emission sources of CH4 were calculated for each ecoregion and Africa as a whole using data from Etiope et al. (2019) (Table 9, Table S11 in Supporting Information S1). These include emissions from onshore seeps (gas-oil seeps and mud volcanoes), diffuse exhalation of CH4 associated with petroleum fields (microseepage) and geothermal manifestations mainly from volcanoes and geothermal sites, but excluding submarine seeps (see Ciais et al., 2022). The North African desert ecoregion contributes 46% of the estimated total African geological CH4 emissions of 1.01 TgCH4 yr−1 (see Figure S3 in Supporting Information S1 for the spatial distribution). Semi-arid and Sub-humid savanna ecoregions contribute 30% and 20%, respectively, while the forest ecoregion only contributes 5% of the estimated geological CH4 emissions across Africa.

Table 9. Geological, Inland Water and Coastal CO2, CH4, N2O, and Net GHG Emissions and Sinks
CO2 (Tg yr−1) CH4 (Tg yr−1) N2O (Gg yr−1) CO2eq (GWP100) (Tg yr−1) C (Tg yr−1)
Geological sourcesa 18.3 (3.9/32.7) 1 (1/1) 45.7 (31.3/60.1) 5.8 (1.8/9.7)
Atmospheric fluxes
Lakesb 12.1 (12.1/12.1) 2.2 (2.2/2.2) −0.2 (−0.2/−0.2) 71.4 (71.4/71.4) 5 (5/5)
Reservoirsc 16.2 (6.8/26.1) 2.1 (1.2/3.1) 6.6 (2.7/8.6) 74.7 (39.9/111) 5.7 (1.9/9.4)
Riversb 1,175 (990/1,360) 4.6 (3.9/5.2) 17.3 (14.8/19.8) 1,302.6 (1,099.3/1,505.8) 322.4 (271.7/373.1)
Estuary Emissions (Tidal systems and lagoons)d 21.6 (12.7/32.4) 0 (0/0.1) 2.8 (2.5/3.2) 23.3 (13.4/37.3) 5.9 (2.5/9.6)
Coastal Wetland Emissions (Mangroves, Salt marshes, Seagrasses)d −118.8 (−149.1/−82) 0.1 (0.1/0.3) 0.1 (0.1/0.3) −116 (−147.1/−73.4) −32.4 (−45.8/−22.5)
Net aquatic atmospheric fluxes 1,106.2 (872.5/1,348.6) 9 (7.4/10.9) 26.6 (19.8/31.8) 1,356.1 (1,076.9/1,652.2) 306.6 (235.2/374.7)
Carbon stock change
OC burial—inlande −131.9 (−24.1/−212.6) 0 (0/0) 0 (0/0) −131.9 (−24.1/−212.6) −36 (−6.6/−58)
OC burial—coastald −20.9 (−20.9/−20.9) 0 (0/0) 0 (0/0) −20.9 (−20.9/−20.9) −5.7 (−5.7/−5.7)
Net aquatic carbon stock change −152.8 (−45/−233.5) 0 (0/0) 0 (0/0) −152.8 (−45/−233.5) −41.7 (−12.3/−63.7)
Lateral fluxes
DICf −55.7 (−55.7/−55.7) 0 (0/0) 0 (0/0) −55.7 (−55.7/−55.7) −15.2 (−15.2/−15.2)
DOCg −71.4 (−71.4/−71.4) 0 (0/0) 0 (0/0) −71.4 (−71.4/−71.4) −19.5 (−19.5/−19.5)
POCg −64.6 (−64.6/−64.6) 0 (0/0) 0 (0/0) −64.6 (−64.6/−64.6) −17.6 (−17.6/−17.6)
Coastal Margin C inputsd −458.3 (−187/−729.7) 0 (0/0) 0 (0/0) −458.3 (−187/−729.7) −125 (−51/−199)
Net aquatic lateral fluxes −650 (−378.6/−921.3) 0 (0/0) 0 (0/0) −650 (−378.6/−921.3) −177.3 (−103.3/−251.3)

2.4.2 Weathering Uptake of Atmospheric CO2

We extracted estimates of weathering CO2 uptake and the weathering dissolved inorganic carbon (DIC) release from gridded products provided by Lacroix et al. (2020) for the African ecoregions (Table 9, Table S12 in Supporting Information S1). The method quantifies weathering and depends on surface runoff and temperature, lithology types and soil shielding, and is based on a modified version of the weathering model of Hartmann et al. (2009). Weathering on the continent induces a flux of −12.2 Tg Cyr−1 of CO2, accounting for around 7% of the global weathering consumption. The sink estimate for the continent is comparable with the previous estimate of −11.7 Tg Cyr−1 of Ludwig et al. (1998). The carbon uptake from the atmosphere and carbon originating from the rock material add up to a total of −15.2 Tg Cyr−1 DIC exported to freshwaters and the ocean. Lacroix et al. (2020) reported that there was a general underestimation of catchment DIC exports for African catchments, for example, a 20% underestimation compared to measurements for the Congo basin.

In Africa, the lowest consumption rates (0–0.1 tC km−2 yr−1) were recorded over eastern and southern Africa, while larger amounts (0.5–5 tC km−2 yr−1) of CO2 were consumed in central Africa and parts of East Africa. The Semi-arid savanna ecoregion, which consists, to a large degree, of metamorphics, unconsolidated and silicoclastic sediment lithological classes, accounts for the highest weathering rates per area and the largest part of the continent's weathering drawdown and DIC release (Table 9, Table S12 in Supporting Information S1), owing to rather high runoff rates ranging from 50 to 250 mm yr−1. Weathering rates in warm and runoff-abundant tropical forest areas are strongly reduced due to shielding by old and highly weathered soils (Hartmann et al., 2014), whereas weathering in the dry semi-arid savanna and desert is limited by precipitation and runoff, which is predominantly less than 25 mm yr−1.

2.4.3 Inland Water Emissions

Emissions of CO2, CH4, and N2O from rivers and lakes were taken from the regional estimates by Borges et al. (2015), Borges, Deirmendjian, Bouillon, Okello, et al. (2022) which provide average annual emissions of 990–1,360 TgCO2yr−1, 3.9–5.2 TgCH4 yr−1 and 14.8–19.8 GgN2O yr−1 from African rivers, and annual emissions of 12.1 TgCO2 yr−1 and 2.2 TgCH4 yr−1 from African lakes, but explicitly excluded reservoirs (Table 9). Moreover, they suggest that African lakes can be a minor sink of 0.2 GgN2O yr−1 (Borges, Deirmendjian, Bouillon, Okello, et al., 2022). For reservoir emissions, we used numbers provided in the synthesis of regionalized inland water emission estimates by Lauerwald et al. (2023a) for the RECCAP2 initiative. These estimated emission amount to 16 (7/26) TgCO2 yr−1, 2.1 (1.2/3.1) TgCH4 yr−1 and 6.6 (2.7/8.6) GgN2O yr−1 (Lauerwald et al., 2023a). Summing up these estimates, we get to total emission fluxes of 1.11 (0.87/1.35)3 PgCO2 yr−1, 9 (7/11) TgCH4 yr−1, and 0.027 TgN2O yr−1 from African inland waters (Table 9). It is noteworthy that rivers contribute 98% of inland water CO2 emissions, but only about half of inland water CH4 emissions. To quantify DOC and POC, we summarized data from Zscheischler et al. (2017), and freshwater burial was quantified from Mendonça et al. (2017).

2.4.4 Fluxes From Estuaries and Coastal Wetlands

Emissions of CO2, CH4, and N2O from various coastal ecosystems in Africa were estimated using available empirical data scaled to the total surface area of each of the coastal ecosystems (Table 9). These systems include tidal systems and deltas, lagoons, mangroves, salt marshes and seagrasses. Organic carbon burial and coastal margin (non-riverine) C inputs were also estimated. However, although the coastal margin C sink is likely to be substantial, methodology is not yet resolved enough to calculate at the regional scale. To deal with this highly uncertain estimate, we therefore included the (rough) estimate in Table 9 for reporting purposes, but for the final budgets we set the mean value to zero and the 95th quantile as our best estimate. Hereby, the coastal margin sink is not represented in the final budgets, but the uncertainty has been accounted for.

2.5 Trade Fluxes

2.5.1 Carbon in Crop and Wood Trade

The transfer of physical and embodied carbon to and from Africa represents a relatively small percentage when compared to the rest of the world (Peters et al., 2012). We consider the physical flows of carbon via trade in biomass that includes crops and harvested wood products for three different periods, including 1961–1984, 1985–2008, and 2009–2019, based on inventory data from the Food and Agricultural Organization of the United Nations database (FAOSTAT, 2021). Ftrade is considered a carbon flux source by the region if it imports more than it exports or a carbon flux sink if it does not.

Africa was a net importer of crops during all three periods (Table 10). Carbon imports through crops increased more than six-fold in the 1985 to 2008 period from the 1961 to 1984 period and almost doubled from the 1985–2009 to 2010–2019 periods. From 1961 to 2009, Africa was a small net exporter of carbon through wood. During the RECCAP2 period, however, Africa's wood carbon imports exceeded the exports, although the amount of carbon entering the region was still relatively small in contrast to global carbon trade.

Table 10. Crop and Wood Trade Fluxes (±Inter-Annual Variability) in TgCO2 yr−1 and Tg Cyr−1
Period 1961–1984 1985–2009 2010–2019
TgCO2yr−1 TgCyr−1 TgCO2yr−1 TgCyr−1 TgCO2yr−1 TgCyr−1
Crop export −13.6 ± 2.3 −3.7 ± 0.6 −14.9 ± 4.0 −4.0 ± 1.1 −29.1 ± 10.8 −7.9 ± 2.9
Crop import 22.6 ± 13.2 6.1 ± 3.6 73.6 ± 23.9 19.9 ± 6.5 137.2 ± 45.3 37.2 ± 12.2
Crop Net flux 9.0 ± 13.4 2.4 ± 3.6 58.6 ± 24.2 15.8 ± 6.5 108.7 ± 46.6 33.2 ± 12.6
Wood export −3.9 ± 0.7 −1.1 ± 0.2 −7.7 ± 3.3 −2.1 ± 0.9 −9.9 ± 3.3 −2.7 ± 0.9
Wood import 1.6 ± 0.6 0.4 ± 0.2 4.2 ± 2.3 1.1 ± 0.6 9.9 ± 3.6 2.7 ± 1.0
Wood Net flux −2.3 ± 1.0 −0.6 ± 0.3 −3.4 ± 4.0 −0.9 ± 1.1 0.05 ± 4.9 0.3 ± 1.3
  • Note. Positive values represent imports (source) and negative values represent exports (sink).

2.6 Anthropogenic Emissions of Greenhouse Gases From Inventory Data

We summarize the GHG emission estimates provided by the UNFCCC and International Energy Agency acquired through Climate Watch (2022). Total fossil fuel emissions increased from 1.23 PgCO2-eq to 1.74 PgCO2-eq from the 1990–2009 to 2010–2019 period (Table 11). Fossil fuel emissions contributed 42% of the total anthropogenic emissions, while LUC contributed about 32% during RECCAP2. We therefore notice that the proportional contribution of fossil fuel emissions has increased since RECCAP1 (39% and 35% contribution for fossil fuels and LUC, respectively). Of the 23% contribution of agriculture (including livestock) to the total emissions, methane emissions are responsible for 15%. Waste includes the national reported data of solid waste disposal, wastewater treatment and discharge, and the incineration and open burning of waste as per the IPCC guidelines. Emissions reported here for Agriculture include those from enteric fermentation, manure management, agricultural soils, prescribed burning of savannas, and field burning of agricultural residues. For a comprehensive analysis and comparison of inventory data to atmospheric inversions for Africa, see Mostefaoui et al. (2024).

Table 11. Anthropogenic Greenhouse Gas Emissions for the 1990–2009 (R1) and 2010–2019 (R2) Periods
Period Anthropogenic emissions (PgCO2-equivalent yr−1)
Fossil fuels (including industrial processes) Waste Agriculture LUC Total incl LUC Bunkers (Tg CO2-eq yr−1)
CO2 R1 0.83 ± 0.11 0.98 ± 0.02 1.81 ± 0.13 37.1 ± 3.83
R2 1.28 ± 0.06 1.20 ± 0.07 2.48 ± 0.12 41.6 ± 1.69
CH4 R1 0.35 ± 0.04 0.13 ± 0.02 0.44 ± 0.05 0.06 ± 0.02 0.99 ± 0.08 0.04 ± 0.01
R2 0.38 ± 0.02 0.16 ± 0.01 0.61 ± 0.03 0.06 ± 0.00 1.21 ± 0.04 0.02 ± 0.01
N2O R1 0.06 ± 0.02 0.01 ± 0.00 0.28 ± 0.03 0.04 ± 0.01 0.36 ± 0.05 0.24 ± 0.03
R2 0.08 ± 0.00 0.02 ± 0.00 0.36 ± 0.01 0.05 ± 0.00 0.46 ± 0.01 0.28 ± 0.02
Total R1 1.23 ± 0.12 0.15 ± 0.02 0.73 ± 0.06 1.09 ± 0.03 3.15 ± 0.16 37.4 ± 3.83
R2 1.74 ± 0.06 0.19 ± 0.01 0.97 ± 0.03 1.31 ± 0.07 4.15 ± 0.12 41.9 ± 1.69

2.6.1 Emissions From Different Fossil Fuel Energy Sources

We used the Greenhouse Gas from Energy Database Highlights data set (IEA, 2023) to evaluate the greenhouse gas emissions from different energy sources (Figure 9). The data in Table 12 show that fuel combustion from coal, gas and oil increased substantially from 1985 to 2009 to 2010–2019 while the increasing trend for fugitive emissions seems to slow down for the RECCAP2 period but still contributing almost the same amount of emissions as for RECCAP1. Emissions from bunkers add a relatively small amount of emissions to the total estimate, with emissions increasing for aviation bunkers and decreasing for marine bunkers from 1985 to 2008 to 2009–2019.

Details are in the caption following the image

Fossil fuel (and biofuel) emissions by fuel type.

Table 12. Emission Estimates (TgCO2-Eq yr−1) for Different Fossil Fuel Energy Sources
Energy source 1985–2009 2010–2019
Coal - Fuel combustion 276.51 ± 59.43 399.06 ± 18.29
Oil - Fuel combustion 298.85 ± 62.37 522.65 ± 38.78
Gas - Fuel combustion 95.03 ± 44.38 233.78 ± 30.50
Fugitive emissions 337.91 ± 57.02 340.63 ± 20.10
Marine bunkers (CO2 only) 19.65 ± 3.91 18.55 ± 0.97
Aviation bunkers (CO2 only) 15.34 ± 3.50 23.85 ± 1.01

2.7 Results of Top-Down Atmospheric Inversions

2.7.1 CO2 Inversions

For the land CO2 fluxes, we used a set of four CO2 inversions that used data from the global surface in situ network: CAMS v20r2 (Chevallier et al., 2005, 2019), sEXTocNEET_v2021 (Rödenbeck et al., 2003, 2018), Carbon Tracker Europe CTE2021 (Van Der Laan-Luijkx et al., 2017), University of Edinburgh or UoE (L. Feng et al., 2016) and one inversion driven by both in-situ and satellite column-averaged dry air mole fraction of atmospheric CO2 from OCO-2 and GOSAT: CMS-Flux (J. Liu et al., 2021), all with different priors, algorithms and transport and re-analyses fields, described in the global carbon budget 2021 (Friedlingstein et al., 2022) (Figure 10). Inversions were all adjusted for fossil fuels, cement and river fluxes (see GCB—Friedlingstein et al., 2022).

Details are in the caption following the image

Annual land CO2 fluxes (represented as year +0.5) over Africa (PgC yr−1).

Previous synthesis studies showed that the net terrestrial carbon balance of Africa is a small CO2 sink (Ciais et al., 2011; Valentini et al., 2014; Williams et al., 2007). However, the inversions are subject to large uncertainties, especially in the tropics, because of the lack of observations and the difficulties of representing tropical convection and related vertical mixing (Gaubert et al., 2019; Schuh et al., 2019). Using satellite CO2 column retrievals (Palmer et al., 2019) identified northern tropical Africa as being responsible for the majority of the pan-tropical net carbon seasonal cycle, with the largest emissions found over western Ethiopia and western tropical Africa during March and April.

In RECCAP1, the spread of the net exchange carbon according to four inversions was 1 PgC yr−1 for five years' annual means (2001–2004). Based on our collected CO2 inversions, the standard deviation was 0.25 PgC yr−1 for both 2001–2004 and for 2000–2009, and 0.30 PgC yr−1 for 2010–2019 (Table 13). For the 2000–2009 period, the average land flux (sink) was −0.14 PgC yr−1 ± 0.25 PgC yr−1 with three out of four inversions showing moderate CO2 uptake throughout the decade. In contrast, the same four inversion models find the 2010–2019 period to be a carbon source (0.11 ± 0.27 PgC yr−1) to the atmosphere, likely as a result of the 2015/2016 El-Niño with most inversions showing a net source in 2016 with an average flux of 1 PgC yr−1(Table 13). This source is in line with previous studies that identify increased respiration rates associated with the increased surface-temperature in 2016 (Gloor et al., 2018; J. Liu et al., 2017). For the full set of five available inversion models used for the 2009–2019 period, this source is estimated at 0.27 ± 0.3 PgC yr−1 as the CMS-flux inversion model estimates net emissions over most of this period. Within Africa, this source is mostly driven by emissions from the sub-humid savanna (0.27 ± 0.19 PgC yr−1). The CMS-Flux inversion is driven by GOSAT and OCO-2 data and shows a larger source than the in situ inversions alone. This source is driven by satellite observations of high CO2 over northern tropical Africa during the dry season and might be overestimated (Gaubert et al., 2023).

Table 13. Inverse Model Ensemble Summary of Posterior Land Fluxes for CO2 (PgC yr−1)
2000-2009 (4 inversions) 2010-2019 (4 inversions) 2010-2019 (5 inversions)
Mean Stdev Range Mean Stdev Range Mean Stdev Range
African continent −0.14 0.25 −0.35/0.37 0.11 0.27 −0.07/0.29 0.27 0.3 −0.07/0.93
Desert/Shrubland 0 0 −0.01/0. 0 0 −0.01/0.01 0 0 −0.01/0.01
Forest −0.05 0.05 −0.13/0.07 −0.03 0.07 −0.16/0.06 −0.05 0.06 −0.16/0.06
North-Africa desert 0 0.01 −0.04/0.02 −0.01 0.01 −0.04/0.01 −0.01 0.01 −0.04/0.01
Semi-arid savanna −0.03 0.05 −0.07/0.01 0.05 0.06 −0.01/0.15 0.07 0.06 −0.01/0.15
Sub-humid savanna −0.06 0.16 −0.23/0.29 0.09 0.15 −0.1/0.25 0.27 0.19 −0.1/0.98
  • Note. A positive value means a source to the atmosphere.
  • Value for 2009–2019 for all five available inversions are also shown (column 3), but for assessing change since the previous decade it is more appropriate to compare data with only 4 inversions.

2.7.2 CH4 and N2O Inversions

We present an inter-comparison of six surface-based atmospheric inversion models for CH4 over Africa and four inversions with assimilation of GOSAT observations with different transport models and inversion techniques CT-CH4/SURF (Tsuruta et al., 2017), NICAM-TM/4DVar (Niwa et al., 2017), NIES-TM-FLEXPART (Maksyutov et al., 2021; F. Wang et al., 2019), TM5-CAMS (Bergamaschi et al., 2010, 2013; Pandey et al., 2016; Segers & Houweling, 2018), TM5-4DVAR (Bergamaschi et al., 2013, 2018). The comparison reveals a significant model estimate range difference of over 15 TgCH4 yr−1 in annual mean estimates for Southern Africa (Table 14). The inversion results from surface based ensemble mean estimates for North Africa between 2009 and 2017 was 25.94 ± 3.03 TgCH4 yr−1, and for Southern Africa, it was 52.08 ± 5.05 TgCH4 yr−1 (Table 14). These values are slightly larger than the mean methane emissions during the previous period 2000–2008, which were 23.02 ± 3.76 TgCH4 yr−1 for North Africa, and 49.37 ± 3.81 TgCH4 yr−1 for Southern Africa. This is nearly 5% for North Africa and 12% for Southern Africa of the global total methane estimate of 557 TgCH4 yr−1 (F. Wang et al., 2019).

Table 14. Inversion Estimates Include the Model Means, Variance, and Ranges for CH4 and N2O
CH4 2000–2008 2009–2017
(6 surface-based inversions) Mean Model variance Range Mean Model variance Range
Africa 72.39 2.91 68.56–75.53 78.02 3.88 73.04–82.90
North Africa 23.02 3.76. 19.01–27.84 25.94 3.03 22.86–30.25
Southern Africa 49.37 3.81 45.56–54.99 52.08 5.05 45.73–60.03
CH4 2000–2008 2009–2017
(GOSAT inversions) Mean Model variance Range Mean Model variance Range
Africa 80.80 6.45 73.16–87.11
North Africa 23.14 2.29 21.20–26.34
Southern Africa 57.66 5.68 51.31–63.85
2000–2008 2009–2016
N2O Mean Model variance Range Mean Model variance Range
TgN 3.26 0.19 3.40–3.53 3.44 0.14 3.29–3.61
TgN2O 5.1182 0.2983 5.338–5.5421 5.4008 0.2198 5.1653–5.6677

GOSAT based inversions show similar estimates to the surfacebased inversions. Mean estimates of four GOSAT-based inversions were 23.14 ± 2.29 TgCH4 yr−1 for Northern Africa, and 57.66 ± 5.68 TgCH4 yr−1 for Southern Africa for the years 2010–2017 (Table 14). Although Africa's contribution to global methane emissions is relatively small, it is important to monitor the continent's emissions as they may increase in the future due to population growth, urbanization, and the development of oil and gas production. Agriculture and wetlands are responsible for more than 80% of net methane emissions in Africa.

The spatial mean estimations of N2O concentrations in Africa, as reported by five inversion models, have shown a relatively small discrepancy with a mean value of 3.26 ± 0.19 TgN yr−1 during the years from 2000 to 2008 (Table 14). This value has slightly increased to 3.44 ± 0.14 TgN yr−1 from 2009 to 2016. The data from these models showed similar results over these two time periods, with a small increase in the average N2O concentrations.

3 Synthesis of the African Region Greenhouse Gases Budget

We summarized the estimates and trends for the African GHG flux components and carbon stocks for the RECCAP2 period (Table 15). We present separate total estimates for each of the gases (CO2, CH4, N2O) and calculated the Carbon (Pg C yr−1) and GHG budgets in CO2 equivalents using the GWP100 values from the IPCC sixth assessment (IPCC, 2021). We employed both bottom-up (BU) and top-down (TD) approaches as described by Ciais et al. (2022) and compare these estimates below. Uncertainty estimates, calculated as the 5th and 95th percentiles, are provided in brackets where possible. Uncertainty in the net fluxes was difficult to calculate as some flux estimates were reported with standard deviations and other flux estimates only had minimum (min) and maximum (max) values (or 5th and 95th quantiles). For this reason, we converted all standard deviations to a 5th and 95th quantiles using the equations; min = mean − 1.645 * sd; max = mean + 1.645 * sd. We then produced a min and max net flux estimate by summing across these min and max values. When summing across positive and negative fluxes, we summed the smallest fluxes and not the smallest numbers. For example, if the min NPP estimate was −8.18 and the max NPP estimate was −17.44 PgC, and the min Rh was 4.8 and the max Rh was 17.2, we summed −8.18 and 4.8 and −17.44 and 17.2. This is still a very crude way of assessing uncertainty and results in very large uncertainty values, but until we have more data on all fluxes, it is the best uncertainty estimates we are able to provide at present.

Table 15. Synthesis of the Estimates (With Uncertainties) and Trends of GHG and Carbon Stocks (Pg) and Fluxes (Pg yr−1) for Africa Over the RECCAP2 Period (Specific Periods Depicted by Footnotes)
Carbon stocks CO2 CH4 N2O Carbon budget

GHG budget

(CO2 equivalents)

Estimate

(PgC)

Trend

(PgC yr−1)

Above ground biomass
Satellite based modelsa

84

(71/95)

TRENDY model ensemblea

56

(48/64)

aDGVMb 59.54
Belowground biomass: Peat 36.9 −0.012
Belowground biomass: Soils
Soilgridsc

87.7

(77/99)

TRENDY model ensemblea 148 ± 60
aDGVMb 76.77
Total Carbon stocks 208.6
GHG fluxes

Estimate

(PgCO2 yr−1)

Estimate

(TgCH4 yr−1)

Estimate

(TgN2O yr−1)

Estimate

(PgC yr−1)

Trend

(PgC yr−1)

Estimate

(PgCO2eq yr−1)

Trend

(TgCO2eq yr−1)

GPP
Satellite based modelsd

−90.5 ± 9

(−105.3/−75.6)

−24.7 ± 2.5

(−28.7/−20.6)

−0.03

−90.5

(−105.3/−75.6)

−0.12
TRENDY model ensemblea

−103.0 ± 12.4

(−123.5/−82.6)

−28.1 ± 3.4

(−33.7/−22.5)

−0.09

−103.0

(−123.5/−82.6)

−0.35
aDGVMb

−49.2

(−49.2/−49.2)

−13.4

(−13.4/−13.4)

−0.11

−49.2

(−49.2/−49.2)

−0.42
Autotrophic respiration (Ra)
TRENDY model ensemblea

56.1 ± 9.9

(39.7/72.4)

15.3 ± 2.7

(10.8/19.8)

0.05

56.1

(39.7/72.4)

0.19
aDGVMb

4.4

(4.4/4.4)

1.2

(1.2/1.2)

0.02

4.4

(4.4/4.4)

0.06
NPP
TRENDY model ensemble a

−47.0 ± 10.3

(63.9/30)

−12.8 ± 2.8

(17.4/8.2)

−0.04

−47

(63.9/30.0)

−0.16
aDGVMb

−44.8

(−44.8/-44.8)

−12.2

(−12.2/−12.2)

−0.06

−44.8

(−44.8/−44.8)

−0.23
Heterotrophic respiration (Rh)
TRENDY model ensemblea

40.4 ± 13.9

(17.6/63.2)

11.0 ± 3.8

(4.8/17.2)

0.03

40.4

(17.6/63.2)

0.09
aDGVMb

32.3

(32.3/32.3)

8.8

(8.8/8.8)

0.05

32.3

(32.3/32.3)

0.19
Wild fire emissions
FREMv2.1a

3.2

(3.2/5.5)

6.8

(4.9/9.1)

0.08

(0.08/0.42)

1.0 ± 0.1

(1.0/1.6)

−0.01

3.5

(3.5/5.8)

TRENDY model ensemblea

3.2 ± 2.1

(−0.3/6.6)

0.9 ± 0.6

(−0.1/1.8)

−0.002

3.2

(−0.3/6.6)

aDGVMb 4.2

1.2

(1.2/1.2)

4.2
Land use change emissions
TRENDY model ensemble a

1.7 ± 0.6

(0.8/2.7)

0.5 ± 0.2

(0.2/0.7)

1.7

(0.8/2.7)

Net ecosystem production

−1.5

(−4.2/3.4)

6.8

(4.9/9.1)

0.08

(0.08/0.42)

−0.35

(−1.05/1)

−1.3

(−3.9/3.5)

Biofuel emissionsa

0.9 ± 0.2

(0.6/1.2)

0.2 ± 0.05

(0.2/0.3)

0.01

0.9

(0.6/1.2)

Crop trade fluxesa

0.1 ± 0.05

(0.03/0.19)

0.03 ± 0.01

(0.01/0.05)

0.1

(0.03/0.2)

Wood trade fluxesa

0 ± 0.005

(−0.008/0.008)

0 ± 0.001

(−0.002/0.002)

0

(−0.008/0.008)

Lateral fluxes (aquatic)a

−0.19

(−0.19/−0.65)

−0.05

(−0.05/−0.18)

−0.19

(−0.19/−0.65)

Aquatic atmospheric fluxesa

1.11

(0.87/1.35)

9

(7.4/11)

0.03

(0.02/0.03)

0.31

(0.25/0.37)

1.36

(1.08/1.65)

Organic C buriala (freshwater/coastal)

−0.15

(−0.04/−0.23)

−0.04

(−0.01/−0.06)

−0.15

(−0.05/−0.23)

Geological fluxesa

0.02

(0/0.03)

1.01

(1.01/1.01)

0.01

(0.002/0.01)

0.05

(0.03/0.06)

Termitesa

1.4

(1.3/1.5)

0.001

(0.001/0.001)

0.04

(0.04/0.04)

Herbivoresa

17.6

(9.2/21.7)

0.013

(0.007/0.016)

0.48

(0.25/0.59)

10.8
Emissions from soile

−1.5 ± 3

(−6.4/3.5)

1.1 ± 0.9

(−0.4/2.6)

Net ecosystem exchange

0.3

(−2.4/4.7)

34.4

(17.3/47.7)

1.24

(−0.25/3.07)

0.16

(−0.52/1.36)

1.5

(−0.2/5.1)

Fossil fuelsa

1.28 ± 0.11

(1.1/1.45)

14.2 ± 0.8

(12.9/15.5)

0.30 ± 0

(0.29/0.31)

0.36 ± 0.03

(0.31/0.41)

1.74

(1.53/1.96)

Bunkersa

0.04 ± 0.002

(0.04/0.04)

0.001 ± 0

(0.001/0.001)

0.001 ± 0

(0.001/0.001)

0.01 ± 0

(0.01/0.01)

0.04

(0.04/0.04)

Agriculturea

22.5 ± 1.1

(20.7/24.2)

1.33 ± 0.04

(1.26/1.41)

0.02 ± 0

(0.02/0.02)

1.0

(0.97/1.04)

Wastea

6.0 ± 0.3

(5.4/6.5)

0.07 ± 0.004

(0.06/0.07)

0.004

(0.004/0.005)

0.18

(0.16/0.2)

Net bottom-up total (NBP)

1.6

(−0.9/5.8)

77 ± 2.2

(56.4/93.9)

2.9 ± 0.1

(1.4/4.9)

0.6 ± 0.2

(−0.1/1.7)

4.5

(−3.3/14.1)

Atmospheric inversions (top-down)

0.4

(−0.26/1.06)a

78.02 ± 3.88

(73.04/82.9) f

5.40 ± 0.22

(5.17/5.67) g

0.17 ± 0.27

(−0.02/1.62)

4.0

(3.1/4.9)

  • Note. Estimate units for CH4 and N2O in blue italics are Tg yr−1. Where more than one estimate is provided for a component the value considered as the “best estimate” was used for calculating the net balances and is provided in bold.
  • a 2010–2019.
  • b 2009–2018.
  • c 2009–2019.
  • d 2009–2015.
  • e Valentini et al. (2014).
  • f 2009–2017.
  • g 2009–2016.

Total CH4 fluxes for Africa over the RECCAP2 period amount to 77 ± 2.2 (56.4/93.9) Tg C yr−1. This BU estimate is very close to the TD estimate of 78.02 ± 3.88 (73.04/82.9) from the atmospheric inversion models. An estimate of 66 ± 35 TgCH4 yr−1 was reported for RECCAP1 (Valentini et al., 2014). For N2O, the RECCAP2 BU estimate of 2.9 ± 0.1 (1.4/4.9) TgN2O−1 is much lower than the estimate from the atmospheric inversions at 5.401 ± 0.22 (5.165/5.668). The RECCAP1 estimate was 3.3 ± 1.3 TgN2O yr−1. As the large majority of N2O emissions for Africa are from agricultural sources, we would expect this flux to be increasing over time. Given the lack of certain component fluxes in our bottom-up estimates and the large uncertainty associated with our estimates, a considerable effort should be directed at improving observations and estimates for CH4 and N2O fluxes in Africa.

Considering the carbon in CO2 and CH4, we find that the BU approach estimates Africa to contribute 0.6 ± 0.2 (−0.1/1.7) PgC yr−1 to the global carbon cycle when we include non-terrestrial fluxes such as fossil fuels. Within this BU net carbon balance, terrestrial fluxes contribute 0.16 (−0.52/1.36) PgC yr−1 with the rest being produced through anthropogenic emissions from fossil fuels, agriculture and waste. However, the TD approaches estimate a much lower African contribution at 0.17 ± 0.27 (−0.02/1.62). Similarly, the calculated balance of fluxes from all three gases (in CO2 equivalents) adds to a total of 4.5 (−3.3/14.1) PgCO2eq yr−1 of which NEE contributes 1.5 (−0.2/5.1) PgCO2eq yr−1 for the BU approaches. The TD approaches estimate the African contribution of GHG emissions at 3.98 (3.13/4.85) PgCO2eq yr−1. The estimate for RECCAP1 (Valentini et al., 2014) was −2.7 ± 4.3, but they did not include key aquatic fluxes which make significant contributions. The differences between the estimates from the BU and TD approaches are not unexpected as BU approaches often omit some flux components due to the challenges in observation and lack of data. In particular, the coastal ocean margin sink (Kwon et al., 2021) could not accurately be quantified and was omitted from the final budget, and models of above and below ground biomass change require further validation. The large uncertainty values of the TD approaches are also a consequence of the sparse surface observations, which makes it difficult to constrain the inversion models.

Nevertheless, we find increasing trends of carbon and GHG emissions in the net balance estimates from both BU and TD approaches. Given the large uncertainties associated with these balances, it is difficult to definitively state that Africa is a source of carbon emissions, although it does appear to be likely. If we consider the contribution of N2O and CH4 in the total GHG net emission estimate, Africa does however categorize as a net source. Certainly, we do see that Africa's carbon and GHG budget remains close to carbon neutral and still contributes a small percentage to the global budget relative to other regions. However, it is concerning that the sink capacity in Africa is decreasing.

4 Conclusion

For the RECCAP2 synthesis, it is important to highlight the advances in several component estimates since the RECCAP1 period. Particularly, we incorporated the most recent methodology for biomass estimation through the use of novel L-VOD passive microwave data (Diouf et al., 2015) and LiDAR-based biomass data (Potapov et al., 2021). Fire emission estimates were improved through the use of a top-down regional product (FREMv2.1) derived specifically for Africa. Empirical data from the continent on livestock emission factors were used to adjust the livestock methane flux estimates, while new termite biomass data and emission factors shed more light on methane emissions from insects. Peatland loss rates are reported for the first time in the African GHG budget and we expect further development on this topic in the near future.

We also made a concerted attempt to calculate lateral fluxes, both from crop and wood trade, and from rivers. However, much of the data is based on coarse methods that used Tier 1 inventory data and/or taken from global models with insufficient Africa-specific observation data. Although lateral trade fluxes represent a relatively small contribution to the net estimates, future efforts should be directed at improved methodology and the inclusion of embodied carbon in products. Similarly, for carbon transport in rivers, we advocate for increased observations and empirical studies that are specific to Africa.

The information from this African budget is key to assessing which aspects of the greenhouse gas cycle are most important to be managed, and what sorts of management are possible in the quest to achieve net zero. Our budget indicates that shifts to C-neutral energy sources can potentially remove up to 30% (1.74 (1.53/1.96) PgCO2eq yr−1) of the current anthropogenic emissions, but emissions from LUC (1.7 (0.8/2.7) PgCO2eq yr−1) are more difficult to reduce. Both agricultural intensification, and expansion of agricultural land will continue to increase GHG fluxes in the short term, and the impact on the GHG budget depends on the degree to which climate-smart agricultural practices can be rolled out. This key component requires more direct attention because even with the availability of novel state-of-the-art satellite products, categorization of land use and land cover is still coarse, irregular and difficult to verify (Tubiello et al., 2023).

As natural ecosystems are increasing their C-sink capacity, and currently more than compensating for the LUC emissions (CO2 fertilization estimated as −2.02 + −0.88 PgCO2eq yr−1 by the TRENDY model ensemble), there is hope that nature-positive investments in Africa can help balance the global GHG budget. The IPCC AR6 scenarios for limiting warming to 1.5° include substantial carbon-capture in African ecosystems, 2.3 Pg annually by 2050, involving over 700 million ha of land (Forster et al., 2018). Key fluxes that are targeted are the fuelwood emissions (0.91 PgCO2eq yr−1), and the above-ground biomass (highly uncertain), as well as climate-smart agricultural practices. There is no evidence yet that this is possible within the socio-ecological context, with evidence emerging that estimates of potential above-ground biomass stocks are unrealistic, and some will have negative biodiversity and social outcomes (Armani, 2022; Bond et al., 2019). This RECCAP2 GHG budget sets a baseline against which to assess the effectiveness of policies and highlights the key fluxes that need better quantification to support financing these interventions and assessing their consequences.

Currently, the ability to accurately monitor C stock changes at large scales in Africa is limited, as the remotely sensed data sets have not been well parameterized for these ecosystems. This will improve rapidly due to private-public partnerships as C offset projects are scrutinized and verification procedures provide the motivation for improved C monitoring. Soil carbon stocks likewise, need attention in the DGVM modeling community: the TRENDY models all predict large increases in soil carbon reserves in the past few decades, but the causes of this are unclear. With better quantification, it will be easier to access funding to drive ecosystem-based mitigation activities.

A key flux highlighted here is the 0.48 (0.248/0.585) PgCO2eq yr−1 contributed by livestock methane emissions. Our paper demonstrates how sensitive this value is to incorrect emissions factors and to varying livestock production systems, and highlights that there is a growing body of evidence on the continent to enable better parameterization of this important flux. It is also important to note that only 60% of this methane flux represents a net increase above what would have been emitted by the wildlife of Africa before they were replaced with livestock (Hempson et al., 2017). Options for reducing the livestock methane flux in African ecosystems need to be sensitive to the social contexts involved, but policies enabling mixed livestock-wildlife systems might prove important.

As one of the significant fluxes in Africa, fire contributed between 46% and 65% to the global fire emission estimate. We have shown that wildfire emissions decreased from the RECCAP1 period, but much of this appears to be a consequence of land conversion that manifests as an alternative source of GHG emissions to the atmosphere. Further decreases in fire emissions in Africa have been advocated to help mitigate climate change (Tear et al., 2021), but only 12% of the current emissions are considered a net source, and fire is a process that maintains functionality in a large proportion of Africa's ecosystems (e.g., grasslands and savannas).

To conclude, we show that Africa's sink capacity is decreasing and that the continent most likely switched from a small net sink to a small net source during the 2010–2019 period. Although we have improved many of the component estimates since the previous RECCAP period, we still have large uncertainties in our estimates. What is clear is that Africa has an increasing GHG emissions trend and it deviates from the mitigation aims of the Paris Agreement towards net-zero emissions. Forecasts of a growing population associated with increasing emissions from fossil fuel burning and land conversion will inevitably increase Africa's relative contribution to the global GHG estimates in the next decade. For Africa to assist with increasing international carbon trade demand from countries that are under pressure to meet their carbon dioxide reduction targets (see Jones, 2023; Yang et al., 2023), there will have to be a distinctive shift in the continents' development trajectory towards carbon-neutrality. This will require (a) enabling policy environments, (b) financial and technical support, and (c) global commitment to addressing the socio-economic challenges that will likely multiply as climate change continues to impact this region. We suggest a directed attempt to increase the GHG observation network of Africa for all BU components of the GHG budget, but especially with regard to LUC and biomass estimates. Importantly, a protocol for accountability within national pledges should be accompanied by enabling African countries to observe and report more consistently in a standardized way for centralization of data in inventories.

Acknowledgments

YE and SA were funded by the Oppenheimer Generations Research and Conservation: the Future Ecosystems for Africa Program. TT was funded by the Swedish National Space Agency (Dnr: 2021-00144; 2021-00111), FORMAS (Dnr. 2021-00644) and the EU-Aid funded CASSECS Project (Dnr: FOOD/2019/410-169). RL acknowledges funding from French state aid, managed by ANR under the “Investissements d'avenir” programme (ANR-16-CONV-0003). SL was funded by South African Research Chair Initiative (SARChl) (# 64796). CR was funded by the SECO project (Dnr: NE/T01279X/1). Nicola Stevens was funded by Trapnell Fund, Linacre College—Oxford.

    Data Availability Statement

    SMOS-IC L-VOD data product is available from the Centre Aval de Traitement des Données SMOS (CATDS, 2024) (https://www.catds.fr/Products/Products-over-Land/SMOS-IC); The X-VOD data product (INRAE BORDEAUX Soil Moisture and VOD PRODUCTS, 2024) can be downloaded at https://ib.remote-sensing.inrae.fr/index.php/tag/amsr2-xvod-dataset/; GlobBiomass data and the ESA CCI (Santoro et al., 2018) biomass data is freely available for download at https://globbiomass.org/wp-content/uploads/GB_Maps/Globbiomass_global_dataset.html and https://climate.esa.int/en/projects/biomass/data/, respectively. The NCEO product (Rodríguez-Veiga & Balzter, 2021) is available from https://doi.org/10.25392/leicester.data.15060270.v1; The McNicol data product (McNicol & Ryan, 2018) is available at https://datashare.is.ed.ac.uk/handle/10283/3059.

    Soilgrids can be downloaded from https://www.isric.org/explore/soilgrids (Hengl et al., 2017b). The modeled GPP data derived by Tagesson et al. (2021) are available at: https://doi.org/10.17894/ucph.b2d7ebfb-c69c-4c97-bee7-562edde5ce66 (Tagesson, 2020). TRENDY v9 simulations for the Global Carbon Budget 2020 (Friedlingstein et al., 2020b) can be obtained from https://www.wdc-climate.de/ui/entry?acronym=DKRZ_LTA_891_ds00012. The HYDE database is accessible through the data portal at https://doi.org/10.17026/dans-25g-gez3 (Klein Goldewijk, 2017). The aDGVM was forced with CCAM regionally downscaled GCM daily input data available from the Global Change Institute, University of the Witwatersrand upon request ([email protected]). The FREM fire emissions inventory data can be provided upon request to Martin Wooster ([email protected]). GFED4.1 data (Randerson et al., 2017) is freely available at https://doi.org/10.3334/ORNLDAAC/1642. Emission estimates from the International Energy Agency (IEA, 2022, 2023) is available at https://www.iea.org/data-and-statistics/data-product/world-energy-statistics and https://www.iea.org/data-and-statistics/data-product/greenhouse-gas-emissions-from-energy-highlights. The termite dataset and associated information (Zhou et al., 2022) are available from https://doi.org/10.5061/dryad.vt4b8gtvk.The gridded dataset of Etiope et al. (2018) is available for download at https://doi.org/10.25925/4j3f-he27. The Lacroix et al. (2020) data used to estimate fluxes from weathering are archived by the Max Planck Institute for Meteorology and are available upon request ([email protected]).

    Data for inland water flux estimates are available at figshare: Lauerwald et al. (2023b) (https://doi.org/10.6084/m9.figshare.22492504) and https://doi.org/10.5281/zenodo.6025626 (Borges, Deirmendjian, Bouillon, & Morana, 2022). DOC and POC estimates were based on data extracted from Zscheischler et al. (2017) and data on freshwater OC burial is available as a supplementary file to Mendonça et al. (2017). Data used for the coastal margin C input estimates as well as data for the atmospheric inversions are available from the Max Planck Institute for Biogeochemistry (2019) GeoCarbon Data Portal at https://www.bgc-jena.mpg.de/geodb/projects/Home.php. Estimates for crop and wood trade are based on data from the Food and Agricultural Organisation of the United Nations (FAOSTAT, 2021) available freely from https://www.fao.org/faostat/en/#data. Anthropogenic emission estimates presented in this paper are available from https://www.climatewatchdata.org/ (Climate Watch, 2022).