Abstract
An optimal network design was carried out to prioritise the installation or refurbishment of greenhouse gas (GHG) monitoring stations around Africa. The network was optimised to reduce the uncertainty in emissions across three of the most important GHGs: CO2, CH4, and N2O. Optimal networks were derived using incremental optimisation of the percentage uncertainty reduction achieved by a Gaussian Bayesian atmospheric inversion. The solution for CO2 was driven by seasonality in net primary productivity. The solution for N2O was driven by activity in a small number of soil flux hotspots. The optimal solution for CH4 was consistent over different seasons. All solutions for CO2 and N2O placed sites in central Africa at places such as Kisangani, Kinshasa and Bunia (Democratic Republic of Congo), Dundo and Lubango (Angola), Zoétélé (Cameroon), Am Timan (Chad), and En Nahud (Sudan). Many of these sites appeared in the CH4 solutions, but with a few sites in southern Africa as well, such as Amersfoort (South Africa). The multi-species optimal network design solutions tended to have sites more evenly spread-out, but concentrated the placement of new tall-tower stations in Africa between 10ºN and 25ºS. The uncertainty reduction achieved by the multi-species network of twelve stations reached 47.8% for CO2, 34.3% for CH4, and 32.5% for N2O. The gains in uncertainty reduction diminished as stations were added to the solution, with an expected maximum of less than 60%. A reduction in the absolute uncertainty in African GHG emissions requires these additional measurement stations, as well as additional constraint from an integrated GHG observatory and a reduction in uncertainty in the prior biogenic fluxes in tropical Africa.
Disclosure statement
The authors declare that no competing interests are present.
Supplemental data
Supplemental data for this article can be accessed here.
Acknowledgements
The authors would like to thank Alistair Manning for useful tea-time discussions on biomass burning emissions.
Additional information
Funding
Notes on contributors
Alecia Nickless
AN wrote and implemented the code for calculating the posterior covariance matrices and performing the incremental optimisation procedure in Python, reshaped and processed the prior fluxes in Python and R, produced all figures using R, interpreted the results, and was responsible for the development of the manuscript. All co-authors contributed towards and reviewed the manuscript. RJS conceived the original concept and interpreted results. AV produced the FLEXPART footprints. JB and AB contributed to site identification and creation of the candidate site list. JA provided the LPJ output. UK, VK, and KP contributed to the FLEXPART runs and data management. MR processed the CH4 and N2O emission datasets and contributed towards interpretation of the results. VJ contributed towards project management. WK contributed towards the initial concept and interpretation of the results.
Robert J. Scholes
AN wrote and implemented the code for calculating the posterior covariance matrices and performing the incremental optimisation procedure in Python, reshaped and processed the prior fluxes in Python and R, produced all figures using R, interpreted the results, and was responsible for the development of the manuscript. All co-authors contributed towards and reviewed the manuscript. RJS conceived the original concept and interpreted results. AV produced the FLEXPART footprints. JB and AB contributed to site identification and creation of the candidate site list. JA provided the LPJ output. UK, VK, and KP contributed to the FLEXPART runs and data management. MR processed the CH4 and N2O emission datasets and contributed towards interpretation of the results. VJ contributed towards project management. WK contributed towards the initial concept and interpretation of the results.
Alex Vermeulen
AN wrote and implemented the code for calculating the posterior covariance matrices and performing the incremental optimisation procedure in Python, reshaped and processed the prior fluxes in Python and R, produced all figures using R, interpreted the results, and was responsible for the development of the manuscript. All co-authors contributed towards and reviewed the manuscript. RJS conceived the original concept and interpreted results. AV produced the FLEXPART footprints. JB and AB contributed to site identification and creation of the candidate site list. JA provided the LPJ output. UK, VK, and KP contributed to the FLEXPART runs and data management. MR processed the CH4 and N2O emission datasets and contributed towards interpretation of the results. VJ contributed towards project management. WK contributed towards the initial concept and interpretation of the results.
Johannes Beck
AN wrote and implemented the code for calculating the posterior covariance matrices and performing the incremental optimisation procedure in Python, reshaped and processed the prior fluxes in Python and R, produced all figures using R, interpreted the results, and was responsible for the development of the manuscript. All co-authors contributed towards and reviewed the manuscript. RJS conceived the original concept and interpreted results. AV produced the FLEXPART footprints. JB and AB contributed to site identification and creation of the candidate site list. JA provided the LPJ output. UK, VK, and KP contributed to the FLEXPART runs and data management. MR processed the CH4 and N2O emission datasets and contributed towards interpretation of the results. VJ contributed towards project management. WK contributed towards the initial concept and interpretation of the results.
Ana López-Ballesteros
AN wrote and implemented the code for calculating the posterior covariance matrices and performing the incremental optimisation procedure in Python, reshaped and processed the prior fluxes in Python and R, produced all figures using R, interpreted the results, and was responsible for the development of the manuscript. All co-authors contributed towards and reviewed the manuscript. RJS conceived the original concept and interpreted results. AV produced the FLEXPART footprints. JB and AB contributed to site identification and creation of the candidate site list. JA provided the LPJ output. UK, VK, and KP contributed to the FLEXPART runs and data management. MR processed the CH4 and N2O emission datasets and contributed towards interpretation of the results. VJ contributed towards project management. WK contributed towards the initial concept and interpretation of the results.
Jonas Ardö
AN wrote and implemented the code for calculating the posterior covariance matrices and performing the incremental optimisation procedure in Python, reshaped and processed the prior fluxes in Python and R, produced all figures using R, interpreted the results, and was responsible for the development of the manuscript. All co-authors contributed towards and reviewed the manuscript. RJS conceived the original concept and interpreted results. AV produced the FLEXPART footprints. JB and AB contributed to site identification and creation of the candidate site list. JA provided the LPJ output. UK, VK, and KP contributed to the FLEXPART runs and data management. MR processed the CH4 and N2O emission datasets and contributed towards interpretation of the results. VJ contributed towards project management. WK contributed towards the initial concept and interpretation of the results.
Ute Karstens
AN wrote and implemented the code for calculating the posterior covariance matrices and performing the incremental optimisation procedure in Python, reshaped and processed the prior fluxes in Python and R, produced all figures using R, interpreted the results, and was responsible for the development of the manuscript. All co-authors contributed towards and reviewed the manuscript. RJS conceived the original concept and interpreted results. AV produced the FLEXPART footprints. JB and AB contributed to site identification and creation of the candidate site list. JA provided the LPJ output. UK, VK, and KP contributed to the FLEXPART runs and data management. MR processed the CH4 and N2O emission datasets and contributed towards interpretation of the results. VJ contributed towards project management. WK contributed towards the initial concept and interpretation of the results.
Matthew Rigby
AN wrote and implemented the code for calculating the posterior covariance matrices and performing the incremental optimisation procedure in Python, reshaped and processed the prior fluxes in Python and R, produced all figures using R, interpreted the results, and was responsible for the development of the manuscript. All co-authors contributed towards and reviewed the manuscript. RJS conceived the original concept and interpreted results. AV produced the FLEXPART footprints. JB and AB contributed to site identification and creation of the candidate site list. JA provided the LPJ output. UK, VK, and KP contributed to the FLEXPART runs and data management. MR processed the CH4 and N2O emission datasets and contributed towards interpretation of the results. VJ contributed towards project management. WK contributed towards the initial concept and interpretation of the results.
Ville Kasurinen
AN wrote and implemented the code for calculating the posterior covariance matrices and performing the incremental optimisation procedure in Python, reshaped and processed the prior fluxes in Python and R, produced all figures using R, interpreted the results, and was responsible for the development of the manuscript. All co-authors contributed towards and reviewed the manuscript. RJS conceived the original concept and interpreted results. AV produced the FLEXPART footprints. JB and AB contributed to site identification and creation of the candidate site list. JA provided the LPJ output. UK, VK, and KP contributed to the FLEXPART runs and data management. MR processed the CH4 and N2O emission datasets and contributed towards interpretation of the results. VJ contributed towards project management. WK contributed towards the initial concept and interpretation of the results.
Karolina Pantazatou
AN wrote and implemented the code for calculating the posterior covariance matrices and performing the incremental optimisation procedure in Python, reshaped and processed the prior fluxes in Python and R, produced all figures using R, interpreted the results, and was responsible for the development of the manuscript. All co-authors contributed towards and reviewed the manuscript. RJS conceived the original concept and interpreted results. AV produced the FLEXPART footprints. JB and AB contributed to site identification and creation of the candidate site list. JA provided the LPJ output. UK, VK, and KP contributed to the FLEXPART runs and data management. MR processed the CH4 and N2O emission datasets and contributed towards interpretation of the results. VJ contributed towards project management. WK contributed towards the initial concept and interpretation of the results.
Veronika Jorch
AN wrote and implemented the code for calculating the posterior covariance matrices and performing the incremental optimisation procedure in Python, reshaped and processed the prior fluxes in Python and R, produced all figures using R, interpreted the results, and was responsible for the development of the manuscript. All co-authors contributed towards and reviewed the manuscript. RJS conceived the original concept and interpreted results. AV produced the FLEXPART footprints. JB and AB contributed to site identification and creation of the candidate site list. JA provided the LPJ output. UK, VK, and KP contributed to the FLEXPART runs and data management. MR processed the CH4 and N2O emission datasets and contributed towards interpretation of the results. VJ contributed towards project management. WK contributed towards the initial concept and interpretation of the results.
Werner Kutsch
AN wrote and implemented the code for calculating the posterior covariance matrices and performing the incremental optimisation procedure in Python, reshaped and processed the prior fluxes in Python and R, produced all figures using R, interpreted the results, and was responsible for the development of the manuscript. All co-authors contributed towards and reviewed the manuscript. RJS conceived the original concept and interpreted results. AV produced the FLEXPART footprints. JB and AB contributed to site identification and creation of the candidate site list. JA provided the LPJ output. UK, VK, and KP contributed to the FLEXPART runs and data management. MR processed the CH4 and N2O emission datasets and contributed towards interpretation of the results. VJ contributed towards project management. WK contributed towards the initial concept and interpretation of the results.