Understanding carbon storage in sub-Saharan Africa
Posted on Wednesday 22 July 2026
Luke Misra reflects on tree planting schemes in South African savannas.
The largest emitter of CO2 on the African continent, ~440 Mt yr-1, South Africa is often considered the most industrialised nation in sub-Saharan Africa. It is little surprise, therefore, that tree planting schemes become such an attractive means of carbon offset for government bodies. Being cost-effective and easily scalable are draws, but the intuitiveness and salability of the message are the most appealing factor: more trees equals more carbon trapped. Even to a non-scientific audience, that message flies. And when confronted with large areas of savanna, a traditionally low carbon storage system, and international agreements which are legally binding, the natural response from policy makers is to plant up these grasslands and increase the carbon stock.
Schemes such as AFR 100, a continent-wide, World Bank-backed project, often fall into the same trap. At the heart of these schemes is a misunderstanding of what a savanna is. Global mapping tools used by AFR100 often use a simple definition of “forest” based purely on tree canopy percentage. Under these definitions, ancient, healthy savannas are frequently misclassified as “degraded forests” simply because they have wide-open spaces and fewer trees. Because they are flagged as “degraded,” they are aggressively targeted for mass tree-planting.
Furthermore, the value of savannas is often overlooked, not just from a carbon perspective where carbon estimates for savannas routinely exclude below-ground carbon due in part to difficulties in measurement and in part due to struggles with implementing methodologies from tropical forests where biomass is visibly stored aboveground compared to savannas where biomass below ground is often the significant piece of the pie. This also benefits habitats and grazing lands, while supporting entire South African communities built on ecotourism revenue.
It is perhaps unfair to talk about the misclassification of savannas without also mentioning where some of this misunderstanding around believing that savannas are degraded woodlands may originate. If we want a more comprehensive understanding of the ecosystem history of areas like Kruger National Park, it’s also important to understand the context in which the baseline conditions for parks like Kruger were established. The pre-European history of southern African savannas is poorly understood: woody cover, herbivore abundance and fire regimes prior to colonial settlement were largely undocumented, and the work to interpret local knowledge is still developing. At the time of the park’s establishment in the early twentieth century, these conditions were treated as the baseline, and have largely been treated that way for a century since. Yet Kruger was only established in 1926, 81 years after the Great Trek brought Voortrekker settlement to the Transvaal, and 274 years after Dutch colonisation of the Cape. There is also evidence of pastoral and agricultural communities in the region dating back to at least 500 AD. So rather than reflecting pristine wilderness, the baseline conditions established at Kruger’s founding likely captured a landscape recently emptied of people and livestock through land dispossession, with much of its fauna already reduced by disease and hunting. As conservation efforts have since helped herbivore populations recover, the resulting decline in woody cover has often been perceived as degradation, rather than as a return towards a more natural state. It is therefore important to contextualise current tree cover within historical variation ranges as opposed to a snapshot in its history.
But how do we actually explore the variation in historic tree cover ranges and what the associated carbon stocks might have been through time?
Previous work yielded pollen records from six water basins across the park, revealing which plants grew around these sites historically. Our current challenge, however, is interpreting not just past landscapes, but also the historical carbon stocks. To understand the past, we must first look to the present. This requires establishing the contemporary relationship between tree cover and carbon, which we can then apply to historical data. Let me first briefly explain how we assess the modern landscape. LiDAR (Light Detection and Ranging) is a remote sensing method, utilising laser pulses and measuring the time taken for the pulses to return to provide a high-resolution measurement of landscapes at a scale that traditional measurement methods would not be possible. NASA provides a service for this through their Global Ecosystem Dynamics Investigation (GEDI) programme. What this has provided us with is vegetation height and cover estimates for the whole park. GEDI also provides biomass transects. Simply put, to measure the above-ground carbon, we need the biomass, which we can then multiply by a biomass to carbon conversion constant, giving us the carbon in the above-ground vegetation. Our problem is that these transects do not necessarily line up with our study sites. To overcome this, we used a machine learning method called a Random Forest Model, which is able to make predictions by asking a series of yes or no questions of the available data. These predictions of site biomass can then simply be multiplied by a carbon conversion factor to give us site carbon; this has allowed us to begin to answer our question. Now that we understand this relationship, it can be applied to the past pollen records to inform us of carbon stock change through time.
It is worth noting that this work is not a total dismissal of tree-planting schemes. Rather, an acknowledgement that there is a risk of mistaking savanna for degraded land in the first place. Should tree-planting schemes continue to use tree-canopy percentage as a proxy for ecosystem health, and if carbon accounting keeps ignoring what’s stored below ground, then the risk is that some of Africa’s oldest, most biodiverse grassland systems will be converted into plantations.
