Estimating the extent of tropical forests in Africa around 1900

Accurate empirical estimates of historical forest extent and associated deforestation rates are crucial for quantifying tropical carbon cycles and formulating conservation policies. Until now, they have been lacking.
Published in Ecology & Evolution
Estimating the extent of tropical forests in Africa around 1900

Share this post

Choose a social network to share with, or copy the shortened URL to share elsewhere

This is a representation of how your post may appear on social media. The actual post will vary between social networks

Find our paper in Nature Ecology & Evolution here:

Previous estimates of deforestation rates in Africa since the 1900s range between 35 and 55%. These numbers – widely cited in the literature – often serve as basis for carbon mitigation strategies and conservation planning of forest systems. Our recent analysis, however, shows that only 21.7% of closed-canopy forests in tropical Africa has actually been converted since the 1900s. While this number is lower than the previous estimates, it is still considerable. Moreover, this rate represents a continental average, with strong spatial heterogeneity. For example, West and East Africa forests have been reduced by 80 and 90 percent, respectively. By contrast, in some areas of Central Africa, forests have in fact encroached onto savannas, resulting in net expansion. This finding has strong implications for savanna conservation.

The idea for the paper arose from our expertise with African savannas. As savanna ecologists, we were puzzled by the high deforestation estimates from Africa often cited by forest ecologists. These estimates suggested forests were historically very widespread in Africa, dominating in areas that savanna ecologists know to have high rates of endemicity in the savanna flora and where paleo-analyses have shown that savanna actually dominated. This sense – that historical forest extents have been overestimated – prompted us to adopt a systematic, data-based approach to estimating deforestation rates.  

As part of my post-doc with Carla on the long-term dynamics of biome distributions, I was already synthesizing paleodata from around Africa. One day, while discussing our frustrations about the pervasive (and, we felt, misguided) assumption that forests were historically widespread in Africa, she suggested that we should try to actually do something about it: perhaps we could use our dataset to reconstruct an empirical estimate of the extent of forest before widespread colonization.

Core of the water-sediment interface
Core of the water-sediment interface, taken with a kajak-brinkhurst corer, in a lake of the Republic of Congo. These sediments represent the last 150 years. [Picture by Carla Staver, August 2015]

However, this represented a challenge. Because paleodata are relatively rare in Africa in general, especially in parts of West Africa and the Congo Basin, particularly the Democratic Republic of Congo, spatially extensive reconstruction was tricky. To address this paleodata scarcity problem, I suggested that we also include information from historical maps and documents dating back to around 1900, which sometimes include description of vegetation. For example, we used the map of ‘the great forest region’ that Henry M. Stanley designed during the relief expedition of Emin Pacha, to draw the boundary between closed-canopy forest and savanna in the Eastern part of the Congo Basin.

We then needed to integrate data from both the paleo-proxies (pollen, phytolith, d13C and charred macro-remains) and the old maps. Paleo-proxies carry precise information about vegetation characteristics and thus biome type (forest vs savanna), but represent only several kilometers around a given paleosite. By contrast, historical maps are spatially explicit but the location of savanna-forest boundary may not be as accurate due to the wide range of names given to vegetation types (e.g. open forest, dry forest, grassland, savanna, woodland, bushland).

Lake and coring platform
Coring platform in a lake located in forest and savanna mosaics of the Republic of Congo. Cores of lacustrine sediments are being taken, and paleo-proxies are currently analyzed to reconstruct past vegetation and fire. [Picture by Carla Staver, August 2015]

Moreover, to combine these data sources, we needed an appropriate statistical tool, taking into account their various strengths and weaknesses. This tool also needed to spatially model the forest extent given other variables such as rainfall, soil types and distance to rivers. That’s when Marta joined us: as a specialist in quantitative and statistical ecology, she was very keen to apply her skills to this study.

This experience was not only about bringing together researchers with complementary skills. This paper is the first Marta and I have authored by women only, and we are very proud of it. As members of the Women in Science at Yale mentoring program, we are all very passionate and involved in promoting careers of women in STEM.

  To conclude, we bring our skills together to build this data-driven reconstruction of historical tropical forest and savanna distributions in Africa since 1900, a first such estimate at the continental scale. We hope that these revised estimates will inform carbon mitigation initiatives and conservation planning in both biomes. We also hope they can serve as a basis to stop misguided reforestation strategies that target old-growth savannas, and instead help target areas that have been truly deforested.

Julie Aleman, Marta Jarzyna and Carla Staver

Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in