20 000 years of african vegetation change

Submitted by editor on 16 May 2020.

By Leanne Phelps

Understanding the effects of climate-disturbance-ecosystem interactions is essential for producing accurate predictions of vegetation change. Yet large uncertainties exist in future projections of biodiversity and ecosystem change worldwide, especially when the effects of disturbance dynamics – e.g. changes in human land use, grazing and browsing regimes, and CO2 – are not considered. Grassy biomes, which are most extensive in Africa, are disproportionately affected by disturbance dynamics. As a result, projections of vegetation change are subject to particular uncertainty on the world’s largest tropical landmass. These uncertainties have sweeping ramifications for land management strategies.

In order to address this issue, the primary aim of our study was to reconstruct the effects of climate-disturbance-ecosystem interactions on vegetation envelopes of African forest and grassy biomes over the past 20 000 years. We achieved this by applying a multivariate envelope approach to subfossil pollen data and climate model outputs. In addition, we aimed to highlight potential links between concurrent changes in land cover and land use by proposing three hypotheses about the role that anthropogenic disturbance played in the observed trends. By quantifying changes in vegetation envelopes, we also helped to clarify continental-scale vegetation trends during the African Humid Period (ca 14 700 – 5500 BP), which was an interval of spatial and temporal variability in hydroclimatic conditions across much of Africa.

Our work provides the first continental-scale vegetation reconstructions for Africa from the last glacial period until recent times. Although our reconstructed trends are broadly in accordance with existing research on the African Humid Period, we described three non-linear vegetation responses to insolation change. First, the climatic envelope of grassy biomes expanded into increasingly diverse climatic conditions during the early to mid-Holocene, whereas that of forest did not. Second, forest retreat during the mid- to late Holocene occurred more slowly than forest expansion during the early African Humid Period (ca 147000 – 10000 BP). Third, the relationship between forest and grassy biomes fundamentally changed during the Holocene, as the density-based overlap between their envelopes became consistently low.

Changes in forest, grassy biomes and sampling coverage from the beginning of the African Humid Period until recent times, reconstructed on a modern climate grid (WorldClim).

These non-linear vegetation responses led us to propose three hypotheses about the potential role that anthropogenic disturbance played in vegetation change: (1) increasing anthropogenic disturbance associated with the spread of agriculture, pyrotechnologies and the Bantu expansion led to land clearances that increased the advantage of grassy biomes after the African Humid Period; (2) the expansion of the African animal production niche preferentially benefitted grassy biome development during the Holocene, and (3) interactions between anthropogenic disturbance and increasing CO2 formed alternative stable states – especially during and after the termination of the African Humid Period.

To achieve our study aims, we applied a novel approach to African vegetation reconstructions that was modified from several existing approaches. For example, our selection of vegetation groups was built upon existing work in plant functional type and biome classification schemes. Climatic vegetation envelopes were reconstructed and quantified for each of the vegetation groups by applying a modified form of niche dynamics, in line with a previous application by Phelps et al. (2020) to reconstruct the African animal production niche. Using these combined methods, we were able to reconstruct changing vegetation envelopes at 100-year intervals and to quantify trends over the past 20 000 years.

Methodological processes used to analysed subfossil pollen records for vegetation reconstructions.

Our methods departed from traditional biome approaches that assign one biome per grid cell. Instead, each grid cell was associated with a percentage for each vegetation type, permitting the existence of unknown vegetation combinations that do not necessarily exist today (non-analogous). Although plant species are likely to maintain their ancestral ecological traits, i.e. biome conservatism, we also excluded pollen taxa that occurred in multiple vegetation groups to filter out those that may change between biomes through time. Finally, by utilizing two different envelope metrics (density-based and binarized) to measure vegetation change, and two different approaches for grouping and reconstructing pollen taxa (direct and indirect), we were able to provide complementary types of information about vegetation change. While our approach does not aim to disentangle the drivers of vegetation change, it is useful for measuring the combined effects of climate-disturbance-ecosystem interactions, and allows us to pose informed hypotheses about the drivers of change.

One challenge involved in our work was finding ways to overcome the limitations of existing methodological approaches and data sources in order to appropriately address our research aims, e.g., reconstructing and comparing broad scale continental trends in forest and grassy biomes using spatially and temporally discontinuous records and variable plant functional type assignments. As described above, we overcame these obstacles by harmonizing different facets of different methodological approaches to maximize the strength of our reconstructions, to best suit the aims of our study, and to maintain the integrity of individual sedimentary records. Of course these reconstructions can and should be advanced in future studies by including more data and/or by applying complementary methodological approaches.

Ideally, future work will investigate different facets of vegetation change and the effects of individual drivers, as well as explicitly test the hypotheses we have proposed about the roles of anthropogenic disturbance.