A team, led by the supervisor, has pioneered a new method using the 17O-excess of phytoliths, silica particles found in plants and used as a proxy for past vegetation, to quantify changes in relative humidity (RH). But essential questions remained before applying this new proxy in lacustrine sediments; this is what we addressed in the present project.
First, we determined the time frame in which the 17O-excess of phytoliths accurately reflects RH. We examined the 17O-excess in rainfall, soil water, grass and tree phytoliths at various sites in Benin and Senegal. We demonstrated that the 17O-excess of phytoliths accurately records RH of the growing season (+/- 6%), corresponding to the rainy season for C4 grasses, dominating in savannas, and to the entire year for evergreen and semi-evergreen trees in tropical dry and moist forests of West Africa.
We also explored in which tree tissues were produced globular granulate morphotypes found in large quantities in forest soils and characteristic of tropical trees. We sampled phytoliths from different parts of trees (bark, wood and leaves) and found that partial dissolution modifies morphotypes assemblages from tissues to soils and sediments. Identifying the source of globular granulate phytoliths proved challenging, as not all tree species produce this morphotype.
We then assessed whether phytoliths originating from riparian vegetation around Lake Guiers (Senegal) introduced a bias in 17O-excess measurements of the total sedimentary phytoliths assemblage. We found that phytoliths from this vegetation, as well as other grass species, recorded RH from the growing season and did not introduce significant bias.
We also examined how phytoliths were transported and deposited into these lake sediments. We compared top-core phytolith assemblages with soil samples from different land-cover types in the lake's catchment area, and with a land-cover analysis at different time period around Lake Guiers. The analysis revealed a shift toward more degraded and open vegetation in the watershed, associated with increased agriculture and population density.
After constraining this new proxy, we applied it to long-term paleo-sequences from Lake Guiers and Lake Ngofouo. At Guiers, there was a substantial drop in RH that coincided with vegetation changes c. 2000 BP, suggesting that a threshold of 65% RH was associated with the vegetation transition. Phytolith analysis of Lake Ngofouo suggests that around 2000 BP, the region was likely covered by a large swamp forest, the RH was high above 80%, with no recorded fires. The swamp forest started shrinking around 1500 BP, leading to increased watershed contributions and erosion episodes. At 1500 BP, fires emerged, causing forest fragmentation and the appearance of savanna, coinciding with a decline in RH. In this case, it seems that the transition from forest to savanna was not associated with a change in RH, but rather with an increased fire activity
These findings have been presented at conferences and discussed during radio interviews. The research resulted in one submitted paper, with three more in preparation. The project beneficiated from the framework, technical support and collaborations of two ANR projects, HUMI-17 (2017-2022) and PAST-17 (2023-2027).