Résumé
Aim This study calibrates the relationship between phytolith indices, modern vegetation structure, and a climate parameter (AET/PET, i.e. the ratio of annual actual evapotranspiration to annual potential evapotranspiration), in order to present new proxies for long-term Quaternary climate and vegetation changes, and model/data comparisons. Location Sixty-two modern soil surface samples from West Africa (Mauritania and Senegal), collected along a latitudinal transect across four bioclimatic zones, were analysed. Methods Two phytolith indices are defined as normalized data: (1) humidity-aridity index [Iph (%) ¼ saddle vs. cross + dumbbell + saddle], and (2) water stress index [fan-shaped index (Fs) (%) ¼ fan-shaped vs. sum of characteristic phytoliths]. Vegetation structures are delimited according to Iph and Fs boundaries. Bootstrapped regression methods are used for evaluating the strength of the relationship between the two phytolith indices and AET/PET. Additional modern phytolith assemblages, from Mexico, Cameroon and Tanzania are extracted in order to test the calibration established from the West African samples. Accuracy of the AET/PET phytolith proxy is compared with equivalent pollen proxy from the same area. Results Characterization of the grass cover is accurately made through Iph. A boundary of 20 ± 1.4% discriminates tall grass savannas from short grass savannas. Water stress and transpiration experienced by the grass cover can be estimated through Fs. AET/PET is accurately estimated from phytoliths by a transfer function: AET/PET ¼)0.605 Fs) 0.387 Iph + 0.272 (Iph-20) 2 (r ¼ 0.80 ± 0.04) in the application domain (AET/PET ranging from 0.1 ± 0.04 to 0.45 ± 0.04). Phytolith and pollen estimate with similar precision (r pollen ¼ 0.84 ± 0.04) the AET/PET in the studied area. Conclusions This study demonstrates that we can rely on the phytolith indices Iph and Fs to distinguish the different grasslands in tropical areas. Moreover, a new phytolith proxy of AET/PET, linked to water availability, is presented. We suggest from these results that combining phytolith and pollen proxies of AET/ PET would help to constrain this climate parameter better, especially when phytolith assemblages are dominated by Panicoideae and Chloridoideae C 4-grass phytoliths, are devoid of Pooideae C 3-grass phytoliths, and occur with a few tropical ligneous woody dicotyledon phytoliths. As AET/PET is a bioclimatic indicator commonly used in vegetation models, such a combination would help to make model/data comparisons more efficient.