Review of Palaeobotany and Palynology 158 (2009) 29–41 Contents lists available at ScienceDirect Review of Palaeobotany and Palynology j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / r e v p a l b o Phytoliths of East African grasses: An assessment of their environmental and taxonomic significance based on floristic data Doris Barboni a,⁎, Laurent Bremond b a b CEREGE (UMR6635 CNRS/Université Aix-Marseille), BP80, F-13545 Aix-en-Provence cedex 4, France Centre for Bio-Archaeology and Ecology (UMR5059 CNRS/Université Montpellier-2/EPHE), Institut de Botanique, 163 rue Broussonet, F-34090 Montpellier, France a r t i c l e i n f o Article history: Received 23 December 2008 Received in revised form 13 July 2009 Accepted 18 July 2009 Available online 25 July 2009 Keywords: phytolith Africa Poaceae silica grassland savannas paleoecology a b s t r a c t Relations between phytolith occurrences, taxonomy, and habitat are assessed for 184 East African grass species through the re-analysis of two qualitative surveys of phytolith types in leaf epidermis. This is done in conjunction with data on grass subfamily, photosynthetic pathway, and requirement for light and moisture compiled from floras and the literature. This survey includes ca 79% of the grass genera listed in the flora of tropical East Africa. It aims to further investigate the potential for grass short cell phytoliths to characterize the environment, and therefore improve reconstructions of past vegetation and climate in Africa. In this analysis, we identified ca 60 phytolith types (within the main categories Rondel, Trapeziform Short Cell, Bilobate, Cross, Polylobate, Saddle, and Trapeziform Sinuate) reported to occur in 10 grass subfamilies (Pharoideae, Bambusoideae, Ehrhartoideae, Pooideae, Danthonioideae, Arundinoideae, Chloridoideae, Centothecoideae, Panicoideae, and Incertae Sedis Streptogyna). These subfamilies include hydrophytic, helophytic, mesophytic and xerophytic species, with C3 or C4 photosynthetic pathways, and with affinities for shade, open, or semi-shade habitats. Analysis of phytolith occurrences shows that few morphotypes are restricted to some species only. However, there are morphological variations (of size and number of lobes) within the main phytolith categories Rondel, Bilobate, and Cross, which could additionally be considered to improve environmental and taxonomical interpretation of phytolith assemblages. The Rondel phytolith with a base diameter N 15 μm was only reported in C3-Pooideae, while the Rondel with a base diameter of b 15 μm occurs in several grass subfamilies (including Pooideae). Bilobates with long shanks between the two lobes are most frequently reported in xerophytic species, while Bilobates with short shanks are most frequently reported in mesophytic grass species. Finally, three-lobed crosses are reported only in Panicoideae and Chloridoideae, all being C4, light-loving species. A correspondence analysis confirms already known relationships between 1) Saddle forms, C4 pathway, open and xeric habitats, 2) Bilobates, Crosses, Polylobates, shaded and hydric habitats, 3) Trapeziform Sinuates and Pooideae. Of major implication for palaeoenvironmental reconstructions in East Africa, we found that the Trapeziform Sinuate phytoliths mark the presence of C3-grasses in the Afromontane zone, whereas the Rondels alone do not because they also occur in many C4 species of the Chloridoideae subfamily. We also establish that collapsed saddles are not diagnostic for Bambusoideae closed-habitat grasses since they occur in xerophytic species of the Chloridoideae, characteristic of several open habitats. In conclusion, this study contributes to better characterize Afromontane vegetation and better discriminate mesic and xeric vegetation types in East Africa. It also brings caution for future phytolith studies that rely on the presence of diagnostic types instead of phytolith assemblage analysis to trace the presence of particular taxa and/or environments in East Africa. © 2009 Elsevier B.V. All rights reserved. 1. Introduction The taxonomic resolution of grass (Poaceae) phytoliths is difficult to tackle because a given taxon produces multiple phytoliths (multiplicity), and a given phytolith may be produced by many taxa (redundancy). Within the Poaceae, however, subfamilies Pooideae ⁎ Corresponding author. Tel.: +33 442 97 17 66. E-mail address: barboni@cerege.fr (D. Barboni). 0034-6667/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.revpalbo.2009.07.002 (C3, high elevation grasses) on the one hand, Panicoideae (mainly C4, tall grasses) on the other hand, as well as Chloridoideae (C4, short grasses) produce in high amount one category of short cell phytoliths (Twiss et al., 1969), which makes the relative abundance of each category as well as their ratio (also called phytolith indices) informative on the grass subfamily dominance (Diester-Haass et al., 1973; Twiss, 1992; Alexandre et al., 1997). On this basis, modern phytolith assemblages collected from soil humic horizons were calibrated in terms of vegetation physiognomy and climate parameters (Fredlund and Tieszen, 1994; Alexandre et al., 1997; Fredlund 30 D. Barboni, L. Bremond / Review of Palaeobotany and Palynology 158 (2009) 29–41 and Tieszen, 1997; Bremond et al., 2005a,b; Bremond et al., 2008). Another method for interpreting phytolith data relies on the presence of phytolith types exclusively observed in (i.e. diagnostic of) a given taxon at a regional scale. This method requires knowing the phytolith content of a regionally representative plant reference collection, and involves a higher level of morphological discrimination among phytoliths than the previous one (Piperno and Pearsall, 1998; Mindzie et al., 2001; Pearsall et al., 2003; Strömberg et al., 2007). Grass short cell phytoliths may be classified according to their threedimensional shape into seven main categories: Rondel, Trapeziform short cell, Bilobate, Cross, Polylobate, Trapeziform Sinuate, and Saddle (adapted from Twiss et al., 1969; Mulholland, 1989). However, one may observe differences in size, outline and number of lobes, etc. within each category. Hence, while studying Pliocene samples from Ethiopia, more than 60 different types of grass short cell phytoliths could be identified and counted separately (D. Barboni, unpublished data). Such variety of shapes within the main categories suggests that taxonomical and/or environmental resolution of short cell phytoliths could be improved to infer more precision while interpreting fossil phytolith assemblages preserved in rock strata. Also, it is suggested that greater taxonomical identification of grass phytoliths could lead to discriminating the shadeloving basal grasses from the light-loving crown grasses of the phylogeny tree, and therefore improve the environmental interpretation of phytoliths (Strömberg, 2005). This is especially true when arboreal phytoliths are under-represented despite the abundance of trees in the vegetation (Barboni et al., 2007; Bremond et al., 2008). Misrepresentation of arboreal versus grass phytoliths occurs elsewhere (Albert et al., 2006), thus seeking information on the tree cover directly from phytoliths of shade-loving grasses may be worthwhile. Phytoliths are non-symmetrical silica bodies and their precise identification under the microscope is time-consuming. This study, therefore, aims to assess how far grass phytoliths should be morphologically discriminated (beyond the main categories) to improve the paleoenvironmental significance of East African fossil phytolith assemblages. It also aims to test the potential of East African grass short cell phytolith (main categories and derived types) to relate to grass taxonomy, and grass photosynthetic pathways, and to environmental parameters such as moisture and light availability in the environment. We used published descriptions and micrographs of phytoliths available for East African and South Arabian grass species (Palmer and Tucker, 1981, 1983; Palmer et al., 1985; Palmer and Gerbeth-Jones, 1986; Ball, 2002) to look for patterns of occurrences. In total, phytolith data were obtained for 184 species, which represent ca 20% of the species and ca 79% of the genera of Poaceae listed in the flora of tropical East Africa (Clayton, 1970, 1974, 1982). Also, for the first time here, the phytolith morphological diversity of East African grasses is assessed and tested against the newly revised grass systematics (Grass Phylogeny Working Group, 2001). 2. Materials and methods Short-cell phytoliths from 142 grass species were described and photographed by Palmer and colleagues (Palmer and Tucker, 1981, 1983; Palmer et al., 1985; Palmer and Gerbeth-Jones, 1986) who analyzed the most characteristic Poaceae species of a wide range of East African vegetation types and habitats (Palmer, 1976). Palmer and colleagues (Palmer and Tucker, 1981, 1983; Palmer et al., 1985; Palmer and Gerbeth-Jones, 1986) considered species from the region dealt with in the publication of the flora of tropical East Africa (FTEA) i.e. the region including (at that time) Uganda, Kenya, and Tanzania (Clayton, 1970, 1974, 1982). In addition, phytoliths from 42 grass taxa photographed under optical microscope by Ball (2002) were investigated (hereafter referred to as Ball's dataset). The grass taxa represent some common genera from the Southern Arabian Peninsula. They were sampled in the Dhofar region of Oman (T. Ball, personal comm., and S. Gazanfar, Kew Botanical Garden, personal comm.). Altogether Palmer and Ball's datasets consist of 184 grass species. These species are from very diverse habitats ranging from coastal saline lowlands to humid highlands, including dry open savannas, moist forests, and various riparian and lacustrine vegetation types. Given the ranges of habitats and the geographical extend (four countries) represented by this dataset, it is therefore unlikely that the sampling of grasses by Palmer and colleagues (Palmer and Tucker, 1981, 1983; Palmer et al., 1985; Palmer and Gerbeth-Jones, 1986) and Ball (2002) biases the patterns of phytolith occurrence identified here. Phytolith descriptions and micrographs of the leaf epidermis were methodically surveyed. Phytoliths that were not described by the authors but visible on the micrographs were also reported. Phytoliths types were named (or renamed) following the International Code for Phytolith Nomenclature, i.e. considering their three-dimensional shape (ICPN Working group, 2005) (Table 1). For each grass species, all phytolith types were reported, in order to obtain a table of presence/absence data (Appendix A, Supplementary data). Palmer and colleagues (Palmer and Tucker, 1981, 1983; Palmer et al., 1985; Palmer and Gerbeth-Jones, 1986) retained 28 different types of phytoliths. However, examining the scanning electron micrographs (SEMs), which display phytoliths as they occur on top of the leaf epidermis, allowed us to recognized 14 additional types of phytoliths (Table 1, Fig. 1, Appendix A). We made our own record for saddle types that we distinguished according to the length of the convex sides (the so-called “axe-edges”, Prat, 1948). We interpreted the “Intercostal tall and narrow” phytolith type as Rondels with concave base or Stipa-type bilobates (sensu Fredlund and Tieszen, 1994) (e.g. in Arundinaria alpina and Phyllorhachis sagittata, Palmer and Tucker, 1981). The type named “Tall and narrow” in Palmer's dataset is unclear, and our attributions to types of the nomenclature vary from one species to the other. Phytoliths classified as “Tall and narrow” resemble Trapeziform short cells in Brachypodium flexum, Short Saddle in Eleusine floccifolia, and Rondel in Paspalum conjugatum (Appendix A). Phytoliths with circular and elliptical bases were identified as Rondels and differentiated according to the outline of the base; their top (which can be acute, keeled, ridged, or with spikes) being oriented towards the interior of the leaf is not visible on the SEMs of the grass epidermis. We considered phytoliths with oblong slightly constricted base as Stipa-type bilobates (sensu Fredlund and Tieszen, 1994) or “shovel” bilobates (bilobate in top view, but top concave in side view, sensu Strömberg, 2003), because the 2D view does not allow precise identification. Palmer and colleagues described Bilobate phytoliths according to the length and width of the connecting part between the lobes (called shank), without providing measurements, but we observed that Bilobates reported with long shanks, in general, exhibit shanks longer than half of the lobes. Unlike SEM photomicrographs, light microscope micrographs of the extracted phytoliths allow the observation of silica bodies in the three dimensions. From Ball (2002), we could therefore distinguish 38 phytolith types within the five broad originally described categories (Table 1, Fig. 1, Appendix A). For each grass species, we compiled information on their subfamily (Grass Phylogeny Working Group, 2001), their photosynthetic pathway, and their requirement for light and moisture (Watson and Dallwitz, 1992 onwards) (Appendix A). Ecological characters were assigned at the genus level or at the species level when a genus includes species with opposing characters. In this case, we sought information on web sources, essentially the Aluka digital library of African plants (http://www.aluka.org), the grass database of the FAO (http://www.fao.org/AG/AGp/agpc/doc/GBASE/Default.htm) and the literature (Carr, 1998). In most cases the grass species were assigned several modes for moisture-requirement, which revealed their large tolerance to moisture levels. For example Paspalum conjugatum can show the three adaptations: helophytic, mesophytic, and xerophytic (Watson and Dallwitz, 1992 onwards). D. Barboni, L. Bremond / Review of Palaeobotany and Palynology 158 (2009) 29–41 31 Table 1 Phytolith types identified in 184 grass species from East Africa (Palmer and colleagues 1981, 1983, 1985, 1986; 142 species, 28 phytolith types) and the Southern Arabian Peninsula (Ball, 2002; 42 grass species, 5 phytolith types). Phytoliths original names Phytolith names following ICPN (2005) and types we further differentiated using the micrographs Coding for CA Palmer and colleagues (Palmer and Tucker, 1981, 1983; Palmer et al., 1985; Palmer and Gerbeth-Jones, 1986) Acutely-angled Rondel acutely angled on bottom/top view or Cross angular lobes Crenate-vertical Bilobate with extra-lobe on shank Crescent-shaped Rondel, base with crescentic outline or Saddle Cross-shaped Cross 4-lobed Dumbbell ends concave Bilobate concave lobes Dumbbell ends rounded Bilobate round lobes Dumbbell ends rounded + end knob Bilobate round lobes with knob on end lobes Dumbbell ends straight Bilobate truncated/slightly concave lobes Dumbbell middle narrow + long Bilobate shank narrow + long Dumbbell middle narrow + short Bilobate shank narrow + short Dumbbell middle wide + long Bilobate shank wide + long Dumbbell middle wide + short Bilobate shank wide + short Dumbbell middle saddle-shaped Bilobate in top view, in side view top is concave (“shovel-bilobate”) Elliptical Rondel elliptical base Elongated cross-shaped Bilobate with very concave lobes Elongated-sinuous Trapeziform sinuate Elongated-smooth Trapeziform sinuate Figure-eight Bilobate round lobes, possibly Stipa-type Intercostal tall & narrow Rondel with concave base or Stipa-type bilobate Narrowly saddle-shaped Seems to refer to saddles with short convex edges Nodular Polylobate Oblong Trapeziform sinuate Round Rondel round base Saddle-shaped Saddle (includes symmetrical, long, & short saddles) Saddle-shaped with lateral lobes Saddle with small lobes on the concave edges Square Trapeziform short cell Tall & narrow Unclear. Includes long saddles, Rondels, & Trapeziform sinuates Widely saddle-shaped Seems to refer to saddles with long convex edges Acutely-angled Cren-vert. Crescent Cross Du-concave Du-round Du-knob Du-straight Du-n.l Du-n.s Du-w.l Du-w.s Du-saddle Elliptical El-cross El-sinuous El-smooth Figure-eight Intercostal Narrowly s. Nodular Oblong Round Saddle-shaped Saddle-lobate Square Tall, narrow Widely s. Additional types we identified in atlases of SEMs by Palmer and colleagues (numbers refer to volume and plate) I-21f, IV-54d,68d Bilobate with lobes staggered in top/end view I-23e, II-23f,29f,52d, III-81f, IV-1e,39d Bilobate Stipa-type or “shovel” III-12e,28d,81d, IV-75e,80d Cross 3-lobed I-10e,31c, II-30e, IV-42d,57e,78d,86d Polylobate, ends rounded, wide + short shank IV-49e,75c Polylobate, ends rounded, narrow + short shank I-31c, II-48e,49e, III-9e,18e, IV-20c Polylobate, ends truncated/concave, wide + short shank I-33e,40d, IV-35d,72d Polylobate, ends trunc./conc., narrow + long shank II-44d Polylobate, ends trunc./conc., wide + short shank, alternate lobes II-35f,52e Rondel base outline irregular II-41e Rondel tabular oblong e.g.: III-22c Saddle symmetrical, concave and convex edges of the same length e.g.: III-76d,29d Saddle short convex edges e.g.: III-31e Saddle long convex edges e.g.: II-37f Trapeziform sinuate BI-28 BI-29,31,34 CR-51 Pol-A Pol-B Pol-C Pol-D Pol-E RO-7 RO-5 SA-62 SA-63 SA-64 SI-53 Ball (2002) Round–trapezoid short cell Bilobate short cell Crossbody short cell Polylobate short cell Saddle short cell Rondel Bilobate Cross Polylobate Saddle Rondel Bilobate Cross Polylobate Saddle Additional types we identified on Ball's micrographs (numbers refer to volume and plate) e.g. slide 293 Rondel elliptical base, top keeled e.g. slide 225 Rondel elliptical base, top truncated e.g. slide 3 Rondel elliptical base, perfect cylinder e.g. slide 3 Rondel base outline irregular, top flat e.g. slide 185 Rondel sinuous base e.g. slide 185 Rondel sinuous base, cylindrical e.g. slide 68 Trapeziform cubic e.g. slide 108 Trapeziform ridged top or keeled e.g. slide 40 Bilobate assymetrical lobes e.g. slide 138 Bilobate with extra-lobe on shank e.g. slide 138 Bilobate round lobes, thick shank e.g. slide 68 Bilobate round lobes, narrow + short shank e.g. slide 50 Bilobate round lobes, short + thick shank e.g. slide 142 Bilobate angular lobes e.g. slide 71 Bilobate concave lobes e.g. slide 71 Bilobate truncated lobes e.g. slide 314 Bilobate with lobes staggered in top/end view e.g. slide 204 Bilobate Stipa-type e.g. slide 204 Bilobate trapeziform, flat/keeled top, base lobate, base ≫ top RO-1 RO-3 RO-6 RO-7 RO-9 RO-11 TR-15 TR-18 BI-20 BI-22 BI-23 BI-rd,ns BI-24 BI-25 BI-26 BI-27 BI-28 BI-29 BI-31 (continued on next page) 32 D. Barboni, L. Bremond / Review of Palaeobotany and Palynology 158 (2009) 29–41 Table 1 (continued) Phytoliths original names Phytolith names following ICPN (2005) and types we further differentiated using the micrographs Additional types we identified on Ball's micrographs (numbers refer to volume and plate) e.g. slide 366 Bilobate trapeziform, lobes falling down the ridge/rectangular top, base concave e.g. slide 174 Bilobate trapeziform, round lobes, base ≫ top at lobes only e.g. slide 174 Bilobate parallelepipedal round/concave lobes, base N top at the lobes, but same length for base and top e.g. slide 216 Bilobate parallelepipedal, alate e.g. slide 174 Bilobate parallelepipedal, lobes slightly larger than shank e.g. slide 18 Cross tabular, 4 round lobes e.g. slide 51 Cross tabular, 4 angular lobes e.g. slide 142 Cross trapeziform e.g. slide 142 Cross tall trapezoid, lobes hanging down the top e.g. slide 71 Cross 3-lobed irregular e.g. slide 185 Polylobate parallelepipedal, concave lobes e.g. slide 185 Polylobate base N top at lobes only e.g. slide 185 Polylobate flat/rectangular top with ridge, base with round/angular lobes e.g. slide 185 Polylobate thin shanks, flat/rectangular top with ridge, base with staggered lobes e.g. slide 314 Polylobate base ≫ top at side lobes only, lobes round/acute e.g. slide 116 Saddle symmetrical e.g. slide 116 Saddle short convex edges e.g. slide 116 Saddle long convex edges e.g. slide 225 Saddle collapsed, i.e. very concave on one side in side view, but saddle-like in base view Coding for CA BI-32 BI-33 BI-35 BI-36 BI-paral CR-46 CR-47 CR-48 CR-50 CR-51 NewPO-1 NewPO-2 Pol-55 Pol-56 Pol-57 SA-62 SA-63 SA-64 SA-65 We give here the original phytolith names and equivalences following the ICPN (2005). Additional phytolith types that we distinguished in this study using the original micrographs are also listed here with reference to the volumes and plates (e.g. vol I, plate 21f). Coding is for reading the correspondence analysis plots. In order to evaluate the relationships between phytolith occurrence data, taxonomy, and ecology of grasses we first analyzed the occurrence of each of the main phytolith categories and derived phytolith types, and then ran a correspondence analysis (CA) on the matrix of presence/ absence data with grass species in lines, and ecological characters and phytoliths in columns. In the CA, grasses are the observations (relevés or samples), ecological characters and phytoliths the observed elements (species). Original phytolith data (names and occurrences as given by the authors) and phytolith data that we obtained from observing the micrographs were pooled into the same data matrix, but coding for phytoliths allows distinguishing the two types of variables. We used CANOCO software for the CA (Ter Braak and Smilauer, 1998). 3. Results 3.1. Relations between grass subfamilies, photosynthesis, light- and water-requirement The grass species investigated for their short cell phytolith content belong to 10 subfamilies of Poaceae (Table 2). The most represented are Chloridoideae (67 species), Panicoideae (64 species), and Pooideae (26 species), but Pharoideae, Bambusoideae, Ehrhartoideae, Danthonioideae, Arundinoideae, and Centothecoideae subfamilies are represented as well (Palmer and Tucker, 1981, 1983; Palmer et al., 1985; Palmer and Gerbeth-Jones, 1986). In total, 60 species have a C3 photosynthetic pathway, and 120 have a C4 photosynthetic pathway. Four Panicoideae can be both C3- and C4-species (Watson and Dallwitz, 1992 onwards). Among C3-species, 24 grow in shaded habitats, 25 in open habitats, and 11 may be found in both shaded or open habitats. To our knowledge, among the surveyed C4-species, none are true shade-loving, but rather semi-shade loving species. There are no common species to Palmer's and Ball's datasets, but duplicates may be present since species names are not available for Chloris, Tripogon, Digitaria, Echinochloa, Panicum, Setaria, and Sporobolus in Ball's dataset (2002). There are obvious relations between grass subfamilies, photosynthetic pathway, light- and moisture-requirement (Table 2). For example, most C3-Panicoideae (8/13) are found in shaded-habitats, while most C4-Panicoideae (32/47) are found in open-habitats. Among C3-grasses, those favoring open-habitats are mostly of the Pooideae subfamily (high elevation grasses in the tropics), while C3-grasses favoring shaded- Fig. 1. Drawing of the principal phytolith types identified in 184 grass species from East Africa (Palmer & Tucker, 1981; 1983; Palmer et al, 1985; Palmer & Gerbeth-Jones, 1986) and the Southern Arabian Peninsula (Ball, 2002). Drawings with an arrow and a “?” are the expected 3D shape. Magnification: × 640. 1. Rondel acutely angled on bottom/top view or Cross angular lobes (Phytolith names following ICPN, 2005); 1a. Coelachne africana, (Palmer and colleagues 1981, 1983, 1985, 1986: Volume IV: pages 6, 46, and 47); 1b. Heteranthoecia guineensis, (IV:6, 48, 49); 1c. Brachypodium flexum, (II:7, 38, 39); 2. Bilobate with extra-lobe on shank; 2a. Oryza punctata, (I:21, 54, 55); 2b. Sporobolus africanus. (III:28, 120, 121); 2c. Urochloa mosambicensis, (IV:30, 118, 119); 2d. Oplismenus burmannii, (IV:21, 92, 93); 2e. Cynodon dactylon (Ball, 2002); 3. Rondel with concave base or Saddle; 3a. Arundinella nepalensis, (IV:2, 34, 35); 3b. Sporobolus africanus. (III:28,120, 121); 3c. Sporobolus africanus. (III:28, 120, 121); 3d. Phalaris arundinacea. (II:13, 60, 61); 3e. Puelia olyriformis, (I:17, 42, 43); 4. Cross 4-lobed; 4a. Eragrostis macilenta, (III:12, 68, 69); 4b. Olyra latifolia, (I:8, 18, 44, 45); 5. Bilobate concave lobes and 11. Bilobate shank wide + long; Enneapogon cenchroides, (III:3, 42, 43); 6. Bilobate round lobes and 9. Bilobate shank narrow + long; 6a and 9a. Zonotriche inamoena, (IV:5, 44, 45); 6b and 9b. Anthephora elongata, (IV:10, 58, 59); 6c and 10. Bilobate shank narrow + short; Holcolemma inaequale, (IV:18, 84, 85); 7. Bilobate round lobes with knob on end lobes; Dinebra retroflexa (III:9, 58, 59, 11), Entolasia imbricata, (IV:17, 80, 81); 8. Bilobate truncated/slightly concave lobes and 12. Bilobate shank wide + short; Anthephora elongata, (IV:10, 58, 59); 13. Bilobate in top view, top is concave in side view (“shovel-bilobate”); Ehrharta erecta var. abyssinica, (I:23, 60, 61); 14. Rondel elliptical base; 14a. Asthenatherum glaucum, (I:24, 64, 65); 14b. Brachypodium flexum, (II:7, 38, 39); 15. Bilobate with very concave lobes; 15a, b. Maltebrunia leersioides, (I:6, 20, 52, 53; II:7. 38, 39); 16. Trapeziform sinuate; 16a. Lolium temulentum, (II:4, 28, 29); 16b. Helictotrichon elongatum, (II:12, 56, 57); 17. Trapeziform sinuate smooth; 17a. Colpodium chionogeiton, (II:3, 22, 23); 17b. Vulpia bromoides, (II:6, 34, 35); 18. “Figureeight”, Bilobate round lobes (18b), possibly Stipa-type (18a); 18a. Oreobambos buchwaldii, (I:16. 38, 39); 18b. Olyra latifolia, (I:8, 18, 44, 45); 19. Sinuate trapeziform; 19a. Psilolemma jaegeri, (III:17, 86, 87); 19b. Koeleria capensis, (II:13, 58, 59); 20. Rondel round base; Deschampsia caespitosa var. oliveri, (II:12, 54, 55); 21. Saddle with small lobes on the concave edges; Acrachne racemosa, (III:5, 46, 47); 29. Bilobate with lobes staggered in top/end view; 29a. Pennisetum purpureum, (IV:23, 100, 101); 29b. Humbertochloa greenwayi, (I:12, 22, 56, 57); 29c. Hylebates chlorochloe, (IV:19, 86, 87); 30. Bilobate Stipa-type or “shovel”; 30a. Phyllorhachis sagittata, (I:22, 58, 59); 30b. Secale africanum, (II:8 42, 43); 30c. Aira caryophyllea, (II:10, 48, 49); 30d. Stipa dregeana, (II:16, 70, 71); 30e. Crypsis schoenoides, (III:27, 118, 119); 30f. Arundinella nepalensis, (IV:2, 34, 35); 30g. Crytococcum multinode, (IV:14. 72, 73); 31. Cross 3-lobed; 31a. Apochiton burtii, (III:5, 48, 49); 31b. Crypsis schoenoides, (III:27, 118, 119); 31c. Setaria plicatilis, (IV:26, 108, 109); 31d. Stenotaphrum dimidiatum, (IV:28, 112, 113); 32. Polylobate, ends rounded, wide + short shank; 32a. Olyra latifolia, (I:8, 10, 18, 44, 45); 32b. Crinipes abyssinicus, (I:7, 25, 66, 67); 32c. Snowdenia petitiana, (IV:27, 110, 111); 33. Polylobate, ends rounded, narrow + short shank; 33a. Eriochloa meyerana, (IV:17, 82, 83); 33b. Setaria plicatilis, (IV:26, 108, 109); 34. Polylobate, ends truncated/concave, wide + short shank; 34a. Acritochaete volkensii, (IV:8, 52, 53); 34b. Crinipes abyssinicus, (I:7, 25, 66, 67); 34c. Gastridium phleoides, (II:15, 66, 67); 35. Polylobate, ends truncated/concave, narrow + long shank; Elytrophoms globularis, (I:11, 12 26, 68, 69); 36. Polylobate, ends truncated/concave, wide + short shank, alternate lobes; Agrostis schimperiana, (II:14, 62, 63); 37. Rondel base outline irregular; 37a. Deschampsia caespitosa var. oliveri, (II:12, 54, 55); 37b. Stipa dregeana, (II:16, 70, 71); 38. Rondel tabular oblong; Phalaris arundinacea, (II:13, 60, 61); 39. Saddle symmetrical, concave and convex edges of the same length; Humbertochloa greenwayi, (I:12, 22, 56,57); 40. Saddle short convex edges; 40a. Eragrostiella bifaria, (III:11, 66, 67); 40b. Rendlia altera, (III:25, 112; 113); 40c. Eragrostis macilenta, (III:12, 68, 69). 41. Saddle long convex edges; Eragrostis macilenta, (III:12, 68, 69). D. Barboni, L. Bremond / Review of Palaeobotany and Palynology 158 (2009) 29–41 33 34 D. Barboni, L. Bremond / Review of Palaeobotany and Palynology 158 (2009) 29–41 Table 2 Number of grass species from East Africa analyzed for their phytolith content by Palmer et al. (1981, 1983, 1985, 1986) and by Ball (2002) grouped by subfamily, photosynthesis, light- and moisture-requirements. Photosynthesis/light-requirement Sub-families No. of species C3 Shade Moisture-requirement C4 S/O Open C3/C4 Shade S/O Open Palmer and colleagues (Palmer and Tucker, 1981, 1983; Palmer et al., 1985; Palmer and Gerbeth-Jones, 1986) Pharoideae 1 1 Bambusoideae 5 5 Incertae Sedis (Streptogyna) 1 1 Ehrhartoideae 6 3 2 1 Pooideae 26 2 7 17 Danthonioideae (°) 5 1 3 1 Arundinoideae (°) 3 1 1 1 Chloridoideae 49 4 43 Centothecoideae 3 3 Panicoideae 43 8 2 3 9 20 Total 142 24 11 25 14 65 Ball (2002) Arundinoideae Chloridoideae Panicoideae Total 3 18 21 42 6 6 3 18 12 33 ? Hy Me Xe S/O 1 2 2 He 1 1 3 3 3 8 1 1 10 1 5 1 3 25 2 4 5 21 44 33 3 37 110 1 1 1 3 4 15 20 35 14 3 2 27 11 57 3 18 17 38 S/O: species reported in both shade- and open-habitats; ?: habitat unknown; C3/C4: species reported with both photosynthetic pathways; °: only C3 species according to GPWG (2001). Hy: hydrophytic, He: Helophytic, Me: mesophytic, Xe: xerophytic. NB: some species were assigned several modes for moisture requirement according to their tolerance to various moisture levels. Fig. 2. Proportion of grass species with bilobate phytoliths with short and long shanks as a function of species moisture-requirement (obtained using data from Palmer et al., 1981, 1983, 1985, 1986). habitats occur in all the subfamilies surveyed here except Arundinoideae. C3 non-Pooideae grasses (i.e. essentially low elevation grasses) mostly include helophytic species in open-habitats (e.g. riverbanks), and mesophytic species in shaded-habitats (e.g. in forest understories). C4lowland grasses mostly favor open-habitats (98/118), while few species are found in semi-shade (10/118). The later may be helophytic/ mesophytic Panicoideae (9 species) or mesophytic Chloridoideae (4 species). C4-lowland grasses found in open-habitats are mostly xerophytic and mesophytic/xerophytic Chloridoideae, or mesophytic and helophytic/hydrophytic Panicoideae. 3.2. Morphological differences within main phytolith categories Analyzing the occurrences of phytolith types in each of the 184 grass species (Appendix A) show that few types are restricted to some species only. Bilobates with alate, deeply concave lobes (“Elongated crossshaped”, Table 1, Fig. 1:15a) are reported only in Maltebrunia leersioides Fig. 3. Range of sizes for Rondel phytolith types in various subfamilies of Poaceae from East Africa and Southern Arabia. Sizes were measured directly on the micrographs published in Palmer et al. (1981, 1983, 1985, 1986) and Ball (2002). D. Barboni, L. Bremond / Review of Palaeobotany and Palynology 158 (2009) 29–41 (Palmer and Tucker, 1981, p.53), Ehrhartoideae, C3, shade species. Rondel tabular oblong (Table 1, Fig. 1:38) is observable only in Phalaris arundinacea (Palmer and Tucker, 1983, p.60), a C3-Pooideae species found in damp soils and swamps. Saddles “with lateral knobs” and bilobates “with knobs on end lobes” (Table 1, Fig. 1:7, 21) (Palmer et al., 1985, p.47, p.58) exclusively occur in Chloridoideae, C4, open-habitat species. Paying greater attention to morphological variations within the seven main phytolith categories shows that: (1) In the category Bilobate, the relative length of the shank increases with increasing water-requirement of the grass species (Fig. 2). 35 Among species with short shank bilobates (Fig. 1:6c, 10), 36% are mesophytic and less than 14% are xerophytic. On the contrary, long shank bilobates (Fig. 1:11) are reported more frequently in xerophytic than in mesophytic grasses. (2) In the Cross category, 3-lobed crosses (Fig. 1:31) are reported only in Panicoideae and Chloridoideae species (10 species in total, which are all light-loving C4-species), while the 4-lobed crosses (Fig. 1:4) are reported in seven different subfamilies (72 species in total, ca 40% of the dataset). (3) In the Saddle category, Saddles with long convex edges (Fig. 1:41) are reported only in Chloridoideae (10/67 species). Other Saddle types are redundant. Saddles with short convex Fig. 4. Proportions (and numbers) of species within each grass subfamily showing at least one of the eight main categories of short cell phytolith in their epidermis. Data synthesized from Palmer et al. (1981, 1983, 1985, 1986) and Ball (2002). 36 D. Barboni, L. Bremond / Review of Palaeobotany and Palynology 158 (2009) 29–41 edges (Fig. 1:40) are frequently observed in Chloridoideae (18 species over 67), but also in Arundinoideae (Phragmites mauritianus), Bambusoideae (Oreobambos buchwaldii, Oxytenanthera abyssinica), and in the Incertae Sedis Streptogyna crinita. It is not possible to see on the SEMs if collapsed saddles (saddles with one very concave side in side view, but saddle-like in base/ top view) occur in any of the East African Bambusoideae species (Palmer and Tucker, 1981), but the collapsed saddle type is observable in five Chloridoideae species, namely Chloris gayana, Chloris sp., Dactyloctenium aegyptium, Ochthochloa compressa, and Sporobolus sp. from Southern Arabia (Ball, 2002). (4) In the Rondel category (Fig. 1:3, 14, 20, 37, 38), although there is an overlap of sizes, the base diameter of Rondel phytoliths tends to be bigger in Pooideae (5.4 µm to 19.5 μm) than in Chloridoideae (4.5 μm to 12.5 μm), Panicoideae (5.7 μm to 14.1 μm), Bambusoideae (Arundinaria alpina) and Danthonioideae (Asthenatherum glaucum syn Centropodia glauca), but shape and outline are similar (Fig. 3). Apart from the size, Rondel types from Pooideae and Chloridoideae show no other visible morphological differences. Three Chloridoideae species seem to have only Rondel phytoliths in their epidermis (Urochondra setulosa, Sporobolus tenuissimus, and Sporobolus sp.-slide 119 in Ball 2002), but these show no particularity (bases are round, elliptical, crescent, and tops are flat, spiked, or keeled). Rondels with crescent-shaped base (Fig. 1:3a) occur in three species of Pooideae as well as in two species of Chloridoideae (Sporobolus ioclados and Sporobolus sp.-slide 119, Ball, 2002), and therefore do not help discriminating these two grass subfamilies. The three other categories of phytolith types show no apparent differences in their morphology that would be characteristic of a given subfamily. In the Polylobate phytolith category, no relation is observed between the shape of the lobes and any grass type or subfamily, or between the length of the shanks and moisture. Trapeziform Sinuates and Trapeziform short cells reported in Pooideae are morphologically identical to those produced by Bambusoideae, Chloridoideae, and Panicoideae. 3.3. General pattern of occurrence of main phytolith types in East African grass subfamilies The seven main short cell phytolith categories are redundant to several grass subfamilies (Fig. 4, Appendix A). However, a general pattern is evident. For Panicoideae, phytolith categories most frequently reported are Bilobate (in 61 over 64 species), Polylobate (42/64), and Cross (40/64). Arundinella nepalensis, Coelachne africana, and Heteranthoecia guineensis are the only Panicoideae species for which Bilobates or Polylobates are not reported. Instead, A. nepalensis is reported with Saddles, C. africana and H. guineensis with acutely angled phytoliths of the Cross category (Fig. 1:1). For Chloridoideae, phytoliths most frequently reported are Saddles (mainly symmetrical and short) (44/67), and Bilobates (31/67). Chloridoideae devoid of Saddles (15/67) show Cross, Bilobate or Rondel phytoliths in their epidermis. For Pooideae, phytoliths most frequently reported are Trapeziform Sinuates (22/26). Fourteen species of Pooideae are reported with “oblong” phytoliths, which are classified according to the nomenclature as Trapeziform Sinuate or Rondel (ICPN Working group, 2005) (see Appendix A for details). Pooideae species without Trapeziform Sinuates are reported with Rondels, Stipa-type Bilobates, or Polylobates in their epidermis. Phytoliths of the Trapeziform Sinuate category are also observed in Puelia olyriformis (Bambusoideae), Psilolemma jaegeri (Chloridoideae), Pennisetum cenchroides (syn Cenchrus ciliaris Lin.), Cyrtococcum multimode, Oplismenus burmannii, Paspalum conjugatum, and Tricholaena teneriffae (Panicoideae). For the three surveyed species of Arundinoideae, Saddles are reported in Phragmites mauritianus, Bilobates in Stipagrostis uniplumis, Saddles and Bilobates in Aristida sp., and Polylobates in A. adscensionis. For Bambusoideae, Bilobates are reported in all species (5/5), Saddles and “tall and narrow” phytoliths that may be classified as Bilobates with round lobes and/or Stipa-type Bilobates are reported in three species (3/5). For Ehrhartoideae, Palmer and Tucker (1981) reported “tall and narrow” phytoliths. However, we observed Bilobates with round lobes (6/6 species) and/or Stipa-type bilobates, 4-lobed Crosses (3/6), Saddles (2/6), and a phytolith with crescent-shaped base that may either belong to the Rondel or Saddle phytolith category in Phyllorhachis sagittata. Reported phytoliths in Centothecoideae (three species) and Pharoideae (one species) are Bilobates (with round or straight lobes) and Crosses. For Danthonioideae, three species among five have Polylobates, the other two are described with Rondels or Bilobates. Finally, three phytolith categories are mostly reported: Bilobates preferentially occur in Panicoideae, Bambusoideae, Arundinoideae, Danthonioideae, Ehrhartoideae Centothecoideae and Pharoideae subfamilies; Saddles preferentially occur in Chloridoideae and Bambusoideae subfamilies; Trapeziform Sinuates preferentially occur in Pooideae. It is worth noting that most surveyed grass Fig. 5. Relative proportions of C3 versus C4 species with various phytolith categories in their epidermis. Proportions are normalized by the proportions of C3 and C4 species. For example, four C4-species over 120 (that is ~ 3%) are reported with Trapeziform sinuate against 25 C3-species over 60 (that is ~ 41%). The proportion of species with Trapeziform sinuate phytoliths is therefore greater in C3 (~ 93%) than in C4 (~ 7%) grasses. Data synthesized from Palmer et al. (1981, 1983, 1985, 1986) and Ball (2002). D. Barboni, L. Bremond / Review of Palaeobotany and Palynology 158 (2009) 29–41 37 Fig. 6. Scatter plot obtained from the analysis of correspondences between grass subfamilies, photosynthetic pathways, and habitat requirement for light and water on one hand, and of silica bodies on the other hand for the 142 East African grass species surveyed by Palmer et al. (1981, 1983, 1985, 1986). Phytolith descriptors of Palmer et al. are reported here, except for those we recognized on SEMs. For coding and equivalence with the phytolith international nomenclature (ICPN, 2005), see Table 1. subfamilies are those exhibiting the greatest variety of phytolith types: Pooideae (26 species surveyed) amount to 20 different phytoliths, Panicoideae (43 species) to 29, and Chloridoideae (49 species) to 27 different phytolith types. Subfamilies for which fewer species were surveyed (6 at the most for Pharoideae, Centothecoideae, Bambusoideae, Danthonioideae, Ehrhartoideae and Arundinoideae) amount to a maximum of 12 different phytoliths types. Thus, under-representation of some grass subfamilies most likely leads to under-representation of phytolith types. 3.4. Relation between phytoliths and photosynthesis Plotting the occurrence of the main short cell phytolith categories (normalized by the number of species with such phytoliths in their epidermis) relatively to the photosynthesis pathway of the grass species shows that 93% of the species in which the Trapeziform Sinuate phytolith type occurs are C3-grasses (Fig. 5). The Trapeziform Sinuate is reported in the epidermis of most Pooideae (22/26 species, exclusively C3), in some C3-shade-loving species of Panicoideae (2/ 64) and Bambusoideae (1/5), and in few C4-light-loving species of Panicoideae (3/64 species) and Chloridoideae (1/67). On the other hand, 75% of the species in which the Saddle type occurs are C4grasses. Relations are less clear for the other phytolith categories: the Bilobate, Stipa-bilobate, Cross and Polylobate, although mainly occurring in C4-grasses, can be observed in C3-grasses. Rondel and Trapeziform short cell phytoliths, although mainly found in C3grasses are also observed in C4-grasses. 3.5. Correspondence between phytolith types and the environment In the ordination diagram of the CA (Figs. 6 and 7, and Table 1 for coding), the main phytolith categories (e.g. Saddle) occur in the same space as their related types (e.g. Saddle with long convex edges), which suggests that distinguishing Saddles (according to the size of the convex edges), Bilobates and Polylobates (according to the shape of lobes and length of the shanks) adds little to discriminating among grass subfamilies and environments, and that interpreting phytolith assemblages as proxy of environment can be done using the main phytolith categories only. Correspondence Analysis on the East African dataset of Palmer and colleagues (Palmer and Tucker, 1981, 1983; Palmer et al., 1985; Palmer and Gerbeth-Jones, 1986) indicates that 27.7% of the variance included in the phytolith × taxonomy × environment matrix is explained by the two first axes, and that these axes relate to environmental gradients (Fig. 6). The explained variance is relatively low because of the large data matrix. Axis-1 distinguishes Pooideae and C3 grasses from all other grass subfamilies. It is defined by phytolith types of the categories Trapeziform Sinuate and Rondel at its positive end, and by Bilobates, Polylobates, Crosses, and Saddles at its negative end. Axis-2 distinguishes Chloridoideae, Arundinoideae and Pooideae from Pharoideae, Centothecoideae, Ehrhartoideae, Bambusoideae, Danthonioideae and Panicoideae. It is essentially defined by Saddles at its negative end and by Bilobates, Crosses, and Polylobates at its positive end. Axis-2 represents both a gradient of moisture and light requirement, with C4-grasses occurring preferably in open, xeric 38 D. Barboni, L. Bremond / Review of Palaeobotany and Palynology 158 (2009) 29–41 Fig. 7. Scatter plot obtained from the analysis of correspondences between grass subfamilies, photosynthetic pathways, and habitat requirement for light and water on one hand, and of silica bodies on the other hand for the 42 East African grass taxa surveyed by Ball (2002). a) Dispersion of the phytolith morphotypes and b) dispersion of the grass taxa. See Table 1 for coding and equivalence with the phytolith international nomenclature (ICPN, 2005). We reported here the specimen numbers given by T. Ball (2002) for the grasses not identified at the species level. environments, and C3-grasses in shaded and mesic habitats. Pooideae occur towards the negative end of Axis-2 like Chloridoideae and Arundinoideae, therefore indicating their affinity to open rather than shaded-habitats. The pool of C3 grasses, however, occurs towards the positive end of Axis-2, indicating that a large number of C3 grasses (other than the Pooideae) occur within the subfamilies Pharoideae, Centothecoideae, Ehrhartoideae, Bambusoideae, Danthonioideae and Panicoideae in shaded and/or wet habitats. In the CA scatter plot, most phytolith types occur in association with a particular subfamily or environment, except Rondels (“Round”, “Oblong”, “Crescent”, “Elliptical”, RO-7, RO-5) and Bilobates Stipa-type or “shovel” (BI-29, 31,34), which occur between the C4-Chloridoideae–Arundinoideae pole and the Pooideae pole, thus showing no particular affinity for any subfamily or environment, or instead, showing their redundancy in grasses of different subfamilies. Correspondence Analysis on the Southern Arabian grass dataset of Ball (2002) shows that Axis-1 distinguishes hydrophytic and helophytic, semi-shade loving grasses (Panicoideae and Arundinoideae with essentially Bilobates and Crosses in their epidermis) from the mesophytic and xerophytic, light-loving species favoring open-habitats (Chloridoideae with Rondels or Saddles in their epidermis). Axis-2 separates among the Chloridoideae those with Rondels (Sporobolus and Urochondra species) from those with Saddles in their epidermis (Chloris, Dactyloctenium, and Ochthochloa species) (Fig. 7). 4. Discussion 4.1. Limitations of our survey First, because top and base of phytoliths can differ markedly, such as short cell phytoliths from Phragmites that may be Rondel-like in side view but saddle-shaped in bottom view (Piperno and Pearsall, 1998), a 3D-image is necessary to fully characterize a given phytolith type (Pearsall et al., 2004). This could not be done for all the 184 species included in this survey, since SEMs only show one side of the short cells. SEMs of the leaf epidermis show the base of phytoliths, but not their top which is oriented towards the interior of the leaf. Our attempt to characterize as many types as possible among the main phytolith categories was therefore limited. Second, although surveying as much as 184 grass species, the reference collection made from Palmer's and Ball's datasets is nonetheless restricted to a small portion of the grass species occurring in East Africa and listed in the flora of tropical East Africa (ca 20%) (Clayton, 1970, 1974, 1982). Consequences are that relations between phytolith occurrence, grass subfamilies, photosynthesis, and environment identified here could somewhat be altered when other grass species are considered. Also, the grass species which contributed to fossil phytolith assemblages may be different from the ones investigated here. Third, our study is qualitative only. The data that we used (Palmer and Tucker, 1981; Palmer and Tucker, 1983; Palmer et al., 1985; Palmer and Gerbeth-Jones, 1986; Ball, 2002) did not allow investigation the frequency of phytolith types in grass species (e.g. Carnelli et al., 2004), but the phytolith morphological diversity. Using quantitative rather than qualitative data surely would have improved the discrimination of grass subfamilies (Mulholland, 1989; Strömberg, 2003). However, even with quantitative data a modern plant reference collection may not give strong clues for interpreting fossil phytolith assemblage, since phytolith production varies with the age of the plant (Henriet et al., 2006) and silica available in the soil (e.g. Henriet et al., 2008). To our knowledge, studies looking at differential production of phytolith types according to environmental parameters are still needed. In these conditions, calibrating modern phytolith D. Barboni, L. Bremond / Review of Palaeobotany and Palynology 158 (2009) 29–41 assemblages with current vegetation physiognomic and floristic characters appears the most suitable technique for interpreting fossil phytolith assemblages (e.g. Fredlund and Tieszen, 1997). Despite these limitations our survey highlights several points that can complement paleoenvironmental reconstructions from fossil phytolith assemblages. 4.2. Further assessment of taxonomic and environmental significance of the main phytolith types The general pattern of phytolith type occurrences in East African grass subfamilies evidenced by our study is in agreement with the global pattern first identified by Twiss et al. (1969), except for Chloridoideae. We observed that widespread East African Chloridoideae species such as Sporobolus consumilis, S. spicatus and Urochondra setulosa do not have Saddle but Rondel phytolith types in their epidermis. This is in agreement with previous work showing that Rondel types represent ca 95% of the phytoliths produced by some Tanzanian Sporobolus species (Bamford et al., 2006). Hence, because of inherent phytolith redundancy, Rondel phytolith types in tropical East Africa may be interpreted as Pooid or Chloridoid phytoliths. Our study shows, however, that phytolith assemblages with Rondel, Trapeziform short cell, and Saddle phytolith types most likely result from the input of Chloridoideae and Panicoideae species (C4-lowland grasses), while assemblages including Rondel, Trapeziform short cell, and Trapeziform Sinuate phytolith types most likely result from the input of Pooideae species (C3-highland grasses). Rondel phytoliths recovered from soils/sediments from East African lowlands have a greater chance to originate from C4-Chloridoideae than from C3Pooideae species, provided that input of phytoliths from higher elevation is insignificant. Our study also highlights that Trapeziform Sinuate phytoliths (rather than Rondel or Trapeziform short cell) attests to the presence of Pooideae (high elevation C3-grasses). Trapeziform Sinuate phytoliths are reported in some other (few) non-Pooideae species, but are the least redundant phytolith type. In agreement with phytolith abundance data obtained from the analysis of modern soils, Trapeziform Sinuate phytoliths may be confidently used as proxies for Pooideae, and to attest to the presence of C3-high elevation grasses in East Africa (Bremond et al., 2008). Our results show that 93% of East African C3grasses have Trapeziform Sinuate phytoliths in their epidermis, and 75% of C4 grasses have Saddles (Fig. 5). However, the ratio of Trapeziform Sinuate versus Saddle phytoliths calculated using the data of Bremond et al. (2008) is less correlated (r2 = 0.53) with elevation (and with C3/C4 grass occurrence) than the original Ic index (r2 = 0.90), which is the ratio of Trapeziform Sinuate + Rondel + Trapeziform phytoliths over Saddle + Bilobate + Polylobate + Cross phytoliths. It may be that redundancy and multiplicity of phytoliths involves less bias on reconstructed grass species representation when abundances rather than occurrences are taken into account. This further suggests the need for investigating grass production of phytolith types through comparisons between phytolith assemblages from the top of soils and vegetation floristic and physiognomic characteristics. Apart from the Trapeziform Sinuate phytolith type, no other phytolith main type is qualitatively diagnostic for the presence of a given subfamily, or any other group of grass taxa in phytolith assemblages. 4.3. Morphological discrimination among the phytolith types for refining environmental interpretation of phytolith assemblages Morphological discrimination among Bilobate, Cross, and Rondel main phytolith types could improve the paleoenvironmental interpretation of East African phytolith assemblages. The length of the shank between the lobes of Bilobate phytoliths decreases with increasing moisture-requirement of grasses, in agreement with the phytolith morphological pattern identified from 85 grass species from China and 39 Southern USA (Lu and Liu, 2003). The East African grass dataset used here shows that mesophytic grasses with short shank bilobates are onethird Panicoideae species (29%), one-third Chloridoideae species (29%), and species of other subfamilies such as Bambusoideae (17%), Ehrhartoideae (12%), Centothecoideae, and Danthonioideae (ca 6%). The length of the shank of the Bilobate type is therefore linked to water availability rather than to grass phylogeny. The 3-lobed Cross phytoliths only occur in C4-light-loving Panicoideae and Chloridoideae grasses. Identifying and counting separately 3-lobed Cross phytoliths may therefore improve the identification of C4-open-habitat grasses from East African phytolith assemblages. Rondel phytoliths with base diameter N15 μm were only reported in Pooideae species, while Rondels with base diameter of b15 μm were observed in several grass subfamilies including Pooideae, Bambusoideae, Chloridoideae, Danthonioideae, Ehrhartoideae, and Panicoideae. Considering the size of Rondel phytoliths may thus confirm the presence of Pooideae C3-high elevation grasses in the environment. To our knowledge, phytolith analysts usually do not consider this morphological criterion. Within the Saddle category, the four types we distinguished (Saddle with symmetrical convex edges, Saddle with long convex edges, Saddle with short convex edges, and Collapsed saddle) are mainly reported in the Chloridoideae subfamily. We expected the Saddle type with short convex edges (confusedly named “long saddle” by some authors in reference to the length of the “seat”) to be characteristic for Bambusoideae, as shown for Asian bamboos (Kondo et al., 1994; Lu et al., 2006). However, in East Africa, this type is not exclusive to Bambusoideae. It also occurs in many open-habitat Chloridoideae species (18 widespread species such as Eleusine floccifolia), in the common helophytic Phragmites mauritianus, and in the semi-shadeloving Streptogyna crinita found in dry forests. The Saddle type with long convex edges (confusingly named “short saddles” by Kondo et al., 1994; Lu et al., 2006) only occurs in Chloridoideae species and is absent from the East African Bambusoideae species. The Collapsed saddle type could not be identified in the surveyed East African Bambusoideae species, since the scanning electron micrographs from Palmer and Tucker (1981) do not allow the observation of phytoliths in the three dimensions. The Collapsed saddle type was found in several Chloridoideae species. It is therefore not restricted to the Bambusoideae, as assumed by Strömberg (2004) and Strömberg et al. (2007) when interpreting North American and European Cenozoic phytolith assemblages, and as shown for South American subtribes Guaduineae and Chusqueinae (Piperno and Pearsall, 1998). This result is in agreement with the insignificant amounts of collapsed saddles (b1% of the count of Saddle phytoliths) in soil samples under bamboo stands of Arundinaria alpina on Mt Kenya (Bremond et al., 2008). Thus the Collapsed saddle type is not characteristic for Bambusoideae worldwide. Few phytolith types are reported in one grass species or in a restricted number of taxa. The Bilobate type with alate, deeply concave lobes is reported only in Maltebrunia leersioides (Palmer and Tucker, 1981, p.53), Ehrhartoideae, C3, shade species. Assumed to be characteristic for Ehrhartoideae (Piperno, 2006), this type, however, is not observable in other African species of Ehrhartoideae (Palmer and Tucker, 1981). The Rondel tabular oblong is observable only in Phalaris arundinacea (Palmer and Tucker, 1983, p.60), a C3-Pooideae species found in damp soils and swamps. The Saddle “with lateral knobs” and Bilobate “with knobs on end lobes” (Palmer et al., 1985, p.47, p.58) exclusively occur in Chloridoideae, C4, open-habitat species. Investigations of phytolith assemblages from modern soils collected below well described vegetation will confirm the taxonomical significance of these phytolith types. 4.4. Grass short cell phytoliths and openness of the environment Due to over-representation of grass taxa that are both xeric and light-loving, the CA diagram shows that grasses with different water and light requirement cannot be distinguished using the phytolith 40 D. Barboni, L. Bremond / Review of Palaeobotany and Palynology 158 (2009) 29–41 occurrence data investigated here (Fig. 6). The phytolith signal of shade-loving East African grasses is uncharacteristic and overlaps with the phytolith signal of moisture-loving (hydro-, helo-, and mesophytic) species. Thus, it seems not possible, so far, to discriminate shade-loving from light-loving grasses based on their phytolith signal. Our results are in agreement with the use of Iph (the phytolith index defined by Diester-Haass et al., 1973), or the proportion of Saddle versus Bilobate, Cross, and Polylobate phytolith types as a proxy for moisture in the environment in West Africa (Alexandre et al., 1997; Bremond et al., 2005b). Calibrations of modern soil phytolith assemblages versus openness of the environment and moisture availability are required to decouple the combined effect of these two environmental factors on the grass species composition and resulting surface soil phytolith assemblages. It is commonly assumed that evolution has led “crown” grasses (those which evolved most recently) to favor open-habitats and “basal” grasses (those at the base of the phylogeny) to favor closedhabitats. However, the East African grasses dataset shows that several species of the lately evolved subfamilies, such as the Panicoideae, are shade-loving grasses commonly found in forests and woodlands (Palmer and Gerbeth-Jones, 1986; Watson and Dallwitz, 1992 onwards). Since the same phytolith types occur in both shade- and light-loving Panicoideae, and in both Bambusoideae (shade-loving species) and Chloridoideae (light-loving species), we doubt that using the position of grass species on the phylogeny tree is a valid approach to interpreting assemblages of fossil grass phytoliths in terms of openness of the environment (Prasad et al., 2005; Strömberg, 2005). 5. Conclusions We conclude that some morphological variations within the main phytolith categories Bilobate, Cross, and Rondel could improve the environmental and taxonomical interpretation of phytoliths, but that most of the environmental information is already captured by the main phytolith categories. Except for the Trapeziform Sinuate phytolith almost exclusively reported in Pooideae and C3-grasses, no other phytolith type is qualitatively diagnostic for the presence of a given subfamily, or any other group of grass taxa in phytolith assemblages. On the interpretation of phytoliths in terms of C3–C4, we therefore conclude that Trapeziform Sinuate phytoliths are the only type that can confidently infer the presence of C3-grasses. Using occurrence data, we were not able to discriminate the phytolith signal of non-Pooideae C3-grasses occurring at low and mid-elevation, because it closely resembles that of mesophytic and helophytic C4-grasses. On the interpretation of phytoliths in terms of moisture availability in the environment, our results indicate that East African xerophytic grasses preferentially have Saddle and Rondel (b15 μm) or both phytolith types in their epidermis, as well as long shank bilobates. Mesophytic grasses, on the contrary tend to have Bilobate with short shanks, Cross, and Polylobate phytoliths. In East Africa, Sporobolus and Urochondra are Chloridoideae species reported with Rondel phytoliths (b15 μm), morphologically similar but smaller in size to Rondel phytoliths in Pooideae (N15 μm). We are not able to provide an interpretation of phytoliths in terms of shade–light availability in the environment, since the phytolith signal overlaps with that of moisture. Also, we found that phytoliths of East African crown and basal grasses are morphologically similar. To our knowledge it is therefore not possible to infer the presence of basal grasses from phytoliths. This precludes the use of grass phylogeny to interpret phytoliths in terms of openness of the habitat. Finally, if our study revealed that there are morphological differences within the main grass short cell phytoliths that could improve the taxonomical and environmental resolution of phytoliths, it also shows, by comparison with the literature, that abundance rather than occurrence data are most likely to infer greater precision to the interpretation of fossil phytolith assemblages. Acknowledgments We thank R. Bonnefille and A. Alexandre for commenting on an earlier version of this manuscript. DB is grateful to Tim White and the ‘Revealing Human Origin Initiative’ for fostering phytolith research in Africa through fruitful discussions, and financial support to the Phytolith Analytical Working Group. This material is based upon work supported by the U.S. National Science Foundation under NSF Award #BCS-0321893. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.revpalbo.2009.07.002. References Albert, R.M., Bamford, M.K., Cabanes, D., 2006. Taphonomy of phytoliths and macroplants in different soils from Olduvai Gorge (Tanzania) and the application to Plio-Pleistocene palaeoanthropological samples. Quaternary International 148, 78–94. Alexandre, A., Meunier, J.-D., Lézine, A.-M., Vincens, A., Schwartz, D., 1997. Phytoliths indicators of grasslands dynamics during the late Holocene in intertropical Africa. Palaeogeography, Palaeoclimatology, Palaeoecology 136, 213–219. Ball, T.B., Baird, G., Woolstenhulme, L., al Farsi, A., Ghazanfar, S., 2002. Phytoliths Produced by the Vegetation of the Sub-Tropical Coastal Region of Dhofar, Oman. CD Distributed by The Society for Phytolith Research. Bamford, M.K., Albert, R.M., Cabanes, D., 2006. Plio-Pleistocene macroplant fossil remains and phytoliths from Lowermost Bed II in the eastern palaeolake margin of Olduvai Gorge, Tanzania. Quaternary International 148, 95–112. Barboni, D., Bremond, L., Bonnefille, R., 2007. Comparative study of modern phytolith assemblages from inter-tropical Africa. Palaeogeography, Palaeoclimatology, Palaeoecology 246, 454–470. Bremond, L., Alexandre, A., Hély, C., Guiot, J., 2005a. A phytolith index as a proxy of tree cover density in tropical areas: calibration with Leaf Area Index along a forest–savanna transect in southeastern Cameroon. Global and Planetary Change 45, 277–293. Bremond, L., Alexandre, A., Peyron, O., Guiot, J., 2005b. Grass water stress estimated from phytoliths in West Africa. Journal of Biogeography 32, 311–327. Bremond, L., Alexandre, A., Wooller, M.J., Hely, C., Williamson, D., Schafer, P.A., Majule, A., Guiot, J., 2008. Phytolith indices as proxies of grass subfamilies on East African tropical mountains. Global and Planetary Change 61, 209–224. Carnelli, A.L., Theurillat, J.-P., Madella, M., 2004. Phytolith types and type-frequencies in subalpine–alpine plant species of the European Alps. Review of Palaeobotany and Palynology 129, 39–65. Carr, C.J., 1998. Patterns of vegetation along the Omo River in Southwest Ethiopia. Plant Ecology 135, 135–163. Clayton, W.D. (Ed.), 1970. Flora of tropical East Africa—Gramineae, Part 1. Crown Agents for Oversea Governments and Administrations, London, pp. 1–176. Clayton, W.D. (Ed.), 1974. Flora of tropical East Africa—Gramineae, Part 2. Crown Agents for Oversea Governments and Administrations, London, pp. 177–449. Clayton, W.D. (Ed.), 1982. Flora of tropical East Africa—Gramineae, Part 3. A.A., Rotterdam, Balkema, pp. 450–850. Diester-Haass, L., Schrader, H.J., Thiede, J., 1973. Sedimentological and Paleoclimatological Investigations of Two Pelagic Ooze Cores off Cape Barbas, North-West Africa. Meteor Forsh-Ergebnisse C16, pp. 19–66. Fredlund, G., Tieszen, L.T., 1994. Modern phytolith assemblages from the North American Great Plains. Journal of Biogeography 21, 321–335. Fredlund, G., Tieszen, L.T., 1997. Calibrating grass phytoliths assemblages in climatic terms: application to the late Pleistocene assemblages from Kansas and Nebraska. Palaeogeography, Palaeoclimatology, Palaeoecology 136, 199–211. Grass Phylogeny Working Group., 2001. Phylogeny and subfamilial classification of the grasses (Poaceae). Annals of the Missouri Botanical Garden 88, 373–457. Henriet, C., De Jaeger, N., Dorel, M., Opfergelt, S., Delvaux, B., 2008. The reserve of weatherable primary silicates impacts the accumulation of biogenic silicon in volcanic ash soils. Biogeochemistry 90, 209–223. Henriet, C., Draye, X., Oppitz, I., Swennen, R., Delvaux, B., 2006. Effects, distribution and uptake of silicon in banana (Musa spp.) under controlled conditions. Plant and Soil 287, 359–374. ICPN Working group, Madella, M., Alexandre, A., Ball, T.B., 2005. International code for phytolith nomenclature 1.0. Annals of Botany 96, 253–260. Kondo, R., Childs, C., Atkinson, I., 1994. Opal Phytoliths of New Zealand. Manaaki Whenua Press. 85 pp. Lu, H.Y., Liu, K.B., 2003. Morphological variations of lobate phytoliths from grasses in China and the south-eastern United States. Diversity and Distributions 9, 73–87. Lu, H.Y., Wu, N.Q., Yang, X.D., Jiang, H., Liu, K.B., Liu, T.S., 2006. Phytoliths as quantitative indicators for the reconstruction of past environmental conditions in China I: phytolith-based transfer functions. Quaternary Science Reviews 25, 945–959. Mindzie, C.M., Doutrelepont, H., Vrydaghs, L., Swennen, R.L., Swennen, R.J., Beeckman, H., de Langhe, E., de Maret, P., 2001. First archaeological evidence of banana cultivation in central Africa during the third millennium before present. Vegetation History and Archaeobotany 10, 1–6. D. Barboni, L. Bremond / Review of Palaeobotany and Palynology 158 (2009) 29–41 Mulholland, S.C., 1989. Phytoliths shape frequencies in North Dakota grasses: a comparison to general patterns. Journal of Archaeological Science 16, 489–511. Palmer, P.G., 1976. Grass cuticules: a new paleoecological tool for East African lake sediments. Canadian Journal of Botany 54, 1725–1734. Palmer, P.G., Gerbeth-Jones, S., 1986. A scanning electron microscope survey of the epidermis of East African grasses, IV. Smithsonian Contribution to Botany, vol. 62. Washington, 86 pl., 120 pp. Palmer, P.G., Gerbeth-Jones, S., Hutchison, S., 1985. A scanning electron microscope survey of the epidermis of East African grasses, III. Smithsonian Contribution to Botany, vol. 55. Washington, 98 pl., 136 pp. Palmer, P.G., Tucker, A.E., 1981. A scanning electron microscope survey of the epidermis of East African grasses, I. Smithsonian Contribution to Botany, vol. 49. Washington, 48 pl., 84 pp. Palmer, P.G., Tucker, A.E., 1983. A scanning electron microscope survey of the epidermis of East African grasses, II. Smithsonian Contribution to Botany, vol. 53. Washington, 52 pl., 72 pp. Pearsall, D.M., Chandler-Ezell, K., Chandler-Ezell, A., 2003. Identifying maize in neotropical sediments and soils using cob phytoliths. Journal of Archaeological Science 30, 611–627. Pearsall, D.M., Chandler-Ezell, K., Chandler-Ezell, A., 2004. Maize can still be identified using phytoliths: response to Rovner. Journal of Archaeological Science 31, 1029–1038. Piperno, D.R., 2006. Phytoliths. A Comprehensive Guide for Archaeologists and Paleoecologists. AltaMira Press (Rowman & Littlefield), Lanham, New York, Toronto, Oxford. ix +238 pp. Piperno, D.R., Pearsall, D.M., 1998. The silica bodies of tropical American grasses: morphology, taxonomy, and implications for grass systematics and fossil phytolith identification. Smithsonian Contributions to Botany 85, 1–40. Prasad, V., Strömberg, C.A.E., Alimohammadian, H., Shani, A., 2005. Dinosaur coprolites and the early evolution of grasses and grazers. Science 310, 1177–1180. 41 Prat, H., 1948. General features of the epidermis in Zea mays. Annals of the Missouri Botanical Garden 35, 341–351. Strömberg, C.A.E. 2003. The origin and spread of grass-dominated ecosystems during the Tertiary of North America and how it relates to the evolution of hypsodonty in equids. Ph D thesis, UC Berkeley, Berkeley, CA. 779 pp. Strömberg, C.A.E., 2004. Using phytolith assemblages to reconstruct the origin and spread of grass-dominated habitats in the great plains of North America during the late Eocene to early Miocene. Palaeogeography, Palaeoclimatology, Palaeoecology 207, 239–275. Strömberg, C.A.E., 2005. Decoupled taxonomic radiation and ecological expansion of open-habitat grasses in the Cenozoic of North America. Proceedings of the National Academy of Sciences of the United States of America, vol. 102, pp. 11980–11984. Strömberg, C.A.E., Werdelin, L., Friis, E.M., Sarac, G., 2007. The spread of grass-dominated habitats in Turkey and surrounding areas during the Cenozoic: phytolith evidence. Palaeogeography, Palaeoclimatology, Palaeoecology 250, 18–49. Ter Braak, C.J.F., Smilauer, P., 1998. CANOCO Reference Manual and User's Guide to CANOCO for Windows. Microcomputer Power, Ithaca, NY, USA. Twiss, P.C., 1992. Predicted world distribution of C3 and C4 grass phytoliths. In: Mulholland, S.C. (Ed.), Phytoliths Systematics Emerging Issues. Advance Archaeological Museum Science, pp. 113–128. Twiss, P.C., Suess, E., Smith, R.M., 1969. Morphological classification of grass phytoliths. Proceedings of Soil Science Society of American 33, 109–115. Watson, L., Dallwitz, M.J., 1992 onwardss. The Grass Genera of the World: descriptions, Illustrations, Identification, and Information Retrieval; Including Synonyms, Morphology, Anatomy, Physiology, Phytochemistry, Cytology, Classification, Pathogens, World and Local Distribution, and References. http://delta-intkey.com/grass/Version: 28th November 2008.