The variability of Amazonian palm phytoliths

The Arecaceae (palm) family isone of themostcommon and economicallyimportant plant groups in Amazonia, and play large roles in ecosystem functioning and carbon dynamics. The modern dominance of palms may be a result of environmental gradients, past climate variability and legacies of past human activities. To understand the roles people and environment have played in shaping the abundance and widespread occurrence of Amazonian palm species, past vegetation reconstructions using palaeoecological proxies that can identify palm species or genera are needed. Phytoliths are siliceous microfossils abundantly produced in palms and are used to reconstruct past vegetation composition.The aim of thispaperwas to assess thevariability ofAmazonianpalm phyto-lith morphotypes and to determine which morphological characteristics can be used to increase the taxonomic resolution of palm identi ﬁ cations. Phytolith size, sphericity, spinule traits and number of (sur ﬁ cial) projections were measured for 24 Amazonian palm species. Differences were compared using PCA and ANOVA or Tukey tests,followedbyKruskal-WallisorDunntests.Resultsshowpotentialforphytolithsof Aiphanesaculeata,Attalea maripa, Bactris simplicifrons, Dictyocaryum fuscum, Euterpe precatoria, Iriartea deltoidea, Oenocarpus bacaba, Socratea exorrhiza and the genera Geonoma, Bactris, Euterpe and Oenocarpus to be identi ﬁ ed based on their size, shape or number of projections. This will allow future phytolith analyses to reconstruct past palm abundances with an increased taxonomic resolution. We have formulated our ﬁ ndings into a ‘ quick guide ’ to identify palm phytoliths and test hypotheses regarding the drivers of modern palm distributions and abundances. © 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http:// creativecommons.org/licenses/by/4.0/).


Introduction
Palms are a common component of the Amazon rainforest: 20 palm species are among the 227 'hyperdominant' tree species that account for more than half of the individuals in the forest (Ter Steege et al., 2013;ter Steege et al., 2020). Their high abundances mean that palms play a large role in Amazonia's carbon dynamics (Fauset et al., 2015;Muscarella et al., 2020). Abiotic factors such as climate, soil, topography, hydrology, dispersal and physiological adaptations have shaped palm distributions in Amazonia (Eiserhardt et al., 2011;Kristiansen et al., 2009;Vormisto et al., 2004). Also, animal abundances, particularly of seed dispersers, affect palm abundances and distributions (Eiserhardt et al., 2011) (Goulding et al., 2007). It is hypothesized that modern palm abundances across Amazonia are a legacy of past human activities, particularly during the pre-Columbian era (Levis et al., 2017), because palms were wellexploited (e.g., Morcote-Rios and Bernal, 2001). For example, Bactris gasipaes, the peach or pehibaye palm, was domesticated (Clement, 1988). Still today, people in Amazonia use at least 134 of 225 different palm species for food, construction, tools and cultural practices (Macia et al., 2011). Existing data from the prehistoric period reveal variability in the extent to which ancient people may have influenced modern palm abundance in Amazonia. In some regions, palms that grew near human settlements increased in abundance (Maezumi et al., 2018;McMichael et al., 2015), while decreasing in others (Åkesson et al., 2021). Palm abundance has also increased or remained largely unchanged in areas where human impacts were minimal (Heijink et al., 2020;McMichael et al., 2021;Piperno et al., 2021).
To understand the roles people and environment have played in shaping the abundance and widespread occurrence of Amazonian palm species, past vegetation reconstructions using palaeoecological proxies that can identify palm species or genera are needed. Phytoliths (silica microfossils) are used to reconstruct past vegetation composition and are produced abundantly in most palm species (Piperno, 2006). Phytoliths are usually not dispersed by wind, but instead provide a local signal of the vegetation (±30 m) in both soils and lake sediments (Piperno, 1989;Åkesson et al., 2021). Phytolith analysis can be carried out in areas where pollen degrades, such as terrestrial soils (McMichael et al., 2012), or used to complement pollen reconstructions in lake sediments (Åkesson et al., 2021;Maezumi et al., 2018). Palms (ARACACEAE) usually make either spheroid or conical phytoliths with an echinate texture (Tomlinson, 1961;Piperno, 1989Piperno, , 2006. ARACACEAE phytoliths from South America can be visually separated from other monocotyledon families (BROMELIACEAE, CANNACEAE, MARANTACEAE, ORCHIDACEAE, STRELITZIACEAE, and ZINGIBERACEAE) based on differences in the morphotypes produced and the morphological variation of spheroid echinates (Benvenuto et al., 2015;Chen and Smith, 2013;Piperno, 2006Piperno, , 1989. But the strong overlap of morphological variables (such as size) between palm species makes it difficult to increase taxonomic resolution to the species level (Neumann et al., 2019).
However, a detailed morphometric analysis on spheroid echinate and conical palm phytoliths from South America is missing, because studies either measured a limited number of species (Benvenuto et al., 2015;Delhon and Orliac, 2010) or used a descriptive approach for a large reference collection (Huisman et al., 2018;Morcote-Ríos et al., 2016;Watling and Iriarte, 2013). Perhaps because there is no general agreement on what diagnostic features lower taxonomic resolution, most studies count only 1 conical morphotype and few spheroid morphotypes of palms in fossil records (Åkesson et al., 2021;Piperno et al., 2021;Whitney et al., 2014). Increasing the taxonomic resolution of palm phytoliths will improve vegetation reconstructions, making it possible to study the drivers of modern palm distributions in Amazonia.
The aim of this paper is to answer the following research questions: (1) What is the variability of phytolith morphology in 24 Amazonian palm species?
(2) Which morphological characteristics of phytoliths can be used to increase taxonomic identification of palms in palaeoecological reconstructions?
To standardize and improve the taxonomic resolution of phytolith studies in Amazonia, we investigate here whether a detailed morphometric analysis of phytoliths and the relative abundance of morphotype for each species may further differentiate palm taxa.

Material and methods
Reference material of 24 palm species was collected from the Leiden Herbarium and Hortus Botanicus Amsterdam (the Netherlands). Species focused on were hyperdominant (11 species) or useful for people (19 species), and 3 species had not been previously analyzed (Table 1). Specimens were sampled for leaves, woody parts, fruits, or seeds, depending on availability. All samples were processed according to standard laboratory procedures for wet processing (Piperno, 2006). Organic matter and carbonates were removed from the phytolith samples using 33% H 2 O 2 , KMnO 4 − and 10% HCl. Phytoliths were extracted by heavy liquid flotation using Bromoform at a specific gravity of 2.3 (McMichael et al., 2021). Phytoliths were mounted on microscope slides using Naphrax and identified using immersion oil at 1000 × magnification under a Zeiss Axioscope microscope The relative abundance of morphotypes for each sample was calculated by counting a total of 300 palm phytoliths per slide (Table S1). Phytoliths were counted using morphological descriptions from the literature (e.g., Morcote-Ríos et al., 2016;Huisman et al., 2018), using the latest nomenclature (Table S2). For each sample, at least 30 phytoliths and for each morphotype at least 3 replicates per species were photographed and morphological characteristics were measured using Image J software (Abràmoff et al., 2004). The following numerical variables were measured for conical phytoliths: length (μm), width (μm), number of projections, length of spines (μm), width of spines (μm), distance between spines (μm) and sphericity (length/width). For spheroid phytoliths, the numerical variables measured were maximum diameter (μm), number of surficial projections, length of spines (μm), width of spines (μm), distance between spines (μm), sphericity (maximum diameter/perpendicular diameter) and perpendicular diameter (μm) according to Benvenuto et al. (2015).

Data analysis
Principal Components Analysis (PCA) was used to simultaneously assess the variability of all morphological measurements across species, to see if species would separate based on morphology and which variables contributed mostly to this separation. PCA were run separately within general morphotypes (e.g. within conical phytoliths and within spheroid phytoliths). Pearson correlation tests were used to see if measured characteristics were correlated within each morphotype separately. Because several variables were correlated, and most variables were not normally distributed, we conducted pairwise comparisons between species, for each variable separately. For all variables, mean differences between the measured morphological characteristics of species were tested using ANOVA and Tukey tests, or Kruskal-Wallis and Dunn tests (when residuals were not normally distributed). We adjusted p-values with the Benjamini-Hochberg method because multiple comparisons were made on the same dataset. The pairwise comparisons were used to assess for each variable separately the amount of variation between and within a species, and the ranges of minimum and maximum values. The minimum sample size needed to accurately reflect mean values of each morphotype and species was calculated (Ball et al., 2016). Analyses were performed in R Studio Version 1.4.1103, colors were chosen using the 'viridis' package (Garnier et al., 2021) and the PCA was performed using the 'princomp' function of the 'stats' and 'factoextra' packages (Kassambara and Mundt, 2017). The full reproducible code is available in the Supplementary Material.

Multivariate analysis of spheroid phytoliths
The first axis of the PCA (PCA1; eigenvalue = 3.9; 56% of the total variance) separated the larger phytoliths from the smaller ones (Fig. 3). For example, larger SPH_ACU phytoliths from Euterpe and Oenocarpus are on the positive PCA1 axis, along with the large SPH_SYM of Oenocarpus bacaba. Species that consistently produced smaller phytoliths, such as Attalea speciosa, Elaeis oleifera and Mauritiella armata, are on the negative PCA1 axis. Attalea maripa produced large phytoliths and is displayed on the positive PCA2 axis. The second axis (PCA2; eigenvalue = 1.3; 19% of the total variance) separated the phytoliths based on the number of surficial projections. Therefore, morphotypes SPH_SYM and SPH_TUB occupied space on the negative side of PCA2, whereas SPH_ACU phytoliths are on the positive PCA2 axis. The previously established morphotypes, shown as different colors in Fig. 3, are separated by the PCA, except for the overlap between SPH_TUB and other morphotypes.

Pairwise comparisons between species producing spheroid phytoliths
Most species producing spheroid phytoliths overlap in size, shape, spinule traits and number of surficial projections (Fig. S1, S2). Species differences between spinule traits resembled patterns of maximum and perpendicular diameter for all spheroid morphotypes except SPH_TUB phytoliths (Supplementary Material "Correlation"). Morphotype SPH_ACU was rare, therefore, for robust statistical analysis the sampling size of these morphotypes should be increased to 203 phytoliths (Supplementary Material "Robustness"). Also, the species that produced a mixture of small SPH_ECH and large SPH_ACU required large sampling sizes of at least 140 phytolith.
On average, the maximum diameter of ELL_ECH phytoliths (7-18 μm) of Attalea maripa was larger than other species. The average maximum diameter of SPH_ECH phytoliths (11-18 μm) of Attalea maripa was also larger than other species, but overlapped with the range of Euterpe catinga, Mauritia flexuosa and Oenocarus bacaba (Fig. S1). The average surficial projections of SPH_ECH phytoliths of Mauritiella armata (14) exceeded that of other species. For SPH_SYM phytoliths, Oenocarpus bacaba produced larger phytoliths ranging from 8 to 22 μm, which was significantly higher than other species, which were typically smaller than 15 μm, except M. flexuosa. The maximum diameter of SPH_ACU of Euterpe (7-37 μm) and Oenocarpus species (14-24 μm) was larger than other species, who typically did not exceed 20 μm. SPH_TUB morphotypes were produced by Attalea maripa with a maximum diameter that typically ranged from 9 to 17 μm, whereas the other species were typically smaller than 13 μm, except for Oenocarpus bacaba, who ranged from 7 to 15 μm (Fig. S1).

Multivariate analysis of conical phytoliths
The first axis of the PCA (PCA1; eigenvalue = 3.3; 47% of the total variance; Fig. 5) separated the phytoliths with a larger width, length and spine width from the smaller ones. For example, larger CON_FEW phytoliths from Iriartea deltoidea and Socratea exorrhiza are on the negative PCA1 axis, as well as large CON_ECH from Bactris. Species that consistently produced smaller phytoliths, such as Geonoma undata and Aiphanes aculeatea, are on the positive PCA1 axis. The second axis (PCA2; eigenvalue = 1.2; 17% of the total variance) separated the phytoliths based on sphericity and spine length. Therefore, CON_BAS is mostly on the negative side of PCA2 as well as the oblong CON_FEW of Iriartea deltoidea, and circular CON_FEW and CON_ECH phytoliths are occupying the positive PCA2. There is some overlap between CON_FEW and CON_ECH morphotypes and between CON_ECH and CON_TAB morphotypes. But overall, previously established morphotypes are separated by the PCA.

Pairwise comparisons between species producing conical phytoliths
Most species that produce conical phytoliths overlap in size, shape, and spinule traits (Supplementary material "Correlation"). But species differences between size were found for all conical morphotypes. Differences between shape and spinule traits stood out for CON_FEW, and differences in projections for CON_ECH. Morphotype FUS_ECH was rare, therefore, for robust statistical analysis the sampling size of this morphotype should be increased to 45 phytoliths (Supplementary material "Robustness"). Also, several species produced phytoliths with a wide range of length and width, resulting in a larger required sampling size for statistcal comparisons.
On average, CON_FEW phytoliths that were circular and larger than 12 μm were produced by Socratea exorrhiza and Wettinia hirsuta, and CON_FEW phytoliths that were oblong and larger than 12 μm were produced by Iriartea deltoidea (Fig. 6-de). A similar pattern was found for spine length and width: the size of Iriartea deltoidea, Socratea exorrhiza and Wettinia hirsuta typically exceeded 1 μm, whereas the spine length and width of other species was typically 1 μm or smaller (Fig. S3). Dictyocaryum fuscum produced smaller CON_FEW phytoliths, ranging from 8 to 10 μm, whereas other species were typically larger than 10 μm.

Species specific morphotypes
The woody parts sampled of Geonoma undata and the seed phytoliths of A. aculeata and Dictyocaryum lamarckianum produced phytoliths that were not seen in other species (Fig. S4-mr, Table S1). Geonoma undata produced phytoliths that were tabular with a variety of 2D shapes (oblong, reniform, ovate, fusiform), an echinate texture and a length and width < 10 μm (Fig. S4-pr). The woody samples of A. aculeata (3%) and Euterpe precatoria (outside of phytolith count) also contained some small, irregularly shaped phytoliths, but not as much as the woody samples of Geonoma undata (>40%). The seed phytoliths of A. aculeata had a flattened shape like CON_TAB phytoliths, but lacked an echinate texture, appearing more as an irregular ornated spheroid with a size of ±5 μm (Fig. S4-mn). Such phytoliths comprised 73% of the phytolith count, along with CON_ECH (4%) and CON_TAB (23%). The seed phytoliths of Dictyocaryum lamarckianum only contained phytoliths that appeared irregularly ornated spheroids. The seeds of Euterpe precatoria and Dictyocaryum fuscum and the long seeds of Dictyocaryum lamarckianium did not contain such irregular phytoliths. Leaf phytoliths of Euterpe precatoria contained the morphotype EUTERPE SPHEROID GRANULATE (EUT_SPH) with a size range between 28 and 38 μm (Fig. S4-aj,ak, Table S1). This morphotype was not found in other species and was counted only once in the phytolith count (Table S1).

Discussion
We have assessed the variability of Amazonian palm phytolith morphotypes to determine which traits can be used to increase their taxonomic resolution and identification in palaeoecological reconstructions. Several species differed in the size of the phytoliths they produced, and to a lesser extent in shape and spinule traits. But PCA results showed considerable overlap between subfamilies, unlike other studies (Benvenuto et al., 2015). Corresponding with previous research, several genera of palms could be separated based on phytolith morphology or size (Huisman et al., 2018;Morcote-Ríos et al., 2016;Piperno et al., 2019;Piperno and McMichael, 2020). Spheroid and conical phytoliths differ in size, shape and number of projections and PCA results confirmed the morphotypes that were described for Amazonian and Andean palm phytoliths (Huisman et al., 2018;Morcote-Ríos et al., 2016;Piperno and McMichael, 2020). Species produced these morphotypes in a different frequency: some species, such as Iriartea deltoidea, only produced 1 morphotype. Considering different morphotypes, the frequencies in which they are produced, and the different size ranges and means will allow identifying certain taxa: A. aculeata, Attalea maripa, Bactris simplicifrons, Dictyocaryum fuscum, Euterpe precatoria, Iriartea deltoidea, Oenocarpus bacaba, Socratea exorrhiza and the genera Geonoma, Bactris, Euterpe and Oenocarpus. Therefore, the influence of past human activities and environmental changes through time on palm abundances can be robustly assessed for those taxa. Differences in the modern distribution or edaphic factors may be used to rule in or out species or genera that share phytolith morphotypes (of similar size and shape).

Spheroid phytoliths: Size matters
PCA confirmed the morphological differences between the previously established spheroid morphotypes (Morcote-Ríos et al., 2016). Size measurements can be used to increase taxonomic resolution of palm species that produce spheroid phytoliths. Spinule traits were often correlated with size, meaning that larger phytoliths produced spines that were longer, wider and had more distance between spines (Supplementary material "Correlation"). Therefore, focusing on diameter differences between species also implicitly considers spinule traits. We have summarized our findings into a 'Quick Guide' (Fig. 7) to identify palm species producing spheroid phytoliths, for future phytolith research in tropical South America. Future work can refine and modify the guides and include more species.
For SPH_SYM and SPH_TUB morphotypes, separating based on size will allow identifying the larger phytoliths of Oenocarpus bacaba and Attalea maripa. For SPH_ACU morphotypes, phytoliths >20 μm are likely produced by Euterpe and Oenocarpus species (Morcote-Ríos et al., 2016;Piperno et al., 2019;Piperno and McMichael, 2020). Euterpe oleraceae also produces SPH_ACU > 20 μm . Mauritiella armata produced small SPH_ACU phytoliths, and phytoliths that could be small SPH_ACU or SPH_ECH phytoliths with many projections. This might explain why in other studies, Mauritiella armata was not reported to produce SPH_ACU morphotypes (Morcote-Ríos et al., 2016). Small SPH_ACU phytoliths (<15 μm) were also produced by Attalea speciosa, and they can also be produced by Chelyocarpus, Manicaria, Prestoea or Syagrus species (Morcote-Ríos et al., 2016;Piperno and McMichael, 2020). Attalea maripa produced few SPH_ACU phytoliths that resembled large versions of SPH_ECH phytoliths. Because SPH_ECH phytoliths are between 4 and 18 μm long, we categorized these larger phytoliths as SPH_ACU. But other phytolith researchers might categorize them as (large) SPH_ECH. Lowering taxonomic resolution based on the morphotype of SPH_ECH or ELL_ECH alone is difficult. Differences in the combination of phytoliths produced may help to increase taxonomic resolution for species when globular phytoliths overlap in size. For example, if large SPH_ECH phytoliths are identified in the absence of SPH_SYM phytoliths, it is likely not produced by Mauritia flexuosa. Also, Attalea maripa is the only species that produces large SPH_ECH, ELL_ECH and SPH_TUB phytoliths (Fig. 7). But given the small number of species studied, such interpretations cannot be made with absolute certainty. Also, using relative frequencies can be difficult in fossil samples when multiple species may have contributed to the phytolith assemblage.
Other studies also identified spheroid phytoliths with rounded projections in Oenocarpus bacaba, and smaller spheroid phytoliths with rounded projections were seen in Oenocarpus minor (Piperno et al., 2021). Future research should study more specimens of Oenocarpus to confirm if this morphotype is indicative of the genus. We found no SPH_ACU phytoliths for Oenocarpus bacaba and only found SPH_ACU in Oenocarpus mapora during extended scans, in contrast to other studies , this is likely because this morphotype is produced in rare amounts. A previous study (Morcote-Ríos et al., 2016) reported that M. flexuosa produced only SPH_SYM morphotypes, whereas we also identified SPH_ECH phytoliths. Hyospathe elegans has been reported to produce only ELL_ECH phytoliths (Morcote-Ríos et al., 2016), whereas in this study, SPH_ECH phytoliths were also identified. These differences highlight the need to study the variation of phytolith traits within multiple individuals of the same species. Another explanation might lay in the eye of the phytolith researcher. To clarify our categories, we have included Figs. S1-S4. Phytolith preservation in fossil samples should also be studied for Amazonian soils and lake sediments, to take a possible preservation bias of phytolith size into account (Albert et al., 2009). 4.1.2. Conical phytoliths: Size and shape matter PCA confirmed the morphological differences between the 4 conical morphotypes (Huisman et al., 2018). Separating 4 morphotypes will increase taxonomic resolution because species differ in the morphotypes they produce: Iriartea deltoidea and Socratea exorrhiza produced only CON_FEW, and Bactris gasipaes produced only CON_FEW and CON_ECH. CON_BAS morphotypes can be used to identify Geonoma species (Huisman et al., 2018;Morcote-Ríos et al., 2016). We have summarized our findings into a 'Quick Guide' (Fig. 8) to identify palm species producing conical phytoliths, for future phytolith research in Amazonia. Future work can refine and modify the guides and include more species.
If CON_FEW phytoliths are separated based on size and shape differences, the larger phytoliths of Iriartea deltoidea and Socratea exorrhiza and the smaller phytoliths of Dictyocaryum fuscum may be identified. Because Wettinia hirsuta barely produced CON_FEW phytoliths, this species is excluded from Fig. 8, but it's morphotype is like Socratea exorrhiza. We have focused on size and shape differences between CON_FEW morphotypes, because spinule traits reflect size differences and the standard error of the mean overlaps between the number of projections.
If CON_ECH phytoliths are separated based on size differences, the larger phytoliths of Bactris simplicifrons and smaller phytoliths of Aiphanes aculeata may be identified. Although Bactris simplicifrons produces a wide range of CON_ECH phytoliths, it was the only species that produced CON_ECH phytoliths >17 μm. Also, Bactris species produced CON_ECH phytoliths with fewer projections, allowing possible identification of this genus, but more Bactris species should be studied to determine this.
Separating CON_TAB phytoliths based on size and shape differences may allow identification of A. aculeata, because the mean and median of CON_TAB phytoliths of other species was greater than 10 μm, despite overlap in range. Also, a combination of small CON_ECH and CON_TAB phytoliths likely indicates the presence of A. aculeata. Although this was not measured, the top projections of CON_TAB phytoliths produced by Wettinia hirsuta seemed less echinate compared to other species (Fig. 6u). Future research should include other Wettinia species to study if this texture is characteristic of the genus.
All Geonoma species differed in the maximum diameter of CON_BAS phytoliths, but because the ranges overlap and we included only 3 of the 17 Geonoma species in Amazonia (Henderson et al., 2019), we have not used this size difference to separate within the Geonoma genus. Phytolith morphotypes produced by previously assessed species were mostly corresponding to the literature (Huisman et al., 2018;Morcote-Ríos et al., 2016;Piperno et al., 2019;Piperno and McMichael, 2020). But there were some differences, such as in the occurrence of reniform phytoliths (Table S1). Phytolith morphotypes produced in rarity might vary more between reference samples and fall outside the phytolith count.

Other phytolith morphotypes
EUT_SPH phytoliths seem to be indicative of Euterpe species (Huisman et al., 2018). Because of the low abundance but large size of EUT_SPH phytoliths, we suggest doing an extended scan at 200X magnification to confirm the presence or absence of this morphotype, outside of the phytolith count. The woody phytoliths of Geonoma undata may be species-specific or typical for woody phytoliths, as A. aculeata and Euterpe precatoria also produced some woody phytoliths (Fig. S4-r). The seed phytoliths of A. aculeata and Dictyocaryum lamarckianum were not found in other seeds (Fig. S4-mo). The rare FUS_ECH phytoliths produced by several species (Table S1) were similar to woody and seed phytoliths of Ceroxylon alpinum, that were reported as malformations (Huisman et al., 2018). This research was not performed with different plant parts per specimen. Sampling more plant parts to create a nested design will reveal if plant parts produce different phytoliths, and if certain morphotypes or malformed phytoliths are typical for seeds or woody parts. Also, the effects of environmental variability on plant growth and subsequent phytolith size must be better studied for positive species-specific identifications to be made based on phytolith size itself.
Another explanation for variation among phytoliths is phylogenetic differences (Faurby et al., 2016). For example, larger CON_FEW morphotypes were restricted to Iriarteae species and CON_BAS phytoliths were restricted to Geonomateae species. Also, CON_ECH phytoliths with 7-9 projections only occurred only in Bactritinidae species and Aiphanes produced smaller phytoliths than species from other tribes. Finally, only Euterpeae and Oenocarpus species produced large SPH_ACU phytoliths (> 20 μm) . But phylogenetic differences alone are unlikely to explain variation in phytolith production, because spheroid phytoliths are produced by several tribes (Euterpeae, Cocoseae and Lepidocaryeae) (Huisman et al., 2018;Morcote-Ríos et al., 2016). Also, variation in size, shape or number of projections varied widely within a genus. Studying the phytoliths of Astryocaryum and Desmoncus species, who are closely related to Bactris and Aiphanes, might shine light on the role of phylogenetic relatedness in the morphology of conical phytoliths.

Future palaeoecological research
The quick guides of spheroid and conical phytoliths (Figs. 7, 8) can be used to in future phytolith analyses to reconstruct the distribution and abundance of specific palm species and genera through time, improving our ability to understand how people and the environmental context have shaped the widespread occurrence of Amazonian palm species. Our results allow separating the past histories of palm species used for their trunks (such as Iriartea deltoidea, Socratea exorrhiza) from palm species used for food (such as Euterpe precatoria), which is useful, because people may have locally depleted species used for wood but increased the abundance of palm species they ate. Understanding the role of past human activities, such as agroforestry, on modern palm abundance and occurrence, will help to understand the Fig. 8. A quick guide to identify conical palm phytoliths under the microscope, using morphological differences in number of projections, size and shape of 11 species and previously established morphotypes (Huisman et al., 2018;Morcote-Ríos et al., 2016). temporal and spatial scale of past human impact in Amazonia, which is a highly debated (Levis et al., 2017;Lombardo et al., 2020;McMichael et al., 2017;Piperno et al., 2021Piperno et al., , 2019. Given the large number of palm species used in Amazonia (currently 134), expanding this dataset of phytolith traits from reference material with species such as Astryocaryum murumuru or Phytelephas macrocarpa will be of great value to those studying Amazonian phytoliths (and the role of past human activities). Because this dataset only contains 30 phytoliths per sample, expanding this dataset is necessary to allow more robust statistical analysis. Because measuring and counting phytolith is labour intensive, the automatic classification of phytolith morphometrics may be a promising area for future research.

Conclusion
Although only 24 Amazonian palm species were included in this research, results show potential for phytoliths of the following species to be identified on the basis of their size, shape or number of projections: A. aculeata, Attalea maripa, Bactris simplicifrons, Dictyocaryum fuscum, Euterpe precatoria, Iriartea deltoidea, Oenocarpus bacaba, Socratea exorrhiza and the genera Geonoma, Bactris, Euterpe and Oenocarpus. This will allow future phytolith analyses to reconstruct past vegetations with an increased taxonomic resolution, especially of hyperdominant and useful palms. We recommend that more specimens of some of these species should be studied to determine if phytolith morphology is sensitive to environmental variability. However, it appears that important economic species of Oenocarpus and Euterpe, O. bataua and Emesaya precatoria, consistently produce SPH_ACU phytoliths larger in size (>20 μm) than in some congenerics and other SPH_ACU-producing taxa studied so far, as plants sampled from different areas of Amazonia showed little phytolith size difference when compared with our results (Morcote-Ríos et al., 2016;Piperno and McMichael, 2020). The growing number of vegetation inventories being carried out in lowland Amazonia and knowledge of where different species are distributed due to edaphic or other preferences may also allow greater taxonomic resolution (Piperno et al., 2021). The quick guides of Figs. 7 and 8 can be used to test hypotheses regarding the abundance and occurrence of Amazonian palms and the possible influence of past human activities.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.