Data collection

We used data of 4,454 arboreal palm and tree species distributed in 1,946 plots across the Amazon basin and Guiana Shield (Amazonia) compiled by the Amazon Tree Diversity Network (ATDN). We started with the data available in ter Steege et al. [6]. We then updated this data set with the currently adopted taxonomic nomenclature of arboreal species and their estimated population sizes for Amazonia available in ter Steege et al. [37], reducing the number of species from 4,962 [6] to 4,454. We used estimated population sizes as a measure of abundance. In the ATDN inventory plots, individuals of trees and palms with ≥ 10 cm DBH were sampled mostly in one-hectare plots located in five main types of mature lowland forests in Amazonia: terra firme (non-flooded), white-sand (campina and campinarana), and seasonally or permanently flooded terrain (várzea, igapó, swamp).

Ter Steege et al. [6] showed that only 227 hyperdominant arboreal species dominate Amazonian forests and represent half of the total arboreal individuals in Amazonia. After correcting for synonyms [37], 222 species remained for our analysis; their estimated population sizes varied from approximately 298 million to 4.7 billion individuals [for more details of estimated population sizes see ter Steege et al. [37]].

Of the 85 arboreal species known to have domesticated populations [26], we identified 80 in our study (Annona muricata, Annona squamosa, Crescentia cujete, Elaeis oleifera and Sapindus saponaria were excluded from ATDN datasets after corrections). The list of domesticated species followed Clement et al. [33] and Levis et al. [26]. The degree of domestication was based on magnitude of phenotypic (and sometimes genotypic) variability between cultivated and wild populations following Clement [33] and Levis et al. [26]. Domesticated species were assigned to use categories as with any other useful species.

We used 29 Amazonian ethnobotanical scientific papers (16), books (9), book chapters (3) and a doctoral dissertation (1) published between 1926 and 2013 in various parts of Amazonia (Fig 1; this reference list is provided in the S1 Text in S1 Appendix) to identify the uses of arboreal species for daily-life or commercial purposes (S1 Table). The ethnobotanical studies covered different regions and ethnic groups, including indigenous people and local communities (Fig 1). Among these, 19 are local studies and 10 are compendia of several studies (Fig 1c and S2 Text in S1 Appendix), including large-scale compilations of local ethnobotanical studies and other large ethnobotanical books and compendia (e.g., the book by Richard E. Schultes, Plants of the Gods). The identification of a use by one culture or community does not mean that all cultures and communities use the species similarly, but does demonstrate a use. We grouped subspecies or varieties mentioned in the studies into the corresponding species, and accepted species with “cf.” identification as belonging to the named species. We only considered studies that adopted botanical nomenclature with specimens identified at the species level and we excluded those that only presented common names.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1.

(a) Location of the 29 ethnobotanical studies of useful arboreal species in Amazonia. The citation of the ethnobotanical study (see the corresponding references in S1 Appendix) is given for each location in the map. Ethnobotanical studies are classified in two categories of spatial coverage: local studies (red dots) and compendia (asterisks). Black asterisks represent studies conducted in a country, and blue asterisks represent studies in the State of Pará, Brazil, and the Orinoco River basin, Venezuela. The number of asterisks represents the number of compilations in a given country. The studies of Patiño [38] and Revilla [39] are not represented on the map, since they cover the entire Amazon; (b) Cumulative number of useful arboreal species documented in Amazonia, ordered by the period of publication (1926–2013); (c) Species accumulation curve showing the 29 studies ordered by contribution of new species (black dots) and the total number of useful species that each study contributed to the dataset (red dots, and blue and black asterisks). The highest asterisk values correspond to Corrêa [40], de la Torre et al. [41] and Revilla [39], carried out in Brazil, Ecuador and for all of Amazonia, respectively. Base map source: Natural Earth, Forest-GIS and Eva et al. [42].

https://doi.org/10.1371/journal.pone.0257875.g001

All uses recorded from the literature review were classified into ethnobotanical categories based on Prance et al. [43] and Macía et al. [44]: food, medicine, manufacturing, construction, thatching and firewood (S1 Table). For each of the arboreal species we assigned a main use category, which was determined by the most cited use category in the references. Assigning a main use may help us have a more comprehensive understanding of what is the most common use for each species. Our final dataset (available in the Data availability section) included the currently adopted taxonomic nomenclature of the species, all use categories mentioned for each species and the references of the plant use information. We constructed a collector curve to assess the cumulative number of useful arboreal species recorded during the last century in the ethnobotanical studies (Fig 1b and 1c). We also performed a literature review of archaeological, archaeobotanical and ecological studies about long-term use of almost all hyperdominant species or genera with incipiently and semi domesticated populations (S2 Table).

Data analysis

All analyses were performed in R software [45]. To test if useful species tend to be the most abundant species, we compared the mean population sizes of useful and non-useful species with a one-way ANOVA. We also investigated if mean population sizes differ between useful and non-useful species among phylogenetically related species (see S1 Text in S1 Appendix for more details). We used Generalized Linear Mixed Models (GLMM) (lme function of nlme package [46]; S1 Fig), with “Family” and “Genera” as random factors. We adopted this analytical framework to correct for the assumption that more closely related species may have more similar population sizes [15]. We also evaluated the chance that a given species is useful as a function of its population size using a logistic regression (see S1 Text in S1 Appendix for more details). Mean population size was log₁₀ transformed before all analyses to normalize variable distribution [47].

To assess whether the hyperdominant non-useful species (species in which the chance to be useful ranges from 81.2 to 93%) are indeed non-useful, we performed a complementary literature search to find at least one use for each of those species (S7 Table in S2 Appendix) in Google Scholar, ScienceDirect and Web of Science, yet we did not use this complementary search in our analysis and results. The search used the scientific name of each plant species with the keywords “plant + use”, “ethnobotanical + studies”, “ethnobotany”, “local + knowledge”.

To assess if population sizes of species differ among the seven use categories, we used bootstraps to estimate means and standard deviations for (i) species used in one use category (called ‘single use’), (ii) species used in more than one use category (‘multiple uses’), and (iii) species that may be used in more than one category, but with a main use category (‘main use category’). The number of use categories varied from zero (no use) to six (all categories). Bootstrapping was applied to account for the large differences in sample sizes (number of species) among the use categories [48].

To assess whether population sizes of species differ among useful and non-useful categories, and the degrees of domestication (incipiently, semi and fully domesticated species), we also performed bootstrap analyses. We identified significant differences by the lack of overlap between standard deviations [49]. For these analyses, we used the function groupwiseMean (R companion package) with a confidence interval of 95% and 9999 randomizations [48]. To understand if larger species population sizes are associated with broader spatial distributions across Amazonia, we evaluated the relationship between the population sizes of species and the number of plots in the ATDN network where they occur across Amazonia using a linear model (LM) after log₁₀ transformation to normalize both variables.

Do abundant species have a higher probability than rare species to be useful?

The population sizes of useful plant species were much higher than those of non-useful species (Fig 2), and include many hyperdominant species, which are more useful than expected by chance. High numbers of individuals of useful plants are commonly observed at local spatial scales in Amazonia [8, 51], including oligarchic forests dominated by useful species, sometimes of commercial importance [22, 23]. These patterns highlight the great usefulness of modern Amazonian forests at scales from the local to the regional, suggesting that the usefulness of a given plant species may be shared among many societies, despite their sociocultural differences. The greater abundance of useful species also holds true within genera and families (S1 and S2 Figs), although closely related species are expected to have similar evolutionary histories and abundances [15]. This implies that the probability of utility increases with abundance across all evolutionary groups despite evolutionary forces that drive patterns of abundance of phylogenetically related species.

Although our model suggests that the probability of a species being useful increases with population size, humans also use rare species (Annona coriacea and Pouteria collina) (Fig 3). Our finding of rare species being useful may indicate that these species have specific or preferred properties, such as those responsible for edibility or medicinal use [5], or growth form, such as the tall-stemmed palms, used preferentially as materials for souvenirs in northwestern Amazonia irrespective of their abundance [4]. Another possibility for explaining this finding is that many of these species are regionally rare but may be locally abundant, such as Bactris major [52] and Rhizophora mangle [53].

The opposite pattern is also true; we found that some hyperdominant species are not being used by humans. The existence of two hyperdominant species (Eschweilera atropetiolata and Pouteria elegans) without a reported use (S7 Table in S2 Appendix) may be explained by several factors, such as: (i) these non-useful species have undesirable traits for people (e.g., fruit taste); (ii) cultural aspects may influence people’s choice and preferences of uses [54]; (iii) uses were not recorded by ethnobotanists [55]; or (iv) we could not find uses for these species in the literature. Lack of data on usefulness may have occurred in our study, although this is less probable for hyperdominant species because they are more easily found in the landscape and their uses have a greater chance to be recorded.

How does population size vary among use categories and among one or multiple use categories?

Useful species generally had greater population sizes than non-useful species, even if they are used for a particular purpose (Fig 4 and S3 Fig in S2 Appendix), supporting the positive relation between plant use and abundance. Among use categories, the high abundance of species used for thatching is mostly attributed to the use of palms with remarkable economic and cultural importance [56], possibly a reflection of the many uses that most of these species have (on average palm species have four uses while tree species have two uses). Their ancient uses [57] might also have led to an increase of their populations, which is inferred by archaeobotanical studies in Amazonia [58–60]. Contrary to what we expected for those species whose uses involve the extraction of trunks, i.e., suppression of individuals, we found a high overall abundance of species used for manufacturing, construction and firewood. One possible explanation is that physical properties of species for timber and technological uses, such as mechanical resistance or durability, are shared by many species and are more easily substituted by people [5], although there are preferences for some species [61]. Another explanation is that the suppression of individuals locally is probably of low impact (of a few individuals) and does not lead to a decrease in the overall abundance of these species regionally [62], and therefore it is not noticeable in the data used. Moreover, even the cutting of palms and trees by indigenous people may contribute to forest regeneration. The Hi-Merimã people, an isolated Arawá group in the South Amazonas, harvest the fruits of bacaba (Oenocarpus mapora), açaí (Euterpe precatoria), buriti (Mauritia flexuosa), patauá (Oenocarpus bataua) and sorveira (Couma macrocarpa) by cutting the trees and palms, and the seeds are left in the surroundings of the camps when the individuals are felled [63]. This practice, in addition to the great territorial mobility of Hi-Merimã people, as of other Arawá people, result in an enrichment of useful plants after leaving a territory [63].

Population sizes of arboreal species were positively associated with their number of uses (S4 Fig). This suggests that the wider geographic distribution of the most abundant species might influence the diversity of plant uses due to the diversity of preferences and ethnobotanical knowledge across cultures in Amazonia [43, 44, 54]. Moreover, it is likely that the more uses a species has, the more likely it is to be actively managed because of the benefits gained from this effort [64]. Our results support the idea that exchange networks among indigenous groups in Amazonia [24] may have dispersed many plants with multiple uses [22, 65], such as for food, handicrafts, spiritual purposes and a range of other daily life activities, used not only for utilitarian or economic purposes but also with symbolic and cultural value [66]. Indeed, genetic data have suggested geographical dispersal by people in Amazonia of multi-purpose species, such as Brazil nut [28], peach palm (Bactris gasipaes) [67] and cacao (Theobroma cacao) [68]. Similar studies for other useful hyperdominant species will allow us to understand better the relation between (past) human populations and the abundance and distribution of arboreal useful species.

How does population size vary with degree of domestication?

Near archaeological sites, legacies of pre-Columbian and post-Columbian indigenous peoples are detectable in modern forest composition by looking at richness and abundance of domesticated species [26, 69, 70]. Our analysis found that the largest population sizes across Amazonia were generally of incipiently domesticated species (Fig 5). This suggests that incipiently domesticated species were and are managed and propagated by indigenous peoples and local communities, and their populations survive and persist in forests after abandonment and without recent management [26, 70, 71]. Indeed, many incipiently domesticated species are clearly visible in early archaeobotanical records across Amazonia, attesting to ancient food use and probable landscape domestication [57]. Long-term use of almost all hyperdominant species or genera with incipiently domesticated populations have been suggested by archaeological, archaeobotanical and ecological studies (S2 Table) that show an association of past landscape management with the presence of incipiently domesticated species found in archaeological sites, such as the açaí palm (Euterpe precatoria), patauá (Oenocarpus bataua) and buriti (Mauritia flexuosa) [57, 72, 73]. At the Cerro Azul archaeological site in Colombian Amazonia, seeds of inajá (Attalea maripa), açaí (Euterpe precatoria), buriti (Mauritia flexuosa) and patauá (Oenocarpus bataua) appear around 12,000 years ago and were probably used for food [74], as they are today. At the Teotônio site in Brazilian Amazon, seed remains of piquiá trees (Caryocar sp.) and Brazil nut appear in the early and mid-Holocene associated with the lifeways of these populations [75]. The fact that incipient domesticated species are six times more likely to be hyperdominant than expected by chance suggests that the long-term use of their populations is associated with processes of forest enrichment [22], leaving a lasting legacy in modern Amazonian forests, as previously hypothesized [6].

Modern large-scale economic forces may have also driven changes in population abundance of several commercially important species in Amazonia. During the 20^th century, some arboreal species were intensively exploited to supply both local and international markets in Europe and the USA [76]. While some populations of rubber and Brazil nut were enriched through management [27, 70, 77], several other species were depleted by the extraction of individuals through logging [e.g., massaranduba (Manilkara elata) and itaúba (Mezilaurus spp.) for timber, sorva (Couma spp.) for chewing gum and rosewood (Aniba rosiodora) for essential oil [3]]. For these species, logging pressure may have reduced their populations near settlements in the last century. Based on projected forest losses, which include historical deforestation from 1900 to future deforestation up to 2050, these species are qualified as globally threatened under IUCN threat status criteria (critically endangered, endangered and vulnerable species) [78].

Fully domesticated populations have low mean population sizes in forests, similar to non-useful species, and do not include any hyperdominant species (Fig 5). This suggests that people do not try to manage these species in mature forests, such as the forests inventoried by the ATDN. Instead, fully domesticated species are adapted and kept in agroecosystems created and maintained by people [22, 34, 79]. A higher number of fully domesticated individuals in Amazonian landscapes is thus expected in homegardens and swiddens where they are still cultivated [33, 34, 36].

Historical people-plant relationships: Contemporary perspectives

The positive relation between plant use and abundance we found offers information for two currently debated hypotheses [80]: people use the most abundant plants or people have increased plant abundance through long-term management. One of the main hypotheses in ethnobotany is that at the landscape scale the availability of plant species influences their uses due to ease of gathering and transportation [81]. Therefore, the most abundant plant species are more frequently used by people of numerous cultures [82], as observed in this study and across tropical forests in different parts of South America [5, 8, 10, 83]. Although it is reasonable to think that people may have taken advantage of plants that are naturally very abundant due to their adaptation to natural conditions, there is evidence from historical and current studies that indigenous people have influenced the distribution and abundance of useful plants [22, 34]. People’s practices have probably contributed, intentionally or unintentionally, to the enrichment of many useful arboreal species locally [12, 25] and regionally [26]. Yet, questions about the influence of long-term management on the abundance of useful palms and trees in present-day Amazonian forests remain open and more interdisciplinary approaches may help us understand pre-Columbian legacies in Amazonian forests [84].