GuiaTreeKey, a multi-access electronic key to identify tree genera in French Guiana

Abstract The tropical rainforest of Amazonia is one of the most species-rich ecosystems on earth, with an estimated 16000 tree species. Due to this high diversity, botanical identification of trees in the Amazon is difficult, even to genus, often requiring the assistance of parataxonomists or taxonomic specialists. Advances in informatics tools offer a promising opportunity to develop user-friendly electronic keys to improve Amazonian tree identification. Here, we introduce an original multi-access electronic key for the identification of 389 tree genera occurring in French Guiana terra-firme forests, based on a set of 79 morphological characters related to vegetative, floral and fruit characters. Its purpose is to help Amazonian tree identification and to support the dissemination of botanical knowledge to non-specialists, including forest workers, students and researchers from other scientific disciplines. The electronic key is accessible with the free access software Xper², and the database is publicly available on figshare: https://figshare.com/s/75d890b7d707e0ffc9bf (doi: 10.6084/m9.figshare.2682550).


Introduction
Th e tropical rainforest of Amazonia is one of the most species-rich ecosystems on earth, with an estimated 16000 tree species and often more than 200 species of trees per hectare (ter Steege et al. 2013). Due to this high diversity, the botanical identifi cation of Amazonian trees is very diffi cult and often requires the consultation of taxonomic specialists. Taxonomists usually specialize in only one or few families or genera (Bacher 2012;Joppa et al. 2011) with few botanical experts, including generalist taxonomists and parataxonomists (Schmiedel et al. 2016), able to identify specimens of various families to the genus level. As a result, the number of specimens incorrectly named or unnamed is still very high in many forest inventories and more widely in the world's plant collections (Goodwin et al. 2015). Also, traditional dichotomous keys impose constraints on identifying tree samples because they rely on a hierarchical and fi xed organization of characters that hampers the identifi cation of a sample when one or several characters are not observed at the time of collection. Moreover, traditional keys often focus on Linnaean characters (fl owers and fruits, Rejmánek and Brewer 2001), and only few tools aid the identifi cation of plant species based on vegetative characters (Belhumeur et al. 2008). Th e identifi cation of sterile samples is therefore diffi cult even though sterile samples are much more common than fertile ones (the proportion of sterile specimens commonly reaches 90-95%, Aymard et al. 2009;Martinez and Phillips 2000) and taxonomists are often unwilling to review sterile material. Furthermore, printed keys are static and are not frequently revised and republished to refl ect taxonomic changes. Here we attempt to modernize botanical identifi cation in Amazonia, by developing a user-friendly electronic key to help tree identifi cation in French Guiana.
French Guiana is a French overseas department of about 85000 km² located in the eastern Guiana shield; it is home to approximately 2000 tree and palm species belonging to 404 genera (updated checklist by Molino et al. 2009). We introduce an original multi-access electronic key for the identifi cation of tree genera occurring in French Guiana terra-fi rme forests based on a set of 79 morphological characters related to vegetative, fl oral and fruit characters that can be selected in any order. In addition, almost all characters and genera are described and illustrated. Its purpose is to help the identifi cation of tree samples and to disseminate botanical knowledge to non-specialists.

General spatial coverage
French Guiana is bordered to the east and south by Brazil and to the west by Suriname ( Figure 2). About 90% of the region is covered by evergreen rainforest occurring principally on terra-fi rme soils of granitic or sedimentary origins. Th e relief is fairly fl at with a mean altitude of 140 m and few peaks exceeding 800 m. Th e climate is equatorial, characterized by a mean annual temperature of 26°C and annual rainfall varying from 2000 mm in the south and west to 4000 mm in the northeast. Th e rainy season usually occurs between May and August and the dry season between December and January (Guitet et al. 2014). Th is key covers French Guiana, but the geographical usefulness is by no means restricted to French Guiana: 99% of the genera included in this key are also present in Brazil (Refl ora -Virtual Herbarium. Available at: http://refl ora.jbrj.gov.br/refl ora/ herbarioVirtual/ Accessed on 21/3/2016), more than 90% in Suriname and Guyana, and more than 80% in the Venezuelan states of Amazonas and Bolìvar (Funk et al. 2007). Th e electronic key may thus also be used in and easily extended to these other regions, although users should keep in mind that these regions also include many other tree genera not covered by this key.
Th e characters are grouped into four main sections: 'leaves', 'other vegetative characters', 'fl owers', and 'fruits and seeds'. A substantial proportion of characters (33 of 79) is related to leaves which are almost always observable. In addition, many vegetative characters rarely used in classical dichotomous keys are suggested (e.g., presence of latex, type of trichomes, leaf base venation). Almost all genera and characters are defi ned and illustrated with more than 9000 photographs (mainly herbarium specimens). Among the 79 characters, 74 are qualitative and 5 are quantitative. Qualitative characters are scored by the presence or absence of the character (e.g., opposite leaves), while quantitative characters are scored by the minimum and maximum number of modalities potentially observable for each genus (e.g., number of calyx segments ranging from three to fi ve): the user may thus select the exact number of modalities observed in the sample.

Genera-characters matrix file description
Th e electronic key consists of a genus × character matrix consultable with Xper², a software dedicated to taxonomic descriptions and computer-aided identifi cation (Ung et al. 2010). Xper² is a user-friendly management system for creating interactive identifi cation keys available on Windows, Mac or Linux in French, English or Spanish versions. It is free software and the botanical keys can be installed locally in order to be used without an internet connection, which is not allowed in the most recent version of the program (Xper 3 ).

PANE 1: Characters box:
In the left pane, the characters are listed and organized by categories and sub-categories (i.e. 'leaves', 'other vegetative characters', 'fl owers', and 'fruits and seeds', Box a). Th e user is invited to describe his/her sample using the characters listed in this pane. During the identifi cation process, the user can access a summary of the characters that have been selected (Box b). PANE 2: Defi nition and illustration box: Th e middle pane displays the defi nition and illustration of characters and retained genera.
PANE 3: Results box: Th e right pane displays the results in real time. It lists the genera that fi t the selected characters. Genera are listed in alphabetic order and they are combined with their family name. A botanical description and photographs of each genus may be displayed in pane 2 by clicking on the genus.

Examples of identification using GuiaTreeKey
In this section, we provide several examples of identifi cation using the electronic key (Figures 4-7).