Three decades of tree inventories from French Guiana now open to all

More than three decades of forest inventory records from French Guiana have been compiled into a single, openly accessible dataset covering over 100 hectares of lowland tropical forest and roughly 90,000 trees. The collection, known as GUYADIV, brings together data gathered between 1986 and 2018 from 242 plots scattered across the territory, and its creators describe it as a foundation for studying how tree diversity and function vary across tropical communities at multiple spatial scales. The dataset, published in the journal Annals of Forest Science, is available through the IRD DataSuds repository.

The scale of the effort is considerable. Of the 90,409 trees recorded in the database, 98 percent were identified to species level, spanning 1,393 species, 370 genera and 84 families. Most of the plots sit in old-growth terra firme forests that have experienced little or no recent disturbance, although a handful capture logged forests, patches of swamp forest, sites of ancient pre-Columbian occupation and monodominant Spirotropis forests. The network stretches across a broad elevational range, from 30 to 800 metres above sea level, though most plots lie below 400 metres, and spans much of the north-south climatic gradient, with annual rainfall ranging from roughly 2,000 millimetres to nearly 4,000 millimetres.

GUYADIV was initiated in 1986 by a small core team of botanists specialising in tree identification and ecology, with a third joining in 1993. Data collected before 1993 were computerised in a database management system developed that same year, and from 1994 onwards the records have been managed in a relational system continuously refined since. Over the years the network expanded through contributions from researchers affiliated with national and international institutions. Among the main contributors, the founding botanist initiated most of the plots, contributing to 95 percent of them and to 65 percent of tree identifications, while a second contributed to 48 percent of the plots, including all those censused twice, and to 48 percent of identifications.

The inventory design falls into two categories. Exhaustive plots cover the entire tree community within a delimited area; 157 of these are represented in the dataset, covering 98 hectares and comprising 76,416 trees, with individual plot sizes ranging from 0.04 to 1.4 hectares. In each, trees were identified as precisely as possible, individually tagged, their diameter at breast height measured, and their positions recorded using a plot-specific local coordinate system. Non-exhaustive plots, of which there are 85, sample a subset of trees around regularly spaced points using the Point-Centered Quarter method and its variants, capturing 13,993 trees. In these, the space around each point is divided into four quadrants and the closest tree in each is censused, meaning the design covers only a fraction of the stems within a given area.

Botanical identification was a central and demanding part of the work. Common and well-recognised species, representing 50 to 80 percent of the trees depending on the plot, were identified in the field using binoculars, usually by two botanists working together. Samples were taken from the remaining 20 to 50 percent of trees, collected either by climbing with a rope, using a pole pruner, or shooting branches from the ground with an elongated single-barrel shotgun. Fertile specimens of uncommon or unknown species were systematically sought, sometimes by revisiting plots years after the initial census. Representative specimens were deposited as herbarium vouchers at the French Guiana herbarium, which houses nearly 40,000 specimens collected from trees in the region.

Of the 1,393 species recorded, 1,168 are formally published species while 225 carry provisional names, such as Pouteria sp. B or Eugenia FG-9. These provisionally named species are either ones that could not be accurately identified without fertile material, cryptic species within species complexes, or in many cases new species never formally described. Overall, 1,589 trees, or 1.8 percent, remained unidentified at the species level, while 4,374 trees, or 4.8 percent, were identified by provisional names. Taxonomic and nomenclatural updates were incorporated during 2025 based on the checklist of tree species of French Guiana, recent literature and the French national taxonomic reference system.

The diameter threshold defining a tree varied by research objective. It was 10 centimetres for 78 percent of the trees censused, but dropped to 1 centimetre in some plots in the Spirotropis area and to 2 centimetres in others at Piste de Saint-Élie and Paracou. Palms are included because their stipe is ligneous and they occupy the same structural position as woody trees. During initial censuses, the phenological stage of each tree was recorded, distinguishing sterile, flowering and fruiting stages. Overall, 96.5 percent of trees with a diameter of at least 10 centimetres were recorded at the sterile stage, a figure the authors note is noteworthy given that many inventories took place during periods known to be those of flowering and fruiting at the community level.

Most plots were inventoried only once, though some were re-inventoried several years later. In total, 20,652 trees from the Piste de Saint-Élie and Paracou sites were re-inventoried, and for the 23,681 trees re-inventoried the dataset identifies newly recruited individuals and records dead trees, with information on the type of death when available. Because the point of measurement was not permanently marked and measurements were taken at approximately 1.3 metres above the ground, the authors caution that repeated measurements should be interpreted carefully. As a consistency check, they calculated a mean annual diameter increment for non-palm trees of 0.12 centimetres per year, within the range of 0.09 to 0.17 centimetres per year reported for a nearby undisturbed plot dedicated to forest dynamics studies.

The dataset has already contributed to numerous studies at local, regional, continental and global scales, partly through the integration of its 1-hectare plots into major international networks such as the Amazonian Tree Diversity Network and Forestplots.net. The botanical vouchers collected during inventories have also contributed to the discovery and confirmation of several hundred tree species in French Guiana and to the establishment of comprehensive checklists of the region’s tree flora. The 1,393 species recorded represent approximately 70 percent of the 1,668 validly named species listed in that checklist, though the species accumulation curve for the complete dataset continues to rise without reaching an asymptote, indicating that current sampling remains incomplete. An earlier estimate proposed roughly 2,200 tree species for the region, while a log-series model fitted to the abundance distribution predicts a substantially larger total of 4,368 species.

Beyond documenting what has been found, the data reveal where future work should go. Among non-monodominant exhaustive plots of roughly 1 hectare with trees of at least 10 centimetres diameter, tree density ranges from 382 to 1,055 stems per hectare, while species richness ranges from 92 to 211 species per hectare. Sampling is concentrated in northern French Guiana, leaving large and relatively inaccessible areas poorly documented, particularly the lower Maroni and Mana basins in the north-west and the middle Oyapock and southern Approuague basins in the east. Standardised measures of local richness show relatively low values in the centre and south of the territory, with higher values concentrated in the north-east, a pattern consistent with north-eastern French Guiana and neighbouring Amapá being among the areas of highest alpha diversity in Amazonia. Because the plots provide snapshots rather than continuous monitoring, the authors note the dataset is not well suited to studying forest dynamics or monitoring biomass over short periods, but it offers a valuable baseline for future long-term studies and a practical guide for planning new inventories, including the most suitable periods for collecting fertile specimens.

Subject of Research: A long-term forest plot inventory dataset documenting tree diversity in French Guiana

Article Title: Thirty-two years of inventory data (1986–2018) from the GUYADIV forest plot network in French Guiana, covering 100 ha and 90,000 trees

Article References: Sabatier, D., Molino, J.-F., Brousseau, L., Smock, J.-L., Pélissier, R., Geniez, C., Fonty, E., Odonne, G., Paget, D., Goret, J.-Y., Madelaine-Antin, C., Riéra, B., Freycon, V., Flores, O., Gonzalez, S., ter Steege, H., Grenand, P., Bánki, O., Birnbaum, P., … Engel, J. (2026). Thirty-two years of inventory data (1986–2018) from the GUYADIV forest plot network in French Guiana, covering 100 ha and 90,000 trees. Annals of Forest Science, 83(1), Article 49. https://doi.org/10.1186/s13595-026-01367-1

Image Credits: AI Generated

DOI: 10.1186/s13595-026-01367-1

Keywords: GUYADIV, French Guiana, tropical forest, tree diversity, forest inventory, biodiversity dataset, Amazonia, terra firme, botanical identification, species richness, open data, forest plots