Recognizing Amazonian tree species in the field using bark tissues spectra
Graphical abstract
Introduction
The identification of tree species in forest inventories is a key activity in maintaining high quality of forest management. In the field, trees are frequently identified to species by para-botanists (‘mateiros’) and voucher specimens are only rarely collected for comparison with herbaria or consultation with plant specialists. In the Amazon region, high tree species diversity (Steege et al., 2016, Cardoso et al., 2017) and poor sampling make it difficult to recognize species with confidence (Nelson et al., 1990, Hopkins, 2007, Gomes et al., 2013). Problems relate to both finding scientific names for species, and also to the grouping of trees into the same species at local scales (Gomes et al., 2013). While the first problem is complex, as the Amazon has many undescribed species (62% of the estimates of ter Steege et al., 2013) and taxonomic uncertainty abound, the second problem, crucial to the sound management of plant populations, can be more easily tackled. Recent studies suggest that near-infrared spectroscopy may increase the quality of grouping trees into local species and, given a robust model, to also obtain a scientific name for them (Durgante et al., 2013, Lang et al., 2015). Hence, near-infrared spectroscopy is a promising technique for increasing the accuracy of local species delimitation and also for their taxonomic assignment, a problem that impairs our understanding of basic diversity patterns for the Amazon (Cardoso et al., 2017).
It is common practice in forest inventories, particularly for forest management, to identify trees only in the field without the collection of herbarium specimens for plant identity confirmation. This activity is most frequently conducted by para-botanists, lacking formal training, who use morphological characteristics, such as trunk shape, bark type and texture, wood color, the presence or absence of exudates, odor and other vegetative traits to identify trees into local species in the field. This is conducted in a cognitive way and the actual characters used are hard to untangle (Procópio and Secco, 2008, Gomes et al., 2013). These field observations allow varying identification of trees to family, genus and species, with uncertain scientific names frequently extracted from vernaculars used during fieldwork. Vernacular names also vary among regions and para-botanists, causing the same species to have different vernaculars, and the same vernaculars are used for different species (Martins-da-Silva et al., 2003, Procópio and Secco, 2008). Hence, the extraction of scientific names from vernaculars is prone to large errors, with negative consequences to the management of plant populations, and this is likely the most frequent method by which inventories for timber exploitation are conducted. The clustering of different species with the same vernacular causes overexploitation of rare species, overestimation of the size of populations of commercial species, and jeopardizes the integrity of commercial transactions of forest products (Procópio and Secco, 2008, Gaui, 2013).
A more reliable process to recognize species is conducted when herbarium samples are collected for morphological comparison with specimens identified and deposited in herbaria, or with the aid of the botanical literature. Rarely all individuals are collected, so the problem of grouping trees into local species may persists if this process is conducted in the field without caution to difficulties caused by poor taxonomy, cryptic variation and ontogeny in local species assignment (Gomes et al., 2013). Moreover, in many plant groups it is frequently necessary to have reproductive material to identify taxonomically a local species from an inventory, and most samples collected in inventories are sterile due to irregular phenological patterns (Newstrom and Frankie, 1994), and also due the low importance that the problem of local species and taxonomic assignments receives in forest inventories in general (Gomes et al., 2013; Procópio and Secco, 2008; Martins-da-Silva et al., 2003).
The problem of species identification in the Amazon is also increased by the taxonomic quality of the reference collections. Herbaria with plants from the Amazon are the most important reference to obtain a taxonomic name for a local species from a forest inventory. However, different plant taxonomists and specialists may identify plants in a single herbarium collection. Hence, different people may have different levels of training and see different things (Gomes et al., 2013), and even specialists in a single plant group may differ in their taxonomic concepts. This is exemplified by the fact that duplicates of the same samples in different herbaria have distinct taxonomy, amounting to more than 50% of incongruence for the name of some tropical species (Hopkins, 2007, Goodwin et al., 2015).
Therefore, it is necessary to improve the processes by which trees are assigned into local species in forest inventories, and also the quality of reference collections with better sampling and taxonomic definitions. Some new technologies promise to greatly change the way plant species are defined and may be recognized. Molecular data is causing a revolution in taxonomic practice (Hollingsworth et al., 2009), and is likely to be in the near future the main reference for plant identification despite current limitations of plant molecular barcoding (Kress et al., 2009, Gonzalez et al., 2009). However, the use of molecular data for assigning trees into species at local scales, a non-trivial problem in Amazonian forests (Gomes et al., 2013), is difficult to operationalize. For this task, when species taxonomic definition is not at stake, technologies using chemometrics data, like visible and infrared spectra (NIR and SWIR) of plant tissues, are showing excellent results in the discrimination of plant species in tropical forests (Pastore et al., 2011, Durgante et al., 2013). Visible and infrared spectroscopy is a fast and non-destructive tool (Pasquini, 2003) that together with multivariate techniques identifies, quantifies and characterizes organic samples (Tsuchikawa, 2007). This tool has been used to discriminate plant species using absorbance extracted from dried leaves and wood. In the Amazon, recent studies show high predictive power of NIR spectra from dried leaves to discriminate closely related species in Lecythidaceae (Durgante et al., 2013), 111 species of 34 families from a large-scale plot (Curty, 2014), species of Burseraceae at different stages of development (Lang et al., 2015), and species with samples from different Amazon regions (Botelho, 2017). Absorbance data from wood has been used to discriminate endangered species from other species with similar wood characteristics, and to assist in the monitoring of illegal logging (Pigozzo, 2011, Braga, et al., 2011, Pastore et al., 2011, Bergo et al., 2016). NIRs technology is on the list of new forensic methodologies for detecting illegal logging (Dormontt et al., 2015).
The majority of studies on leaf and wood spectroscopy were performed under laboratory conditions. The development of field methods with portable equipment is necessary to improve the recognition of tree species in forest inventories with high reliability, a problem that can be tackled otherwise only by extensive botanical sampling (Gomes et al., 2013). Portable devices have been used in remote sensing to study forest canopy vegetation through leaf reflectance (Asner, 1998), as well as to discriminate species in tropical forests by spectra from fresh leaves (Clark and Roberts, 2012, Asner et al., 2014). Local species assignment during forest inventories would greatly benefit from the possibility of using information from the bark of trees, lowering costs and improving accuracy. The trunk is easier to access in forest inventories than leaves and the possibility of using spectral data from the bark would diminish the need for herbarium samples once a spectral model is build, greatly facilitating studies and inventories that conduct frequent censuses of trees in permanent plots. Here, we ask whether spectra obtained from the external or internal bark of trees, collected with a portable spectrometer in the field, would permit to discrimination of 11 selected species in an Amazonian forest.
Section snippets
Study area
The study was carried out at the ZF-2 Experimental Station of the National Institute for Amazonian Research (INPA), located 90 km north of Manaus - AM, Brazil (2°38′38″S and 60°09′49.9″W). The station has a total area of 21,000 ha, covered by tropical rainforest (Higuchi and Santos, 1997), characterized by high floristic diversity (Higuchi et al., 1998, Carneiro, 2004, Gaui, 2013). The trees were selected in the 12 permanent plots of 1 ha of BIONTE Project that has been monitored since 1989 (
Results
Rhytidome and phloem have different spectral signals and are clearly separated in a reduced space of two dimensions for the majority of the tested species. These ordinations also indicated that spectra from the rhytidome present greater variation than those collected from the phloem (Fig. 1). Therefore, we considered two separate datasets, rhytidome and phloem, to test the utility of each data type in species discrimination.
Both datasets indicated high potential for discriminating species in
Discussion
Here we asked whether VIS-NIR-SWIR spectral data collected from the bark of trees in the field would permit the discrimination of species. Both inner (phloem) and outer bark (rhytidome) spectral data permitted the discrimination of the 11 selected tree species with high accuracy (>=94% of correct tree assignment when using the average of three spectra per tree per tissue). Although our taxonomic sampling was small and further tests are still required, our results indicate a great potential of
Acknowledgments
This contribution is part of the Master’s thesis of Hadlich H.L. at the Instituto Nacional de Pesquisas da Amazônia (INPA), which was supported by a fellowship from the CNPq. We appreciate the financial support by FAPEAM, CNPq, INCT – Madeiras da Amazônia and NGEE-Tropics. We thank the forest management laboratory at INPA for all logistical support, especially Dr. Adriano Lima, Tatiana Gaui and all those who helped the field collections.
References (50)
Biophysical and biochemical sources of variability in canopy reflectance
Remote Sens. Environ.
(1998)- et al.
Forensic timber identification: It's time to integrate disciplines to combat illegal logging
Biol. Conserv.
(2015) - et al.
Species spectral signature: discriminating closely related plant species in the Amazon with near-infrared leaf-spectroscopy
For. Ecol. Manage.
(2013) - et al.
Use of near-infrared spectroscopy as an indicator of emerald ash borer infestation in white ash stem tissue
For. Ecol. Manage.
(2016) - et al.
Discrimination of taxonomic identity at species, genus and family levels using Fourier Transformed Near-Infrared Spectroscopy (FT-NIR)
For. Ecol. Manage.
(2017) - et al.
Variables selection methods in near-infrared spectroscopy
Anal. Chim. Acta
(2010) - et al.
Study on difference between epidermis, phloem and xylem of Radix Ginseng with near-infrared and infrared spectroscopy coupled with principal component analysis
Vib. Spectrosc.
(2011) FieldSpec 3 user manual
(2010)Functional and biological diversity of foliar spectra in tree canopies throughout the Andes to Amazon region
New Phytol.
(2014)- et al.
NIRS identification of Swietenia macrophylla is robust across specimens from 27 countries
IAWAI J.
(2016)
Amazon plant diversity revealed by a taxonomically verified species list
Proc. Natl. Acad. Sci.
Species-level differences in hyperspectral metrics among tropical rainforest trees as determined by a tree-based classifier
Remote Sens.
Anatomia das plantas com sementes
Variability in southern yellow pine bark from industrial sources
Botânica: morfologia interna das plantas: anatomia
Lipophilic Extractives of the Inner and Outer Barks of Eucalyptus globulus
Holzforschung
Widespread mistaken identity in tropical plant collections
Curr. Biol.
Local plant species delimitation in a highly diverse Amazonian forest: do we all see the same species?
J. Veg. Sci.
Identification of Amazonian trees with DNA barcodes
PloS One
Wood and Bark Composition of Picea orientalis (L.) Link
Holzforschung
Cited by (23)
Practical recommendations and limitations for pushbroom hyperspectral imaging of tree stems
2023, Remote Sensing of EnvironmentChemical and nutritional characterization of Ambelania duckei (Apocynaceae) an unexplored fruit from the Amazon region
2023, Food Research InternationalCitation Excerpt :In addition, the same extract showed anti-trypanosomal activity in vitro against the protozoan Trypanosoma cruzi (Arias et al., 2021). Despite the above knowledge related to A. duckei little is known about the nutritional properties of the fruits, since the literature found for the species A. duckei is mainly focused on its botanical aspects (Hadlich et al., 2018), which makes this study of the edible fruits of this species significant. However, reports describe the species of the genus Ambelania as an edible fruit in the Peruvian Amazon (Rutter, 1990).
Prediction of dry matter, carbon and ash contents and identification of Calycophyllum spruceanum (Benth) organs by Near-Infrared spectrophotometry
2022, Microchemical JournalCitation Excerpt :In several articles published in the specialized scientific literature, the efficiency of NIR technique in analysis of forest species has been verified. The technique has been a viable alternative in the evaluations of wood quality [11,12], species identification [7], discrimination of genotypes of the same species [8], prediction of chemical and physical attributes [13,12], among others. C. spruceanum wood is used in various ways in several areas, for example in tool handles, picture frames, window frames, among others, and its bark is used in teas in folk medicine to assist in the treatment of wounds and mycoses.
Quantifying Tropical Plant Diversity Requires an Integrated Technological Approach
2020, Trends in Ecology and EvolutionCitation Excerpt :In a wide-ranging study, 1449 canopy species in the Andes–Amazon region were classified to species level with an accuracy of >85% [35]. Other research has demonstrated the utility of using bark and branch tissue in addition to leaf tissue for spectroscopic identifications [37,39]. One recent study across a number of Amazonian taxa found that species level identifications made with branch samples had an accuracy >90%, which could be increased to 94% if leaf tissue was included [37].