Advanced search
Start date
(Reference retrieved automatically from Web of Science through information on FAPESP grant and its corresponding number as mentioned in the publication by the authors.)

Near-infrared spectrometry allows fast and extensive predictions of functional traits from dry leaves and branches

Full text
Costa, Flavia R. C. [1] ; Lang, Carla [1] ; Almeida, Danilo R. A. [2] ; Castilho, Carolina V. [3] ; Poorter, Lourens [4]
Total Authors: 5
[1] INPA, Coordenacao Pesquisa Biodiversidade, Manaus, Amazonas - Brazil
[2] Univ Sao Paulo, ESALQ, Av Padua Dias 11, BR-13418900 Piracicaba, SP - Brazil
[3] EMBRAPA, Boa Vista, Roraima - Brazil
[4] Wageningen Univ, Forest Ecol & Forest Management Grp, POB 47, NL-6700 AA Wageningen - Netherlands
Total Affiliations: 4
Document type: Journal article
Source: Ecological Applications; v. 28, n. 5, p. 1157-1167, JUL 2018.
Web of Science Citations: 2

The linking of individual functional traits to ecosystem processes is the basis for making generalizations in ecology, but the measurement of individual values is laborious and time consuming, preventing large-scale trait mapping. Also, in hyper-diverse systems, errors occur because identification is difficult, and species level values ignore intra-specific variation. To allow extensive trait mapping at the individual level, we evaluated the potential of Fourrier-Transformed Near Infra-Red Spectrometry (FT-NIR) to adequately describe 14 traits that are key for plant carbon, water, and nutrient balance. FT-NIR absorption spectra (1,000-2,500 nm) were obtained from dry leaves and branches of 1,324 trees of 432 species from a hyper-diverse Amazonian forest. FT-NIR spectra were related to measured traits for the same plants using partial least squares regressions. A further 80 plants were collected from a different site to evaluate model applicability across sites. Relative prediction error (RMSErel) was calculated as the percentage of the trait value range represented by the final model RMSE. The key traits used in most functional trait studies; specific leaf area, leaf dry matter content, wood density and wood dry matter content can be well predicted by the model (R-2=0.69-0.78, RMSErel=9-11%), while leaf density, xylem proportion, bark density and bark dry matter content can be moderately well predicted (R-2=0.53-0.61, RMSErel=14-17%). Community-weighted means of all traits were well estimated with NIR, as did the shape of the frequency distribution of the community values for the above key traits. The model developed at the core site provided good estimations of the key traits of a different site. An evaluation of the sampling effort indicated that 400 or less individuals may be sufficient for establishing a good local model. We conclude that FT-NIR is an easy, fast and cheap method for the large-scale estimation of individual plant traits that was previously impossible. The ability to use dry intact leaves and branches unlocks the potential for using herbarium material to estimate functional traits; thus advancing our knowledge of community and ecosystem functioning from local to global scales. (AU)

FAPESP's process: 16/05219-9 - Monitoring forest landscape restoration through Light Detection and Ranging (LiDAR).
Grantee:Danilo Roberti Alves de Almeida
Support type: Scholarships in Brazil - Doctorate