Search

The study of wood chemistry to explain the response of trees to climate variation is efficient for retrospective analysis of changes induced in wood during the process of adaptation to seasonal variations. Hyperspectral imaging (HSI) in near infrared could be useful for prediction of several wood chemical components. The objective is understand how water constraints influence wood formation. Wood disks of E. grandis clone from six treatments (100% and 65% of rainfall crossed with potassium and sodium fertilizations) were sampled. HSI were obtained on transversal section of the disks using a line scan hyperspectral camera from 1000-2500 nm with a pixel size of 625 x 625 µm. For calibration, were selected 60 wood solid samples from an Eucalyptus collection, which were measured the total extractive values. HSI or these samples were acquired from transversal section with the same camera. We regressed the mean spectra for each sample with extractives values by Partial Least Square Regression. Due to the model, the total extractives for each pixel to produce an image for the wooden disks were predicted. Based on them, the total extractive distribution according to growth conditions were compared. revealing that trees under stress conditions show a higher heterogeneous chemical profile from the pith to bark. Eucalyptus demonstrates sensitive precipitation variations. with rapid growth and considerable increase in DBH during rainy seasons and the reverse in drought. Consequently, is possible to discriminate different patterns of chemical compound distribution according to the growth conditions.

The aim is to study tree development submitted to different water and mineral constraints. The objective is to correlate spatial distribution of wood chemistry with wood density secondary growth. We sampled wood discs from 54 trees in rainfall exclusion design with Eucalyptus grandis at Esalq-USP. ltatinga Station.brazil. We felled the 5 years old trees growing under different water availability. For each disk. we acquired an image with a near infrared (nir) hyperspectral camera (hs1, specim. pixel size 625 x 625µml. The challenge here was to transfer our previous nir calibrations for total extractives built with a benchtop spectrometer (vector. bruker) to a hsi camera by using standard sample set measured ,>n the two devices. An efficient model were built for the his camera with a prediction error of 10.3% compared to the prediction error of 11.60% for our previous calibration with benchtop spectrometer. Then, based on spectra from his image. we predicted extractive contents and built images. These allow us to compare their spatial distributions according to the growth conditions. In the absence or absence or fertilization, trees with higher water stress showed a higher heterogeneous distribution. from pith to bark, for the total extractive contents and a higher average mean value. In perspective. these data will allow us to study and reline the knowledge on cambium activity according to climatic variations by crossing variability of the chemical properties, x-ray micro-density and anatomy of wood, and diameter growth rate measured by electronic dendrometers.

Most near-infrared spectroscopy (NIRS) studies for wood density use multiple spectral acquisitions taken manually at several points from pith to bark. Recently the use of near-infrared hyperspectral imaging (NIR- HSI) has shown good performance in predictions of wood properties allowing build high spatial resolution predictive images. The combined use of NIR-HSI and X-ray microdensity imaging (X-ray MDI) could be more practical for evaluation of spatial variation of wood density. The aim was to develop a wood density calibration model for NIR-HSI (625 µm pixel size) from X-ray MDI (30 µm pixel size) and evaluate the spatial variation of wood density along stem cross-section in Eucalyptus grandis trees. Wood discs were collected from 18 trees of 6 years old, submitted to two different water availability, located in Sao Paulo. Brazil. The challenge here was to match the pixels of X-ray MDI to pixels of NIR-HSI (in 2 mm width radial region) to transfer accurate values of wood density to each, pixel of NIR-HSI. The R2 of the model was 0. 72, while the root mean squared error was and 4.8 x 10-2 g cm-3 for the validation group. The prediction model allows comparing their spatial distributions according to the growth conditions. Trees with higher water stress showed higher wood density. Treatments showed a similar spatial variation or wood density. increasing from pith to bark. In perspective, these data will allow evaluating the spatial distribution of wood density inter and intra-tree rings in relation to meteorological variability.