Stacked Machine Learning for Timber Identification Using Laser-Induced Breakdown Spectroscopy (LIBS)

Area of Science:

• Analytical Chemistry
• Machine Learning
• Forestry Science

Background:

• Accurate timber identification is crucial for enforcing trade regulations and preventing illegal logging.
• Traditional methods for wood identification can be time-consuming and require specialized expertise.
• Laser-induced breakdown spectroscopy (LIBS) offers a rapid, non-invasive analytical technique.

Purpose of the Study:

• To develop and validate a novel wood species identification method using LIBS and stacked machine learning.
• To assess the performance of the proposed method against traditional classification techniques.
• To identify key elemental markers for differentiating tropical timber species.

Main Methods:

• Analysis of 700 wood samples from 18 tropical timber species using a handheld LIBS analyzer.
• Development of a stacked machine learning model integrating three Support Vector Machine (SVM) classifiers with a Partial Least Squares Discriminant Analysis (PLS-DA) meta-learner.
• Application of Principal Component Analysis (PCA) for dimensionality reduction of spectral data.

Main Results:

• The stacked LIBS-ML model achieved a high classification accuracy, indicated by a Cohen's kappa value of 0.8671 on the validation set.
• The developed stacking methodology significantly outperformed traditional flat classifiers.
• Elemental analysis identified calcium, magnesium, and barium as key indicators for species differentiation, correlating with environmental factors.

Conclusions:

• The combination of LIBS and stacked machine learning provides a powerful tool for rapid and accurate wood species identification.
• This technique has significant potential to support sustainable forest management and combat illegal timber trade.
• Elemental composition derived from LIBS can offer insights into the geographical origin and environmental conditions of timber species.