A probabilistic approach to estimating timber harvest location.

Area of Science:

• Ecological science
• Forestry
• Analytical chemistry

Background:

• Accurate timber harvest location is vital for enforcing regulations against illegal logging.
• Stable Isotope Ratio Analysis (SIRA) uses natural isotope ratios in wood to verify timber origin.
• Current SIRA models for predicting timber provenance are often too simple and limited in scope.

Purpose of the Study:

• To develop a novel analytical pipeline for SIRA data to predict timber harvest locations within continuous, large geographic areas.
• To move beyond binary location predictions towards probabilistic outcomes for timber provenance.
• To incorporate an active learning tool for optimizing future reference data collection.

Main Methods:

• Utilized Gaussian processes to create robust isoscapes from reference wood samples.
• Integrated species distribution data with isotope data to calculate location probabilities.
• Developed an active learning component to guide efficient data acquisition for model improvement.

Main Results:

• The new pipeline predicts timber harvest locations with an accuracy of up to 520 km.
• The approach outperforms existing state-of-the-art methods for timber provenance determination.
• Incorporating species distribution data enhanced prediction accuracy by up to 36%.

Conclusions:

• The developed SIRA pipeline offers a significant advancement in predicting timber harvest locations with high accuracy and probabilistic outputs.
• This method has the potential to revolutionize global efforts in combating deforestation and protecting forest resources.
• The active learning tool efficiently optimizes future data collection, maximizing model performance gains.