Support Vector Machines Show Promise in Predicting Tree Yield

A recent study published in the Journal of Engineering has found that Support Vector Machines (SVMs) can accurately predict the yield of Pinus taeda (L.) plantations after four years. The research, conducted by a team at Virginia Polytechnic Institute and State University (Virginia Tech), integrated UAV LiDAR-derived individual tree crown metrics and distancedependent competition indices as input to machine learning models, including SVM and Random Forest (RF). The study's findings demonstrate the potential of LiDAR technology in accurate tree- and stand-level yield predictions.

Key Takeaways:

• The study investigated the integration of UAV LiDAR-derived individual tree crown metrics and distancedependent competition indices as input to machine learning models.
• Support Vector Machines (SVMs) achieved the highest individual tree-level yield accuracy (normalized RMSE (nRMSE): 9.59 %, R2: 0.59) among the models tested.
• When aggregated at the stand level, SVM underpredicted total stem volume by -1.50 %, while RF overpredicted by 1.53 %.
• The SVMreduced model underpredicted stand volume by 0.90 %, while the RFreduced overpredicted by 0.71 % at the stand level.
• The study concluded that LiDAR technology has the potential for accurate tree- and stand-level yield predictions, with broader uses in yield, carbon, and biomass modeling.

Statistics:

• Normalized RMSE (nRMSE) for SVM: 9.59 %
• R-squared (R2) for SVM: 0.59
• Normalized RMSE (nRMSE) for RF: 10.86 %
• R-squared (R2) for RF: 0.48
• Difference in total stem volume predicted by SVM and RF: -1.50 % and 1.53 %, respectively

Sources:

• Predicting the Yield Of Pinus Taeda (L.) Using Uav Lidar Data In Random Forest and Support Vector Machine Models. Forest Ecology and Management, 2025;594.