Researchers from Beijing Forestry University and the Chinese Academy of Sciences have developed a new deep-learning architecture designed to predict hourly light intensity within tree canopies, a development that could transition forestry from traditional schedules to precision management.
The model, known as CLIP-TLNet, uses a hybrid approach involving a Transformer encoder to capture canopy structure and a bidirectional Long Short-Term Memory decoder to manage temporal dynamics. According to the study published in the journal Plant Phenomics, the system achieved a mean absolute percentage error of 6.8%.
To develop the model, researchers gathered high-resolution LiDAR data via unmanned aerial vehicles over triploid poplar plantations in Shandong Province, China. This data was synchronized with hourly light-intensity measurements recorded over seven consecutive days.
The team also introduced a regional canopy complexity index to account for structural variations. This metric, which measured values between 1.94 and 2.23 in poplar canopies, serves as a key indicator of light transmission. Canopies with an index below 1.9 are considered sparse, allowing for more than 40% light transmittance, while those with an index above 2.1 are dense, with transmittance falling below 15%.
In comparative testing, CLIP-TLNet reduced root mean square error by 33.6% compared to the second-best existing CNN-LSTM model. Additionally, researchers found that the complexity index is vital to accuracy, as its removal resulted in a 20.6% increase in prediction error.
The technology supports several practical pruning strategies, such as thinning overly complex regions to create light channels and allocating resources based on which canopy layers contribute most to light interception. The model specifically identified that lower-canopy light intensity frequently drops below 25,000 lux between 11:00 and 14:00, suggesting that strategic pruning could boost these levels by an estimated 15% to 30%.
While the framework offers a more scientific approach to forest management, the researchers noted that improper pruning could lead to reduced carbon sequestration and increased vulnerability to pests.
