A sophisticated machine learning technique is revolutionizing urban ecological research by providing precise measurements of tree growth across seasons. Researchers led by Professor Bing Xu have developed the Seasonal Tree Height Neural Network (STHNN), a cutting-edge model that achieves remarkable accuracy in estimating tree heights with an R² of 0.80 and a mean absolute error of just 1.58 meters.
The innovative approach integrates LiDAR and satellite data with advanced machine learning algorithms, addressing critical challenges in urban forest monitoring. By analyzing a comprehensive dataset from 2018 to 2023, the team demonstrated the model's ability to track tree height variations across different seasons and geographic regions.
A key innovation of the study was using SHAP (SHapley Additive exPlanations) for feature optimization, which streamlined the model by eliminating 23 non-essential variables from an initial set of 52. This refinement not only enhanced predictive accuracy but also reduced computational demands, making the approach more efficient and scalable.
The research revealed that Shenzhen's urban trees typically range between 6 and 14 meters in height, with notable seasonal variations. Crucially, winter canopy heights consistently measure lower than summer canopies, highlighting the importance of seasonal dynamics in urban forest management.
Beyond its immediate scientific significance, the STHNN model offers transformative potential for urban planning and ecological conservation. By providing a precise, data-driven method for tracking tree growth, the technology could help cities worldwide develop more effective strategies for green space allocation, biodiversity preservation, and climate change mitigation.
