Leaf-it to AI

Cities worldwide must adopt innovative strategies to enhance resilience and sustainability in re­sponse to escalating urbanization and climate change. Urban heat islands (UHIs), a phenomenon of increased temperatures and reduced livability, present a significant challenge in this context. Trees are crucial in mitigating UHIs by providing shade and other ecosystem benefits. However,  manual censuses are both expensive and labor-intensive. This limits the availability of precise and frequent data at the city scale, especially for cities in the global south. Our work introduces an automated approach to the tree census process by generating a detailed tree inventory by integrating diverse remote sensing data to locate and measure tree morphol­ogy while considering the Leaf Area Index (LAI) we aim to improve urban forestry management. First, we combine segmentation techniques with multi­spectral analysis to delineate individual trees from high-resolution aerial imagery (RGB + NIR). Second, we then measure the height, the crown diameter, and the mean LAI for each tree from Aerial Lidar Scan (ALS). Our contribution offers a novel method for quantifying canopy density in complex urban settings, enabling accurate shade predictions. We present our tree dataset for Amsterdam and the machine learning method we trained to predict LAI from genus and morphology. This research proposes a tool for urban planners and environmental scientists contributing to the field of urban ecology and sustainable city planning.