Supplementary Material for "Automated detection of an insect-infested keystone vegetation phenotype using airborne LiDAR"

Ecologists, foresters and conservation practitioners need "biodiversity scanners" to effectively inventory biodiversity, audit conservation progress and track changes in ecosystem function. Quantifying biological diversity using airborne LiDAR remains challenging, especially for small invertebrates. However, insect aggregations can drastically alter landscapes and vegetation, and these "extended phenotypes" could serve as environmental landmarks of insect presence in LiDAR data. To test the feasibility of this approach, we studied the symbiotic ants that alter canopy shapes of whistling thorn acacia, a keystone tree species of the black cotton soils of east African savannas. We demonstrate a protocol of LiDAR data collection, training data preparation (including a customizable tree-segmentation algorithm) and convolutional neural network-based classification for the detection of ant-infested, within-species acacia tree phenotypic variations. Surveying ant occupancy of 402 hectares of 9,680 acacia trees took 1,000 work hours, while surveyed patterns of ant distribution was replicated by trained classifier based on an hour-long airborne LiDAR collection. We suggest that large scale surveys of insect occupancy (or insect-vectored disease) can be automated through a combination of airborne LiDAR and machine learning.生态学家、森林管理者与保护实践者亟需“生物多样性扫描仪”，以高效开展生物多样性清查、保护进度评估及生态系统功能变化追踪。利用机载激光雷达（airborne LiDAR）量化生物多样性仍存在诸多挑战，针对小型无脊椎动物的相关工作尤为困难。不过，昆虫聚群会显著改变景观与植被结构，这类“延伸表型（extended phenotypes）”可作为激光雷达数据中昆虫存在的环境标识物。为验证该方法的可行性，我们针对会改变哨刺金合欢（whistling thorn acacia）冠层形态的共生蚂蚁展开研究——该树种是东非稀树草原黑棉土生境中的关键树种。我们提出一套完整研究流程，涵盖激光雷达数据采集、训练数据制备（含可自定义的树木分割算法）以及基于卷积神经网络（convolutional neural network）的分类模型构建，用于检测受蚁侵染的种内金合欢树表型变异。本次普查耗时1000个工时，完成了402公顷范围内9680棵金合欢树的蚂蚁侵染情况调查；而基于1小时机载激光雷达采集数据训练得到的分类器，可复现本次普查得到的蚂蚁分布格局。我们认为，结合机载激光雷达与机器学习技术，可实现昆虫侵染情况（或昆虫媒介疾病）的大规模自动化普查。