Abstract. Green roofs are roofs incorporating a deliberate layer of growing substrate and vegetation. They can reduce both indoor and outdoor temperatures, so are often presented as a strategy to reduce urban overheating, which is expected to increase due to climate change and urban growth. In addition, they could help decrease the cooling energy demand of buildings thereby contributing to energy and emissions reductions, and provide benefits to biodiversity and human well-being. To guide the design of more sustainable and climate resilient buildings and neighbourhoods, there is a need to assess the existing status of green roof coverage and explore the potential for future implementation. Therefore, accurate information on the prevalence and characteristics of existing green roofs is required to estimate any effect of green roofs on temperatures (or other phenomena), but this information is currently lacking. Using a machine-learning algorithm based on U-Net to segment aerial imagery, we surveyed the area and coverage of green roofs in London, producing a geospatial dataset. We estimate that there was 0.19 km² of green roof in the Central Activities Zone (CAZ) of London, (0.81 km²) in Inner London, and (1.25 km²) in Greater London in the year 2019. This corresponds to 1.6 % of the total building footprint area in the CAZ, and 1.0 % in Inner London. There is a relatively higher concentration of green roofs in the City of London (the historic financial district), covering 3.1 % of the total building footprint area. The survey covers 1463 km² of Greater London, making this the largest open automatic survey of green roofs in any city. We improve on previous studies by including more negative examples in the training data, by experimenting with different data augmentation methods, and by requiring coincidence between vector building footprints and green roof patches. This dataset will enable future work examining the distribution and potential of green roofs in London and on urban climate modelling.

中文翻译：

---

使用航拍图像分割的伦敦绿色屋顶的开源自动调查

摘要。绿色屋顶是结合了一层有意生长的基质和植被的屋顶。它们可以降低室内和室外温度，因此通常作为减少城市过热的策略提出，预计由于气候变化和城市发展，过热会增加。此外，它们可以帮助减少建筑物的冷却能源需求，从而有助于减少能源和排放，并为生物多样性和人类福祉带来好处。为了指导设计更具可持续性和气候适应性的建筑和社区，有必要评估绿色屋顶覆盖的现有状况并探索未来实施的潜力。所以，要估计绿色屋顶对温度（或其他现象）的任何影响，需要有关现有绿色屋顶的普遍性和特征的准确信息，但目前缺乏此信息。我们使用基于 U-Net 的机器学习算法对航拍图像进行分割，调查了伦敦绿色屋顶的面积和覆​​盖范围，生成了地理空间数据集。我们估计有 0.192019 年，伦敦中央活动区 (CAZ)、伦敦市中心 (0.81 km ² ) 和大伦敦 (1.25 km ^{2 ) 的绿色屋顶面积为2}平方公里。这相当于总建筑占地面积的 1.6%在 CAZ，在伦敦市中心 1.0%。伦敦金融城（历史悠久的金融区）的绿色屋顶相对集中，占建筑总占地面积的 3.1%。调查范围1463 km ²大伦敦，使其成为任何城市中最大的开放式绿色屋顶自动调查。我们通过在训练数据中包含更多负面示例、尝试不同的数据增强方法以及要求矢量建筑足迹和绿色屋顶斑块之间的重合来改进以前的研究。该数据集将使未来的工作能够检查伦敦绿色屋顶的分布和潜力以及城市气候建模。