| dc.description.abstract | Green spaces are crucial in cities to mitigate climate change impacts from urbanization, such as the urban heat island effect, air pollution, and urban flooding. Nature-based solutions can help resolve these problems in dense urban areas with limited space. This research aims to find vertical and horizontal spaces suitable for installing green roofs and planting trees, using Amsterdam as a case study. For this purpose, a novel, scalable, and reproducible methodology has been developed that combines GIS and deep learning methods. To find suitable roofs for installing green roofs, aerial imagery was combined with DSM and slope data, and a dataset with buildings (BAG) suitable for green roof installation based on height, roof slope, and age was used as ground truth to train a segmentation model using the YOLOv8 model. A geospatial analysis was conducted using datasets of streets, public trees, and parking spaces to find suitable streets for planting trees. The method for finding suitable roofs proved extremely capable, as the Intersection over Union between the predicted and actual suitable roofs was 0.83, and 90% of the suitable roofs were found correctly. Existing green roofs were only identified as suitable for 66.8%, as these buildings often have a larger slope or have only partial green roofs. Additionally, the method for finding suitable streets is very fast in giving a first impression of where trees or planters can be placed, with 64.3% of streets used in the analysis being suitable. | |
