An Open-Source Automatic Survey of Green Roofs in London using Segmentation of Aerial Imagery: Dataset

This archive contains code and data to go with the paper *An Open-Source Automatic Survey of Green Roofs in London using Segmentation of Aerial Imagery*.

This archive contains geospatial data, as well as the code used to generate the geospatial data.

The geospatial data consists of georeferenced polygons identifying areas which are covered by green roofs in London (GBR) generated from 2019 aerial imagery.

The data is described in detail in the manuscript *An Open-Source Automatic Survey of Green Roofs in London using Segmentation of Aerial Imagery*. See abstract below.

GeoJSON format:

GeoJSON is a format for encoding geospatial data, see https://geojson.org/.

GeoJSON can be read using GIS programs including ArcGIS, QGIS, OGR.

Contents:

`geospatial_data/buffered_polygons_2021.zip` a zip archive containing a geojson file. It is the estimated locations of green roofs in London in 2021 and is the main result, which can be opened in any GIS program after being unzipped.

`geospatial_data/buffered_polygons_2019.zip` a zip archive containing a geojson file. It is the estimated locations of green roofs in London in 2019 and is a secondary result, which can be opened in any GIS program after being unzipped. The predictions were made with the same model as the 2021 results.

`geospatial_data/labelled_area.zip` a zip archive containing a geojson file. Identifies the area which was hand-labelled.

`geospatial_data/manual_2021.zip` a zip archive containing a geojson file. Manually labelled green roof from 2021 imagery.

`geospatial_data/manual_2019.zip` a zip archive containing a geojson file. Manually labelled green roof from 2019 imagery.

`segmentation_code` contains the code used to produce the segmentation from the aerial imagery.

`analysis_code` contains the code used to produce the plots and tables for the paper.

Imagery availability:

Unfortunately the aerial imagery and building footprint data cannot be shared directly, as you will require the proper license. Both can be found at [Digimap](https://digimap.edina.ac.uk) provided your institution has the license.

Abstract:

Green roofs can mitigate heat, increase biodiversity, and attenuate storm water, giving some of the benefits of natural vegetation in an urban context where ground space is scarce. 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 needed, but this information is currently lacking. Segmentation algorithms have been used widely to identify buildings and land cover in aerial imagery. 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 \cite[]{simpson_charles_2022_6861929}. We estimate that there was 0.23 km^2 of green roof in the Central Activities Zone (CAZ) of London, (1.07 km^2) in Inner London, and (1.89 km^2) in Greater London in the year 2021. This corresponds to 2.0% of the total building footprint area in the CAZ, and 1.3% in Inner London. There is a relatively higher concentration of green roofs in the City of London, covering 3.9% of the total building footprint area. Test set accuracy was 0.99, with an f-score of 0.58. When tested against imagery and labels from a different year (2019), the model performed just as well as a model trained on the imagery and labels from that year, showing that the model generalised well between different imagery. We improve on previous studies by including more negative examples in the training data, and by requiring coincidence between vector building footprints and green roof patches. We experimented with different data augmentation methods, and found a small improvement in performance when applying random elastic deformations, colour shifts, gamma adjustments, and rotations to the imagery. The survey covers 1558 km^2 of Greater London, making this the largest open automatic survey of green roofs in any city. The geospatial dataset is at the single-building level, providing a higher level of detail over the larger area compared to what was already available. This dataset will enable future work exploring the potential of green roofs in London and on urban climate modelling.