potential-vertical-green-spaces

Aims and Objectives

Nature-Based Solutions are able to provide natural, multifaceted solutions to problems many cities around the world face. Since urban space is scarce, potential vertical spaces (facades and roofs) have to be considered for greening. This project aims to find potential vertical spaces in dense urban areas that are suitable to be greened, using deep learning and geospatial analysis.

The Project

The project contains two Jupyter Notebooks that can be run seperately.

• Facades.ipynb estimates the greening potential on building facades using image segmentation. The output is a geopackage of sample points containing a facade Greening Potential Score (GPS).

• Roofs.ipynb estimates the greening potential on rooftops using geospatial analysis. The output is a geopackage of all suitable buildings (filtered) and all buildings (unfiltered) containing the roof GPS. Note: as of this moment, the study area is limited to Amsterdam only.

Setup

To run the project in Google Colab, clone the complete project to your Google Drive folder (e.g. /My Drive/MyFolderName). Then, open one of the Jupyter Notebooks in Google Colab. Written instructions are also included within each notebook that explain each step of the process. It is recommended that for hardware acceleration a GPU is set and high RAM usage is turned on. To change this in Google Colab, go to Runtime > Change runtime type > Hardware accelerator to select a GPU (V100 GPU recommended) and turn on high RAM usage.

Facades.ipynb

Requirements

Facades.ipynb uses the following dependencies:

Pillow==10.0.1
segment-geospatial==0.10.2
groundingdino-py==0.4.0
leafmap==0.31.2
localtileserver==0.10.1
geopandas==0.13.2
ipython==7.34.0
mercantile==1.2.1
networkx==3.2.1
numpy==1.25.2
osmnx==1.9.1
pandas==1.5.3
Requests==2.31.0
scipy==1.11.4
Shapely==2.0.2
tqdm==4.66.2
vt2geojson==0.2.1
area==1.1.1

Instructions

1. Open Facades.ipynb and make sure to set your working directory to your installed Google Drive path. Also make sure that work_on_drive is set to True.

working_dir = '/My Drive/MyFolderName'
work_on_drive = True

2. Specify the desired input parameters.

city_name = 'Diamantbuurt, Amsterdam'

access_token = 'MLY|XXXXXXXX|xxxxxxxxxxxxxxxx'
distance = 50
num_sample_images = float('inf')
begin = None
end = None
save_roads_points = True
save_streetview = False

• city_name The (neighborhood of the) city to analyze. When a named area processing is chosen (step 3), the OpenStreetMap boundary of the (neighborhood of the) city is used. When bounding box processing is used, this is only used to name the output folders.
• access_token The Mapillary street view access token.
• distance Distance between sample points in meters.
• num_sample_images Maximum number of sample points per processing area (bounding box or named area). For a bounding box selection, the area may be split into smaller processing areas. For named areas, this is the total number of sample points.
• begin First index of a smaller analysis range of the sample points per processing area.
• end Last index of a smaller analysis range of the sample points per processing area.
• save_roads_points Choose whether to save the temporary road network and sample points data.
• save_streetview Choose whether to download the analyzed street view images.

3. Optional: specify an area selection using a bounding box. To use this option, simply draw a rectangular polygon on the provided Leafmap insert and continue as usual. When no rectangle is drawn, the place name is used as the selected study area.

4. These inputs are used to calculate the facade GPS for the facades in the study area. The runtime may vary based on the size of the study area.

5. When done, a city_name_features.gpkg geopackage is automatically created inside /results/city_name/ which contains the analyzed sample points and the facade GPS found at these points.

Roofs.ipynb

Requirements

Roofs.ipynb uses the following dependencies:

pypdl==1.3.0
rasterio==1.3.9
localtileserver==0.10.1
rasterstats==0.19.0

Instructions

1. Open Roofs.ipynb and make sure to set your working directory to your desired Google Drive path. Also make sure that work_on_drive is set to True.

    working_dir = '/My Drive/MyFolderName'
    work_on_drive = True

2. The notebook automatically downloads the required buiding dataset (BAG), Digital Surface Model (DSM) tiles, Digital Terrain Model (DTM) tiles, and municipal boundaries if they are missing. These may also be added manually.

3. These inputs are used to calculate the height, slope, area, and age of each building in the study area. Unsuitable buildings are filtered out. For the remaining buildings, a roof GPS is calculated.

4. When done, filtered_BAG_Amsterdam.gpkg and unfiltered_BAG_Amsterdam.gpkg are automatically created inside /roofs/ which contains the roof GPS for the potential suitable buildings (filtered) and all buildings (unfiltered).

Acknowledgements

This project was made by F.E. Peranović as part of the thesis for the master Geographical Information Management and Applications (GIMA) at Utrecht University, under supervision of Dr. S.M. Labib.