Bologna, like many European cities, faces growing challenges from climate change, biodiversity loss, pollution, and unequal access to public space. Problems like Urban Heat Islands (UHI) and Urban Heat Waves (UHW) particularly threaten vulnerable communities, straining public health and reducing quality of life.
To address these issues, the TALEA project, supported by the European Urban Initiative – Innovative Actions (EUI-IA), introduces TALEA Green Cells (TGCs): modular, adaptable green units that reconnect fragmented green areas, regenerate underutilized spaces, and create local climate refuges.
This repository presents a transparent methodology for planning TGCs placement in Bologna, relying on interpretable AI optimization techniques. The workflow combines open-source urban and demographic data with optimization-based modeling to explore how the distribution of Green Cells shifts under factors such as population density, urban indices or land use.
- Green units under the maintenance of the Bologna Municipality, including a wide range of categories, like flowerbeds, gardens, parks, ponds, school and sports green areas, roadside decorative greenery, etc.
- Mapping of the private green areas within the Bologna cityscape.
- Mapping of the trees planted in the Bologna cityscape, containing the coordinates of each element and the information about the type of tree.
The placement of TGCs is modeled as a Combinatorial Optimization Problem (COP) and relies on the impact of different factors on micro and macro scale and on how fragmented each area is.
The macro scale is referred to the statistical areas, where three indices are considered:
- population density
- existing green
- UHEI
The micro scale is referred to the grid of 100x100 m cells, and is modeled differently according to four different approaches:
| Standard | Difference | Inverse UHEI | NDVI |
|---|---|---|---|
| It aims to maximize a utility function, which is directly proportional to new green areas and inversely to existing ones. | Based on the same concept of the Standard one, it considers the impact of the existing green in a differential way. | It maximizes the impact of the FVC on the Inverse UHEI, directly w.r.t. new TGCs and inversely to the existing vegetation. | It maximizes the impact of the FVC on the NDVI, according to a quadratic relation, inversely proportional to the actual NDVI of the cell. |
Note
Further details on models, indices and sources of this study can be found in the report and in the documentation.
To reproduce our results, ensure Docker is installed on your system. Once Docker is installed, to run the docker the following scripts should be executed from the terminal while in the Dockerfile directory:
-
To build and run the docker
docker-compose up --build -d -
To exec docker commands from your terminal
docker exec -it TALEA bash
There are two modalities to run the pipeline in the docker shell:
-
one for running the entire pipeline, including the grid creation step, necessary for the first run, after changing the size and for changing something in the geometry of data
run_pipeline
-
one for running only the further processing steps, including the instance creation and the solver, used for changing parameter configuration
run_model
Tip
Run the --help option on each command for further details on usage and parameters configuration.
This project relies on spatial data, which is best understood when visualized and exploring them in tables often leads to ambiguity and poor clarity. Therefore, a simple Graphical User Interface (GUI) has been implented to display GeoJson files.
Furthermore, any step of the process is explained in a report and a documentation that follows the ReadTheDocs format, including further details on data, models and the results obtained.
