Back Arrow

Ilots de chaleur urbains

algos site ui
schéma technique si sote django
lcz heatmap
carte lcz
radar  chart
grenoble map


Two applications have been developed, one using Django and the other using Flask:

The Flask application is designed to narrate data-driven stories with the goal of increasing awareness among citizens of Grenoble regarding the urban heat island phenomenon. It offers interactive and personalized graphs that are tailored to the user's location. The application utilizes data provided by the Open Data team of the metropolis, aiming to encourage various local authorities within the metropolis to make their datasets available as open data.

The Django application was developed and made accessible to the municipalities within the Grenoble metropolis with the purpose of assessing the specific urban heat islands in each municipality. This estimation serves to alert scientists from the metropolis or the University of Grenoble Alpes (UGA) in the event of unforeseen results predicted by the two implemented algorithms. The first algorithm utilizes numerical data as input, while the second algorithm relies on satellite images. These algorithms determine the classification of each municipality from the 5 classes that were collaboratively established with the metropolis's scientists. Each class represents a temperature difference range between a reference city (used for comparison) and Grenoble, considering various criteria selected by the metropolis's scientists.

Flask Application:

You can find the beta version of the application at the following link:

Django Application:

The second application created with Django was intended for the local authorities of the Metropolis, where we needed a part to manage user access, their registration, as well as an administration page to manage their rights. That's why I chose to use Django, as it greatly facilitates these functionalities. In addition to that, I implemented two types of algorithms once the user accesses the site.

The first algorithm used is an Extra Tree Classifier, which is used for numerical data.


The second algorithm takes a satellite image of a neighborhood as input to predict its urban heat island value. I use MobileNet for feature extraction from the image and an XGBClassifier algorithm for prediction. The latter gives me a macro f1 score of 100%.

I invite you to check the following link for more details on the steps taken(in French).