My participation to Hackaviz
Hackaviz is an annual open data visualization challenge hosted by Toulouse DataViz. I participated to this 2025 edition. It is organized by the awesome folks at Toulouse DataViz. Given the data from the competition, I created an interactive tool to explore water levels and rainfall in the Toulouse region. Here’s a breakdown of the steps I took to create the visualization.
🎯 My goals#
Before starting I wanted to define my goals which were:
- create interaction between at least two graphs (I wanted to learn that skill using Plotly and Streamlit and experimenting).
- having fun doing it.
- Submit something that looked finished and thought through.
🔍 The Exploration Phase#
Like most data science projects, it all started with some basic yet important steps to better understand the data:
- Running
df.describe()across variables helped flag odd min/max values and high variances. - Simple line plots by date exposed trends, gaps, and a few strange outliers.
- I used these first steps to detect missing data, normalize station values, and prepare for a more guided experience.
🚀 The Final App#
My app is named Toulouse Water level & Rainfall Explorer. The goal to help users understand how water levels and rainfall interact in Toulouse using real open data. Here are a few links to my creation:
The visualization is composed of two plots, a water level linechart and the associated rainfall map. With the first graph, you can select a range of points to filter the dates used for the rainfall map. Also, you can use filter from the sidebar to select a date range for the water level observations in Toulouse and adjust the top N rainfall stations to display on the map.
🧠 Lessons & Learnings#
Things I am happy with:
- avoided classic errors with outliers of the data.
- delivered a working and interactive app.
- my solution can handle many point (nearly 150 years of data can be shown on my first plot).
- done and learned how to do interaction between two Plotly plots in a Streamlit app (completed my personal goal).
Things I can improve:
- my capacity to create more appealing visualizations (using better visual elements and improve the design).
- as many data dates were missing for the rainfall, I could color the first graph points with blue for rainfall data available and stay in grey otherwise (instead of acceleration).
- create a multi-langue version.
A big thanks to the Toulouse DataViz association for organizing Hackaviz 2025 and providing such rich datasets. Hackaviz was a great opportunity to engage with real-world data and explore new visualization ideas.
Let me know what you think, I hope it was useful for you. See you later, Vincent