For a few years now, everything and its opposite has been said about the inexorable rise of Artificial Intelligence. In fact, it is already changing the game in the Digital domain of course but also in many other sectors: Economy, Defense, Industry, Commerce, Media/Advertising… So, for better or for worse?
New digital Eldorado and umpteenth industrial revolution for some, liberticidal instrument of manipulation of the masses or even real horseman of the Apocalypse for others… what is it really?
The future will tell, but in the meantime, far from ideological debates, Artificial Intelligence can certainly be considered as a tool with an unprecedented potential put at the service of our technological societies. What Man will do with this tool will certainly depend on the geopolitical context and the choices that will be made in the next few years.
For its part, vorteX-io, is part of those who bet that AI could be put at the service of the Environment, Progress in general and Science in particular, and this, in the interest of everyone.
The use of Artificial Intelligence in the field of hydrology was both a desire and a strategic orientation for vorteX-io from the start. The challenge is to use the power and flexibility of this type of tool to concretely improve the prediction and anticipation of continental floods based on the data from concrete measurements made in situ by the micro-stations of the vorteX-io network.
When we talk about AI, we have to keep in mind a fundamental point: AI consumes data, it is even fond of it. But to achieve the formidable efficiency for which it is known, it needs to be fed with high quality data in large quantities.
The vorteX-io network is constantly being enriched with new micro-stations deployed in the field and having an unparalleled measurement frequency, making this constellation of sensors a true generator of hydrological data (water level, surface speed, images, etc.)
The “AI for Earth 2” hackathon
From this strong will to exploit the richness of these data, and thanks to the expertise and the precious and benevolent help of Simon Chambers (manager ofAxsysnav) was born the idea of organizing a hackathon*.
The concept: challenge experts in the field of Artificial Intelligence to develop hydrological prediction models.
In this perspective, Axsysnav has identified the online platform FruitPunch AI to co-organize the event, in the framework of the project SUN Space Hubs Network, an EU-funded support program for the most promising European companies in the space technology sector.
This hackathon was therefore submitted to the “AI for Good” community, composed of all the AI actors interested (researchers, engineers, students, etc.) to volunteer their talent for useful projects for the good of all and for the Planet.
In partnership with vorteX-io, FruitPunch AI has successfully organized a challenge/hackathon named “AI for Earth – Inland Floods Prediction Challenge’ which took place between June and the end of August 2022.
The goal for the participating Data Scientists was to accurately predict the evolution of the water level of some French rivers from the in-situ measurements made by several vorteX-io micro-stations and the French Vigicrues public hydrological network, with the final objective of anticipating future floods and allowing the relevance and efficiency of the vorteX-io network.
Two teams worked for two months on this exciting challenge and, each with its own methodology, they developed awesome complementary flood prediction algorithms. The initial dataset provided was composed of water level measurements from the vorteX-io micro-stations installed on the Garonne river, completed by measurements from the corresponding Vigicrues stations available on the same sector.
The first team based its approach on a regression model while the second team focused its work on neural networks. The challenge has been very fruitful as very encouraging results have been achieved on both sides and now only need to be implemented and improved by vorteX-io’s R&D teams.
The results of the challenge
Let’s be clear, at the end of an event of this type, no winner is designated, there is nothing to win, no prize, that’s not the purpose here. The outcome is both purely symbolic and honorary for the participants and very concrete regarding the exploitation of the results.
This valuable preparatory work thus enabled vorteX-io to evaluate in a very pragmatic way the validity of the predictive models developed by the data scientists during the hackathon.
It lays the groundwork for the next steps that will soon enable us to further improve the quality of services offered by MAELSTROM®, our platform dedicated to hydrology.
An example of AI-based flood prediction
(time series from the Marmande micro-station)
- Orange line: actual water level measurement
- Blue line: Water level forecast calculated with 6 hours of anticipation (ARIMA model)
The results reached by the 2 teams were fully up to the expectations of each of the stakeholders of this hackathon. The predictions made with the time series of data studied allowed the é teams to predict the occurrence of floods with a very high accuracy between 6h and 24h in advance.
Not surprisingly, these results also highlighted the superiority of models based on measurements from several sensors located upstream of the point of interest. The vision of a constellation of in-situ sensors in real time promoted by vorteX-io is hereby validated once again.
VorteX-io has set itself the ambitious goal of harnessing hydrological data from its own network and from external sources to create river models and their digital twins.
Because water is both precious and devastating, and rivers and streams are its main continental vectors. As such, they must be considered as real assets (in the patrimonial sense of the term) of the territories they irrigate.
Our approach therefore has the dual purpose of facilitating water governance based on meaningful hydrological data and improving the safety of people and property in the many areas subject to natural hazards such as floods and droughts. No doubt that in a few months, the power of Artificial Intelligence will allow us to achieve even more precise results.
In conclusion
Like the huge majority of tools ever discovered, created and used by humans, AI has positive aspects but also darker sides. It remains to be seen whether we will have the wisdom to tip the scales in the right direction.
Acknowledgements and useful links
For specialists (or not), the detailed results of the AI for Earth 2 Challenge are available here:
https://www.fruitpunch.ai/blog/ai-based-early-warning-system-for-river-floods
We take advantage of this article to renew our thanks to :
- Axsysnav
- SUN (Space Hubs Network)
- FruitPunch AI
And to the members of the 2 participating teams.
⬆︎ hackathon : (noun)event during which specialists meet for several days around a collaborative computer programming or digital creation project.
(source: Dictionnaire le Robert)
Subscribe to our newsletter
The latest news about hydrology, vorteX-io, MAELSTROM® and more!
Don't hesitate to subscribe to keep posted.
I subscribe
