Flood Prediction In Tigre, Argentina

People in areas like Tigre, the Paraná delta, and more have a problem: they are victims of frequent flooding.

In some periods, the level of the river is continually rising, and the flooding of the Paraná can coincide with “sudestadas”, which directly affects the delta and leaves it trapped due to the “hydraulic plug effect” of the river. However, what affects them most is the intense private suburbanization. Neighbors usually find out about river flooding through social networks and WhatsApp messages. People go to information and not information to people. This dynamic is not efficient or effective and can leave people on the sidelines.

Our goal is to reverse this. We seek to ensure that all people who may be potentially affected by floods can be aware of them promptly, thus allowing them to plan their days accordingly.

To do this, we constructed a dataset with climatic, tidal, and alert variables from Tigre’s government to predict using Machine Learning models when Tigre will have an alert of the flood. Finally, we developed an application using Streamlit to study which variables are more relevant for flood prediction.

Objectives

===

• Develop a prediction model for river flood alerts in Tigre, Buenos Aires, Argentina.
• Build a dataset with climatic, tidal, and alert variables for Tigre, since there is no such thing as an existing one.
• Develop an application that allows the user to select different Machine Learning models, manipulate variables used in the training of the algorithm, and study their influence when seeing the probability that a flood alert exists or not.

All the documentation and code can be found at: GitHub Repository