About the challenge

Large-scale hydrodynamic models generally rely on fixed-resolution spatial grids and model parameters as well as incurring a high computational cost. This limits their ability to forecast flood crests and issue time-critical hazard warnings accurately. In this task, we should build a fast, stable, accurate flood modelling framework that can perform at scales from large scale. Specifically, we will provide the input data and ground truth data in Pakistan flood 2022. Supervised or unsupervised methods based on machine learning methods should be designed to solve the 2-D shallow water equations. Finally, based on this model, a flood forecast model should be achieved in the event of the Pakistan flood in 2022.

Requirements

What to Build

Create a flood modelling framework that is able to solve the shallow water equations in 2D

What to Submit

initial pitch presentation video
code or link to the repository

Hackathon Sponsors

Prizes

€12,300 in prizes

1st: 800€ + Beyond Fellowship

800€ in books, GPU resources, and journal subscriptions, all related to ML and EO

AND

A Beyond Fellowship Scholarship for one student at the Future Lab at the Technical University of Munich for a period of 3 to 6 months and collaborate with us on all possible innovative AI4EO topics. Beyond Fellows will be reimbursed their travel costs from their home organization to Munich, Germany, up to 1,000 Euros and they will receive a monthly stipend of 1,750 Euros.
(The Beyond Fellowship can not be granted to members of the Technical University of Munich)

Judges

Xiaoxiang Zhu
Prof. Dr. / TUM

Zhitong Xiong
Dr. / TUM

Judging Criteria

Idea
Is it a unique approach?
Result
Did the team achieve their goal?
Final-Pitch Performance
How good was the presentation? Were the idea and result comprehensibly presented?

Online	Public
€12,300 in prizes	73 participants

ML4Earth - Physics-aware Machine Learning