‘Superior capabilities’: Chinese AI can make flooding forecast for every river on Earth
- The model overcomes existing forecasting challenges, as it does not need historical records to make predictions
- 95 per cent of small and medium-sized river basins have limited data, but information collected from other parts of the world was used to test the model
In a paper published in peer-reviewed journal, The Innovation, on May 6, the team led by researchers at the Chinese Academy of Sciences (CAS) detailed how their model outperformed others.
“We pre-trained the model on several continental scales using basins with historical monitoring data,” said Ouyang Chaojun, corresponding author and professor at the CAS Institute of Mountain Hazards and Environment. He explained this can make flow predictions within basins lacking streamflow records.
In the paper, the researchers wrote: “Our proposed model achieved state-of-the-art performance in cross-region streamflow forecasting tasks relative to other machine learning models and classic hydrological models.
“Streamflow and flood forecasting remains one of the long-standing challenges in hydrology.”