top of page
  • Writer's picturedanieleproietto92

The Google AI effectively predicts floods seven days in advance.

Google has recently announced a revolutionary breakthrough in the field of extreme event prediction, particularly floods. Thanks to advanced artificial intelligence models, the company can forecast the onset of floods seven days in advance, providing valuable time for evacuation and securing at-risk communities.

The devastating impact of floods

Floods represent one of the most common and devastating natural disasters worldwide. Each year, they cause financial damages of approximately 50 billion dollars and put the lives of 1.5 billion people at risk. Climate change is exacerbating the situation, with a doubling of the flood-related disaster rate since 2000.

Artificial Intelligence as a Rescue Tool

Google's commitment to researching and developing early warning systems based on artificial intelligence represents a concrete response to this growing threat. The machine learning models developed by the company can analyze large volumes of data, including meteorological, historical, and geographical data, to identify patterns preceding a flood event.

How does Google's new model work?

The dataset includes information from 5,680 watersheds, totaling 152,259 years of data (60GB of disk data). Google's model is based on AI and utilizes LSTM networks to simulate sequences of flow data, employing an encoder-decoder approach to analyze past meteorological data and future forecasts.

The first LSTM network is fed with historical data from the past 356 days at the time of prediction (NASA IMERG, NOAA CPC Global Unified Gauge-Based Analysis of Daily Precipitation, and ECMWF ERA5-land reanalysis), while the second LSTM acquires seven days of forecast data (ECMWF HRES). Geographic and geophysical attributes of the watersheds (HydroATLAS) are integrated into this data.

The model's output is a probabilistic prediction of river flow at a given moment, providing valuable information for flood risk assessment.

A globally applicable model

An important aspect of Google's model is its global applicability. Firstly, the datasets used are public, but the most interesting aspect is that the model can be applied even in areas where there is no or limited flow monitoring network (such as regions with very low GDP). In these countries as well, it will be possible to alert the population in case of danger.

AI as a Defense Tool

Extreme events due to climate change pose a serious threat, but artificial intelligence offers new hopes for risk mitigation. Google's model is an example of how artificial intelligence can be used to save lives and protect communities.

Read more:


bottom of page