GraphCast
A computer designed AI to predict accurate weather predictions.
what GraphCast is.
Graph Cast is a state-of-the-art AI model. It's able to make medium-range weather predictions in up to 10-days in advance. Graph Cast is said to be more accurate than your average weather forecaster. GraphCast can also analyze more serious weather evens, these include extreme temperatures, cyclones and atmospheric rivers.
How it works
GraphCast uses machine learning through Graph Neural Networks (GNNs), trained with decades of weather data collected globally. It uses this data to find patterns in complex weather data, which it can predict weather variables such as wind-speed, temperature, and pressure. It takes over 160 million data points every 6-hours, by breaking the earth’s surface into a 3-dimensional grid each of which measures 5 or more data points.
​
At its simplest GraphCast requires just two sets of data. The state of weather 6 hours ago alongside with the state of the weather as of now. Then, it will use this data to build models of weather changes over a 6-hour period. Scale this with 4 decades of weather data, this provides GraphCast with over 58,000 weather scenarios. That is how weather is predicted.
​
This approaches almost 10-trillion data points, this is more than any human could ever analyze. This is what makes machine learning and AI a valuable tool when dealing with substantial amounts of data.
Where its used.
GraphCast was created by Googles research unit. Its already being used by weather agencies including the European Centre for Medium-Range Weather Forecasts. They are now running a live experiment of out model's forecasts on their website.
Positives and negatives.
positives of GraphCast.
- produces more accurate weather predictions.
- can help to save lives by rolling out with new data every minute.
- extremely reliable.
negatives of GraphCast
- it could take peoples jobs.
- it requires internet to run, leaving places with little to no internet unable to use the AI.
