The weather technology company ClimaCell has launched an historical data archive for the purpose of training AI-based weather models.
The archive known as Weather for AI (WAI) makes available hyper-local historical weather data from around the world.
Weather-sensitive industries and businesses can use the archive to train machine-learning models aimed at providing them with a better understanding of how weather events can impact their business operations.
Until now businesses have mostly relied on data from satellites, radar and weather stations to train their AI models. But according to ClimaCell this data is often not suitable for their purpose, either because it isn’t local enough or because the data source is too big and complex to be of practical value.
In contrast, ClimaCell says its WAI product provides ultra-high resolution datasets going back years, “and can be quickly and fully customized to train AI models by desired location, coverage, resolution, as well as specific weather and air quality parameters”.
The data available in WAI is drawn from a global network of Weather of Things (WoT) virtual sensors – wireless signals, connected cars, airplanes, street cameras, drones, and other Internet of Things (IoT) devices.
Through AI-driven modeling techniques the data is used to create a hyper-local global grid of less than 500m.
“WAI’s hyper-accuracy and unique historical reanalysis of gridded data will catalyze real time, data driven action,” said ClimaCell CEO and co-founder Shimon Elkabetz.
