ClimateAi Has U.S. Patent Granted for GenAI-Based Approach Applied to Weather Forecasting

Meredith Mejia • March 21st, 2024.

Patent granted for groundbreaking approach using deep generative models to dramatically improve future climate impact predictions at a local level

San Francisco, March 21, 2024 — ClimateAi, a pioneer in applying artificial intelligence to climate risk modeling, today announced the award of a new U.S. patent covering an innovative machine-learning approach that increases the resolution and accuracy of local weather forecasts, to capture regional extremes in a fraction of the time and computational resources previously required.

This patent, titled “Increasing the accuracy and resolution of weather forecasts using deep generative models,” represents ClimateAi’s seventh patent in the field of weather and climate modeling. It is related to ClimateAi’s peer-reviewed paper, “Increasing the Accuracy and Resolution of Precipitation Forecasts Using Deep Generative Models,” from 2022.

As climate change accelerates — bringing more frequent and severe precipitation, as well as rising temperatures to locations around the world — the challenge lies in accurately predicting regional weather patterns characterized by intense extremes, especially in low-income countries vulnerable to its impacts. Global forecasts can integrate large amounts of data and several weather models, but they often lack precision and are susceptible to errors because minor overlooked details can lead to significant deviations on larger scales. On the other hand, regional forecasting demands costly and time-consuming supercomputers operated by trained local experts, creating accessibility barriers for less affluent nations.

This newly patented ClimateAi system takes a data-driven approach to correcting and downscaling (i.e. increasing the resolution of) global weather/climate forecasts using AI deep generative networks, to correct for biases in current weather models. ClimateAi’s patented approach uses a subset of machine learning known as generative adversarial networks (GANs), trained on coarse global weather forecasts. GANs — the models also used in AI image generators like Midjourney — are able to identify errors in these global projections and subsequently downscale them to a high resolution suitable for local and regional scales. The output local forecasts are accurate up to a few weeks — a key forecasting time period — with comparable high resolution and quality to expensive supercomputers’ regional forecasts.

“This development offers a promising new outlook for forecasting extremes in lower-income countries lacking the means to afford high-resolution local forecasting technology,” ClimateAi CEO Himanshu Gupta says. “This breakthrough enables the provision of precise local forecasts for precipitation and extreme weather, eliminating the conventional constraints imposed by costly forecasting systems to improve lives and livelihoods.”

ClimateAi intends to deploy this technological breakthrough as part of its impact-driven work offering low-cost or free extreme weather forecasting to low-income nations. The GANs generate multiple visuals depicting different potential scenarios with equal probability over various timescales, offering a valuable tool for those in weather-dependent trades and industries on local levels. The patented approach can enable low-income regions to access precise, high-resolution forecasts essential for climate adaptation in agriculture, infrastructure planning, and beyond.

Those seeking further information can reference Patent No. 11,880,767 B2 from the U.S. Patent and Trademark Office.

For any media inquiries, please contact media@climate.ai.

About ClimateAi

ClimateAi is pioneering the application of artificial intelligence to help businesses and governments build climate resilience. By applying AI to climate risk modeling, its ClimateLens platform provides short and long-term insights into weather and climate impact, helping businesses identify the actions needed today to adapt to the climate change disruptions of tomorrow, as well as new opportunities that may arise as a result. Clients include Advanta, Nuveen (a TIAA company), Suntory, and Oatly.