Top Management College in Kolkata | PGDM College in India Praxis

While both Pangu-Weather and GraphCast are AI-based models, they have different focuses and strengths – and both contribute to the ongoing advancements in weather forecasting through AI technologies

 

A likely turbulence is brewing in the global weather forecasting model that is expected to become a US$4 billion market in the next few years. While Google has developed a system using Generative AI, that can forecast the weather 10 days in advance, Chinese telecom company Huawei has come out with its Pangu-Weather, an AI-based system that uses deep network for fast and accurate numerical weather forecasting with a granularity of one hour. More important than the market size, accurate forecasting has a colossal impact on taking proactive steps to mitigate the economic losses from extreme weather.

To give an indication of the enormity of such losses, Swiss Re, in a report estimated that global insured losses from natural catastrophes in 2023 exceeded US$100 billion for the fourth consecutive year. Last year alone, the US witnessed a staggering $92.9 billion in weather-related damages, making it a record-breaking year for destruction and financial losses. China has revealed that 2024’s floods, droughts, earthquake and severe cold snaps caused direct economic losses of $3.28 billion to the country.

Pangu forecasts by the hour

The Pangu-Weather outperforms NWP (numerical weather prediction) methods in terms of accuracy for medium-range forecasts, offering higher forecast accuracy than NWP methods in all time ranges, from one hour to one week, at a speed of 10,000 times faster. NWP computer models process current weather observations to forecast future weather.

The Chinese system is part of Huawei Cloud’s Pangu series of pre-trained AI models that cover computer vision, natural language processing, multimodal understanding, scientific computing, and weather forecasting. The model uses the spatial resolution 0.25° x 0.25° across 13 pressure levels, with a time granularity of 1 hour, and was trained on a subset of the ERA5 dataset. Huawei Cloud has also developed other AI models for weather forecasting, such as Pangu Meteorology Model, which use deep learning-based methods to improve forecast accuracy.

However, Google’s model GraphCast is claimed to have performed better than the Huawei system in most tests. The model known by its acronym SEEDS (Scalable Ensemble Envelope Diffusion Sampler), a generative AI technology for weather forecast ensemble generation. SEEDS is based on denoising diffusion probabilistic models, a state-of-the-art generative AI method pioneered in part by Google Research.

Google GraphCast forecasts 10 days in advance

Huawei’s Pangu weather forecasting model and Google’s GraphCast are both innovative AI-based models that aim to improve weather forecasting. However, they have different strengths and approaches.Pangu-Weather, developed by Huawei, focuses on daily forecasts starting from ECMWF initial conditions. It is available on GitHub, and its live forecasts are shared to build trust in its performance and support the weather and machine learning communities.

GraphCast outperforms European models

Pangu-Weather is designed to enhance transparency and accessibility, with a software solution that allows the public to access its results with a single command. On the other hand, Google’s GraphCast is a machine-learning global weather model that predicts. weather conditions up to 10 days in advance with remarkable accuracy and speed.

It outperforms the European Centre for Medium-Range Weather Forecasts (ECMWF) model in more than 90% of over 100 verification targets, including hurricane tracks and extreme temperatures. GraphCast can also offer meteorologists accurate warnings, much earlier than standard models, of conditions such as extreme temperatures and the paths of cyclones

Google DeepMind, the developer of GraphCast, emphasises the importance of sharing data, methods, and results to accelerate the progress of science that can ultimately benefit society. By making GraphCast open source, Google DeepMind enables scientists and forecasters around the world to benefit from this advanced AI model.

While both Pangu-Weather and GraphCast are AI-based weather forecasting models, they have different focuses and strengths. Pangu-Weather is designed for daily forecasts and enhanced accessibility, while GraphCast excels in long-term predictions, accuracy, and speed. Both models contribute to the ongoing advancements in weather forecasting through AI technologies.

 

Know more about our Top Ranked PGDM in Management, among the Best Management Diploma in Kolkata and West Bengal, with Digital-Ready PGDM with Super-specialization in Business AnalyticsPGDM with Super-specialization in Banking and Finance, and PGDM with Super-specialization in Marketing.

Leave a Reply

Your email address will not be published. Required fields are marked *