Weather Forecasting: An Application of Artificial Intelligence, Cloud, IOT &...
summary
Operational system fuses ~178 global models with dynamic bias correction to deliver 15-minute updates and hyperlocal forecasts on a 500 m grid, ingesting ~24+ TB/day and serving ~26 billion API calls/day. The IBM GRAF/MPAS model runs globally at 3 km resolution, updated hourly with five-minute outputs, ported to a Power9 + NVIDIA V100 cluster using OpenACC and deployed on 676 GPU nodes with 4 GPUs per node. Results show finer storm structure (Sri Lanka wind stream), aided 400,000 farmers and supported energy and aviation forecasting.
AI-written summary · about
chapters (8)
- 0:00Introduction to session
- 1:48Weather Company overview
- 4:43Data sources for forecasts
- 8:37Data scale and cloud infrastructure
- 12:57AI in weather forecasting
- 15:29IBM Graph Model explained
- 20:33Graph Model benefits and use cases
- 24:51Q&A session
related talks
17mAI Weather Forecasting
28mHow Human Expertise and AI drive Scalable Forecasting Success at
09mGen AI Workflows for Climate Risk Modeling in Food Systems
32mPredicting Grid Faults Before They Happen: AI for Weather-Driven Grid Resilience
35mBuilding a successful inventory planning solution with Amazon Forecast
21mForecasting the Future: An Introduction to Machine Learning for...
20mFoundations for AI+Science
42mRobust Automated Forecasting in Python and R