Breaking the Cap: Using Amazon Web Services and Modern Web Practices in an Operational Weather-forecasting Framework

The advent of cloud computing has enabled an on-demand approach to high-performance computing. Bolt Forecast leverages this power in operations, using Amazon EC2 instances to generate forecast statistics for visualization and delivery using modern web technologies. However, a remaining science challenge lies in the mining and presentation of large weather-forecast data sets (‘big data’), particularly when communicating risk and probability to the consumer. 

We first present a methodology for forecast-data generation using Amazon Web Services (‘the cloud’). Next, we investigate emerging application methods, such as machine learning and principle-component analysis, that reduce big data sets into tractable data products ready for delivery to clients. Finally, we investigate cutting-edge web-visualization methods to best interpret output from ensemble forecast systems for use in personal and private sector applications.