A high resolution version of operational HWRF: forecast ability and computational challenges presented by an operational environment

In recent years, there has been much interest in using high-resolution hurricane model simulations for hurricane forecasting. Indeed, observational data, analytical theory and some computational work has suggested that forecasts may improve at higher resolution. However, while much study has gone into producing high-resolution research models, little effort has been put into modifying existing operational models for high resolution or modifying high-resolution research models for an operational environment. The operational environment imposes necessary artificial limitations including short time deadlines, specific machine configurations, limited storage space, and complete lack of human intervention. These limitations are typically contrary to the needs of a high-resolution model.

In this study, we analyze a high resolution version of the operational HWRF model, on a cluster with a nearly identical setup as the clusters used for running the operational forecasts. This experiment serves two purposes: we analyze the forecasting abilities of the current operational HWRF when run at high resolution, and we examine computational challenges that lie ahead in bringing this and other high resolution models to an operational environment.