General Information

The SBU-Ensemble of conventional numerical models: WRF, MPAS, gSAM, and AI models: PanguWeather and Graphcast, (hereafter the Ensemble), is a multi-model, automated forecasting system administered by the Coastal Meteorology and Atmospheric Prediction (CoMAP) research group based in the School of Marine and Atmospheric Sciences, Stony Brook University. The ensemble produces daily, multi-model weather forecasts of the northeast United States for the purposes of education, research, and accessible weather information for the general public, free of charge.

Weather forecasts are supported by numerical models that apply physical law over vast spatial scales in order to determine future states of the atmosphere. In general, numerical models are inaccessible to the general public as they require substantial computational resources and a moderate knowledge of programming to run, but most are familiar with their end product, namely, local precipitation and severe weather forecasts. Perhaps less well known to the public, is the variation in forecasts that occur with small changes to model settings. To illustrate this, CoMAP ran a decade-long public experiment “SBU-wrf” in which daily forecasts were generated from the Weather Research and Forecasting (WRF) model initialized with either Global Forecast System (GFS) or North American Model (NAM) data, but otherwise identical. The experiment generated comparative plots of meteorological variables that were available online via a user-friendly interface completely free of charge. This resource helped to foster discourse between academia, forecasters, and the general public; raising awareness of the sensitivity of model forecasts to subtle variations in initial conditions and providing accessible meteorological data to all regardless of background or resources.

The SBU-wrf experiment ran on an internal server, aptly named “CHAOS”. This machine was retired in late 2025 and superseded by a more powerful system that inherited its namesake. With these additional computational resources, the CoMAP group sought to expand the SBU-wrf experiment from a dual-configuration intra-model comparison to a multi-model intercomparison that would highlight the variation in forecasts with model choice. Expansion to multiple models will also enable comparisons between conventional numerical models and emerging AI models, which are increasingly of interest due to their low computational cost. The administrators of this second experiment- the “SBU-Ensemble”, endeavour to uphold the principles of SBU-wrf; free and open meteorological data in an accessible portal, with the intention of fostering discussion and improving access to weather forecasts for all.