The NOAA Hazardous Weather Testbed 2008 Spring Experiment: Technical and Scientific Challenges of Creating a Data Visualization Environment for Storm-Scale Deterministic and Ensemble Forecasts

The NOAA Hazardous Weather Testbed 2008 Spring Experiment focused on examining the usefulness of high-resolution storm-scale ensemble and deterministic forecasts from various WRF modeling systems for severe weather forecasting. The model data were generated by the National Severe Storms Laboratory, the University of Oklahoma Center for Analysis and Prediction of Storms, the NCEP/Environmental Modeling Center, and the National Center for Atmospheric Research in this highly collaborative experiment. Successfully ingesting and visualizing the data sets from 14 different model runs (including the 10 member ensemble) used in the experiment created interesting challenges, both from a technical point of view in terms of data flow management and storage, and from a scientific point of view in terms of generating images that display the model data in a meaningful but concise methodology. The suite of deterministic models and ensemble forecasts represented over 100 gigabytes of data ingested daily for the experiment, and management of this ingest system included several locally developed custom scripts, designed to execute automatically during the early morning hours as model data was generated. In concert with these ingest scripts, dozens of other programs created images that were later placed on web pages or visualized using internal software packages (e.g., N-AWIPS). In addition, the use of 3rd party software to create model and forecast evaluation surveys was used for the first time (Survey Monkey). This paper will discuss the carefully orchestrated technical infrastructure that helped to create and manage all of the Spring Experiment imagery, and the design of the imagery displays and the innovative techniques used to view the high-resolution model data, to foster scientific analysis and evaluation of model and forecast performance.