The Optical Plume Anomaly Detection (OPAD) program at the NASA Marshall Space Flight Center is using plume spectroscopy for the diagnosis of the Space Shuttle Main Engines. A challenging part of this program is matching high resolution spectral data with a physicist’s model of spectroscopy to produce estimates of metallic erosion through the plume. This paper describes the discovery process used in doing this. The physicist’s model had to be debugged in order to discover the various instrument characteristics, discover critical elements of the data, and in general perform exploratory analysis to understand the instrument and the data it produces. This model gives us strong prior knowledge, however, this needs to be incorporated with care. We had to use a range of statistical techniques in our analysis, including one-dimensional super-resolution to determine the instrument response function. The paper concludes with a discussion of the role of discovery in building intelligent instruments that turn real-time data into useful information.