A python program was developed to iteratively ingest observed and modelled upper air soundings. This program systematically scans each vertical profile, identifying the first level at which the lapse rate is positive, cataloging height and temperature values at this level as the starting parameters for the inversion. The program then continues scanning until it finds the first subsequent level at which the lapse rate is negative, establishing the ending parameters of the inversion. A line is then drawn between these two points in a two-dimensional height-temperature space which is referred to as an inversion vector. This process repeats until all inversion vectors within the sounding have been detected. Finally, the height and temperature coordinates are scaled to a similar range and the distance for all the inversion vectors is determined. This distance is then used to identify the primary inversion of interest as the one having the greatest distance within the scaled height-temperature space.
Radiosonde and modelled data were collected for a 6-year time frame from four sites: Rapid City, SD (KUNR); Riverton, WY (KRIW); Glasgow, MT (KGGW); and Denver, CO (KDNR). The dataset consists of 16,860 paired modelled and observed soundings. Preliminary results show the HRRR correctly predicts the existence or absence of temperature inversions between the surface and 5 kilometers above ground level 65.97% of the time, with most of these being surface-based inversions. This presentation will discuss the inversion detection method, the frequency of temperature inversions observed, and discrepancies between HRRR and observed data for inversion starting height, depth, and temperature range.

