A Python-Based Tracking Algorithm for Coarse Temporal Resolution WRF-Simulated Supercells

A combination of the Py-ART and TiNT Python libraries are used to track supercells within a high resolution (4 km grid spacing, 50 vertical levels) dynamically-downscaled Weather Research and Forecasting (WRF) dataset containing two 13-year simulations encompassing the entire continental United States. The first simulation represents the modern day (simulating years 2000-2013), while the second simulates the same years using a “business as usual” global warming scenario. The high-resolution nature of this data results in nearly 200 TB of data, emphasizing the need for fast processing. A tailored algorithm is created for this specific dataset to account for the varying temporal resolution of the output variables. Two-dimensional simulated composite reflectively and surface data are available every hour, while three-dimensional variables (including wind, temperature, and other related fields) are available only every three hours. A combination of the hourly simulated reflectivity and three-hourly environmental characteristics are used to infer if an identified cell is indeed a supercell. The three hourly wind and storm motion fields were used for tracking purposes at intermediate hours. Details of how the two radar-focused Python libraries are modified for this algorithm will be described, particularly with regards to ingestion of WRF data into Py-ART and hourly simulated radar data into TiNT. The accuracy of the algorithm is assessed by manually interrogating the simulated data on a random subset of days, as well as manually checking a subset of algorithm-identified supercells. The algorithm will be used to compare differences in modern day and future supercells based on identified storm characteristics, as well as mesoscale inflow and synoptic information. Despite being tuned for this WRF dataset, any gridded simulated radar data is compatible with this algorithm as the design was intended to be generic by nature.