Numerical model simulation

What might this event have looked like, if we could see it through the modern understanding of severe storms? Unfortunately, we can't go back and take measurements with radars or satellites, but we can try to replicate aspects of the storms using computer models---the same type of models used to forecast the daily weather. Information from the 20th Century Reanalysis shown above can be put into the Weather Research and Forecasting (WRF) model to produce a retrospective simulation of the 4 November 1922 tornado-producing storms. Similar techniques have been used to simulate historic blizzards (\citealt*{Michaelis_2013}) and floods (\citealt{Mahoney_2022}), among other phenomena.
The configuration of the WRF model was generally the same as that used for real-time 4-km forecasts by our research group at CSU, along with a 1.33-km nested grid centered over eastern Colorado. Key parameterization choices include the Morrison 2-moment cloud microphysics parameterization, the Mellor-Yamada-Janjic boundary layer parameterization. The simulations required some trial-and-error with respect to the model initialization time, and it was found that initializing with the ensemble mean analysis at 1800 UTC 3 November 1922, from version v2c of the 20th-century reanalysis, produced a simulation with numerous supercell storms in eastern Colorado, whereas other initialization times showed fewer supercells or different storm modes. So the discussion to follow is based on this most successful model run. Using the full ensemble of analyses from the 20CR in WRF would likely reveal even more interesting insights into this event, but that would require more time and computing resources than are available at the moment.
In this simulation, numerous supercells (rotating thunderstorms) initiate across eastern Colorado and move quickly toward the north and/or northeast (Fig. \ref{447193}). Swaths of updraft helicity, which represent the combined strength of the updraft and rotation in the storm, followed tracks broadly similar to the reported tracks of the tornadoes in the November 1922 event. In addition to long-track storms in southeast and northeast Colorado, several other shorter-lived supercells also developed in the simulation. The longest-track storm in the simulation does follow a path similar to the Sugar City-Genoa tornado, and ends up moving into the Yuma-Phillips County area. This track and timing suggests that it's possible that a single supercell was responsible for both tornadoes, as alluded to by Grazulis, but of course this is impossible to know. One major discrepancy in the model simulation, however, is that the storms form in the evening of the 3rd (generally between 6-11pm local time), rather than on the morning of the 4th as was observed. The storms are also not in the exact locations of the observed storms. But precisely simulating the timing and location of the storm tracks would be unrealistic to expect for a simulation driven by the coarse input of the 20th-Century Reanalysis. These differences notwithstanding, the simulation provides insight into how the storms might have developed and behaved during an event in November 100 years ago.