Here is a summary of the contents of this dataset.

spin-upX   

CM1 model output for the 12-h spin-up period (i.e., prior to warm-bubble insertion) of each of the five simulations initialized with five different random number seeds (X = 1, 2, 3, 4, 5).  Each directory contains the CM1 restart files at t = 12 h used to initialize CM1 at the time of warm-bubble insertion, as well as the CM1 namelist files used to perform the spin-up simulations.  In directory spin-up3, CM1 model output files every 1 h also are included.  The output files are in Grid Analysis and Display System (GrADS) format, and the GrADS descriptor (.ctl) files document the variables contained in each GrADS .dat file.  

stormX, stormX-N, stormX-S, stormX-W, stormX-E

CM1 model output for the 25-member ensemble of supercell simulations that begin with the insertion of a warm bubble after 12 h of spin-up time.  Each X (X = 1, 2, 3, 4, 5) corresponds to a different boundary layer realization obtained by varying the random number seed (the values of X correspond to those in the spin-upX output).  The stormX directory contains CM1 output for the default warm bubble location.  The stormX-N, stormX-S, stormX-W, and stormX-E directories contain simulation output in which the convection-triggering warm bubble is placed, respectively, 2 km north, south, west, and east of the default location.  In the Markowski (2020) article, ensemble members 1-5 are storm1, storm1-N, storm1-S, storm1-W, and storm1-E, respectively; ensemble members 6-10 are storm2, storm2-N, storm2-S, storm2-W, and storm2-E, respectively.  And so on and so forth.  The data are in GrADS format.  The GrADS descriptor (.ctl) files document the variables contained in each GrADS .dat file.  The simulations were carried out for two hours, with output saved every 5 min.  CM1 namelist files also are included in each directory.

smallperts

This directory contains three subdirectories: spin-up1-smallperts, storm1-smallperts, and storm1-E-smallperts.  The spin-up1-smallperts directory contains CM1 output from a 12-h spin-up using the same random number seed used for the simulation in spin-up1, but with the amplitude of the initial random temperature perturbations reduced by 80% (maximum temperature perturbation magnitude was reduced from 0.25 K to 0.05 K).  Output are saved every 1 h, and are in GrADS format.  The storm1-smallperts and storm1-E-smallperts directories contain storm simulations initialized by inserting a warm bubble into the boundary layer spun-up in the spin-up1-smallperts directory.  The warm bubble is shifted 2 km to the east in the simulation in the storm1-E-smallperts directory.

ensemblemean

This directory contains ensemble mean fields, in addition to standard deviations.  The data are in GrADS format.  The GrADS descriptor (.ctl) files document the variables contained in each GrADS .dat file.  

other

This directory contains ensemble means of the nine weakly tornadic (WEAKTOR) storms and nine significantly tornadic (SIGTOR) storms, at t = 95 min and 5 min prior to tornadogenesis, respectively (cm1out_ens_subset_weaktor95min.* and cm1out_ens_subset_sigtorB4TG.*).  The directory also contains ensemble mean fields for the WEAKTOR AND SIGTOR storms at the time of convection initiation (cm1out_ens_subset_weaktorCI.* and cm1out_ens_subset_sigtorCI.*), ensemble mean circulation (circ_mean_weaktor95min.*, circ_mean_sigtorB4TG.*), circulation in each individual WEAKTOR and SIGTOR storm (cm1out_circ_weaktor95min.*, cm1out_circ_sigtorB4TG.*), and smoothed tornado probabilities (cm1out_torP.*).  All of the aforementioned files are in GrADS format.  Lastly, the code subdirectory contains a suite of FORTRAN and GrADS programs used in the analysis and figure production of the Markowski (2020) article.  The getmaxes.f code was used to create Fig. 11.  The computetorprobability.f code was used to compute the smoothed tornado probabilities plotted in Fig. 8b.  The circdist.f code was used to create the circ_mean_weaktor95min.* and circ_mean_sigtorB4TG.* files, which contain the ensemble mean circulation fields shown in Fig. 17.  The computeC-multimembers-singletime.f code was used to create the circulation fields analyzed in Figs. 13 and 15.  The computevariance.f code was used to compute the domain-wide variance plotted in Fig. 19.  The produce-ensmean.f and produce-ensmean-subset.f codes were used to create ensemble means of, respectively, all 25 storms and the WEAKTOR/SIGTOR subsets of storms.  The GraDS scripts named make-figureXX.gs should be fairly self-explanatory.