Revisiting the 1976 Jordan Iowa F5 tornado: A case of subtle forcing with extreme sensitivity of WRF simulations to initial conditions

The 13 June 1976 Jordan Iowa F5 tornado was a unique event, both because of its intensity, described by Ted Fujita at the time as the most intense he had ever studied, and because the F5 cyclonically-rotating tornado was accompanied for 25 minutes by an anticyclonically-rotating F3 tornado that took a parallel path just a few miles away. Both tornadoes also took unusual cycloidal paths with very sharp changes in direction from primarily northward movements to east or east-southeast approximately three-quarters of the way into their life cycles. The event was also interesting because it happened under relatively weak synoptic forcing with thunderstorm initiation and tornadogenesis both happening relatively early in the afternoon, earlier than normal for summer Iowa tornadoes.

A series of WRF simulations has been performed using NCEP/NCAR (NNRP) reanalysis output to initialize the runs at 12 UTC on 13 June to try to better understand the formation mechanism for the thunderstorm, and to investigate possible sensitivies. Emphasis is placed on a convection-allowing inner domain running with 4 km horizontal grid spacing, nested within 16 km grid spacing, and 64 km grid spacing outer domains. A control run using the MYJ planetary boundary layer scheme and Thompson microphysics failed to produce any convection in central Iowa, instead developing an area of thunderstorms in northeastern Missouri that traveled northeast into far southeastern Iowa and northwestern Illinois. A series of tests using different planetary boundary layer schemes and microphysics schemes also failed to produce storms in central Iowa. Yet, all of these simulations appeared to capture the larger-scale scenario rather well, showing a warm front retreating northwestward, ushering in very warm and moist air into central Iowa. Simulated dew points agreed well with those observed around the time of the observed tornado. The most striking difference from observations was in the direction of the surface winds, which were oriented mostly from the south in the model runs, instead of south-southeast as was observed. Perhaps because of the difference in winds, surface dew points were up to a few degrees C too low in parts of SW IA in the early afternoon. A series of additional tests were performed altering initialized moisture. It was found that a 20% increase in initialized 12 UTC moisture in the lower troposphere over the Great Plains, which advects into SW Iowa leading to a roughly 1 C increase in simulated afternoon 2-meter dew points, results in the formation of thunderstorms that move northeastward across central Iowa, within about 25 km of those observed and with only a few hour time lag. However, a 30% increase from the control moisture in the lower troposphere does not lead to any central Iowa storms. In that test, it appears the increased moisture allows a more organized mesoscale convective complex to travel across southeast Iowa, with compensating subsidence around it likely suppressing central Iowa convection.

The WRF simulations also show very strong cold pools in all thunderstorms developing on this day. The strong cold pools reflect very dry air aloft, and might argue against tornado formation since the simulated storms appeared to be strongly outflow dominant with rapidly expanding regions of outflow having speeds up to 20 m/s. However, a modified sounding for the tornado region valid at 19 UTC (Brown and Knupp, Mon. Wea. Rev., 1980) also suggests extremely dry air was present above the near surface layer, with relative humidities of roughly 15% at 800 mb. More detailed comparisons of simulated near-storm environmental conditions to those observed will be performed, and further testing will be presented examining sensitivity to the structure of the near surface moisture layer, and other meteorological parameters.