Remote sensing of tornado damage can provide valuable observations for post-event surveys and reconstructions. The tornadoes of 3 March 2019 in the southeastern United States are an ideal opportunity to relate high-resolution satellite imagery of damage with estimated wind speeds from post-event surveys, as well as with the Rankine vortex tornado wind field model. Of the spectral metrics tested, the strongest correlations with survey-estimated wind speeds are found using a Normalized Difference Vegetation Index (NDVI, used as a proxy for vegetation health) difference image and a principal components analysis emphasizing differences in red and blue band reflectance. NDVI-differenced values across the width of the EF-4 Beauregard-Smiths Station, Alabama, tornado path resemble the pattern of maximum ground-relative wind speeds across the width of the Rankine vortex model. Maximum damage sampled using these techniques occurred within 130 m of the tornado vortex center. The findings presented herein establish the utility of widely accessible Sentinel imagery, which is shown to have sufficient spatial resolution to make inferences about the intensity and dynamics of violent tornadoes occurring in vegetated areas.

Tornadogenesis occurs in a variety of storm types, or convective modes, each having a unique climatology and challenges in their detection and warning. Some warnings result in false alarms, meaning no tornado occurred within the warning polygon. We used a mixed-methods approach to assess how convective mode--discrete supercell, cell in cluster, cell in line, or quasi-linear convective system (QLCS)--affects the tornado climatology and National Weather Service (NWS) procedures within three County Warning A reas (CWAs): Memphis (MEG), Nashville (OHX), and Morristown (MRX). We used three data sets: tornadoes (2003-2014) categorized by convective mode, false alarms (2012-2016) categorized by convective mode, and 11 interviews of NWS forecasters. The CWAs had no significant difference in mode frequency when removing replication from multiple-tornado events. However, when outbreaks were included, discrete supercell and QLCS signals were identified in MEG and OHX, respectively. Convective mode, season, and time of day were strongly associated. Tornadic discrete supercells followed a traditional severe weather pattern of spring a nd daytime occurrences, and caused fewer false alarms. More QLCS tornadoes happened at night and in winter. Cells in lines and clusters accounted for larger proportions of events in the false alarm data set than the tornado data set.Forecasters noted challenges in detecting tornadoes in convective modes other than discrete supercells, including short-lived QLCS tornadoes. Key forecaster concerns other than convective mode included storm speed, outbreaks, and lack of ground-truthing at night. Forecasters differed in their motivation to either warn on every tornado or avoid false alarms.

Snowfall presents a hazard to drivers by reducing visibility and increasing safe stopping distances. Some drivers cancel trips if snowfall is occurring or forecast, and traffic volumes often decrease on snowy days. Lake-effect snow is very localized and is thus hypothesized to have a lesser influence on traffic volume than synoptic-scale snow, which usually covers a broader areal extent. We analyze traffic volume in northeast Ohio during 25 snow events and use a matched-pair analysis to determine whether volumes differ between lake-effect and synoptic-scale snowfall in these regions. We also examine the rate at which traffic volume decreases during snow events by time of day and day of week. Results indicate that there is little difference in mean traffic volume decreases when comparing lake-effect and synoptic-scale snow. Hourly trends suggest that traffic volume is most sensitive to snowfall during the midday on weekdays and late afternoon on weekends and least sensitive to snowfall during the overnight hours. Findings presented herein can assist in transportation planning, risk analysis and roadway safety.

No abstract

Tornado watches are issued by the National Weather Service when conditions are favorable for tornado formation. An individual’s response to a tornado watch may affect their ability to seek shelter before a tornado strikes. Here, survey data of Tennessee residents were used to determine common patterns in intended responses to two tornado watch scenarios: one during daytime, and the other at nighttime. Three common patterns were identified for a daytime watch: doing nothing; seeking information using technology; or seeking shelter and praying for safety. The two patterns for a nighttime watch were either to do nothing or to react actively, by seeking further information, shelter, and contacting friends and family. Logistic regressions indicated younger participants, those with prior tornado experience, and those who understood a tornado watch were less likely to intend to seek shelter and pray for safety during the daytime. Older participants and those without strong self-efficacy beliefs were less likely to use technology to find further information. For the nighttime scenario, participants living in East Tennessee and those who believed that bodies of water provide protection from tornadoes were more likely to respond actively, while wealthier participants and those living in single- or multi-family houses were less likely to respond actively. These results show that intended watch response is influenced by many factors, including age, income, self-efficacy beliefs, as well as knowledge of and experience with tornadoes. Additionally, those who do not understand the meaning of a tornado watch may be more likely to seek shelter prematurely.